Research On Multi-feature Fusion Texture Recognition Based On Trace Transform

Texture recognition is an important content in the field of image recognition,trace transform is an excellent invariant feature extraction method for texture images.It has good description ability for texture features in various complex scenarios.In order to further improve the recognition performance and generalization ability of trace transform in various restricted environments,this paper propose a multi-feature fusion image texture recognition method based on trace transform and Bo F(Bag of Feature)model and a texture recognition method based on trace transform and rotational delta modulation coding features.In this paper,a multi-feature fusion image texture recognition method based on trace transform and Bo F model is proposed.Firstly,the fragmented image of the texture image is obtained by the method of key point detection,and then the trace transform feature and SIFT(Scale-invariant feature transform)feature of the fragmented image are extracted.Through the method of feature cross-coding and the method of dynamic identification of energy,the fusion features of trace transformation features and SIFT features are obtained and the feature words are optimized,and the Bo F model is used for feature encoding.Finally,it is fed into a support vector machine(SVM)for training,prediction,and classification,and tested using standard texture datasets.The identification method based on trace transform and rotational delta modulation coding features proposed in this paper.For the first time,the communication speech coding technology is combined with the image transformation technology,the whole image is scanned by the trace rotation,and the sampling information on the trace is delta-modulated and encoded,so as to obtain multi-angle global ordered structural features.Based on the above theory,two feature extraction methods combined with trace transform are proposed.The first one is the feature extraction algorithm based on the trace transform of rotational delta modulation,the second is a feature extraction algorithm based on the trace transform of the rotary block delta modulation.Then,the feature extraction method of sampling and cross-coding on the trace and the feature optimization method such as reducing the gray level in the statistical histogram are designed,and the features obtained by the two feature extraction algorithms and the original trace transform are fused.Finally,the support vector machine is used to train and recognize the sample images.In order to improve the accuracy and robustness of low-quality face recognition in various special scenarios,the two algorithms proposed in this paper are applied to people with different lighting,different shooting angles,different occlusion types,noise,and different degrees of blur face recognition.The experimental results show that the two algorithms proposed in this paper have better performance in low-quality face recognition.