<i>Recognition of cutting region for pomelo picking robot based on machine vision</i>

摘要

<list list-type=continue><list-item> <b><sc>Abstract.</sc></b> <b>Aiming to locate the area for the cutting action execution of pomelo harvest robot, a method basing on the pomelo centroid and the natural growth characteristics of the pedicle to determine the cutting region was proposed. Firstly, the centroid of the pomelo and the largest circumscribed rectangle were obtained after image preprocessing. Then, the vertical line passing the centroid was made on the top edge of the largest circumscribed rectangle. The intersection point was used as the center of the cutting region, the radius of which would be obtained according to the center point and labeled as Rocc. The cut point was located by a circular mark with rocc as its radius. The optimized valued of Rocc and rocc were determined by comparing the influence of the Rocc and rocc radios at different sizes on the relationships between the cutting region and the cutting point. Image segmentation was implemented by OTSU Adaptive threshold Algorithm after choosing 2R-G-B color difference components by analyzing the color space. The maximum connectivity region was extracted from the target region of a mature pomelo which was obtained by morphological denoising the segmented image. 60 pomelo fruit hanging on the trees in orchard were chosen as samples with 200 images collected by upward shot, over shot and side shot, under different shooting light conditions, including direct sunlight and back lighting in sunny day and in cloudy day, respectively. Recognizing model was established based on 150 images and validated by the rest ones. For the 50 images for independent verification, the accuracy rate of the cutting region reorganization achieved 88% when the value of Rocc and rocc were 46 and 4, respectively. These results indicated this method had the potential to locate the execution area for robot and provide a technical basis for the subsequent cutting action development of the cut region identification problem in the pomelo picking robot.</b> </list-item></list>