Auswahl der wissenschaftlichen Literatur zum Thema „Joint coordinates estimation“

To address the problem of low accuracy in joint point estimation in hand pose estimation methods due to the self-similarity of fingers and easy self-obscuration of hand joints, a hand pose estimation method based on coordinate correction and graph convolution is proposed. First, the standard coordinate encoding is improved by generating an unbiased heat map, and the distribution-aware method is used for decoding coordinates to reduce the error in decoding the coordinate encoding of joints. Then, the complex dependency relationship between the joints and the relationship between pixels and joints of the hand are modeled by using graph convolution, and the feature information of the hand joints is enhanced by determining the relationship between the hand joints. Finally, the skeletal constraint loss function is used to impose constraints on the joints, and a natural and undistorted hand skeleton structure is generated. Training tests are conducted on the public gesture interaction dataset STB, and the experimental results show that the method in this paper can reduce errors in hand joint point detection and improve the estimation accuracy.

A new approach for solving the inverse kinematics problem iteratively is presented. The solution is based on the recursive estimation of a kinematics operator which maps the task space coordinates into joint coordinates. The recursive estimation is based on least square approximation. For controlling the robot, the solution (Joint coordinates) must be compensated to achieve an arbitrarily small error (in least square sense) of the desired task space coordinates. The compensation is provided by closed loop feedback of task space coordinates using an optimal control approach.

Trace lines on the outcrop of a rock mass are usually the primary data source for the estimation of rock structure. It is important to obtain the data of trace lines precisely. Photogrammetry is well suited to finish this task. However, this is mainly conducted by commercial software, and not every researcher has easy access to the method of digital photogrammetry. This study aims to provide researchers with a low-cost method of building a photogrammetry-based textured 3D point cloud model (FMBPM) and display the applicability of the method to estimating the rock structure of rock masses. In the FMBPM, a digital single-lens reflex camera with a prime lens and a total station are the necessary hardware employed to capture images and measure the coordinates of feature points. A coordinate transformation means of converting model coordinates to physical coordinates was introduced. A program for calculating a joint orientation based on the coordinates of inflection points on the trace line of the joint was developed. A section of a rock slope was selected as a case to show the procedures and the practicability of the FMBPM. The textured 3D point cloud model of the rock slope was successfully built, and the rock structure of the rock slope was analyzed using the joint disk model generated based on the trace lines extracted from the point cloud model. The results show that: (1) the precision of the point coordinates of the textured 3D point cloud model could achieve 3.96 mm, taking the data of the total station as the reference; (2) the rock structure of the slope is good, according to the value of the rock quality designation; (3) the new method is applicable in engineering practices.

The technology of Three-dimensional human posture estimation is used in sports, motion recognition, and special effects of video media. Among various methods for this, multi-view 3D human pose estimation is essential for precise estimation even in complex real-world environments. But Existing models for multi-view 3D human posture estimation have the disadvantage of high order of time complexity as they use 3D feature maps. This paper proposes a method to extend an existing monocular viewpoint multi-frame model based on Transformer with lower time complexity to 3D human posture estimation for multi-viewpoints. To expand to multi-viewpoints our proposed method first generates an 8-dimensional joint coordinate that connects 2-dimensional joint coordinates for 17 joints at 4-vieiwpoints acquired using the 2-dimensional human posture detector, CPN(Cascaded Pyramid Network). This paper then converts them into 17×32 data with patch embedding, and enters the data into a transformer model, finally. Consequently, the MLP(Multi-Layer Perceptron) block that outputs the 3D-human posture simultaneously updates the 3D human posture estimation for 4-viewpoints at every iteration. Compared to Zheng[5]'s method the number of model parameters of the proposed method was 48.9%, MPJPE(Mean Per Joint Position Error) was reduced by 20.6 mm (43.8%) and the average learning time per epoch was more than 20 times faster.

Workers in material handling tasks often suffer from work-related musculoskeletal disorders (WMSDs) caused by inaccurate work postures or the lifting of excessively heavy loads. Therefore, effective ergonomic assessment of workers is needed to improve worker productivity while reducing the risk of musculoskeletal disorders. This paper proposes a noninvasive method for evaluating posture risks and load analysis in manual material handling tasks. The study focuses on three main aspects: first, using 3D pose recognition technology to extract the 3D coordinates and joint angles of the human body. Second, the REBA method was improved by using fuzzy logic theory to more effectively capture the slow transition features of continuous movement by humans without abruptly altering risk scores, as well as to increase the accuracy and consistency of posture risk evaluation. Third, joint torque and workloads were estimated using biomechanical calculations by integrating pressure insoles and 3D joint coordinate data. Experiments show that this method can effectively evaluate posture risks and workloads in manual material handling tasks, with a correlation coefficient of 0.817 ( p < 0.01 ) between fuzzy logic REBA and REBA and an error rate of 15% in estimating workloads of eight joints. This method can help reduce occupational health risks for workers and industries and improve work efficiency.

