Study On Detection And Grading Of 'Jiro' Persimmon's External Quality Based On Computer Vision

As a non-destructive measurement method, computer vision technology has been widely applied in food detection field. However, the application on'Jiro'persimmon has not yet been reported.'Jiro'persimmons for exporting from Yunnan province were studied. Computer vision technique was used to study the persimmon's external quality involves weight, surface color and surface defects. Under the static condition, gradation for'Jiro'persimmons was realized. The main research results were as follows:1. Computer vision hardware was established for persimmon. Studied on commonly algorithms of image devoicing, edge detection and image segmentation to the image underlying processing, and then selected the underlying image processing algorithms which applied to persimmon.2. In persimmon weight grading, the model which reflects the relationship between the area in maximal diameter and weight of persimmon was set up to measure the weight of persimmon. It's a linear relationship between the area in maximal diameter and weight of persimmon, and its correlation coefficient was 0.975. The verification showed that the maximum fractional error was 6.64%, the minimum one was 0.02%, the average fractional error was 2.69%, and the average accuracy rate was up to 97.5%. The application effect was very notable in persimmon weight detection and grading.3. In persimmon color grading, the main process was to acquire the mean value of hue of persimmon surface by the computer technology and extract its feature, then make the hue interval value to grade the persimmon color and can be divided into different maturity stages. According to the verification result, the average accuracy rate was 92.0%. Persimmon maturity grades were established by analyzing on the hardness and soluble solids of different maturity stages persimmons.4. The image processing and RBF neural network were integrated to identify the persimmon surface defects. The average accuracy rate of training samples for recognizing defects was 90.8%, and the average accuracy rate of testing samples for recognizing defects was 88.3%.5. Persimmon comprehensive quality grading standards were established from the weight, surface color, defects these three aspects. The persimmon identification system software was designed and developed.This research investigated the application potential of detecting and grading persimmon external quality based on computer vision technology. Results showed that the system can be used to detect the quality of'Jiro'persimmon objectively and accurately.