Distance Estimation from Stereo Vision: Review and Results

Total Page:16

File Type:pdf, Size:1020Kb

Distance Estimation from Stereo Vision: Review and Results DISTANCE ESTIMATION FROM STEREO VISION: REVIEW AND RESULTS A Project Presented to the faculty of the Department of the Computer Engineering California State University, Sacramento Submitted in partial satisfaction of the requirements for the degree of MASTER OF SCIENCE in Computer Engineering by Sarmad Khalooq Yaseen SPRING 2019 © 2019 Sarmad Khalooq Yaseen ALL RIGHTS RESERVED ii DISTANCE ESTIMATION FROM STEREO VISION: REVIEW AND RESULTS A Project by Sarmad Khalooq Yaseen Approved by: __________________________________, Committee Chair Dr. Fethi Belkhouche __________________________________, Second Reader Dr. Preetham B. Kumar ____________________________ Date iii Student: Sarmad Khalooq Yaseen I certify that this student has met the requirements for format contained in the University format manual, and that this project is suitable for shelving in the Library and credit is to be awarded for the project. __________________________, Graduate Coordinator ___________________ Dr. Behnam S. Arad Date Department of Computer Engineering iv Abstract of DISTANCE ESTIMATION FROM STEREO VISION: REVIEW AND RESULTS by Sarmad Khalooq Yaseen Stereo vision is the one of the major researched domains of computer vision, and it can be used for different applications, among them, extraction of the depth of a scene. This project provides a review of stereo vision and matching algorithms, used to solve the correspondence problem [22]. A stereo vision system consists of two cameras with parallel optical axes which are separated from each other by a small distance. The system is used to produce two stereoscopic pictures of a given object. Distance estimation between the object and the cameras depends on two factors, the distance between the positions of the object in the pictures and the focal lengths of the cameras [37]. In this project, the internal parameters of the cameras are calculated after taking several pictures from both left and right cameras. The correspondence problem between left and right images is solved using MATLAB. The disparity is then estimated based on the center of the image plane and finally, the distance of the object is calculated using the epipolar triangulation method. The results of this project showed that the distance estimated using stereo vision to ten different objects is relatively accurate. v Two tests are completed in this project. The first one is done for checking the error between manual and stereo vision results, where the error ranged from 0.09% to 1.4%, the second test is done for comparison between passive and active stereo vision methods, this project used the passive method and the average error was equal to 0.78% which is less than the other methods. _______________________, Committee Chair Dr. Fethi Belkhouche _______________________ Date vi ACKNOWLEDGEMENTS Firstly, I would like to extend my appreciation and love to my respective family for all their continuing support, care and consideration. I would like to express our sincere gratitude to Dr Fethi Belkhouche for accepting to be my project coordinator. We are grateful for the help and advice, for without which we would have not progressed that far. I would also like to extend my sincere thanks to Dr. Preetham B. Kumar, Dr. Behnam S. Arad and CPE department. vii TABLE OF CONTENTS Page Acknowledgements……………………………………………………………………. vii List of tables………………………………………………………………………….…xi List of figures …………………………………………………………………………. xii Chapter 1. INTRODUCTION. ……………………..………………………………………….. 1 1.1 Project Background.……………………………………………………..1 1.2 Stereoscopic Measurement Method……………………………………...1 1.3 Stereo Vision Disparity……………………………………….………..…3 1.4 Stereo Vision Corresponding Problem…………………………………..5 1.5 Objectives and Scope Of The Project.……………………………… . ….5 2. CAMERA PARAMETERS………………………………………………………..... 7 2.1 Camera Model and Parameters………………………………………….7 2.1.1 Intrinsic Parameters ……………….………………………7 2.1.2 Extrinsic Parameters.……………………………………..11 2.2 Camera Calibration………………………………………………...…. 13 2.3 Camera Calibration Techniques………………………………………..14 2.4 Lens Distortion …………………………………………………………15 2.5 Epipolar Geometry ……………………………………………………..15 2.6 Image Rectification.…………………………………………………….17 viii . STEREO VISION 3.1 Introduction …………………………………………………………….19 3.2 A Classification of Stereo Vision Algorithms.