Design and experiment on grading system for online non-destructive detection of internal and external quality of apple

Abstract: At present, grading machinery for quality inspection of apple has the disadvantages of complex structure, expensive price and unbalanced internal and external qualities. However, both the internal and external qualities are vital to its value. The grading device for online non-destructive testing of internal and external quality of apple is thus designed in this research pursuant to the experimental analysis under static conditions. The device is composed of dumbbell roller, chain conveyor module, belt drive module, machine vision system module for detection of external quality, near-infrared internal quality testing module, grading module and control system. In the design of machine vision system module for detection of external quality, in order to increase the contrast ratio between the bruised and non-bruised parts of apple, the reflectivity spectra of bruised and non-bruised parts were collected, and it was determined that the largest difference of the reflectivity between the two parts is at the position of 730 nm, and thus a red LED (light emitting diode) light source with a wavelength of 730 nm is selected as the light source of vision module. In order to get the integrated surface information, apple completed the rotation in the process of forward movement, and the self-designed segmentation and synthesis algorithm was utilized to extract and synthesize the images of a single apple under 3 states of motion. Then the images were processed by Gaussian filtering, QTSU binarization and contour extraction. When the apple is judged to have bruising, a rejection instruction is directly sent. When it is determined that there is no bump, the contour extraction image is subjected to circle fitting processing, and the size of the apple is obtained by fitting the circle diameter. The near-infrared internal quality testing module is mainly used to detect the soluble solids of apple, and the modelling effects of arranging the probe respectively on the upper part and the lower part were contrasted so as to determine the best modelling method under static condition. Experiments show that it will be better to arrange the probe at the lower part. Finally, the on-line detection performance of the device was verified by experiments. The accuracy of the device for detection of bumps on apples was 94%, the correlation coefficient for size detection was 0.9646, and the root mean square error was 2.281 mm. Then, an on-line model was established for measuring the content of soluble solids in apple. The correlative coefficient of the calibration set was 0.950 8, the root mean square error of the correction set was 0.342 6%, the correlation coefficient of the prediction set was 0.949 2, and the root mean square error of the prediction set was 0.448 7%. The detection time for a single apple was 0.71 s. The device has the advantages of small size, simple structure and low cost, which is suitable for the needs of farmers and middle-sized and small-sized enterprises.