Dissertations / Theses on the topic 'Forward detection systems'

The process of quantifying the robustness of a given nonlinear system is not necessarily trivial. If the dynamics of the system in question are not sufficiently involved, then a tight estimate of a bound on system performance may be obtained. As the dynamics of the system concerned become more and more involved, it is often found that using the results existing in the literature provides a very conservative bound on system performance. Therefore, the motivation for this work is to develop a general method to obtain a less conservative estimate of a bound on system performance, compared to the results already available in literature. The scope of this work is limited to two dimensions at present. Note that working in a two dimensional space does not necessarily make the objective easily achievable. This is because quantifying the robustness of a general nonlinear system perturbed by disturbances can very easily become intractable, even on a space with dimension as low as two. The primary contribution of this work is a computational algorithm, the points generated by which are conjectured to lie on the boundary of the smallest robust forward invariant set for a given nonlinear system. A well known path-planning algorithm, available in existing literature, is leveraged to make the algorithm developed computationally efficient. If the system dynamics are not accurately known, then the above computed approximation of an invariant set may cease to be invariant over the given finite time interval for which the computed set is expected to be invariant. Therefore, the secondary contribution of this work is an algorithm monitoring a computed approximation of an invariant set. It is shown that for a certain type of systems, this secondary monitoring algorithm can be used to detect that a computed approximation of an invariant set is about to cease to be invariant, even if the primary algorithm computed the set based on an unsophisticated dynamical model of a system under consideration. The work related to computing approximations of invariant sets is tested mainly with the curve tracking problem in two dimensions. The algorithm monitoring whether a computed approximation of an invariant set is about to cease to be invariant is inspired by work related to detecting Lithium-ion (Li-ion) battery terminal voltage collapse detection.

Wireless relay can be utilized to extend signal coverage, achieve spatial diversity by user cooperation, or shield mobile terminals from adverse channel conditions over the direct link. In a two-hop multi-input-multi-output (MIMO) amplify-and-forward (AF) relay system, the overall noise at the destination station (DS) consists of the colored noise forwarded from the relay station (RS) and the local white noise. We propose blind noise correlation estimation at the DS by utilizing statistics of the broadband relay channel over the RS-DS hop, which effectively improves signal detection at the DS. For further performance improvement, we also propose to estimate the two cascaded MIMO relay channels over the source-RS and the RS-DS links at the DS based on the overall channel between the source and the DS and the amplifying matrix applied at the RS. To cancel cross-talk interference at a channel-reuse-relay-station (CRRS), we utilize the random forwarded signals of the CRRS as equivalent pilots for local coupling channel estimation and achieve a much higher post signal-to-interference ratio (SIR) than the conventional dedicated pilots assisted cancellers without causing any in-band interference at the DS. When an OFDM-based RS is deployed on a high-speed train to shield mobile terminals from the high Doppler frequency over the direct link, inter-subchannel interference (ICI) mitigation is required at the RS. By utilizing statistics of the channel between the base station and the train, we develop both full-rate and reduced-rate OFDM transmission with inherent ICI self-cancellation via transmit and/or receive preprocessing, which achieve significant performance improvement over the existing ICI self-cancellation schemes.

Keeping a safe distance from the forward-leading vehicle is an essential feature of modern Advanced Driver Assistant Systems (ADAS), especially for transportation companies with a fleet of trucks. We propose in this thesis a Forward Collision Warning (FCW) system, which collects visual information using smartphones attached for instance to the windshield of a vehicle. The basic idea is to detect the forward-leading vehicle and estimate its distance from the vehicle. Given the limited resources of computation and memory of mobile devices, the main challenge of this work is running CNN-based object detectors at real-time without hurting the performance. In this thesis, we analyze the bounding boxes distribution of the vehicles, then propose an efficient and customized deep neural network for forward-leading vehicle detection. We apply a detection-tracking scheme to increase the frame rate of vehicle detection and maintain good performance. Then we propose a simple leading vehicle distance estimation approach for monocular cameras. With the techniques above, we build an FCW system that has low computation and memory requirements that are suitable for mobile devices. Our FCW system has 49% less allocated memory, 7.5% higher frame rate, and 21% less battery consumption speed than popular deep object detectors. A sample video is available at https://youtu.be/-ptvfabBZWA.