……………...…………20 3.3 Stereo Vision Flowchart………………………………………………...22 3.4 Depth Recovering and Disparity…………………………………………24 3.5 Stereo Matching Strategies…………………………………………….25 3.5.1 Local Dense Stereo Matching Methods……………….……..27 3.5.1.1 Matching Costs.…………………………………..….27 3.5.1.2 Aggregation Area.……………………………………28 3.5.2 Global Dense Stereo Matching Methods………………….......29 3.5.2.1 Horizontal Disparity Smoothness Assumption. ……30 3.5.2.2 Horizontal and Vertical Disparity Smoothness …….32 3.6 3D Reconstruction via Stereo Vision………………………………….32 4. CALCULATIONS AND RESULTS ………………………………………………..34 4.1 Introduction …………………………………………………………….34 4.2 Tools ……………………………………………………………………..35 4.3 Hardware Setup ………………………………………………………......36 4.4 Results ……………………………………………………………………38 4.4.1 Distance Measurement ……………………………………...…38 4.4.2 Comparison between Passive and Active Stereo Vision Camera…41 4.4.3 Result of Right Camera …………………………………………..43 4.4.4 Result of Left Camera ……………………………..…………..…44 ix 4.4.5 Result Of Both Camera ……………………………………...…..45 5. CONCLUSION AND FUTURE WORK……………………………………….…...47 References …………………………………………………………………………….49 x LIST OF TABLES Tables Page 1. Object in Left and Right Images with Disparity and Distances of The Object for Ten Tests………………………………………………………40 2. Comparison Between This Project and Past Projects ………….……..…42 xi LIST OF FIGURES Figures Page 1. Pinhole Camera Projection (A) Single Camera (B) Stereo Camera……………..2 2. Epipolar Geometry……………………….……………………………………...4 3. Illustration of A Pinhole Camera Example ………………………………………8 4. The Central-Projection Model ………………………………………………….9 5. Skew Effect ……………………………………………………………………11 6. Camera Coordinate System Vs Arbitrary Coordinate System ………………….12 7. The Calibration Patterns. (A) Left Images; (B) Right Images ………………...14 8. Epipolar Geometry Between A Pair Of Images ………..……………………….16 9. A Rectified Stereo Pair………………………………………………………..…18 10. Stereo Vision System Flow Chart ………………………………………………23 11. Stereo Triangulation And Geometry……………………………………………..25 12. Stereo Based on Dynamic Programming ……………………………….…….…31 13. Using Measurement Tools To Measure Dimensions For The Object ………..…35 14. Using iPhone Camera for Calibration…. ……………………………………...36 15. The Ten Objects Used in This Project ……………………………………….…37 16. Camera capture (a) left image camera (b) right image camera ……………….…38 17. (a) Relation between rp & xz (b) relation between cp & xz from right camera…43 18. (a) Relation between rp & xz (b) relation between cp & xz from left camera…..44 19. (a) Relation between rp & xz (b) relation between cp & xz from both camera ..45 xii 1 Chapter One Introduction 1.1 Project Background Different strategies are used to find the distance of an object using machine vision. Traditionally, laser domain finder and ultrasonic measurements are widely used. These methods are no longer viable in applications like industrial automation and navigation because they require continuous human monitoring even though they have excellent accuracy. We can evaluate the object distance and its geometry using stereo vision. This project starts by taking stereo images from different scenes. Experiments are conducted, and the outcomes based on the implementation of certain algorithms are discussed [1]. Stereo vision is suitable for many applications for numerous reasons. The most important one is the similarity to binocular vision in humans and animals. Stereo vision consists of two identical cameras, which allow to obtain pictures from two distinctive viewpoints. Depth is determined through finding the disparity in the images of the same 3D points [2]. 1.2 Stereoscopic Measurement Method Stereoscopy is an approach used to represent (3D) scenes from (2D) images. By using two images taken at slightly distinct locations, it can create a phantasm of depth. In 1838, stereoscopic images and viewing devices was invented by Charles Wheatstone. It is viewed as a passive technique if no additional source of light is applied on the scene. For 2 example, some applications require laser beam. There are many applications where passive stereo vision is used such as robotic navigation [3]. Figure 1.1 Pinhole Camera Projection (A) Single Camera (B) Stereo Camera [4] In figure 1.1(a), a single camera has the same image points for all the points in the same projection line. Two points (P and Q) project to the same image point (p ≡q). This is true for each point along the ray OP. By using triangulation and finding the corresponding (homologous) points from the two pictures, we can determine the depth as shown in figure 1.1(b) [4]. There are many challenges for using stereo vision to perform distance measurement. Processing tasks 3 need to be implemented twice, this led to duplicating the effort for stereo vision such as image enhancement and filtering tasks. To overcome this issue, parallel architectures can be used to increase the performance. Many techniques have been proposed to improve performance and power consumption [5]. 