Detect Impurities And Unsound Kernel In Corn Using Machine Vision

In China,corn is one of chief crops in grain production. This computer vision is introduced into the detection and classification of corn, can effectively solve the long period,weak effectiveness, inefficiency, bad accuracy and high labor intensity of traditional corn detection, influenced by issues such as subjective and objective factors, benefit to evaluating the quality and grade of corn kernels accurately.This dissertation concentrated on research of variety identification for impurities and unsound kernel detection and recognition.The main results of the research are as follows:1. Image preprocessing methods for identification of corn were studied.The grayscale threshold method was selected to background segmentation,which can reserve the color information of corn and remove its background.In this investigation,it was found that the red proportion was very different between the background and the object.The setting of its threshold was based on the theory of minimum error probability.The weight-average method was applied to make the original image to the grayscale image.The graycale threscale method was used in the image binary progress.2. Six color features were extracted from the image after background segmentation,including red average(R),green average(G),blue average(B),hue average(H),saturation average(S) and value average(V).Five morphological features were extracted from the binary image using Blob analysis,including length,width,the ratio of length to width,area and parameter.3. The single-feature threshold and the multi-feature threshold were used between sound kernel and impurities, sound kernel and unsound kernel in this investigation.The single-feature threshold method was used to recognize sound kernel and broken corn,either feature of area and parameter can give a perfect identification accuracies with 100%,94%. The multi-feature threshold method was used to recognize sound kernel and impurities, sound kernel and ill speckle,sound kernel and injured kernel.For the recognition between sound kernel and impurities,the H and S-H were selected and the identification accuracy was 100%;the V and S-H were selected and the identification accuracy was 99%;For the recognition between sound corn and ill speckle,the R*G/B and H*S were selected and the identification accuracy was 99.5%;For the recognition between sound kernel and injured kernel, the identification accuracy was 100% when the hole filling was selected and calculated the pixels, the identification accuracy was 95% when the threshold of area feature was 1800.4. Static image was used in this investigation,a corn automation single grained structure was designed to research dynamics image,which can save time to analyze the characteristic in rather than display various samples.