The use of autonomous robots in our society is increasing every day and a robot is no longer seen as a tool but as a team member. The robots are now working side by side with us and provide assistance during dangerous operations where humans otherwise are at risk. This development has in turn increased the need of robots with more human-awareness. Therefore, this master thesis aims at contributing to the enhancement of human-aware robotics. Specifically, we are investigating the possibilities of equipping autonomous robots with the capability of assessing and detecting activities in human teams. This capability could, for instance, be used in the robot's reasoning and planning components to create better plans that ultimately would result in improved human-robot teamwork performance. we propose to improve existing teamwork activity recognizers by adding intangible features, such as stress, motivation and focus, originating from human behavior models. Hidden markov models have earlier been proven very efficient for activity recognition and have therefore been utilized in this work as a method for classification of behaviors. In order for a robot to provide effective assistance to a human team it must not only consider spatio-temporal parameters for team members but also the psychological.To assess psychological parameters this master thesis suggests to use the body signals of team members. Body signals such as heart rate and skin conductance. Combined with the body signals we investigate the possibility of using System Dynamics models to interpret the current psychological states of the human team members, thus enhancing the human-awareness of a robot.
Användningen av autonoma robotar i vårt samhälle ökar varje dag och en robot ses inte längre som ett verktyg utan som en gruppmedlem. Robotarna arbetar nu sida vid sida med oss och ger oss stöd under farliga arbeten där människor annars är utsatta för risker. Denna utveckling har i sin tur ökat behovet av robotar med mer människo-medvetenhet. Därför är målet med detta examensarbete att bidra till en stärkt människo-medvetenhet hos robotar. Specifikt undersöker vi möjligheterna att utrusta autonoma robotar med förmågan att bedöma och upptäcka olika beteenden hos mänskliga lag. Denna förmåga skulle till exempel kunna användas i robotens resonemang och planering för att ta beslut och i sin tur förbättra samarbetet mellan människa och robot. Vi föreslår att förbättra befintliga aktivitetsidentifierare genom att tillföra förmågan att tolka immateriella beteenden hos människan, såsom stress, motivation och fokus. Att kunna urskilja lagaktiviteter inom ett mänskligt lag är grundläggande för en robot som ska vara till stöd för laget. Dolda markovmodeller har tidigare visat sig vara mycket effektiva för just aktivitetsidentifiering och har därför använts i detta arbete. För att en robot ska kunna ha möjlighet att ge ett effektivt stöd till ett mänskligtlag måste den inte bara ta hänsyn till rumsliga parametrar hos lagmedlemmarna utan även de psykologiska. För att tyda psykologiska parametrar hos människor förespråkar denna masteravhandling utnyttjandet av mänskliga kroppssignaler. Signaler så som hjärtfrekvens och hudkonduktans. Kombinerat med kroppenssignalerar påvisar vi möjligheten att använda systemdynamiksmodeller för att tolka immateriella beteenden, vilket i sin tur kan stärka människo-medvetenheten hos en robot.

The thesis work was conducted in Stockholm, Kista at the department of Informatics and Aero System at Swedish Defence Research Agency.

碩士
國立中央大學
資訊工程學系
104
There were lots of deaths caused by traffic accidents of rear-end collision, advanced driver assistance systems (ADASs) has become an important research topic in recent years. To prevent these fatalities, forward collision warning (FCW) systems have been proposed to protect drivers from the danger due to paying no attention to forward traffic situations. Not only FCW systems but also many other ADASs such as lane departure warning (LDW), blind spot detection (BSD), pedestrian collision warning (PCW), etc. have been developed to assist drivers. In this thesis, we present a weather-adaptive forward collision warning system, which would help drivers to avoid collisions to the preceding vehicles or obstacles. In the proposed FCW system, the local features such as horizontal and vertical edges are first calculated. Then edge maps are bi-leveled using a learning thresholding method to adapt the intensity variation of captured images, so that the extraction of edge points is less influenced by bad weather conditions. Third, the preserved edge points are used to generate possible objects . Fourth, the objects are selected based on edge response, location, and symmetry of object candidates to generate vehicle candidates. Three candidate generation schemes are hierarchically designed to extract vehicle candidates in various weather conditions. At last, a method based on principal component analysis (PCA) is proposed to verify the vehicle candidates. PCA is a technique used to extract the important features of a set of vehicle images. Each extracted feature describes a characteristic of vehicle appearance which is defined as a global feature. Depending on the extracted features, a candidate region can be decomposed and reconstructed. The similarity between the original regions and the reconstructed regions are measured to verify the vehicle candidates. Theoretically, PCA method is used to remove the non-vehicle candidates to reduce the false alarm. The proposed FCW system has been test and evaluated on various weather conditions. The average accuracies of the proposed FCW system in clear and bad weather conditions are 98.5% and 71.8%, respectively. In our experiment, the system execution speed of approximately 50 frames per second and camera captured 30 frames per second, so the system can achieve real-time vehicle detection.