1.3 Stereo Vision Disparity The first step in the stereo vision process is to find the disparities between two pictures produced by using two cameras positioned at different locations. This problem is related to 3D reconstruction which will be discussed
Recommended publications
  • A Jackal UGV 79 A.1 Main Characteristics
    POLITECNICO DI TORINO Master of Science in Mechatronic Engineering Master’s Degree Thesis GPS-based autonomous navigation of unmanned ground vehicles in precision agriculture applications Supervisor Candidate Prof. Marcello CHIABERGE Simone CERRATO October 2020 Abstract The global population is growing exponentially and the actual agricultural tech- niques and resources will not be able to feed every person on the Earth in a few years. To account for this serious problem, groups of research are focusing their attention on precision agriculture, because it looks for the improvement of the productivity and efficiency of both agricultural and farming production processes, while reducing the environmental impact, exploiting automation and robotics. The thesis aims to design and develop a solution, based on GPS, for the autonomous navigation problem in precision agriculture, using only few sensors: an Inertial Measurement Unit, a GPS receiver and a depth camera, in order to be cost effec- tive. The proposed goal has been achieved through a system of inter-operating sub-components, that have to share information and collaborate each other in order to provide a complete autonomous navigation. In particular, the main involved entities are: a localization filter, a global and a local path planning algorithms and an obstacle avoidance approach, that have been developed and can cooperate each other by means of the Robot Operating System. Eventually, the proposed solution has been tested in a simulation environment, through different possible scenarios providing good results in each of them. However, it may be considered as a starting point for future improvement in the field of autonomous navigation for precision agriculture.
    [Show full text]
  • Scalable Multi-View Stereo Camera Array for Real World Real-Time Image Capture and Three-Dimensional Displays
    Scalable Multi-view Stereo Camera Array for Real World Real-Time Image Capture and Three-Dimensional Displays Samuel L. Hill B.S. Imaging and Photographic Technology Rochester Institute of Technology, 2000 M.S. Optical Sciences University of Arizona, 2002 Submitted to the Program in Media Arts and Sciences, School of Architecture and Planning in Partial Fulfillment of the Requirements for the Degree of Master of Science in Media Arts and Sciences at the Massachusetts Institute of Technology June 2004 © 2004 Massachusetts Institute of Technology. All Rights Reserved. Signature of Author:<_- Samuel L. Hill Program irlg edia Arts and Sciences May 2004 Certified by: / Dr. V. Michael Bove Jr. Principal Research Scientist Program in Media Arts and Sciences ZA Thesis Supervisor Accepted by: Andrew Lippman Chairperson Department Committee on Graduate Students MASSACHUSETTS INSTITUTE OF TECHNOLOGY Program in Media Arts and Sciences JUN 172 ROTCH LIBRARIES Scalable Multi-view Stereo Camera Array for Real World Real-Time Image Capture and Three-Dimensional Displays Samuel L. Hill Submitted to the Program in Media Arts and Sciences School of Architecture and Planning on May 7, 2004 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Media Arts and Sciences Abstract The number of three-dimensional displays available is escalating and yet the capturing devices for multiple view content are focused on either single camera precision rigs that are limited to stationary objects or the use of synthetically created animations. In this work we will use the existence of inexpensive digital CMOS cameras to explore a multi- image capture paradigm and the gathering of real world real-time data of active and static scenes.