Curve running is a common form of training and competition. Conducting research on posture estimation during curve running can provide more accurate training and competition data for athletes. However, due to the unique nature of curve running, traditional posture estimation methods neglect the temporal changes in athlete posture, resulting in a decrease in estimation accuracy. Therefore, a posture estimation method for curve running motion using nano-biosensor and machine learning is proposed. First, the motion parameters of humans are collected by nano-biosensor, and the posture coordinates are obtained preliminarily. Second, the posture coordinates are established according to the human motion parameters, and the curve running posture data is obtained and filtered to obtain more accurate data. Finally, the Bayesian network in machine learning is used to continuously track the posture, and a nonlinear equation is established to fuse the posture angle obtained by the sensor and the posture tracked by the Bayesian network, to realize the posture estimation of curve running motion. The results show that the proposed estimation method has a good motion posture estimation effect, and the hip joint estimation error, knee joint estimation error and ankle joint estimation error are all less than 5°, and the endpoint displacement estimation offset rate is less than 2%. It can realize accurate motion posture estimation of curve running motion, and has important application value in the field of track training.

The article considers the definition of absolute types of deformations, such as transfer and rotation of a building, using linear spatial intersection. In particular, knowing the distances between strong points and deformation marks, the coordinates of the marks located on the building in the initial and current observation cycles are obtained. Knowing the coordinates of the marks, calculate the displacement vectors of the marks for a certain period of time. The definition of absolute deformations is based on the estimation of the projections of the displacement vectors on the coordinate axes, as well as the direction cosines of the vectors with the coordinate axes. The determination of the direction of displacement is shown for the transference deformation. And how to determine the axis, the distance to this axis and the angle of rotation is shown for rotation deformation. We show for joint deformation of the transfer and rotation how divide it into components and then determine the direction of displacement and the axis of rotation. The results can be used in assessing of absolute deformations buildings of monuments of cultural heritage.

The task of angular orientation and stabilization of a space structure during its assembly in orbit is solved. The structure includes elastic elements that are installed during the assembly process. The elastic elements of the structure have no sensors to obtain information about their deformation parameters. Control algorithms are proposed to ensure the stability of the angular motion of the structure. A nonlinear extended Kalman filter is used to obtain the necessary information. A joint estimation algorithm for the coordinates of the angular motion of the considered mechanical system and the coordinates of the elastic vibration tones, as well as an algorithm for the identification of their unobservable parameters are developed. The results of mathematical modeling of a variant of the mechanical system of a space structure are presented, which confirm the operability and efficiency of the developed algorithms for estimating coordinates and parameters.

To estimate range and angle information of multiple targets, FMCW MIMO radars have been exploited with 2D MUSIC algorithms. To improve estimation accuracy, received signals from multiple FMCW MIMO radars are collected at the data fusion center and processed coherently, which increases data communication overhead and implementation complexity. To resolve them, we propose the distributed 2D MUSIC algorithm with coordinate transformation, in which 2D MUSIC algorithm is operated with respect to the reference radar’s coordinate at each radar in a distributed way. Rather than forwarding the raw data of received signal to the fusion center, each radar performs 2D MUSIC with its own received signal in the transformed coordinates. Accordingly, the distributed radars do not need to report all their measured signals to the data fusion center, but they forward their local cost function values of 2D MUSIC for the radar image region of interest. The data fusion center can then estimate the range and angle information of targets jointly from the aggregated cost function. By applying the proposed scheme to the experimentally measured data, its performance is verified in the real environment test.

L’évaluation du risque de troubles musculosquelettiques en milieu industriel représente un défi en raison de la complexité des processus de fabrication modernes. Ces environnements comprennent divers facteurs influençant l’activité des opérateurs, tels que les éléments organisationnels, managériaux et environnementaux, ainsi que le rythme de travail. Il est crucial d’évaluer les contraintes physiques auxquelles sont soumis les opérateurs pour prévenir ces troubles. Bien que de nombreux systèmes monitorent actuellement les mouvements des opérateurs et évaluent les contraintes posturales pour fournir un aperçu de l’activité physique, ils échouent souvent à analyser les forces physiques subies ou générées par l’opérateur. Par conséquent, il est essentiel de quantifier ces forces afin d’identifier les facteurs de risque physique liés à l’effort. Cependant, les méthodes classiques de mesure impliquent souvent des processus complexes, invasifs et peu pratiques en milieu industriel. Cette thèse relève ces défis en évaluant des approches d’apprentissage pour estimer les contraintes physiques sans recourir à des mesures invasives, ce qui est fondamental pour améliorer les outils et les pratiques ergonomiques. Nous avons commencé par comparer la précision et la robustesse des systèmes de mesure basés sur la vision par ordinateur pour l’évaluation du RULA, en nous focalisant particulièrement sur les évaluations ergonomiques sur site. Notre analyse s’est principalement concentrée sur l’évaluation des systèmes basés sur la vision par ordinateur, y compris ceux dotés d’une ou plusieurs caméras, utilisant des images RVB ou des images de profondeur, et les systèmes qui s’appuient uniquement sur des données visuelles ou qui intègrent des capteurs portables (systèmes hybrides). Ensuite, nous avons développé et évalué plusieurs architectures d’apprentissage conçues pour émuler l’étape de la dynamique inverse dans l’analyse du mouvement. Ces dernières prédisent les couples articulaires à partir des données squelettiques de l’opérateur et son poids et la masse de la charge transportée, offrant ainsi une nouvelle alternative aux mé- thodes classiques de dynamique inverse. Enfin, nous avons examiné la généralisabilité des outils basés sur l’apprentissage profond, tels qu’OpenCap, dans les tâches industrielles. En utilisant le fine-tuning - une technique courante dans l’apprentissage profond pour adapter les modèles à de nouveaux ensembles de données avec des échantillons minimaux - nous avons cherché à adapter les modèles d’apprentissage d’OpenCap à un nouveau type de mouvement et à un nouvel ensemble de marqueurs
Assessing the risk of musculoskeletal disorders in industrial environments is a challenging task, given the complexity of modern manufacturing processes. These environments include various factors influencing operator activity, such as organizational, managerial and environmental elements, as well as the pace of work. Assessing the physical constraints to which operators are subjected is crucial to preventing these disorders. Although many systems currently monitor operator movements and assess postural constraints to provide an overview of physical activity, they often fail to analyze the physical forces experienced or generated by the operator. Consequently, it is essential to quantify these forces in order to identify effort-related physical risk factors. However, conventional measurement methods are often complex, invasive and impractical in industrial environments. This thesis addresses these challenges by evaluating learning approaches for estimating physical stresses without resorting to invasive measurements, which is fundamental to improving ergonomic tools and practices. We began by comparing the accuracy and robustness of computer vision-based measurement systems for RULA assessment, focusing particularly on on-site ergonomic evaluations. Our analysis focused primarily on the evaluation of computer vision-based systems, including those with one or more cameras, using RGB or depth images, and systems that rely solely on visual data or incorporate wearable sensors (hybrid systems). Next, we developed and evaluated several learning architectures designed to emulate the inverse dynamics step in motion analysis. These predict joint torques from the operator’s skeletal data and the weight and mass of the load carried, thus offering a new alternative to classical inverse dynamics methods. Finally, we examined the generalizability of deep learningbased tools, such as OpenCap, in industrial tasks. Using fine-tuning - a common technique in deep learning for adapting models to new data sets with minimal samples - we sought to adapt OpenCap’s learning models to a new type of motion and a new set of markers

