The Visual Information Processing System And Control Method Of Small Agricultural Products Sorting Machine

Currently most agricultural products sorting machine still sorting by physical methods through the mechanical structure, and limited to the single design sorting structure, sorting modes are relatively simple, which typically are the size, weight and so on. When it relates to the maturity, the degree of damage, whether or not peeled, etc. of product, the traditional mechanical sorters do not apply. Therefore, a new type of intelligent sorting machine produced, which firstly collected the data including product feature information by sensors, then presented the data to the processor device to perform recognition processing, lastly used the results obtained by processing of the product to product sorting. But the new intelligent sorting machine usually equipped with expensive professional sensors and PC or industrial computer which can perform a number of operations in real time, thus which means higher prices and bulky, is not suitable for small farmers demand that should be low-cost, small footprint, easy maintenance.This paper designed a small agricultural intelligent sorting machine based on machine vision. When the agricultural products transferred to the bottom of the camera, the nearby photoelectric switch is triggered, and generates an interrupt signal to DSP processing camera to capture images of the current product, then transfer the results to the ARM through the RS-232 serial port, and the ARM chip controls the sorting machine sorting of agricultural products based on the result, at the same time the results of sorting and sorting mode were displayed in the form of a graphical interface, which is a user-friendly. This design replaces the previous computer through the use of DSP + ARM chip method, which greatly reduces the volume and the cost of sorting machine. In the part of experimental design, the experiments using cherry tomatoes, kumquats and dried longan to sorting by size, color, whether or not broken respectively.Before processing target image, noise elimination and the distortion correction processing were done first. Then, the image is sampled at intervals, and a R-G difference method was used to judge the sampling point, the eligible pixels will as seed to seed filling, at last extracting the full target image. The size and color sorting were used the number of pixels that eligible target image to judge. Breakage detection was based on the degree of approximation circle of the target image to judge. Among them, 640 samples of cherry tomatoes were tested, 601 were correct sorting, the correct rate was about 93.91%; Totally 640 samples of Kumquat were tested, 622 were correct sorting, the correct rate was about 97.19%; 792 samples of Dried longan were tested, 682 were correct sorting, the correct rate was about 86.11%. The maximum processing rate for the three materials respectively were Cherry tomatoes 337 kg / h, Kumquat 311 kg / h, Dried longan 228 kg / h.The small agricultural intelligent sorting machine this article proposed, which with stability and DSP embedded system powerful computing capability, to a certain degree already has the ability preliminary sorting, and greatly reduces the size and cost of machine vision and control unit.