Multi-Strategy Image Registration And Fusion Of Visible And Two Wave Bands Infrared Images For Target Recognition

Multi-source image registration and fusion is the key research topic of ground target detection and recognition.At present,there are no specific fusion algorithms and standard fusion rules for multi-source images at home and abroad.In view of this,this project proposes an algorithm of multi-strategy image registration and fusion of visible and two wave bands infrared images.Based on in-depth research of imaging principle and imaging mechanism of visible,mid infrared and far infrared images,imaging characteristics of visible and two wave bands infrared images are summaried and imaging model is created,which laid foundation for the full use of the complementary information of multi-source images.A novel target registration algorithm based on improved structure similarity for two wave bands infrared images is put forward.Improved multi-scale edge structural similarity is calculated through using different scales images and their edge images in Gauss scale space so that target's complete edge structure information in the infrared images is utilized fully.Then the performance of target registration is improved.The experimental results show that the accuracy and correct rate of target registration based on the proposed algorithm are better than other algorithms,and the target similarity between two wave bands infrared images is higher than 0.98.A new target registration algorithm based on feature fusion of visible,mid-infrared and far-infrared images is proposed.Firstly,the similarity between targets in mid-infrared and far-infrared images is calculated based on improved structure similarity.Then,the similarity between targets in visible and far-infrared images is calculated based on Hu invariant moment feature and cosine similarity.On the basis of the above,target registration of visible and two wave bands infrared images is realized by fusing these two kinds of similarity.The experimental results show that the accuracy and correct rate of target registration based on the proposed algorithm are better than other algorithms,and the target similarity among visible and two wave bands infrared images is higher than 0.95.A new target recognition method based on feature fusion of visible,mid-infrared and far-infrared images is proposed.Firstly,typical features of the target in VIS,MWIR and LWIR images are extracted.Then,the fused feature vector is obtained by feature selection based on Fisher discriminant analysis.Finally,target recognition is realized through Support Vector Machine(SVM).The target recognition performance of the proposed method and 5 other target recognition algorithms were tested and compared on VIS,MWIR and LWIR images.Experimental results showed that the proposed method is superior than other target recognition methods in recognition rate,and it has high operation efficiency and robustness.