Abstract This thesis focuses on transceiver optimization for heterogeneous multi-user multiple-input multiple-output (MIMO) wireless communications systems. The aim is to design decentralized beamforming methods with low signaling overhead for improved spatial spectrum utilization. A wide range of transceiver optimization techniques are covered, with particular consideration of decentralized optimization, fast convergence, computational complexity and signaling limitations. The proposed methods are shown to provide improved rate of convergence, when compared to the conventional weighted minimum MSE (WMMSE) approach. This makes them suitable for time-correlated channel conditions, in which the ability to follow the changing channel conditions is essential. Coordinated beamforming under quality of service (QoS) constraints is considered for interfering broadcast channel. Decomposition based decentralized processing approaches are shown to enable the weighted sum rate maximization (WSRMax) in time-correlated channel conditions. Pilot-aided decentralized WSRMax beamformer estimation is studied for coordinated multi-point (CoMP) joint processing (JP). In stream specific estimation (SSE), all effective channels are individually estimated. The beamformers are then constructed from the locally estimated channels. On the other hand, with direct estimation (DE) of the beamformers, only the intended signal needs to be separately estimated and the covariance matrices are implicitly estimated from the received pilot training matrices. This makes the pilot design more robust to pilot contamination. These methods show that CoMP JP is feasible even in relatively fading channel conditions and with limited backhaul capacity by employing decentralized beamformer processing. In the final part of the thesis, a relay-assisted cellular system with decentralized processing is considered, in which users are served either directly by the base stations or via relays for WSRMax or sum power minimization subject to rate constraints. Zero-forcing and coordinated beamforming provide a trade-off between complexity, in-band signaling and spectrum utilization. Relays are shown to be beneficial in many scenarios when the in-band signaling is accounted for. This thesis shows that decentralized downlink MIMO transceiver design with a reasonable computational complexity is feasible in various system architectures even when signaling resources are limited and channel conditions are moderately fast fading
Tiivistelmä Tämä väitöskirja keskittyy lähetin- ja vastaanotinoptimointiin heterogeenisissä monikäyttäjä- ja moniantennijärjestelmissä. Tavoitteena on parantaa tilatason suorituskykyä tutkimalla hajautettuja keilanmuodostusmenetelmiä, joissa ohjaussignaloinnin tarve on alhainen. Erityisesti keskitytään hajautetun keilanmuodostuksen optimointiin, nopeaan konvergenssiin, laskennalliseen kompleksisuuteen sekä signaloinnin rajoitteisiin. Esitettyjen menetelmien osoitetaan parantavan konvergenssinopeutta ja vähentävän signaloinnin tarvetta, verrattaessa tunnettuun WMMSE-menetelmään. Nämä mahdollistavat lähetyksen aikajatkuvissa kanavissa, joissa kanavan muutosten seuraaminen on erityisen tärkeää. Näiden menetelmien osoitetaan mahdollistavan hajautetun ja priorisoidun tiedonsiirtonopeuden maksimoinnin monisolujärjestelmissä sekä aikajatkuvissa kanavissa käyttäjäkohtaisilla siirtonopeustakuilla. Pilottiavusteisten lähetys- ja vastaanotinkeilojen estimointia tutkitaan yhteislähetysjärjestelmissä. Yksittäisten lähetyskanavien estimoinnissa effektiiviset kanavat estimoidaan yksitellen, ja lähetys- ja vastaanotinkovarianssimatriisit muodostetaan summaamalla paikalliset kanavaestimaatit. Suoraestimoinnissa ainoastaan oman käyttäjän effektiivinen kanava estimoimaan erikseen. Tällöin kovarianssimatriisit saadaan suoraan vastaanotetuista pilottisignaaleista. Tämä tekee estimaateista vähemmän herkkiä häiriölle. Hajautetun yhteislähetyksen osoitetaan olevan mahdollista, jopa verrattain nopeasti muuttuvissa kanavissa sekä rajallisella verkkoyhteydellä lähettimien välillä. Viimeisessä osassa tutkitaan välittäjä-avusteisia järjestelmiä, joissa käyttäjiä palvellaan joko suoraan tukiasemasta tai välittäjä-aseman kautta. Optimointikriteereinä käytetään siirtonopeuden maksimointia sekä lähetystehon minimointia siirtonopeustakuilla. Nollaanpakottava sekä koordinoitu keilanmuodostus tarjoavat valinna laskennallisen kompleksisuuden, ohjaussignaloinnin sekä suorituskyvyn välillä. Välittäjä-avusteisen lähetyksen osoitetaan olevan hyödyllisiä useissa tilanteissa, kun radiorajanpinnan yli tapahtuvan signaloinnin tarve otetaan huomioon keilanmuodostuksessa. Tässä väitöskirjassa osoitetaan hajautetun keilanmuodostuksen olevan mahdollista verrattaen vähäisillä laskennallisilla resursseilla heterogeenisissä moniantennijärjestelmissä. Esitetyt menetelmät tarjoavat ratkaisuja järjestelmiin, joissa ohjaussignalointiresurssit ovat rajallisia ja radiokanava on jatkuvasti muuttuva