    [Show full text]
  • Optimised Implementation of Dense Stereo Correspondence for Resource Limited Hardware
    Journal of Computer Science Original Research Paper Optimised Implementation of Dense Stereo Correspondence for Resource Limited Hardware 1,2 Deepambika Vadekkettathu Anirudhan and 3,4 Malangai Abdul Rahiman 1Department of ECE, Karpagam Academy of Higher Education, Coimbatore, Tamilnadu, India 2Department of ECE, LBS Institute of Technology for Women, Trivandrum, Kerala, India 3Karpagam Academy of Higher Education, Coimbatore, Tamilnadu, India 4Pro-Vice Chancellor, APJ Abdul Kalam Technological University, Trivandrum, Kerala, India Article history Abstract: Computer stereo vision is a passive sensing technique that Received: 21-06-2018 helps to recover 3D information of an environment from 2D images. Revised: 16-07-2018 The stereo correspondence is a challenging task that finds out matching Accepted: 17-10-2018 pixels between the stereo image pair based on Lambertian criteria and its result is a disparity space image. The depth of the objects from the Corresponding Author: Deepambika V.A. camera can be calculated from this disparity value by using the principle Department of ECE, LBS of triangulation. For the vision guided robot navigation, the requirement Institute of Technology for of stereo matching algorithms on low power dedicated hardware that Women, Trivandrum, Kerala, can achieve a high frame rate is unambiguous. A new, highly optimized India implementation of correlation based, Sum of Absolute Differences Email: [email protected] correspondences algorithm on a low cost resource limited FPGA is presented here. This System-on-Programmable-Chip architecture based system achieved a higher frame rate of 50 fps with 64 disparity levels without using a microprocessor. On performance evaluation, the disparity map shows a maximum error of 0.308% only.
    [Show full text]
  • Recovery of 3-D Shape from Binocular Disparity and Structure from Motion
    Perception & Psychophysics /993. 54 (2). /57-J(B Recovery of 3-D shape from binocular disparity and structure from motion JAMES S. TI'ITLE The Ohio State University, Columbus, Ohio and MYRON L. BRAUNSTEIN University of California, Irvine, California Four experiments were conducted to examine the integration of depth information from binocular stereopsis and structure from motion (SFM), using stereograms simulating transparent cylindri­ cal objects. We found that the judged depth increased when either rotational or translational motion was added to a display, but the increase was greater for rotating (SFM) displays. Judged depth decreased as texture element density increased for static and translating stereo displays, but it stayed relatively constant for rotating displays. This result indicates that SFM may facili­ tate stereo processing by helping to resolve the stereo correspondence problem. Overall, the re­ sults from these experiments provide evidence for a cooperative relationship betweel\..SFM and binocular disparity in the recovery of 3-D relationships from 2-D images. These findings indicate that the processing of depth information from SFM and binocular disparity is not strictly modu­ lar, and thus theories of combining visual information that assume strong modularity or indepen­ dence cannot accurately characterize all instances of depth perception from multiple sources. Human observers can perceive the 3-D shape and orien­ tion, both sources of information (along with many others) tation in depth of objects in a natural environment with im­ occur together in the natural environment. Thus, it is im­ pressive accuracy. Prior work demonstrates that informa­ portant to understand what interaction (if any) exists in the tion about shape and orientation can be recovered from visual processing of binocular disparity and SFM.
