Recognition and features extraction of sugarcane nodes based on machine vision

Abstract

To achieve machine intelligence cutting of effective sugarcane kinds of fragments with sugarcane bud, machine vision was introduced to identify sugarcane nodes. Through acquiring the S component of HSV color space by threshold, mathematical morphology filtering as template and the anti-phase image of the H-component by threshold was added to get synthesized image. Synthetic image was divided into 64 regions and obtained seven characteristic indicators, such as centroid ratio, roughness ratio and white point ratio, and so on. Then support vector machine was introduced to identify sugarcane nodes and sugarcane internodes. The average recognition rate of sugarcane nodes between internodes was 93.359%. Clustering analysis was introduced to identify sugarcane nodes blocks which were got by support vector machine (SVM) classification. The average recognition rates of the sugarcane numbers and the sugarcane nodes position were 94.118% and 91.522% respectively.