AbstractAs people live longer, the incidence and severity of health problems increases, placing strain on healthcare systems. There is an urgent need for resource-wise approaches to healthcare. We present a system built using open-source tools that allows health and functional capacity data to be collected remotely. The app records performance on functional tests using the phone’s built-in camera and provides users with immediate feedback. Pose estimation is used to detect the user in the video. The x, y coordinates of key body landmarks are then used to compute further metrics such as joint angles and repetition durations. In a proof-of-concept study, we collected data from 13 patients who had recently undergone knee ligament or knee replacement surgery. Patients performed the sit-to-stand test twice, with an average difference in test duration of 1.12 s (range: 1.16–3.2 s). Y-coordinate locations allowed us to automatically identify repetition start and end times, while x, y coordinates were used to compute joint angles, a common rehabilitation outcome variable. Mean difference in repetition duration was 0.1 s (range: −0.4–0.4 s) between trials 1 and 2. Bland-Altman plots confirmed general test-retest consistency within participants. We present a mobile app that enables functional tests to be performed remotely and without supervision. We also demonstrate real-world feasibility, including the ability to automate the entire process, from testing to analysis and the provision of real-time feedback. This approach is scalable, and could form part of national health strategies, allowing healthcare providers to minimise the need for in-person appointments whilst yielding cost savings.

