A Basic Research On Information Perception Of Kiwifruit Picking Robot

China is the largest amount of kiwifruit acreage country in the world. Kiwifruit was mainly relied on hand picking. Hand picking is a time-consuming, laborious and increasingly higher cost work. Therefore, automated picking is an inevitable trend of future development, research on kiwifruit picking robot has important theoretical and practical significance.The primary task of picking robot is to accurately identify the kiwifruit in the natural environment. Based on the complex natural environment for the growth of kiwifruit, this paper investigates the perception of kiwifruit picking robot. Accurate segmentation of the target fruit area and the complex background area was realized. Adjacent, overlapping, and blocked fruit in the fruit area were recognized one by one during day and night. Feature information of fruit was extracted, and fruit picking order is determined finally. The main contents are as follows:(1) The scheme of the robot system was determined. Based on the analysis of kiwifruit cultivation mode and growth environment, we elaborated kiwifruit harvesting robot operating procedures; The robot manipulator basic configuration, structure size and drive form was determined; The hardware components and software implementation of the robot vision were studied.(2) The color characteristics of the kiwifruit was studied. Using6color space commonly used in image processing comparatively analyze the grayscale image, binary image and line profile analysis image of kiwifruit; We classified the background elements in the kiwi image acquired in the natural environment, and analyze their color statistical. The results show that R-G can be able to distinguish the fruit and the background; Then the optimum partition coefficient of n·R-G color characteristics was determined according to the image evaluation method of error segmentation pixel,0.9R-G was selected finally.(3) Accurate segmentation of the target fruit area and the complex background area was realized. Comparative analyze different of image segmentation method, the Otsu method was used for kiwifruit image threshold segmentation. Residual noise in the segmented binary image was analyzed and classified, the method of set area threshold and morphological operation was employed to remove residual noise, the regions of target fruits and backgrounds were successfully separated.(4) Extraction the binary image kiwifruit edge of the area, overlapping and blocked fruit were recognized. Comparative analyze of different edge detection operator; Select Canny operator to extract the image boundary. Using elliptic Hough transform to recognized the adjacent, overlapping, and blocked fruit, and extracted each fruit coordinates in image.(5) This paper puts forward a night of kiwi fruit identification method. Image acquisition direction is optimized, and then fruit picking order is determined. Comparative analyze different acquisition angle images, camera image acquisition upright direction is selected; The color characteristics of the kiwifruit image collected at night was studied,1.1R-G was selected finally. According to the identification test on the of different light Intensities image, in the range of10lux-800lux of the light Intensities, the fruit can be able to properly identified. After Otsu Method was used for threshold segmentation and morphological operation was employed to remove residual noise. The image boundary was extracted by Canny operator and consequent elliptic Hough Transform, which made the target fruit be recognized separately. Also, fruit centroid coordinates were extracted, and fruit picking order is determined finally.