Online detection and classification system of external quality of leaf for dining hall and family

Abstract: In view of the easy dehydration, yellowing and rotting of leaf vegetables, harmful substances produced in the human body, and being time-consuming and difficult to select leaf vegetables artificially, this study focused on development of nondestructive online detection system for leaf vegetables quality detection in small and medium restaurants based on machine vision. This system was composed of hardware modules and software processing system. The hardware modules included leaf vegetables separation unit, transmission unit, LED (light emitting diode) light sources unit, sensing unit, machine vision detection unit and sorting unit. The leaf vegetables separation could divide samples into pieces by designing a roller with a negative pressure. The sensing unit could control imaging acquisition and sorting operation of defective samples. In addition, the circuit of sorting unit included 555 timer, solenoid valve, relays and air compressor. The software was designed by using OpenCv and Visual C++ for realizing the automatic collection, data analysis and result display of the sample information. Finally, combining the characteristics of the corresponding information extraction method, the 320 spinach samples were adopted for verifying the system's performance, and among these samples, the yellow leaf, leaves with insect hole, rotting leaves and normal leaves were 91, 75, 91, and 63 respectively. For yellow leaves and rotting leaves, the RGB (red, green and blue) and HSV (hue, saturation, value) color space transform method were used by setting weight value H∈(60°, 130°), S∈(0, 0.17) and V∈(0, 0.23) to extract the characteristic information of the yellow leaves and rotting leaves of spinach. For leaves with insect hole, the threshold of 2G-R-B in RGB color space was adjusted for completing image graying, binarization and morphological denoising to extract contour feature information of leaves with insect hole, which could realize the discrimination of this kind of leaves and their area calculation. By the color space transformation of RGB to HSV, the threshold segmentation of hue variable with effective suppression of noise was selected to extract the characteristic region of yellowing leaves and rotten leaves. In order to obtain the correct information of insect eye contour, the closed operation of morphological filtering was used to remove the blade handle information. The experimental results show that the sorting system and information extraction algorithm can achieve the discrimination of spinach external quality. Compared with the artificial discrimination, the discrimination accuracies of yellow leaf, rotting leaves free of pests and leaves with insect hole were 96.70%, 92.59% and 84.62% respectively. The overall discrimination accuracy rate was 94.69%, and the selection speed of the device could reach 1 sample/0.84 s. The results also demonstrate that the sorting speed and the sorting accuracy of this system can meet the requirements of the quality separation of leafy vegetables. The test results show that the parameters of the system can meet the requirements of the quality separation. Compared with the quality of the artificially sorting vegetables, the system not only is easy to operate and time-saving, but also has good reliability. This research provides technical support for the practical research and development of leaf quality sorting device.