AbstractBuilding on the two rounds of exposure studies with human milk coordinated by the World Health Organization (WHO) in the mid-1980s and 1990s on polychlorinated biphenyls (PCB), polychlorinated dibenzo-p-dioxins (PCDD), and polychlorinated dibenzofurans (PCDF), five expanded studies on persistent organic pollutants (POPs) were performed between 2000 and 2019. After the adoption of the Stockholm Convention on POPs (the Convention) in 2001, WHO and the United Nations Environment Programme (UNEP) collaborated in joint studies starting in 2004. The collaboration aimed at provision of POPs data for human milk as a core matrix under the Global Monitoring Plan (GMP) to assess the effectiveness of the Convention as required under Article 16. Over time, the number of analytes in the studies expanded from the initial 12 POPs targeted by the Convention for elimination or reduction to the 30 POPs covered under the Stockholm Convention and two other POPs proposed for listing as of 2019. Many of these chemicals have numerous congeners, homologous groups, isomeric forms, and transformation products, which significantly extends the number of recommended analytes.In the studies between 2000 and 2019, 82 countries from all five United Nations regions participated, of which 50 countries participated in more than one study. For the human milk samples of the 2016–2019 period, results are available for the full set of 32 POPs of interest for the Convention until 2019: (i) the 26 POPs listed by the start of the study in 2016; (ii) decabromodiphenyl ether [BDE-209] and short-chain chlorinated paraffins [SCCP] as listed in 2017; (3) dicofol and perfluorooctanoic acid [PFOA] as listed in 2019; (4) medium-chain chlorinated paraffins [MCCP] and perfluorohexane sulfonic acid [PFHxS] as proposed for listing. This is a unique characteristic among the core matrices under the GMP.Four key messages can be derived: These studies are an efficient and effective tool with global coverage as key contributor to the GMP. After collection of a large number of individual samples (usually 50) fulfilling protocol criteria, pooled samples are prepared using equal aliquots of individual samples (physical averaging) and are considered to be representative for a country, subregion or subpopulation at the time of the sampling. The analysis of pooled representative human milk samples by dedicated Reference Laboratories meeting rigorous quality criteria contributes to reliability and comparability and reduces uncertainty of the analytical results. Additionally, this concept is very cost-effective. These studies can be used for regional differentiation based on concentrations of individual POPs between and within the five UN Regional Groups (African Group, Asia-Pacific Group, Eastern European Group, Group of Latin American and Caribbean Countries; Western European and Others Group). For some POPs, a wide range of concentrations with up to three orders of magnitude between lower and upper concentrations was found, even for countries in the same UN region. Some countries had levels within the usual range for most POPs, but high concentrations for certain POPs. Findings of concentrations in the upper third of the frequency distribution may motivate targeted follow-up studies rather than if the observed level of a POP is found in the lower third of frequency distribution. However, the concentration of a POP has also to be seen in context of the sampling period and the history and pattern of use of the POPs in each country. Therefore, results are not intended for ranking of individual countries but rather to distinguish broader patterns. These studies can provide an assessment of time trends, as possible sources of variation were minimized by the survey concepts building on two factors (sampling design; analysis of the pooled samples by dedicated Reference Laboratories). The estimation of time trends based on comparison of median or mean concentrations in UN Regional Groups over the five surveys in five equal four-year periods between 2000 and 2019 provides a first orientation. However, the variation of the number of countries participating in a UN Regional Group in a certain period can influence the median or mean concentrations. Thus, it is more prudent to only use results of countries with repeated participation in these studies for drawing conclusions on temporal trends. The reduction rates in countries should be seen in context with the concentration range: A differentiation of high levels and those in the range of the background contamination is meaningful. If high levels are found, sources might be detected which could be eliminated. This can lead to significant decrease rates over the following years. However, if low background levels are reported, no specific sources can be detected. Other factors for exposure, e.g. the contamination of feed and food by air via long-range transport and subsequent bioaccumulation, cannot be influenced locally. However, only very few time points from most individual countries for most POPs of interest are available, which prevents the derivation of statistically significant temporal trends in these cases. Yet, the existing data can indicate decreasing or increasing tendencies in POP concentrations in these countries. Furthermore, pooling of data in regions allows to derive statistically significant time trends in the UN Regional Groups and globally. Global overall time trends using the data from countries with repeated participation were calculated by the Theil–Sen method. Regarding the median levels of the five UN Regional Groups, a decrease per 10 years by 58% was found for DDT, by 84% for beta-HCH, by 57% for HCB, by 32% for PBDE, by 48% for PFOS, by 70% for PCB, and by 48% for PCDD and PCDF (expressed as toxic equivalents). In contrast, the concentrations of chlorinated paraffins (CP) as “emerging POPs” showed increasing tendencies in some UN Regional Groups. On a global level, a statistically significant increase of total CP (total CP content including SCCP [listed in the Convention in 2017] and MCCP [proposed to be listed]) concentrations in human milk of 30% over 10 years was found. The studies can provide the basis for discussion of the relative importance (“ranking”) of the quantitative occurrence of POPs. This, however, requires a differentiation between two subgroups of lipophilic substances ([i] dioxin-like compounds, to be determined in the pg/g [=ng/kg] range, and [ii] non-dioxin-like chlorinated and brominated POPs, to be determined in the ng/g [=μg/kg] range; both groups reported on lipid base) and the more polar perfluorinated alkyl substances (PFAS); reported on product base [as pg/g fresh weight] or on volume base [ng/L]. For this purpose, results for the complete set of the 32 POPs of interest for the 2016–2019 period were considered. By far, the highest concentrations of lipophilic substances were found for DDT (expressed as “DDT complex”: sum of all detected analytes, calculated as DDT; maximum: 7100 ng/g lipid; median: 125 ng/g lipid) and for chlorinated paraffins (total CP content; maximum: 700 total CP/g lipid; median: 116 ng total CP/g lipid). PCB was next in the ranking and had on average an order of magnitude lower concentrations than the average of the total CP concentrations. The high CP concentrations were caused predominantly by MCCP. If the pooled samples from mothers without any known major contamination source nearby showed a high level of CP, some individual samples (e.g. from local population close to emission sources, as a result of exposure to consumer products or from the domestic environment) might even have significantly higher levels. The lactational intake of SCCP and MCCP of the breastfed infant in the microgram scale resulting from the mothers’ dietary and environmental background exposure should therefore motivate targeted follow-up studies and further measures to reduce exposure (including in the case of MCCP, regulatory efforts, e.g. restriction in products). Further, due to observed levels, targeted research should look at the balance among potential adverse effects against positive health aspects for the breastfed infants for three groups of POPs (dioxin-like compounds; non-dioxin-like chlorinated and brominated POPs; PFAS) regarding potentially needed updates of the WHO guidance. As an overall conclusion, the seven rounds of WHO/UNEP human milk exposure studies are the largest global survey on human tissues with a harmonized protocol spanning over the longest time period and carried out in a uniform format. Thus, these rounds are an effective tool to obtain reliable and comparable data sets on this core matrix and a key contributor to the GMP. A comprehensive set of global data covering all POPs targeted by the Stockholm Convention, in all UN Regional Groups, and timelines covering a span of up to three decades allows to evaluate data from various perspectives. A widened three-dimensional view is necessary to discuss results and can be performed using the three pillars for assessments of the comprehensive data set, namely: analytes of interest; regional aspects; time trends. This can identify possible problems for future targeted studies and interventions at the country, regional, or global level. Long-term trends give an indication of the effectiveness of measures to eliminate or reduce specific POPs. The consideration of countries with repeated participation in these studies provides the best possible database for the evaluation of temporal trends. The continuation of these exposure studies is important for securing sufficient data for reliable time trend assessments in the future. Therefore, it is highly recommended to continue this monitoring effort, particularly for POPs that are of public health concern.

