Research And Implementation Of Key Technology Of Mobile Robot Image Processing

The key problem of image processing in mobile robot visual navigation is road recognition and obstacle detection.The paper is based on computer monocular vision technology to study unstructured road identification and motion obstacle detection.Based on the existing technology,this paper uses the color road edge detection combined with the road area recognition technology,which obtains good results for the unstructured road recognition in the campus environment.The algorithm also has a partly recognition effect on the complex road environment.Three-frame difference combined with pyramid optical flow method can be used to detect moving objects in the road and estimate its position in the image.The content of this paper is divided into four aspects: mobile robot software system design,image road area identification,road edge detection,obstacle detection.Mobile robot software system design,this paper is the main study of mobile robot vision navigation system design.The function of the visual navigation system is to complete the navigation task by synthesizing the scene information obtained from sensors through the integrated processing of road identification,obstacle detection and motion decision making in the system.Visual navigation system design is through modular,multi-threaded and other methods to complete,so that the system has good maintainability and real-time.In the aspect of road area identification,this paper reviews some common algorithms for road recognition,and introduces the road recognition based on region.In this paper,we use color and LBP texture feature as all features of the image.We use the supervised SVM algorithm to training classifier.The trained classifier is used to roughly recognize the image road area.In the aspect of road edge detection,this paper analyzes the effect of five commonly used edge detection operators in road recognition.Combining with the experimental analysis,the paper proposes an improved Canny edge detection algorithm in color space.The improved Canny edge detection is different from the traditional Canny edge detection which use the fixed threshold that we use an adaptive threshold based on maximum interclass variance on each image that is because of the changing of the mobile robot scene.The improved Canny edge detection algorithm is more suitable for mobile robot autonomous navigation scene.In the detection of obstacle detection,this paper mainly studies the motion obstacle detection in the process of mobile robot navigation.This paper introduces three commonly used algorithms of obstacle detection: the background model difference method,the two-frame difference method and optical flow method.Then,according to the hardware resources of the mobile robot,an obstacle detection algorithm that based on the three-frame difference combined with pyramid optical flow is used to detect the moving target with real-time and accuracy in the front of the mobile robot.