    [Show full text]
  • Real Time Distance Calculation Using Stereo Vision Technique
    Real Time Distance Calculation using Stereo Vision Technique Session 2005-2009 Project Advisor Mr. Khan Asmar Azar Submitted by: Ahmed Tassaduq 050620-158 Aisha Ashraf 050620-128 Fatima Zehra Hassan 060820-089 Shajieuddin Hyder Khan 050620-086 Department of Electrical Engineering University of Management and Technology Real Time Distance Calculation Using Stereo Vision Technique Page i A report submitted to the Department of Electrical Engineering In partial fulfillment of the requirements for the Degree Bachelor of Science in Electrical Engineering by Ahmed Tassaduq (050620-158) Aisha Ashraf (050620-128) Fatima Zehra Hassan (060820-089) Shajieuddin Hyder Khan (050620-086) University of Management and Technology October 12, 2009 Real Time Distance Calculation Using Stereo Vision Technique Page ii CERTIFICATE OF APPROVAL It is certified that the work contained in this project report, entitled “Real Time Distance Calculation using Stereo Vision Technique” carried out by Ahmed Tassaduq (050620-158) Aisha Ashraf (050620-128) Fatima Zehra Hassan (060820-089) Shajieuddin Hyder Khan (050620-086) Under the supervision of Mr. Khan Asmar Azar for the partial fulfillment of the degree requirement of Bachelor in Electrical Engineering Approved By ________________ __________________ Dr. Aziz Bhatti Khan Asmar Azar Dean SST Project Advisor Real Time Distance Calculation Using Stereo Vision Technique Page iii Acknowledgements We truly acknowledge the cooperation and help make by our advisor Mr. Khan Asmar Azar, University of Management and Technology. He has been a constant source of guidance throughout the course of this project. We would also like to thank Mr. Saeed-ur-Rehman Turk, Government College University, Lahore for his help and guidance in understanding many important issues.
    [Show full text]
  • Abstract Computer Vision in the Space of Light Rays
    ABSTRACT Title of dissertation: COMPUTER VISION IN THE SPACE OF LIGHT RAYS: PLENOPTIC VIDEOGEOMETRY AND POLYDIOPTRIC CAMERA DESIGN Jan Neumann, Doctor of Philosophy, 2004 Dissertation directed by: Professor Yiannis Aloimonos Department of Computer Science Most of the cameras used in computer vision, computer graphics, and image process- ing applications are designed to capture images that are similar to the images we see with our eyes. This enables an easy interpretation of the visual information by a human ob- server. Nowadays though, more and more processing of visual information is done by computers. Thus, it is worth questioning if these human inspired “eyes” are the optimal choice for processing visual information using a machine. In this thesis I will describe how one can study problems in computer vision with- out reference to a specific camera model by studying the geometry and statistics of the space of light rays that surrounds us. The study of the geometry will allow us to deter- mine all the possible constraints that exist in the visual input and could be utilized if we had a perfect sensor. Since no perfect sensor exists we use signal processing techniques to examine how well the constraints between different sets of light rays can be exploited given a specific camera model. A camera is modeled as a spatio-temporal filter in the space of light rays which lets us express the image formation process in a function ap- proximation framework. This framework then allows us to relate the geometry of the imaging camera to the performance of the vision system with regard to the given task.
    [Show full text]
  • Real-Time Stereo Vision on Fpgas with Scenescan
    Real-Time Stereo Vision on FPGAs with SceneScan Konstantin Schauwecker1 Nerian Vision GmbH, Gutenbergstr. 77a, 70197 Stuttgart, Germany www.nerian.com Abstract We present a flexible FPGA stereo vision implementa- tion that is capable of processing up to 100 frames per second or image resolutions up to 3.4 megapixels, while consuming only 8 W of power. The implementation uses a variation of the Semi- Global Matching (SGM) algorithm, which provides superior re- sults compared to many simpler approaches. The stereo match- ing results are improved significantly through a post-processing chain that operates on the computed cost cube and the disparity map. With this implementation we have created two stand-alone hardware systems for stereo vision, called SceneScan and Scene- Scan Pro. Both systems have been developed to market maturity and are available from Nerian Vision GmbH. Keywords stereo vision, depth sensing, FPGA, semi-global arXiv:1809.07977v1 [cs.CV] 21 Sep 2018 matching 1 Introduction Computer stereo vision is one of the best researched areas in the field of computer vision. Its origins date back to the 1970s, and it has since seen significant scientific advancement. Compared to other approaches for dense depth perception – such as time-of-flight cameras or struc- tured light camera systems – stereo vision has the advantage of being a passive technology. This means that apart from the ambient light, no further active light source is required for performing measurements. Active camera systems struggle in situations with bright ambient light such as outdoors during bright sunlight. In this case, the active 2 K. Schauwecker light source can no longer provide sufficient contrast from the ambi- ent light, and hence measurements become impossible.