3D human pose estimation is a vital task in computer vision, involving the prediction of human joint positions from images or videos to reconstruct a skeleton of a human in three-dimensional space. This technology is pivotal in various fields, including animation, security, human-computer interaction, and automotive safety, where it promotes both technological progress and enhanced human well-being. The advent of deep learning significantly advances the performance of 3D pose estimation by incorporating temporal information for predicting the spatial positions of human joints. However, traditional methods often fall short as they primarily focus on the spatial coordinates of joints and overlook the orientation and rotation of the connecting bones, which are crucial for a comprehensive understanding of human pose in 3D space. To address these limitations, we introduce Quater-GCN (Q-GCN), a directed graph convolutional network tailored to enhance pose estimation by orientation. Q-GCN excels by not only capturing the spatial dependencies among node joints through their coordinates but also integrating the dynamic context of bone rotations in 2D space. This approach enables a more sophisticated representation of human poses by also regressing the orientation of each bone in 3D space, moving beyond mere coordinate prediction. Furthermore, we complement our model with a semi-supervised training strategy that leverages unlabeled data, addressing the challenge of limited orientation ground truth data. Through comprehensive evaluations, Q-GCN has demonstrated outstanding performance against current state-of-the-art methods. The full version of this paper, along with the code and data, can be found at [39].

Recognition of human behavior is critical in video monitoring, human-computer interaction, video comprehension, and virtual reality. The key problem with behaviour recognition in video surveillance is the high degree of variation between and within subjects. Numerous studies have suggested background-insensitive skeleton-based as the proven detection technique. The present state-of-the-art approaches to skeleton-based action recognition rely primarily on Recurrent Neural Networks (RNN) and Convolution Neural Networks (CNN). Both methods take dynamic human skeleton as the input to the network. We chose to handle skeleton data differently, relying solely on its skeleton joint coordinates as the input. The skeleton joints’ positions are defined in (x, y) coordinates. In this paper, we investigated the incorporation of the Neural Oblivious Decision Ensemble (NODE) into our proposed action classifier network. The skeleton is extracted using a pose estimation technique based on the Residual Network (ResNet). It extracts the 2D skeleton of 18 joints for each detected body. The joint coordinates of the skeleton are stored in a table in the form of rows and columns. Each row represents the position of the joints. The structured data are fed into NODE for label prediction. With the proposed network, we obtain 97.5% accuracy on RealWorld (HAR) dataset. Experimental results show that the proposed network outperforms one the state-of-the-art approaches by 1.3%. In conclusion, NODE is a promising deep learning technique for structured data analysis as compared to its machine learning counterparts such as the GBDT packages; Catboost, and XGBoost.

In order to solve the problem that the spatial structure and temporal dynamic structure of skeletal data are not clearly and fully utilized when using hand bone data for action recognition, a spatio-temporal synchronous graph convolution network with combined attention mechanism is designed. Using the method of 2D estimation and triangulation, the feature is projected into a single 3D volume, the 3D heat map is output, and the 3D joint coordinates are obtained by soft-argmax operation on the heat map. The spatial dimension and the temporal dimension of the bone data are separated, the spatial dimension is encoded according to the order of the related nodes, and the same related nodes are encoded in the temporal dimension, and the spatial embedding matrix and the temporal embedding matrix are obtained. The matrix is synchronously added to the spatio-temporal network sequence. The experiment is tested on the SHREC2017 data set, and compared with some representative gesture recognition methods, the results show that the algorithm has achieved good results in hand action recognition.

AbstractThe actual problem of creation of gravity maps which can be used as correctors in map-matching- navigation systems (MMNS) is considered. It is assumed to use information from two sources: available gravity maps and measurement data of various parameters of the Earth’s gravity field (EGF) (acceleration of gravity, deflection of the vertical, horizontal gradients of the potential).Obtaining the required amount of measurement data of a specified accuracy is considered as a problem of planning primary measurements. At the first stage of its solution, the number of measurements necessary to achieve the required accuracy of the parameters of the gravitational map is minimized. It takes into account the accuracy and performance of measuring devices. The second stage assumes that the human operator specifies the position of the points at which to measure the parameters of the EGF with a known accuracy.Joint processing of information is carried out on the basis of the collocation method. The estimation of EGF navigation quality is considered in a specified point and its nearest neighborhood. This approach is reasonable, if in the future it is required to optimally form the trajectory of the MMNS in a specified area or to a specified destination point. It is assumed rectilinear motion of the object from this point with a known initial level of coordinate errors. Position errors are defined as functions of distance from the initial point and direction of movement.

This study developed a Computer-Vision based anatomical joint coordinates estimation model to predict the 3D joint coordinates with the help of Artificial Intelligence image recognition technology during manual lifting tasks based on single camera video inputs. The workflow of the proposed Computer-Vision model includes 2D joint detection and 3D joint reconstruction. The 3D joint error is calculated based on the Euclidean distance between the predicted 3D joint coordinates from the CV-based method and the corresponding joint coordinates of the ground truth from the Visual 3D TM skeletal model. The results indicated that the floor to shoulder lifting height path induced a greater 3D joint error than the floor to knuckle and knuckle to shoulder lifting height paths (p-value = 0.01). The 3D joint error of the hand was the largest than the other estimated joints. This study verified that the proposed Computer-Vision model could predict 3D joint points. Therefore, while the marker-based motion tracking system is inapplicable, the model can be used as an alternative solution for predicting lifting motion.

