Tomato Fruit Maturity Detection Method Based on YOLOV4 and Statistical Color Model

Abstract

The ripening information of tomato fruit in greenhouse environment is closely related to production operation. Currently, greenhouse tomato ripening information is mainly carried out by manual inspection. In this paper, we present a new real-time method of tomato fruit detection and its maturity measurement in natural greenhouse. More than 25,000 tomato fruits were tested in 1005 images by Yolov4 and the identification accuracy was 95%. After construction of single fruit data set, the color proportion analysis of RGB color features was analyzed. Combined with normalization and noise reduction processing of K-means clustering, we proposed the maturity detection method based on the proportion of R component in RGB space. Experiment results of scouting robot showed that the recognition and detection speed is 5-6 frames/s.