Abstract: 【Objective】 To study a watermelon recognition and positioning technology based on machine vision to meet the automation requirements of watermelon handling equipment. 【Methods】 2D image data and 3D point cloud data of watermelon were collected and the OpenCV function library was called to segment the two-dimensional image of watermelon and extract the outer contour of the watermelon; The 3D point cloud data was preprocessed using the PCL function and perform graphic feature matching with the 2D image to extract the centroid points of the top layer watermelon. 【Results】 The validation test results showed that the total recognition rate of the number of watermelons in a single round was 97.62%, the correct recognition rate was 95.58%, and the centroid recognition rate was 92.72%. The total recognition rate of 20 rounds of testing was 98.02%, the total correct recognition rate was 96.53%, the total centroid recognition rate was 94.17%, and the error rate between the total centroid number and the total number of watermelons was 2.36%; The total time for watermelon image processing was 101.8 seconds, which improved the efficiency by 77.8%. 【Conclusion】 The watermelon recognition and positioning method proposed in this article has high recognition rate and low error rate, which can provide technical support and algorithm reference for the visual part of automatic handling equipment.