Various measurement techniques and data processing are applied to determine point displacements and deformation of geodetic networks or buildings. Considering classical measurements and analysis of the network deformation, we should realize that the measurements are not “immediate.” The question arises: what happens if a point (or some points) displaces between particular measurements within one epoch. In such a case, the observation set would consist of the observations before and after point displacement, and such hypothetical observation groups can be regarded as related to two (or more) pseudo epochs. The paper's main objective is to examine some estimation methods that would probably deal with such a problem, namely Msplit estimation (in two variants, the squared and the absolute Msplit estimation) and chosen robust methods, namely Huber’s method (example M-estimation) and the Hodges-Lehmann weighted estimation (basic R-estimation). The first approach can provide two (or more) variants of the network point coordinates (here, before and after point movements), providing information about two (or more) states of the network during measurements. In contrast, the robust methods can only decrease the influence of the outliers on the computed network point coordinates. Thus, estimation results would concern only one network state in such a case. The presented empirical analyses show that the better and more realistic results are obtained by applying Msplit estimation. Huber’s method can also provide acceptable results (describing the network state at the epoch beginning) only if the number of observations conducted after the point displacements is not too high.

The detection of deformations on man-made structures such as bridges and dams are an essential task in engineering geodesy. The classical method uncovering deformations is based on geodetic networks using measurements from total stations or GNSS receivers. Another new approach is the determination of deformations based on terrestrial laser scans leading to large-scale deformation results by point cloud comparisons. In the field of geodetic engineering, mobile laser scanning systems are increasingly used for high-resolution point cloud generation in short measurement times, which leads to the idea to use these for deformation analysis. A crucial part of this measurement strategy is the estimation of the trajectory (position and orientation) of the scanner, which allows a consistent registration of the single scan lines in a global coordinate system (georeferencing). The largest limitation to the accuracy of the resulting point cloud is the accuracy of the estimated trajectory. In most applications, the estimation of position and orientation are based on the fusion of GNSS (Global Navigation Satellite System) and IMU (Inertial Measurement Unit) measurements. Systematic errors, as they often appear in GNSS measurements, are directly transferred to the georeferenced point cloud and therefor limit the potential for deformation analysis. With this paper we address the questions, if the trajectory estimation can be improved by the integration of known landmarks into the trajectory estimation procedure. Using a point cloud generated with an initial GNSS/IMU based trajectory, landmark targets can be observed in the point cloud and integrated into an updated estimate, using a factor graph-based approach. For the evaluation of a potential accuracy increase due to landmark observations, we performed measurements, comparing GNSS/IMU based results with the ones where landmarks are additionally integrated. The experiments show, that the accuracy increases especially in the heading angle, which is reflected in lower residuals to observed reference coordinates, but also in the trajectory covariances of the estimation results.

Abstract This paper presents the application of a new multivariate extreme value model for the estimation of metocean variables. The model requires fewer assumptions about the forms of the margins and dependence structure compared to existing approaches, and provides a flexible and rigorous framework for modelling multi-variate extremes. The method involves a transformation of variables to polar coordinates. The tail of the radial variable is then modelled using the generalised Pareto distribution, with parameters conditional on angle, providing a natural extension of uni-variate theory to multivariate problems. The resulting model is referred to as the semi-parametric angular-radial (SPAR) model. We consider the estimation of the joint distributions of (1) wave height and wave period, and (2) wave height and wind speed. We show that the SPAR model provides a good fit to the observations in terms of both the marginal distributions and dependence structures. The use of the SPAR model for estimating long-term extreme responses of offshore structures is discussed, using some simple response functions for floating structures and an offshore wind turbine with monopile foundation. We show that the SPAR model is able to accurately reproduce response distributions, and provides a realistic quantification of uncertainty.

The challenges of learning a robust 6D pose function lie in 1) severe occlusion and 2) systematic noises in depth images. Inspired by the success of point-pair features, the goal of this paper is to recover the 6D pose of an object instance segmented from RGB-D images by locally matching pairs of oriented points between the model and camera space. To this end, we propose a novel Bi-directional Correspondence Mapping Network (BiCo-Net) to first generate point clouds guided by a typical pose regression, which can thus incorporate pose-sensitive information to optimize generation of local coordinates and their normal vectors. As pose predictions via geometric computation only rely on one single pair of local oriented points, our BiCo-Net can achieve robustness against sparse and occluded point clouds. An ensemble of redundant pose predictions from locally matching and direct pose regression further refines final pose output against noisy observations. Experimental results on three popularly benchmarking datasets can verify that our method can achieve state-of-the-art performance, especially for the more challenging severe occluded scenes. Source codes are available at https://github.com/Gorilla-Lab-SCUT/BiCo-Net.

This paper presents a method for estimating the absolute lock effect in bolted joints during actual machine operation. The absolute evaluation on self-loosening is indispensable for the prediction of the life of locking devices (parts)/method at actual machine operation. The loosening phenomenon of bolted joint can be viewed as the decrease tendency of the initial axial tension or residual axial tension. According to the author’s previous report, there is a good linear relation on logarithmic coordinates (log-log paper) between the decrease of axial tension (measured tension/initial axial tension) and number of operations (or working time or mileage) since last tightening. Based on this relation, and observation of loosening in the early stages of product development, we estimate the rate of axial tension decrease (degree of self-loosening) accurately after prolonged operation since tightening by measuring the initial axial tension behavior by using the proposed regression formula. The trial calculation of the residual axial tension after prolonged operation is made using the industrial vehicle (stacking capacity 16-ton large-sized forklift-truck) at customer’s site by loosening measurement. According to the estimation using only a few day’s data, it is presumed that initial axial tension is kept 55% or more in case of anaerobic adhesive tightening and almost 30% in case of lubrication tightening after ten years. In the working load type loosening, authors propose the L-N diagram (Loosening Life - Numbers of Cycles to Loosening N Diagram) for loosening life estimation like the S-N diagram for fatigue life estimation. Using the loosening frequency diagram and L-N diagram, the degree of loosening damage for every axial tension’s residual ratio is figured out to analyze the loosening life of the locking device/method by Miner’s rule and cumulative loosening damage models.

