Research On Classification And Recognition Algorithm Of Criminal Investigation Infrared Images

With the continuous development and progress of science and technology,there are also new research technologies in the field of criminal investigation,among which infrared imaging technology is the fastest developing.Through this technology,some people and objects that cannot be found due to occlusion can be detected in the field of public security criminal investigation,and some invisible thermal trace information can also be detected.By classifying and identifying the infrared images of the crime scene and nearby places,it can effectively help the police to analyze the case,promote the progress of the case and provide help and reference for the cracking of criminal cases.At present,there are some problems in the classification and recognition of criminal investigation infrared images,such as the difference of suspect’s target attitude,the ambiguity of target and the occlusion of re-recognition target.The main research contents of this paper are as follows:(1)In order to solve the problem of low classification accuracy caused by the difference of target attitude in criminal investigation infrared images,this paper studies and analyzes the characteristics of infrared images and designs a classification algorithm of criminal investigation infrared images based on multi-feature fusion.Firstly,image features are extracted from infrared images,and then different feature extraction methods are analyzed and studied.The image features are expressed by combining HOG features and LBP features,and the classification model is optimized by generalized adaptive optimization algorithm,and finally the high-precision classification and recognition of criminal investigation infrared images are realized.(2)In order to solve the fuzzy problem of criminal investigation infrared images,a classification algorithm of criminal investigation infrared images based on immune optimization mechanism is proposed.The characteristic data extracted from the infrared image of fuzzy criminal investigation presents a linear inseparable distribution.Firstly,the characteristic data are acted as innate immune factors to identify the linear inseparable data areas and delete the data,and immune memory is carried out at the same time.Then the adaptive immune factor processes and presents the new feature data to train the adaptive immune factor to realize the optimal classification of the model,and then improve the classification accuracy of the criminal investigation infrared image.(3)In order to solve the problem of low accuracy of pedestrian re-recognition caused by occlusion in criminal investigation natural scenes,this paper proposes an infrared pedestrian re-recognition algorithm for criminal investigation based on improved Res Net-50.Re-recognition is an important part in the field of criminal investigation image recognition.In this paper,an EBCBAM attention mechanism is proposed to improve the Res Net-50 network,so as to improve the network’s feature expression ability.By combining Triplet and Softmax loss functions and introducing Adam optimization algorithm,the convergence speed is accelerated,and finally the effect of pedestrian re-recognition algorithm in criminal investigation infrared images is significantly improved.A pedestrian re-recognition system is developed and designed by using the algorithm in this chapter.Above all,according to three different image problems,this paper designs different classification and recognition algorithms of criminal investigation infrared images,and uses infrared images to verify the algorithms,which proves that the algorithms designed in this paper can classify accurately and effectively.