Objective The extraction of Litchi chinensis (litchi) phenotypic parameters is essential for postharvest grading. In order to solve the low efficiency problems of litchi postharvest grading, a method of extracting litchi phenotypic parameters by image processing and three-dimensional point cloud technology was proposed, to obtain a variety of phenotypic traits at one time.

Method The Intel Realsense D405 depth sensor was used to obtain the RGB color image and depth image of litchi under four different angles at the same time, and the three-dimensional point cloud of litchi was obtained by mapping. According to the obtained litchi point cloud, the point cloud curvature and normal vector were calculated, and the mechanical damage area of litchi was identified. The RGB color image of litchi was converted in color space, and the red coloration rate of litchi surface was extracted by judging the threshold of red pixels. The 3D model of litchi was established, and the volume value of litchi was estimated by calculating the thickness of the depth image.

Result The mechanical injury of litchi could be identified successfully by the curvature of the three-dimensional point cloud on the surface of litchi, with the average detection accuracy of 94%. Compared with the method based on three-dimensional point cloud, the method of extracting coloration rate by using RGB color improved the detection speed by 90%. The determination coefficients between the coloring rate results based on RGB images and 3D point clouds and the manual detection results were 0.9574 and 0.9205, the average absolute errors were 6.33% and 4.37%, and the mean relative errors were 4.17% and 6.01%, respectively. The determination coefficient between litchi volume measured by depth image extration and hand was 0.890 1, the average absolute error was 1.59 cm³, and the average relative error was 7.94%.

Conclusion The method proposed in this study can improve the accuracy of litchi phenotypic parameters extraction, and provide efficient technical means and data support for postharvest grading of litchi.