Texture recognition model establishment method based on multi-scale integrated feature coding and application

Abstract

The invention discloses a texture recognition model establishment method based on multi-scale integrated feature coding and application, and belongs to the field of image processing, and the model establishment method comprises the steps of establishing a priori guide feature extraction network, a feature fusion network, a multi-scale coding network and a multi-scale integrated learning network; sequentially connecting the networks, and then carrying out training to obtain a texture recognition model, wherein the networks in the model are sequentially used for: firstly, extracting convolution features with relatively high texture structure expression capability by utilizing texture prior information; secondly, fusing the deep features with the shallow features to obtain multi-scale strong semantic information features; thirdly, coding the features of each scale to obtain global texture features robust to abnormal points; and finally, performing ensemble learning by using the obtained multi-scale global texture features to realize robust recognition of multi-scale textures. According to the invention, the robustness and the accuracy of a texture recognition result under the conditions of complex background interference and large-scale change can be effectively improved.

Description

technical field [0001] The invention belongs to the field of image processing, and more specifically relates to a method for establishing a texture recognition model based on multi-scale integrated feature coding and its application. Background technique [0002] Texture is the basic microstructure of an image and an important middle-level feature for image understanding and scene understanding. Therefore, texture recognition, that is, extracting texture features and classifying them accurately, is an important visual task in the field of computer vision. Texture recognition has applications in many vision tasks, such as image retrieval, industrial vision inspection, face analysis, terrain recognition, etc. [0003] Due to the influence of many factors, such as illumination changes, viewing angle changes, rotation, scale, image degradation (blur, noise, background interference, etc.), texture recognition is still a very challenging visual task. At present, a large number o...

AI Technical Summary

Problems solved by technology

If no additional processing is performed on the abnormal features during the encoding process, the discriminative performance of the encoded global features will be affected, thereby restricting the recognition accuracy and robustness of such methods for background anomaly interference.

At the same time, the exi

Images