The Sorting System Of Vegetable Seed Based On Machine Vision

This thesis is based on the seed sorting system as the research background, the soybean seed as the research object. And use the VC, Opencv and MATLAB as research tools. From soybean seed image process, edge extraction process, the adhesive soybean seeds’ division, shape characteristics of the component, color characteristics of the component, mechanical drive running, the process of communication between PC and PLC carried on the thorough research. The study of machine vision screening seed has certain practical significance. The following results were obtained:1.The soybean seed’s picture is collected in light, get most of the features of seed image Under the ordinary incandescent lamp. Choosing the background color to white when do the research, so due to the irradiate radiation cause reflective heavily in the middle of the seed. When extracting the target object using the Otsu, can be easily mixed between the background color and the middle part. Therefore, take some preprecessing to the image, do study in two ways. One way is the compression of grey value, another is extraction B, G, R’s binary image. Choosing the way of three-channel binarization for the analysis of the late color features.2.In order to identify the dry seeds, taking the Roberts operator, Sobel operator Prewitt operator and the Canny operator as the comparison and analysis. Finally choose the appropriate method of seed extract edge to select dry seeds. In the process of selecting, run into the adhesion problem between two or more seeds inevitably, accordingly causing the misjudgement. Using a arithmetic to solve this problem, which using corrosion method first to make sure the adhesive picture depart, then make the after corrosed legume seeds for the expansion of the same number times. The resulting binary image basically maintain the shape characteristics of the original image, there’s no influence on the parameters, such as the area, perimeter and degree of circular. Using this algorithm solved the problem of the adhesion of the pea segmentation.3. Define and extract 18 kinds of features of the soybeans. Not only in terms of shape, but also in terms of color make the corresponding comparison and this improve the recognition rate. To the characteristics of standard, use mathematical processing software Minitab for analysis, determine if it conforms to normal distribution. To conform normal distribution, we use the 3 σ interval define the certified seeds. To inconform normal distribution, we come up with the ±20% offset. In this way can basically filter out the unqualified bean seeds. Finally, complete the seed sorting of the interface using the VC programming software.4. For the seed conveyer, with frequency converter, frequency control of motor speed. After analysis of mechanical driving devices, determine to drive servo motor by PLC high-speed pulses. And finish the robot hand grasping motion of seeds at three locations in theory. PC and PLC’s communication process is studied, achieve the basic PC and PLC communication process. Prove the feasibility of manipulator sorting soybean seeds in theoretically.From the above, this topic research for scientific seed screening automation, provides a series of theoretical basis and technical support. This system overcomes the drawback of artificial detection to effort. And to realize high speed, accurate and nondestructive testing has certain actual application value.