    [Show full text]
  • The Role of Camera Convergence in Stereoscopic Video See-Through Augmented Reality Displays
    (IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 9, No. 8, 2018 The Role of Camera Convergence in Stereoscopic Video See-through Augmented Reality Displays Fabrizio Cutolo Vincenzo Ferrari University of Pisa University of Pisa Dept. of Information Engineering & EndoCAS Center Dept. of Information Engineering & EndoCAS Center Via Caruso 16, 56122, Pisa Via Caruso 16, 56122, Pisa Abstract—In the realm of wearable augmented reality (AR) merged with camera images captured by a stereo camera rig systems, stereoscopic video see-through displays raise issues rigidly fixed on the 3D display. related to the user’s perception of the three-dimensional space. This paper seeks to put forward few considerations regarding the The pixel-wise video mixing technology that underpins the perceptual artefacts common to standard stereoscopic video see- video see-through paradigm can offer high geometric through displays with fixed camera convergence. Among the coherence between virtual and real content. Nevertheless, the possible perceptual artefacts, the most significant one relates to industrial pioneers, as well as the early adopters of AR diplopia arising from reduced stereo overlaps and too large technology properly considered the camera-mediated view screen disparities. Two state-of-the-art solutions are reviewed. typical of video see-through devices as drastically affecting The first one suggests a dynamic change, via software, of the the quality of the visual perception and experience of the real virtual camera convergence,
    [Show full text]
  • Real-Time 3D Head Position Tracker System with Stereo
    REAL-TIME 3D HEAD POSITION TRACKER SYSTEM WITH STEREO CAMERAS USING A FACE RECOGNITION NEURAL NETWORK BY JAVIER IGNACIO GIRADO B. Electronics Engineering, ITBA University, Buenos Aires, Argentina, 1982 M. Electronics Engineering, ITBA University, Buenos Aires, Argentina 1984 THESIS Submitted as partial fulfillment of the requirements for the degree of Doctor of Philosophy in Computer Science in the Graduate College of the University of Illinois at Chicago, 2004 Chicago, Illinois ACKNOWLEDGMENTS I arrived at UIC in the winter of 1996, more than seven years ago. I had a lot to learn: how to find my way around a new school, a new city, a new country, a new culture and how to do computer science research. Those first years were very difficult for me and honestly I would not have made it if it were not for my old friends and the new ones I made at the laboratory and my colleagues. There are too many to list them all, so let me juts mention a few examples. I would like to thank my thesis committee (Thomas DeFanti, Daniel Sandin, Andrew Johnson, Jason Leigh and Joel Mambretti) for their unwavering support and assistance. They provided guidance in several areas that helped me accomplish my research goals and were very encouraging throughout the process. I would especially like to thank my thesis advisor Daniel Sandin for laying the foundation of this work and his continuous encouragement and feedback. He has been a constant source of great ideas and useful guidelines during my Thesis’ program. Thanks to Professor J. Ben-Arie for teaching me about science and computer vision.