碩士
國立臺灣科技大學
電機工程系
102
Advanced driver assist systems (ADAS) are technologies that provide the driver with essential information when driving and lead to an increase in safe driving. On a long stretch of highway, the lane departure and forward vehicle collision are very important issues for drivers. The proposed ADAS focuses on implementing lane tracking, lane departure detection and forward vehicle detection by using the image processing techniques. In this thesis, our algorithms can provide the following basic information: (1) Lane detection, (2) Lane tracking, (3) Forward vehicle detection. In lane detection, our algorithms first combine the edge distribution function (EDF) with lane marking filter to improve the effects of non-lane marking noise and then detect the lane marking correctly. In lane tracking, the proposed method can predict the region of interest (ROI) of lane marking by the properties of continuous video in the next frame. Then, ROI is examined to track lane marking by Hough transform. In addition, no matter whether a vehicle is changing lane or passing exit-ramp, our algorithms can keep track of the lane marking smoothing. In forward vehicle detection, the position of the forward vehicle inside its own lane is determined by the features of rear shapes of vehicle, such as shadow, horizontal edge, vertical edge, symmetry and image density. Through video streams with resolution of 640×360 pixels, the experimental results achieve the average detection rate of lane tracking at 97.00% with the average processing time 5.4 ms per frame and the average detection rate of forward vehicle at 84.86% with average processing time 2 ms per frame.

The scope of this research concerns pedestrian active safety. Several primary safety systems have been developed for vehicles in order to detect a pedestrian and to avoid an impact. These systems analyse the forward path of the vehicle through the processing of images from sensors. If a pedestrian is identified on the vehicle trajectory, these systems employ emergency braking and some systems may potentially employ emergency steering. Methods for assessing the effectiveness of these systems have been developed. But, it appears difficult to determine the relevance of these systems in terms of pedestrian protection. The general objective of this research was to test the response of these systems in many accident configurations. The first step consisted of gathering a sample of a hundred of accidents involving vehicles with pedestrians. These accidents were provided from accident databases of two laboratories LMA and CASR. Data of these accidents were recorded in sufficient detail from in-depth investigation which enables reconstructing the trajectory of the vehicle and pedestrian prior to the collision. The second step was to analyse qualitatively and quantitatively the data of the selected accidents. These accidents were reconstructed to simulate the pre-crash conditions. From this accident reconstruction, factors relevant to the primary safety of pedestrians were deduced. The next step consisted of coupling the vehicle dynamic behaviour with a primary safety system in order to confront these systems to real accident configurations. The potential of these systems is studied by verifying the feasibility of deploying an autonomous emergency manoeuvre during the timeline of the accident and according to the vehicle dynamic capabilities: i.e. verifying the possibilities in terms of crash avoidance. Based on this procedure, three modelling methods were developed: a first method testing a system to each accident configuration and two others using graphs of evaluation from a parametric study realised on a generic system. The results of the three methods were then discussed. Finally, as a perspective, the last study will approach crash mitigation. As a consequence of an active safety system response, the vehicle impact speed is reduced. The effect of speed reduction on variations in impact conditions will be then addressed to measure the potential safety impact of these systems on pedestrian protection.
Thesis (Ph.D.) -- University of Adelaide, School of Mechanical Engineering, 2016.