Excavators are one of the commonly used types of hydraulic machines in earth moving operations. The material handled is often transferred by dump trucks having a payload capacity that should not be exceeded. Payload monitoring systems are needed in order to prevent the possible problems during the delivery, increase the work efficiency, reduce the cost, and obtain the product information automatically without the requirement of truck scales. In this study, we propose a novel approach to estimate the load weight in the bucket of the excavator when the machine links are in motion. We consider the excavator as a three-revolute joint manipulator in vertical plane with the boom, the stick, and the bucket links. We rewrite the dynamic torque equations in a decoupled form as the linear combination of dynamic parameters and functions of joint angles, velocities, and accelerations. We perform least squares estimation to identify these parameters allowing us to predict the no load joint torques for any configuration of the links. We show that the most accurate torque prediction is the difference between the boom torque and the stick torque. We then derive the relation between the joint torques with and without the load, which are functions of the dynamic parameters. Using these equations, we can estimate the load weight. The relation becomes simpler when the links are stationary, since only the gravitational parameters remain present in the torque equations. The relation in dynamic case requires the parameters of the polar coordinates for the center of gravity of the bucket and we show that these parameters can be estimated with the knowledge of the empty bucket mass. We summarize our findings on load weight estimation for different cases including stationary poses and dynamic trajectories on free space and discuss the results. Although the friction is neglected throughout the modeling, the results obtained indicate that the effect of the static friction plays an important role in the accuracy of the estimated payload mass. We show that our dynamic model based solution is very promising, and exhibit only 2% error for high enough velocities.

The sleeping posture is crucial determinant of infant growth and development. Sleeping in the prone position is associated with a higher risk of sudden infant death syndrome. Therefore, medical recommendations advocate placing infants in the supine position during sleep. Furthermore, certain medical conditions, such as cranial deformity, hip dislocation, and torticollis, may manifest as head-turn preferences, wherein infants consistently face a specific direction, either right or left. Detecting and addressing these sleeping postures are critical for preventing accidental infant deaths during sleep and identifying potential underlying health issues. In this study, we present an automatic method for classifying infant sleeping postures into four categories: supine, prone, right lateral, and left lateral, using only videos. Although various methods exist for classifying sleeping postures during infancy, such as those involving acceleration and pressure seat sensors, they often require physical attachments that may cause discomfort to the infants. To address this limitation, we present a contactless approach that employs video images recorded using an infrared camera. The camera was positioned to record the entire infant bedding area without imposing restrictions on the installation angle. We analyzed the video data collected from the home of each participant and classified the sleeping postures of the participants into four categories. Subsequently, the classification accuracy was calculated for each night. The participants of the experiment were two infants under one year of age. To evaluate data accuracy, we excluded instances of data involving individuals other than the participants and data outside the field of view of the camera. “Vision Pose,” a skeleton estimation software capable of detecting joint points in images, was employed for body position analysis. Specifically, we extracted the two-dimensional coordinates of eight joint points: both shoulders, both elbows, both hips, shoulder center, and hip center. We classified the infant sleeping postures by measuring the distance between these joint points. A linear support vector machine was applied to the features, and classification was conducted in two steps. In the initial step, the sleep data were categorized into two groups: supine or prone and right lateral or left lateral. Subsequently, each of these categories was further divided into two classifications, yielding four types of sleeping postures. Our proposed model demonstrates an impressive average accuracy of 92.3% in estimating the four sleeping postures: supine, prone, right lateral, and left lateral. Our study establishes the feasibility of non-contact sleeping posture classification using an infrared camera. This approach holds promising potential for real-life home environments and childcare facilities, where continuous monitoring of infant sleeping postures can significantly contribute to promoting safe sleep practices and early identification of potential health concerns.

Coastal areas share unique intersections of large-scale climate variability and local hydrology, wetland, benthic and pelagic ecosystems, and anthropogenic pressures. Forecasting of harmful environmental conditions for planning, adaptation, and mitigation purposes is both complex and urgently needed. Ecological forecasting is the qualitative or quantitative projection of biogeochemical, organismal or ecosystem state variables and their drivers on timescales that can range from “now” to decades from now. Estimating hypoxia in Chesapeake Bay today, predicting acidity conditions in the Northeast Pacific in a few months, or projecting the depth of the Bering Sea nutricline in 2075 are all ecological forecasts relevant to planning, adaptation, and mitigation efforts. In 2022, the US CLIVAR and Ocean Carbon & Biogeochemistry (OCB) Programs convened a joint workshop to advance the development of US ecological forecasting. The workshop goals were to 1) identify sources of predictability of physical quantities relevant for marine ecosystems along US coastlines; 2) assess the observational needs of forecast systems and limitations due to gaps in understanding; and 3) promote the development of dynamical and statistical models suitable to meet the forecasting requirements. About 80 participants from over 40 US and international institutions joined this hybrid workshop for plenary talks and breakout discussions. Participants represented a diversity of career stages across academic institutions, government agencies, and non-government organizations. By working together, they collectively identified a path forward for a coordinated US ecological forecasting effort as detailed in this report.