Image Processing Of Agricultural Products(Fruits) Based On Illumination Invariance

With the rapid development of computer image processing technology, computervision technology as a core technology is widely used in the field of industrial andagricultural production, and becoming the hot research topic in the field of artificialintelligence. Application and research of the computer machine vision technology inagriculture began in the1970s, and the studies have focused on the quality inspection、grading of agricultural products and fruit-vegetable picking robots. However, theinterference due to the illumination changes in vision systems, making the latter part of theimage segmentation difficult, and it can seriously affect the accuracy an success rate of thequality inspection、 grading of agricultural products and fruit-vegetable picking robots.Therefore, how to eliminate the illumination effect is particularly important in the imageprocessing of agricultural products. This research results have some significance of socialand economic for improving the success and accuracy rate of automatic picking ofagricultural products.This paper first mainly introduces the computer vision technology, the meaning of theprocessing of illumination invariant, and the domestic and international research status;then briefly discusses the general processes and common method of image processing; andfor the illumination changes in the process of image recognition of agricultural products,proposes and validates three algorithms of illumination invariant; finally, develops theillumination invariant system.The main contents of research and the results achieved can be as following:1. This paper briefly discusses the general processes and common method of imageprocessing. It mainly includes image preprocessing、image segmentation and featureextraction.2. This paper focus on the method of dual homomorphic filtering based on RGB colorspace and HSI color space、Gamma correction based on the illumination component andillumination invariant image to eliminate the light effects, and compare the experimentalresults of the three methods.3. Base on MATLAB platform, we develop the illumination invariant image recognitionsystem of the agricultural products initially. This system mainly includes four modules: theimage pre-processing module、the illumination invariant algorithm module、 imagesegmentation and edge detection module and feature extraction module. According to the results of the experiment, the illumination invariant method which isinterchangeable using in the identification system of agricultural products can improve theperformance of the identification system of agricultural products. This work has certainsignificance for the Quality inspection of agricultural products and fruit-vegetable pickingrobots.