Image Complexity Evaluation And Application Based On Multi Feature Fusion

Image complexity evaluation methods are widely used in target recognition,information security,image compression,cognitive psychology and other fields.Research on the evaluation of image complexity is not only helpful to further understand the cognitive process of human vision,but also can promote the development of other related scientific fields.In this paper,the evaluation and application of image complexity are as follows:In this paper,a BP neural network image complexity evaluation scheme is proposed,which combines texture features,color features and multiple information entropy features.Firstly,the image database is established,and then the image features are extracted.Aiming at the problem of redundancy of these features,a dimension reduction method of correlation coefficient corresponding to the classification results is proposed.The selected features are input into BP neural network for training and testing.The experimental results show that the complexity evaluation method proposed in this paper not only conforms to human vision,but also outperforms the existing complexity evaluation methods.In view of the nonconvex optimization problem in the BP neural network optimization algorithm,an improved BP network image complexity evaluation method based on genetic algorithm is proposed,which can use different specifications of neural network training model in different stages of genetic algorithm evaluation to meet the evaluation needs in different scenarios,and overcome the high complexity of the algorithm on the premise of ensuring the effectiveness of the algorithm The problem.The experimental results show that the performance of image complexity evaluation is greatly improved with the iteration.Finally,the GA-BP neural network image complexity evaluation algorithm based on multi feature fusion is applied to coal gangue detection,and good results are achieved.