To solve the problems occurred in the recognition of oil-tea camellia, such as the big error, lack of universality and much time consuming, the paper propose a new algorithm for fruit recognition, where Region Of Interest (RIO), Histogram of Oriented Gradients (HOG) temperature and Least Square Support Vector Machine (LS-SVM) are applied. First, the images are detected from HSV (hue, saturation, value) color information. The HOG temperature, calculated using four regions of interest (ROI), is input to an LS-SVM classiﬁer, which detects the fruit. The performance of the model was veriﬁed by experiments. The vector sizes were effectively reduced and a higher detection speed was achieved without compromising accuracy (relative to conventional approaches). The detection accuracy can respectively achieve 95.5%, 89.4% and 96.7% for isolated fruit, overlapped fruit and background, which is shown the excellent performance of the proposed algorithm.
Digital Object Identifier (DOI)
Zhi-qiang, Wang and Li-jun, Li
"LS-SVM Recognition of Fruit Using in Harvesting Robot Based on RIO-HOG Feature,"
Applied Mathematics & Information Sciences: Vol. 11:
2, Article 18.
Available at: https://digitalcommons.aaru.edu.jo/amis/vol11/iss2/18