Image Segmentation Of Adhesive Rice Based On PCNN And Pit

With the advancement of science and technology,the modern agriculture and food industry has also flourished.As one of the main grains,rice will inevitably have some quality problems during planting,harvesting and processing.In order to ensure China's food security,it is essential to test the quality of the rice.In rice quality testing,rice appearance quality is one of the important contents.The traditional manual detection method is low in efficiency,large in error and subjective.At present,image analysis technology has been increasingly applied to rice quality analysis.However,in the actual rice inspection work,there is often a adhesions phenomenon about the collected rice images.If the adhesive rice can not be segmented correctly,it will lead to the wrong detection results.Based on the analysis of domestic and foreign adhesion particle image segmentation algorithms,the image segmentation of adhesive rice image is studied in view of the characteristics and practical needs of rice appearance quality detection in this thesis.The work of this thesis mainly consists of the following parts.First of all,in the actually analysis of rice's appearance quality,the image segmentation involved that includes the segmentation of rice and background and the segmentation between the adhesive rice in the collected image.Based on the research and analysis of the characteristics of pulse-coupled neural network image segmentation algorithm in this thesis,the improved particle swarm optimization algorithm(PSO)is used to optimize the parameters of PCNN.Then this method is applied to the background segmentation of rice image.Secondly,due to the process of rice quality inspection,all kinds of complicated adhesion situation between tiled and single-layer rice are combined and evolved on the basis of two-grain adhesion.Therefore,the problem of two adhesive rices segmentation is that a basic problem in the actually detection of rice image segmentation.Consequently,based on the theory of vector angle,this thesis focuses on the segmentation of two-grain adhesive rice.Then pit detection algorithm based on the angle between three points and pit matching method with adaptive characteristics is established.The algorithm firstly performs edge detection on the adhesive rice,calculates the angle feature of the point on the edge,selects the a few points with the smallest feature as the candidate pit,adaptively matches the pits and connects the pair of pits to make the adhesive rice is separated.Theoretical analysis and experimental results show the stability and effectiveness of the proposed algorithm.Compared with other similar algorithms,the algorithm in this thesis has fewer parameters.And it is more convenient to take values and runs faster.The algorithm is insensitive to the translation,rotation,and resolution of the adhesive rice image.And the selected pit position is relatively accurate.Finally,by using artificial placement to collect rice sample images for study the adhesion situation is affected by factors such as small rice size and irregular appearance,the collected images could not cover all possible adhesion conditions.It means that the testing of segmentation algorithm is not thorough and complete enough.Aiming at this problem,this thesis designs a rice adhesion simulation image generation algorithm and implements this method based on MATLAB environment.With this tool,it is possible to simulate almost all possible adhesive situations between two grains of rice and generates corresponding simulation images.By using these simulation images to test the performance of the algorithm in this thesis,not only the work efficiency is high,but also the completeness of the test image set can be greatly improved,which will be helpful to deeply and comprehensively analyze the performance of the algorithm.