The Development Of Wolfberry Classifier Based On Machine Vision

With the increase output of the wolfberry,especially the improvement of classification technique of wolfberry,which will not only directly promote the additional value of wolfberry sales,but also effectively promote the brand awareness of Ningxia wolfberry.In this paper,Introduced from the mechanical structure of the sorting machine,the design of circuit,the software analysis are studied and analyzed respectively.In the analysis of mechanical structure,the paper mainly focuses on the mechanical structure of the vibration feeding device and pneumatic grading actuator,the minimum atmospheric pressure required for the classification of the wolfberry,and the force analysis of the wolfberry during the movement.At the same time,it analyzes the parameters and respective advantages of the disk impeller for separating wolfberries one by one and the differential dial for widening the distance between adjacent wolfberries.In the aspect of circuit,BCM2837 is used as the graphics processor of the cameras upper and lower.The image sensor is IMX219 Sony sensor,with high sensitivity of 30 frames/s.Image transmission uses CSI interface with three groups of differential signals,with the transmission speed of 80M-1 Gbps.In addition,the composition of the analog signal modulation circuit of weighting sensor is analyzed.As for the software algorithm of wolfberry classification,it is mainly studied on the size classification of wolfberry.It includes the longitudinal diameter,the diagonal length of the external rectangle,the length of the outer contour,and the projected area of the wolfberry,then compared these aspects between different grades.In the profile method,the profile perimeter method and the perimeter method of the minimum external convex polygon are compared.The perimeter method of the minimum external convex polygon demonstrates better result than the other one.As for the color identity research,the analysis and study of the moldy wolfberry are respectively used RGB model,HSI model and HSV model of the image.It shows that when the component V of HSV model is used to identify the wolfberry,it has a series of virtues,such as,wide separation between the double peak,easy to identify.During the information processing in the weighing platform,the median filter and the multiple cumulative for errors reducing are adopted.The experiment found that the error can be reduced to 0.02g,which can meet requirements of the quality grading.The information fusion and classification of wolfberry adopt VSM-DS evidence theory as basis.Using size,color and quality of wolfberry,compose the information characteristics of the fusion for establishing the VSM classifier.Finally,by the method of DS evidence argument the grading results are given at the decision-making level of the wolfberry grading.we can draw conclusions that the way of VSM-DS evidence theory is more accuracy than the single wolfberry,and the grading accuracy can reach 94%.In addition,the fusion of quality and images information can make effective judgments on the saturation and the fruit flesh thickness of the wolfberry.