Study On Machine Vision System For Coprinus Comatus Picking Robot

Recent years,the robot has been promoted to the national strategic level,driven by various national policies,government departments,the research institutes,Binocular stereoscopic vision has been wid ely used in the field of picking robots.Based on the similarity between binocular cameras and human eyes in sensing objects binocular stereoscopic vision has the advantages of simple system structure,high positioning accuracy and low cost of equipment an d it has been a very active branch of machine vision.This paper is about a visual system study of Coprinus comatus picking robot and the research object is binocular stereoscopic vision.The improved Canny edge detection algorithm to identify the fruits of Coprinus comatus,the binocular camera that calibrated by Zhang Zhengyou calibration method and the feature extraction and matching of the three-dimensional images made by SURF method were used to make a foundation to obtaining the three-dimensional coordinate information of Coprinus comatus.The main works of this paper include:(1)On the basis of reading a large number of domestic and foreign literatures,a complete binocular stereoscopic vision system was built and the reasons of choosing binocular vision were analyzed and described the principle of binocular vision.Main hardwares used in the system are: two models with high-definition cameras,a computer and simple PTZ.(2)In order toidentify the fruit of Coprinus comatus,the improved Canny edge detection operator was used to extract the edge of Coprinus comatus in this paper.The bilateral filter owns better edge preserving ability than Gaussian filtering.The wavelet transform are sued to compress the low frequency coefficients.And the high frequency coefficients are drawn at the same time,which makes the edge information be enhanced.Experiments show that the improved algorithm is superior to the original algorithm and the improved algorithm is the main innovation of this paper.(3)In order to get the camera's internal and external parameters,this paper calibrates the left and right cameras of the binocular stereoscopic vision system by using the classic Zhang Zhengyou calibration method.By calling Open Source Computer Vision Library,Open CV2.11,to verify Zhang Zhengyou calibration method.The experiment shows that the calibration method has the advantages of simple operation,low requirement of equipment and high calibration precision.(4)In order to find the corresponding points in the two im ages of the target object taken by the left and right cameras,calculate the parallax and lay the foundation for the deep recovery and three-dimensional reconstruction,we studied the primitive selection,matching criteria and method in this paper and expe riments based on SURF image matching algorithm are matched.The experiment's results show that this method can achieve 95% matching accuracy for general image matching,but when it was applied to three-dimensional image matching the matching accuracy is below 80%.Therefore,improved algorithm or some other matching algorithm are still needed to improve it's accuracy.