Review on the Application of Machine Vision Algorithms in Fruit Grading Systems

Abstract

China is the largest fruit producer in the world, so fruit grading is one of important tasks in agricultural production. Because manual testing is time-consuming and laborious, it is reasonable to be replaced by machine vision technology. This review included three parts. The first part introduced the architecture of the fruit grading system, including image grasp and processing algorithms. And the system transferred images to computer for processing after captured by acquisition sensor. Then computer performed image processing operations, including preprocessing, segmentation, feature extraction and selection. The second part summarized the machine vision classification algorithms for fruit grading. Classification algorithms are classified as supervised machine learning, unsupervised algorithms, and deep neural network. The supervised machine learning algorithms are Naive Bayes, K-nearest Neighbor, Support Vector Machine and so on. The unsupervised machine learning algorithms included K-means Clustering and Principal Component Analysis. And the deep learning part mainly introduced Artificial Neural Network and Convolutional Neural Network. This part reviewed the characteristics and application potential of the above classification algorithms, and comprehensively compared the merits and demerits of assorted classification algorithms applied in fruit classification. The third one discussed the difficulties and challenges of fruit grading tasks based on machine vision technology and put forward some prospects. Finally, this paper summarized and analyzed the machine vision algorithms in fruit grading systems and concluded that the research of machine learning and deep learning combined with machine vision systems is the main development trend in the field of fruit grading and will be a hot research spot in future.