Abstract:1000-kernel weight of rapeseed is of great significance for rape breeding, seed quality assessment and yield estimation. Aimed at the current problem that Guizhou mainly relies on manual measurement of rapeseed thousand-grain weight cycle and long workload, based on VS2013 and OpenCV, a rapeseed grain counting system has been developed. System functions mainly include image reading, RGB color feature analysis, grain corner detection, grain hole detection, and grain counting. Based on the appearance color, geometric shape and spatial arrangement characteristics of rapeseed, a set of rapeseed counting algorithm is proposed, and the counting algorithm is integrated into the seed counting system. Through debugged and tested the various functions of the system, the functions work normally and the expected design goals is achieved. The results of the grain detection test show that the detection accuracy of the grain counting algorithm reaches 88%. The construction of the system has certain reference value for the realization of Guizhou rapeseed seed counting automation.