碩士
國立臺灣師範大學
資訊工程學系
104
Vision-based driver assistance systems and its related technologies were started to pay attention and develop from about 20 years ago. Visual analyzing the road situation in front of vehicles through camera to assist drivers. Vision-based driver assistance systems for automobile has been gradually consummated. In contrast, vision-based driver assistance systems for motorcycle went unheeded. The quantity of motorcycle and automobile increases year by year, and the quantity of motorcycle is fifty thousand more than automobile per year. Summarizes the above situation causes that automobile traffic accident rate reduces year after year, but motorcycle traffic accident rate rises every year. Daytime forward vehicle detection technology has been matured by degrees, but there is not so much researchers developing and researching at nighttime. By literatures in recent years of nighttime forward vehicle detection technology, many researches confirm the location of vehicle through related technologies about taillight detection. Therefore this study will use taillight detection to confirm the location of vehicle. Because it has to do a forward vehicle deceleration detection, forward vehicle decelerates or not will be determined by the brake-lights activates or not. When the motorcycle turns a corner, this study will adjust Region of Interest (ROI). There will not be traffic accidents with the forward vehicle when the vehicle stop moving as the red traffic light shows. So it hasn’t to do a taillight detection and brake-light detection. Therefore our system needs to detect forward vehicle move or not. The shape of taillight in the recent years is not only traditional circle but also irregular shape or elongated shape, and therefore this study will aim at the characteristic of surrounding light source around the taillight to do a taillight detection. This study will use illumination and threshold to determine brake-light on or off, and this dynamic threshold according to the distance between taillight and camera. The illumination of some activated brake-lights is lower than our determined threshold, and some non-brake of taillights are higher than it. It will lead to failure of brake-light detection. So our system will adjust threshold specifically to increase the accuracy rate of brake-light detection. Our system experiments on sunny day, rainy day, in the tunnel, and on many kinds of roads. The experiment result shows that it will get the higher accuracy rate without considering the consequence of taillight detection with the reflection of red lights. And our system expects that the reflection of red lights can be filter in the future to increase the accuracy rate of taillight detection. In this study, if it doesn’t consider raindrop dripping on the camera lens on rainy day, the accuracy rate of brake-light detection is about 90%

碩士
國立交通大學
電機與控制工程系所
98
In This thesis the technique of image processing and computer vision theorem are applied to lane detection and vehicle detection. In the meantime, the algorithm is also applied on TI-DM642 for driving assistance system (DAS). The system has been working in different environment such as expressway, urban area and tunnel. Furthermore the algorithm is such robust to be verified with all weather condition like sunny day, cloudy day, evening, morning, night and rainy day. In the lane detection, CCD camera is used to grab the front view, and then the algorithm detects the lane making to contribute a real lane model. This model is applied to estimate and narrow the searching area in order to increase the accuracy and reduce the computation. The lane detection system has been verified successfully in expressway and urban road. Moreover, the system has been equipped on Taiwan iTS-1 (the first intelligence car of Taiwan) as vision system and combines with control system to accomplish lane keeping and lane change. The vehicle detection uses the lane model to build basic vehicle parameters and sets the ROI from the result of lane detection. Algorithm will select multiple possible objects those have strong vehicle characteristics. After feature extraction, the vehicles will be verified with the classify rules. The thesis considers kinds of the vehicle features in different weathers condition and overcomes lots of influence from environment like the text on the road, the strong shadows and light, and the reflected light from the road surface. Even with rainfall and windscreen wiper, the system works successfully. The detection has no relationship with the day or night, so it has good performance at smooth light changing or sudden light changing. With the lane model from lane detection, the algorithm could establish the multiple lane boundaries and finish the multiple forward car detection. At same time, the distance of the forward vehicle can be calculated, and the systems will warn driver that he is in the dangerous distance.

碩士
國立東華大學
電機工程學系
101
Based on the inertia lane marking and tracking, this study presents a fast forward vehicle distance warning system at daylight and night environment, which utilizes CCD camera to capture the moving image and detect the lane marking. Following the result of inertia lane detection, forward vehicle could be detected in the region of lane-marking of road image. The mechanism of forward vehicle detection is divided by daylight and night time. In the daylight time, the bottom shadow of forward vehicle is regard as a major feature. Following YCbCr color model, a suitable region of interest could be segmented to detect the location of low luminance of object and recognize as forward vehicle. In the night time, the rear light and high brightness object is the major feature of forward vehicle. Following RGB color model, a suitable region of interest for high brightness object could be recognized as the location of forward vehicle. Finally, the identified location of forward vehicle can calculate actual distance between the host and the forward vehicle by slope approximation method. Ten-meter is regards as a reminder that the distance reaches alerts purpose. In this thesis, the proposed algorithms has been implemented in TI DM648 SoC platform and has successfully tested with very good results.

