Research On Technology Of Pothole Intellgent Repair

At present,the repair of pothole of asphalt pavement especially the repair by automatic pothole repair machinery is low efficient and poor quality with the labor intensity is high and the degree of automation is not high,in this paper,an intelligent pothole repair method based on machine vision was proposed to solve this problem.At first,the characteristics of image recognition in the process of automatic repair of pothole was totally analyzed,and the key technology of the method of edge detection and positioning manipulator working principle,hardware and software of the system testing device,and the installation layout form were given based on machine vision,after that,visual object detection and detection scheme for visual pattern were proposed.Compared with the traditional methods,whose robustness and universality is poor,including histogram based threshold selection method and method of database based on the traditional method.In order to acquire more accurate edge pits coordinates,based on pothole repair platform,a robust method for automated edge detection of potholes which are not tied is proposed to get better accuracy.Multi-scale kernel local normalization is used to restrain shadow and then some blocks whose standard deviations are less than a threshold are selected to be trained in SVM.After the edge of pothole is been first detecting by trained result,Prewitt operator with eight-neighbor is used to process second detection of the pothole.The control mode of motion control card was applied on the motion positioning repair manipulator,in order to detect the degree of repair and to prepare for the post interface,the real-time motion of the manipulator was tracking by vision based on position recognition.The manipulator position tracking includes horizontal position and vertical position tracking,template matching method based on shape was applied on horizontal position tracking,vertical position was estimated with multi-planar image model.Finally the effectiveness of motion control and manipulator position tracking algorithm were all verified by experiments.