    [Show full text]
  • 3D-Con2017program.Pdf
    There are 500 Stories at 3D-Con: This is One of Them I would like to welcome you to 3D-Con, a combined convention for the ISU and NSA. This is my second convention that I have been chairman for and fourth Southern California one that I have attended. Incidentally the first convention I chaired was the first one that used the moniker 3D-Con as suggested by Eric Kurland. This event has been harder to plan due to the absence of two friends who were movers and shakers from the last convention, David Washburn and Ray Zone. Both passed before their time soon after the last convention. I thought about both often when planning for this convention. The old police procedural movie the Naked City starts with the quote “There are eight million stories in the naked city; this has been one of them.” The same can be said of our interest in 3D. Everyone usually has an interesting and per- sonal reason that they migrated into this unusual hobby. In Figure 1 My Dad and his sister on a keystone view 1932. a talk I did at the last convention I mentioned how I got inter- ested in 3D. I was visiting the Getty Museum in southern Cali- fornia where they had a sequential viewer with 3D Civil War stereoviews, which I found fascinating. My wife then bought me some cards and a Holmes viewer for my birthday. When my family learned that I had a stereo viewer they sent me the only surviving photographs from my fa- ther’s childhood which happened to be stereoviews tak- en in 1932 in Norwalk, Ohio by the Keystone View Com- pany.
    [Show full text]
  • PMF: a Stereo Correspondence Algorithm Using a Disparity Gradient Limit Stephen B Pollard, John E W Mayhew and John P Frisby
    [1] PMF: A Stereo Correspondence Algorithm Using a Disparity Gradient Limit Stephen B Pollard, John E W Mayhew and John P Frisby AI Vision Research Unit, University of Sheffield, Sheffield S10 2TN, UK Reprinted, with permission of Pion Ltd, from Perception, 1985, 14, 449-470. ABSTRACT world, whereas this is not true ofincorrect matches. The value of 1 is not in itself critical - all that is required is a limit that The advantages of solving the stereo correspondence problem balances satisfactorily the competing requirements of by imposing a limit on the magnitude of allowable disparity disambiguating power and ability to deal with as wide a range gradients are examined. It is shown how the imposition of of surfaces as possible (section 2). However, we have chosen such a limit can provide a suitable balance between the twin 1 because this seems to be roughly the limit found for the requirements of disambiguating power and the ability to deal human visual system (Burt and Julesz 1980). with a wide range of surfaces. Next, the design of a very simple stereo algorithm called PMF is described. In In SUbsequent sections we describe the design and performance conjunction with certain other constraints used in many other of an algorithm called P:MFthat is based upon this constraint, stereo algorithms, P:MF employs a limit on allowable and several others that have been identified previously1. disparity gradients of 1, a value that coincides with that Finally we briefly discuss how the theory underlying PMF reported for human stereoscopic vision. The excellent relates to other computational theories of how to solve the performance of P1vIF is illustrated on a series of natural and stereo correspondence problem.
    [Show full text]
  • Computer Stereo Vision for Autonomous Driving
    Computer Stereo Vision for Autonomous Driving Rui Fan, Li Wang, Mohammud Junaid Bocus, Ioannis Pitas Abstract As an important component of autonomous systems, autonomous car perception has had a big leap with recent advances in parallel computing archi- tectures. With the use of tiny but full-feature embedded supercomputers, com- puter stereo vision has been prevalently applied in autonomous cars for depth perception. The two key aspects of computer stereo vision are speed and ac- curacy. They are both desirable but conflicting properties, as the algorithms with better disparity accuracy usually have higher computational complexity. Therefore, the main aim of developing a computer stereo vision algorithm for resource-limited hardware is to improve the trade-off between speed and accu- racy. In this chapter, we introduce both the hardware and software aspects of computer stereo vision for autonomous car systems. Then, we discuss four au- tonomous car perception tasks, including 1) visual feature detection, description and matching, 2) 3D information acquisition, 3) object detection/recognition and 4) semantic image segmentation. The principles of computer stereo vision and parallel computing on multi-threading CPU and GPU architectures are then detailed. Rui Fan UC San Diego, e-mail: [email protected] Li Wang ATG Robotics, e-mail: [email protected] Mohammud Junaid Bocus University of Bristol, e-mail: [email protected] Ioannis Pitas Aristotle University of Thessaloniki, e-mail: [email protected] 1 2 Rui Fan, Li Wang, Mohammud Junaid Bocus, Ioannis Pitas 1 Introduction Autonomous car systems enable self-driving cars to navigate in complicated environments, without any intervention of human drivers.
    [Show full text]