Study Of Fruit Classification System Based On Compressed Sensing

Fast and accurate classification of fruit could increase the income of farmersand businesses, while improving the satisfaction of customer and enhancing thecompetitiveness of our country’s fruit industry. Based on the compressed sensingtheory, the signal can be sampled with a lower frequency than the Nyquistfrequency. Relying on these few measured values, the original signal can bereconstructed or the characteristics of original signal could be extracted.Therefore, the compressed sensing theory is applied to the fruit classification inorder to research the application of compressed theory in the field of agriculturalproducts’ quality detection and classification.The compressed sensing algorithm was used for the classification of applein this paper, and the algorithm consists of sparse decomposition and encoding.The algorithms which were Discrete wavelet transform and discrete cosinetransform for image orthogonal spare decomposition were analyzed, and basedon the overcomplete dictionary, the fruit image was processed, theexperimentation showed that the sparse decomposition algorithm based on thecomplete dictionary could get a best sparse representation of original image.Then the processed image was encoded to obtain the measured values that couldrepresent the original image.The apple image was processed with the compressed sensing algorithm.According to the distribution law of measured values, the size classification ofapple was achieved accurately.The R component of the RGB color model was processed with thecompressed sensing algorithm. According to the distribution law of measuredvalues, the color classification of apple was achieved.The R component and I component of the HIS color model were combined,and then the combined image was processed with the compressed sensingalgorithm, according to the distribution law of measured values, the defectclassification of apple was achieved. The experiments result showed that the classification of apple in the size、color and defects can be achieved quickly and accurately on the basis ofcompressed sensing, and the classification results were satisfying, so thecompressed sensing theory has a certain prospect.