Research On Texture Recognition Method Based On Multi-feature Fusion

Texture features are fundamental features of many images and are one of the important visual features for human understanding and cognition of objects.At the same time,the representation of texture features is also a highly challenging problem in computer vision.Texture recognition can support higher requirements and high-precision object recognition,which is also of great significance for object recognition.However,there are still many shortcomings in current texture recognition methods,such as the complexity of training caused by non end-to-end methods,and the low accuracy of texture recognition in complex datasets.Therefore,this article proposes a texture recognition method based on multi feature fusion.The research content and work are as follows:(1)A texture recognition method based on improved residual pooling layer is proposed.This method introduces a global maximum pooling branch on the basis of the original residual pooling layer,and adds global structural observation on the basis of the original texture features,improving the effectiveness of texture recognition;MSFF module is proposed to fuse multi-scale feature maps and enhance the Semantic information of texture features.The experimental results show that the improved residual pooling layer algorithm proposed in this paper has better texture recognition performance compared to existing texture recognition methods such as B-CNN,Deep filter banks,Deep TEN,TEX-Net-LF,locality aware coding,and DRP-Net.(2)A texture recognition model based on multi feature fusion is proposed.Firstly,LBP encoding mapping texture extraction is proposed,using LBP encoding mapping images as auxiliary input and standard RGB images as dual streams to improve the residual pooling layer texture recognition model.The obtained texture features are fused through fusion coefficients;Then,the GLCM algorithm and uniform rotation invariant LBP texture extraction are introduced to extract texture features separately;Finally,the extracted multiple texture features were fused to obtain the final texture features and subjected to texture recognition.The experimental results show that the texture recognition model proposed in this paper based on multi feature fusion is effective and has better texture recognition performance.(3)Designed and implemented a multi feature fusion texture recognition system.The texture recognition system based on multi feature fusion has multiple modules: system management module,data management module,and texture recognition module.It has functions such as system login,system exit,image preview,image storage,texture recognition,and result storage.System testing has shown that this system can complete texture recognition tasks and has certain practicality.