The Study On Online Inspection Of Embryo Integrity Of Embryo Rice

As people's demand for the quality of life grows,more and more attention on the nutritional value of food is attracted.The embryo rice is closely followed because of its rich nutrition,and the market demand for embryo rice is also increasing year by year,therefore,it is very important to ensure the rice production quality.At present,in China,the evaluation of embryo integrity of embryo rice is mainly relied on artificial subjective judgment,which cannot be quantitatively analyzed and is inefficient.In order to meet the inspection requirements to embryo rice,this paper takes the embryo rice as the research object,using machine vision technology,image processing technology and deep learning technology to study the online inspection method of embryo integrity to embryo rice,achieving the automatic online detection of embryo integrity of rice embryo.In order to improve the production quality of the embryo rice,the embryo rice processing equipment can automatically adapt the processing parameters through inspection information.This paper carries out the following research:Firstly,the methods of image retention and image preprocessing are studied.According to the imaging requirements of embryo rice,image acquisition program for embryo rice is designed,which is followed by the completion of equipment layout and selection.According to the quality of embryo rice images and the subsequent image processing requirements,the preprocessing methods of image denoising,image enhancement and image transformation are analyzed and studied.Combined with the actual image processing effects,the image preprocessing algorithm which is suitable for the embryo rice images is selected.Secondly,the image segmentation algorithm of embryo rice is analyzed and studied for the purpose of extracting single rice image of embryo rice.In order to solve the problem of rice target extraction in the image,the segmentation threshold based on OTSU is used to calculate the binary image of the embryo rice.Then,according to the image area segmentation algorithm,the connectivity of the binary image is marked,and the shape parameters of the embryo rice are extracted.In order to solve the problem of image segmentation of touching rice kernels,an image segmentation algorithm based on the background skeleton feature is proposed.Through the skeletonization of the background images,segmentation endpoints are calculated.According to the endpoints matching rules,the optimal segmentation position could be retrieved and the curve is drawn to get segmentation results of touching rice kernels.Thirdly,the embryonic germ integrity identification network model is designed based on the convolutional neural network.According to the image analysis of the embryo rice,two convolution neural network models are constructed respectively for the direction identification of the embryo rice and the classification of the embryo integrity of embryo rice.The Relu activation function and the learning rate attenuation method are used to optimize the network model training.The model takes the single embryo rice kernel image as input.Based on the convolution neural network model,the embryo rice is classified and outputted according to the standard of the embryo integrity category.By studying the effect of embryo position identification and the interested region setup to the recognition accuracy of the network model,the effectiveness of the two convolutional network models is verifiedFinally,the integrity online detection system is designed.It works out the automatic inspection platform and the integrity inspection software system of germ,realizing the online automatic embryo integrity calculation during the processing of retained embryo rice.Based on the own of the indicators of embryo integrity and embryo retained rate,the average embryo integrity is used as the criterion to determine the preservation of embryo in embryo rice,which is used to judge the embryo integrity macroscopically during the processing of embryo rice.Meanwhile,three test indicators are conducted to a comprehensive test to verify the accuracy and effectiveness of the system.