Obstacle Detection For Smart Car Based On Monocular Vision

In this paper, obstacle detection of smart car based on monocular vision is researched,including the method of obstacle recognition and the design of hardware and software.According to the needs of the smart car driving environment, this paper use CMOS imagesensors as monocular vision context-aware components to detect whether there are obstacles inthe front road of car at real-time. If do exist obstacles, the appropriate obstacle avoidancestrategy will be taken, so that the car won’t collide with obstacles in the course of driving andensure the safety of the car itself.Through access to a large volume of literature related to this study, we have understood theresearch status and problems of smart cars at home and abroad. We also learn from theexperience of previous research and technology so that make improvements on some technicalshortcomings. A smart car hardware circuit system has been designed, in order to buildexperimental platform for obstacle detection system. On the basis of extensive reading andintensive study of the obstacle detection method, an obstacle detection method using CMOSimage sensors which is based on the inverse perspective projection transform has beenproposed. According to the composition of hardware circuit and obstacle detection method, wewrite software program of the system, which can make the hardware system and obstacleidentification fully integrated to achieve the ability of obstacle detection and obstacle avoidance,so that the car can drive steadily.In circuit design, in order to make the car body compact, stable and reliable, we adopt thecircuit design principles of modular and integrated circuits. We give fully play to characteristicof modular, such as good expansion and good maintenance, characteristic of integration, such ashigh density component. Design MC9S12XS128as the core of control decision-making circuit,ov6620as the image acquisition circuit, MC33886and FET as the motor drive circuit,photoelectric switches and grating encode as motor measuring circuit, power circuit whichprovides power to control chips, image sensors, motors, steering gear, the serial circuit forsystem debug and BDM interface circuit.In the aspect of obstacle detection, in order to easy to analyze and solve the problem ofobstacle identification, this paper establishes a coordinate system model firstly. During the sametime, in order to eliminate the effect that environment has on the projected point coordinate inthe inverse perspective projection transformation, the image coordinate system is rotatedcompensation. With the coordinate system model and the coordinate system rotationcompensation as the basis of the analysis, getting the optical center displacement of the imagesensor through motion parameters calculating method that is based on road feature points andobtaining the displacement of surface projection point by corner detection and inverseprojection transformation. Through reusing the relation between optical center displacementand road projection point displacement, we can determine whether there are obstacles, so thatachieve the purpose of obstacle detection. Through the actual operation and test of the car, the results show that the obstacle detectionmethod that the paper proposed is effective, and achieve the desired results.