碩士
國立東華大學
電機工程學系
106
The aim of this thesis is to develop the forward vehicle distance detection system. This system can help the driver maintain a safe distance from the forward vehicle and reduce the accidents. This study proposes a novel regression analysis method based on least Wilcoxon instead of traditional weighted least squares method. The least Wilcoxon could reduce interference of imprecise vanishing point and noise to make lane marking more precise and robust. The distance table is established and updated following the building rule of road dash line to measure the distance of forwarding vehicle in the captured image. Comparing the results of the simulation experiments, it showed the regression analysis with least Wilcoxon is closer to the actual road line than that with weighted least squares. The distance table is built by the algorithm could also accurately measure the distance of the forward vehicle and judge the safety for collision avoidance.

碩士
華梵大學
機電工程學系博碩專班
100
During the recent several years, machine vision has played an important role in various bound and it has been widely applied in general industry, medical industry, and home appliances etc. with the advancement of technology level and the improvement on the hardware, machine vision gradually can be applied to vehicle auxiliary safety system. To detect forward vehicle in all-weather condition, the system apply to machine vision principle and image processing technology. And using color image CCD or CMOS camera to capture the image of forward vehicle image and then perform image processing. And use algorithm to detect object. And used digital signal processing device TI-DM642 or notebook to analyze, to process and to compare its efficiency. First, the system through the lane model to establish the basic parameters to set the ROI(Region Of Interest) to find the location of the vehicle candidate object, and then through feature extraction method to distinguish true vehicle or not to calculate the distance between vehicles. In order to overcome the influence of all kinds of weather conditions such as: strong sunshine, shadows, road marking , light reflection on the road, road water, rain and windshield wipers, etc., the combination of conditions on the various features is added on the system. So it obtain very high adaptability of the system for a variety of environments. Therefore, not only on the highways, but also in urban areas, tunnels, sunny day, cloudy day, rainy day, even in the evening, night scenes, the system may be operated to achieve the effect of vehicle auxiliary to enhance traffic safety goals.

碩士
國立臺灣科技大學
電機工程系
107
Advanced driver assist system is a technology that provides drivers with essential information when driving and lead to avoid traffic accidents. Forward vehicle braking or changing lane out of a sudden are the most common dangers which lead to collision accidents. To prevent collision accident, we propose “a real-time forward vehicle detection and signal recognition system in all-weather situations”. We extract forward vehicle by images of dashboard camera, identify signals of vehicle such as braking and turning, and then display them on screen of dashboard camera to notify driver for safety. In this thesis, our system consists of six main parts: (1) Lighting condition recognition, (2) Vehicle candidate extraction, (3) Vehicle verification, (4) Vehicle tracking, (5) Taillight detection and (6) Signal recognition. Vehicle candidate is detected by vehicle characteristics in different illumination environments. In vehicle verification, we use local binary pattern as feature, training our classifier to verify the vehicle with AdaBoost algorithm. To make signal recognition stable we use object tracking algorithm to support vehicle detection. In signal recognition, we record history information of vehicle lights and use its flicking frequency as feature to train classifier, identify vehicle turning by checking whether lights are flickering or not, and then detect third brake light of vehicle to confirm braking. The experimental results show that our system is real-time, and is works in various weather condition. The accuracy of vehicle detection and signal recognition rates achieve 93.1% and 89.4%, which show the stability of our system.

碩士
國立交通大學
電子工程學系 電子研究所
104
This thesis proposes the design and verification of one popular Advanced Driver Assistance System (ADAS) function: Forward Collision Warning System (FCWS). The proposed method apply vehicle detection and feature tracking to each frame in the videos. Feature tracking, which is adopted to conquer the effect of some critical case like non-ideal shadow, raindrop and windshield wiper, can increase the stability of vehicle localization. The proposed system can be applied to the application of single-lane FCWS, which includes main lane and three-lane FCWS which includes left lane, main lane and right lane. In order to be more suitable for the application of FCWS, the proposed system also calculate the distance between located vehicles and camera. The proposed algorithm is implemented on the personal computer. The input video of the system is captured from driving recorders in D1 resolution (720x480). The performance of the proposed single-lane FCWS can achieve 327 fps and that of the proposed three-lane FCWS can achieve 183 fps in average. The test results are compared with ground truth frame by frame. Detection rate of the proposed single-lane FCWS can reach 96.37% at daytime and 87.64% at night. Moreover, detection rate of the proposed three-lane FCWS can reach 94.05% at daytime and 86.90% at night. The proposed system is also implemented on Freescale i.MX6 with a USB webcam to capture the video. Under the D1 (720x480) resolution, the performance of the proposed 3-lane FCWS can achieve 29 fps.