Target identification based on automatic threshold for eggplant picking robot

Abstract

An automatic-threshold-based target identification method of eggplant fruit picking robot is proposed for to improve the intelligence and real time of the vision system. A conclusion is drawn that EXG color factor is the most favorable for eggplant fruit image segmentation through experiment and statistics of the color features of the eggplant fruit in growing environment and their surroundings. Automatic-threshold method is adopted to segment EXG gray scale images. The mean value of the image gray scales is selected as the estimated value of the initial threshold, which is continuously iterated until satisfactory result is achieved. Laplace operator is used to extract the eggplant fruit image edges. According to the requirements for the vision system of fruit-vegetable picking robot, such features are extracted as the contour, area, circumscribed rectangle and cut-out point of the fruit target. In is shown by the experiments that the segmentation efficiency is greater than 81% respectively and the average time used is 0.12s, meeting the requirements of fruit-vegetable picking robot for the vision system.