Research Of Pummelo Quality Detection Method Based On Machine Vision

Pummelo quality detection and classification was important in the process of production, sales and export. Pummelo detection of national standards and industry standards included the shape, color, surface defect, edible rate, weight and other indicators. In this paper, Guanxi Pummelo as the research sample, according to the Pummelo production quality grading test requirements, based on the machine vision technology, hyper spectral imaging technology, automatic weighing, studied the accurate and efficient detection method of pummelo quality index.The research analyzed the methods based on the analysis of different detection methods to verify the application range, detection accuracy, the detection efficiency by experiment, which laid the foundation for the comprehensive quality of the classification and detection of pummelo.The research content and achievements of this paper are as follows:(1) This study compared the method of image processing based on physiological structure characteristics of pummel, including the fast image restoration algorithm based on median filtering noise, the area and edge image segmentation algorithm. This method is used to support the classification and detection of pummelo.(2) According to the image centroid established fruit shape characteristic analysis model, the paper studied the calculation method of spatial moment based on spindle positioning image and the boundary expansion, and Analyzed the fruit shape and fruit diameter calculation method.The paper proposes to shape parameters and symmetry description method based on feature analysis of regional fruit shape as the core. It put forward Describing the shape index of pummelo:C_Ratio, C_Shape, δ. According to the method of harmonic analysis in the reconstruction of the boundary, the research used Fourier descriptor method to characterize the shape, and used the Fourier descriptors reconstruction method to compare the fruit shape detection of malformed fruit.(3) This method divides the region to describe quantitatively the surface color of pummelo.It was proposed according to the calculation method for detecting rate of coloring. The average error rate for Standard color detection was5%.Using this description method to research the detection method of surface defects of pummel. The recognition rate of the experiment was up to90%.(4) This paper analyzed the structural features of a longitudinal section of pummel, It gave calculation method of characteristic parameters, Including the volume, estimated volume ratio and pericarp thickness difference. This paper established two groups fruit edible rate detection modelbased on characteristics of pummelo, the models applied to nondestructive testing of edible rate and fast sampling experiment. The detection accuracy of the experiments was86.7%,94.5%.(5) This paper analyzed the hyperspectral image of the longitudinal section of pummelo. By finding out the sensitive spectral segmentation, it used the method of sensitive spectral segmentation to accurate calculation of fruit flesh area ratio. The method that established the linear relationship between flesh area ratio and edible rate estimated the edible rate, In order to improve the sampling efficiency, based on a precise area of flesh ratio as the edible rate of detection index was proposed, and It provided a fast sampling method of pummelo edible rate accurately. In the experiment of detection the average error is1.34%.(6) In the paper a automatic weighing and grading equipment was designed According to the weight grading indexes of pummelo, and the equipment was made. The paper designed the key components of the weighing system and built the system. The experiments were weighed under different conditions. The research Analyzed and compared the experimental results, according to the results put forward the factors that influence the system weighing accuracy.