Research On The Technology Of Obstacle Detecting In Backing Environment Based On Video Sensor

The development of Parking Distance Control (PDC) is a major trend in future, andthe study on its related key technology--Obstacle detection has great significance. PDC canbring great convenience and safety to the vehicles and drivers. PDC based on ultrasonicdetection principle has currently become a mainstream product. But parking distancecontrol system based on ultrasonic detection principle also has some blind zone whichmake it unable to meet the modern needs of the complex road conditions.In this dissertation, machine vision principle is introducted after reserching onmainstream PDC system has done. To overcome the blind zone of conventional PDCsystem, this paper mainly focuses on the processing of images captured by cameras toidentify mesh-shaped obstacle and Low obstacle in dynamic background.The highlightsand main contributions of the dissertation include:(1) Reserches on mesh-shaped obstacles detectionThis paper mainly do research on mesh-shaped obstacle detection which based on linecharacteristics. For the goal, firstly we analyze the performance of several classical edgeextraction operators. After image edge detection and morphology filtering, the binaryimage is processed using blob analysis to get the characteristic information of mesh-shapedobstacles, including the locations and parameters. Mesh-shaped obstacles are judged by thethreshold and displayed on the GUI．(2) Reserches on low obstacle detectionThis paper does a research on low obstacle detection which based on heightcharacteristics.Meanwhile, to point against the conventional binocular vision algorithmwhich has low Real-time performance and Immunity.A new algorithm based on nonlinearbinocular vision is proposed in this paper. The principle of the algorithm is to determinewhether there is low obstacle by determine the separation barrier of top ridgeprojection.Firstly, we transform color images into gray images.Next,the blocks which onbehalf of separation barrier are extracted by image subtracting.After image Enhancement and morphology filtering, the binary image is processed using blob analysis to get thecharacteristic information of mesh-shaped obstacles, including the locations andparameters. Low obstacle are judged by the threshold and displayed on the GUI．Theseadvantages of the new algorithm were demonstrated by experiment results of speed，robustness and visual effect．...