Leaf Classification System Based On Hybrid Biogeography-based Optimization Algorithm And BP Neural Network

In this thesis,a plant intelligent classification system is proposed to overcome the long time-consuming and inefficient of the traditional classification problems.This computer aided classification system can not only help to improve the accuracy of plant accuracy but also reduce the workload of the staff.The specific content of the thesis is as follows:First,downloading data sets from authoritative websites and preprocessing these images to reduce problems in later processing.Hu moment and Gray-level Co-occurrence Matrix(GLCM)are used to extract the shape and texture features of images.The extracted 11 features are input into the back propagation neural network(BPNN)for training.In order to avoid falling to local optimum,the chaos optimization algorithm and biogeography-based optimization(BBO)algorithm are mixed to optimize the weights and thresholds of neural network.Secondly,in order to verify the superiority of the improve algorithm,CA-BBO algorithm is compared with BBO algorithm and other chaotic biogeography-based optimization algorithm in the middle of the experiment.In the experimental results and analysis phase,CA-BBO algorithm is compared with particle swarm optimization(PSO),BBO,ant colony optimization(ACO),population-based incremental learning(PBIL)intelligent optimization algorithms in training BPNN,and BPNN trained by CA-BBO is compared with several other classification algorithms.Finally,6-fold cross-validation to prevent the occurrence of over-fitting,the average value obtained by running five experiments found that the classification accuracy of "Hu-GLCM+CA-BBO+BPNN" for 15 kinds of 299 leaf images is 80.19±0.81%.By comparing with other algorithms,we find that the performance of this method is the best.