A Vision Method for Rapeseed Amount Measuring

Abstract

An automatic measuring method for rapeseed was developed to get rapeseed quantity information. A high-throughput device was designed for rapeseed color image collection. A watershed algorithm based on range conversion was developed to separate the stocking grains into single one. In order to improve the accuracy of detection, a total of 23 characteristic parameters of rapeseed and impurities were trained in a random forest classifier to establish a classification model for impurity detection. Furthermore, the characteristic will be used for quality grading. The experimental results show that the method can achieve high-speed detection of rapeseed quantity: the detection speed is about 15,000 grains per minute, and the accuracy of impurity detection can reach 91.24%. In this study, the problem of low efficiency and high intensity in manual rapeseed testing was solved and it enjoys high practical value.