Research On Crop-disease Recogintion Technology Optimization Based On Swarm Intelligence Algorithm

The research on Crop-disease recognition and classification by using swarm intelligence algorithm and image processing technology,which helps farmers to quickly prevent crop diseases according to the category of crop diseases.Such targeted control not only greatly reduces pesticide residues in crops,but also greatly reduces the damage to the ecological environment,and greatly increases crop yields.So this kind of research has very important practical significance.In this paper,longan disease is the research object,through the improvement of swarm intelligence algorithm,it is applied to color disease image preprocessing,color disease image segmentation,multi feature extraction of color diseases image and color disease image recognition.The research work done in this paper is summarized as follows:(1)In order to solve the problem of slow convergence and easy to fall into local optimality when the traditional particle swarm optimization(PSO)algorithm is used in the optimization of each link of the image recognition,the corresponding improvement ideas are put forward.The chaotic particle swarm optimization(PSO)algorithm is improved from two aspects:on the one hand,a binary particle swarm optimization(CBPSO)algorithm based on chaos theory is proposed,which can improve the processing speed of the algorithm by transforming the chaotic variables into the discrete binary vector,and improving the limitation of the algorithm to deal with the discrete problem.On the other hand,a chaotic particle swarm optimization(ACPSO)algorithm based on adaptive adjustment is proposed to improve the search efficiency and improve the accuracy of the algorithm by dynamic adjustment of the inertia weight.Experiments show that the improved algorithm is effective in terms of accuracy and stability.(2)Aiming at the problems of the complexity of image background and difficulties in the later the segmentation,an image segmentation algorithm based on chaotic particle swarm algorithm and fuzzy clustering is proposed.First,the color space is converted from RGB color space into HIS color space.Then a hybrid algorithm of chaotic particle swarm optimization and fuzzy clustering(CPSO-FCM)is proposed.Each color component is processed by the algorithm and the corresponding partition graph is obtained.Finally,the color space is converted into RGB color space to achieve the segmentation effect.(3)A segmentation algorithm of color crop-disease image based on watershed algorithm and Otsu is proposed for solving the problems of big noise and blur edge.First,a new color space is proposed for the factors that are not affected by the reflected light.Then,the reconstruction technology of open and close of mathematical morphology is used to rebuild the image,so as to reduce and eliminate the location of the waterline displacement caused by the details and noise interference.Compared with other segmentation methods,the result is better and the performance is obviously improved.(4)Aiming at the problem of recognition of crop-disease image,a feature representation scheme is proposed for multi feature fusion,such as color,texture,shape and local feature.In order to make the extracted features robust and reliable,the new color space and color moment fusion are given for the feature extraction and representation for the color features.For texture features,the fusion of gray level co-occurrence matrix and local binary pattern is used for feature extraction and representation.For shape features,the Hu invariant moment method is used for feature extraction and representation.For local features,feature detection is carried out after 45 degrees of rotation of the original SURF method,and the features are described with the Pyramid model.Finally,a variety of methods(principal component analysis and binary chaotic particle swarm optimization)are used to optimize the selection of the generated features.From the experimental results,the binary chaotic particle swarm optimization(PSO)is better than the principal component analysis(PCA)for the selection of the generated features.(5)First,it is difficult to give parameters precisely in the SVM model,for classification and recognition of color crop-disease image,so the adaptive chaotic particle swarm optimization and particle swarm optimization are combined to get SVM parameters to get ACPSO-SVM and PSO-SVM.Secondly,the extracted image features are used as the original input data of these models and classified,which can effectively solve some challenges(blindness and uncertainty)that classification of crop diseases are facing.The cross validation method is used to test different PSO parameters.The results show that the ACPSO-SVM algorithm of longan algal blotch is lower than that of PSO-SVM algorithm.The classification accuracy of other disease image is higher than that of the PSO-SVM.But in terms of efficiency,the processing time for ACPSO-SVM is only 88.94% of the time required for PSO-SVM,and the ACPSO-SVM algorithm is relatively simple.The proposed algorithm based on adaptive chaos particle swarm optimization has good classification accuracy and is an effective method.