Research On Infrared Target Recognition Based On Feature Fusion

With the rapid development of high-tech technology,computers and infrared imaging technologies have become the focus of research in many fields such as military.However,as the infrared background conditions become more complicated,the interference information and the target characteristics are more diverse,etc.,the accuracy and robustness of infrared target recognition are faced with great challenges.Therefore,only a single feature of selection and extraction is considered to describe the performance of infrared image targets,the recognition of targets is far from being achieved.In this paper,the global color feature,local motion feature and edge of the target are combined to study the recognition of infrared targets.The main work is divided into:(1)Study the development and status quo of infrared target recognition,and analyze how to effectively solve a series of problems caused by difficult background factors such as poor infrared background conditions,interference information and ever-changing target characteristics.(2)Based on the basic principles and methods of infrared target recognition,it mainly includes the infrared target image preprocessing,analyzing the influencing factors of the infrared recognition process and determining the evaluation index of the recognition result.Filtering out more invalid information data for the subsequent target feature extraction process,which minimizes the running time of subsequent series of algorithms and simplifies the operation process,thereby improving the real-time processing of recognition and the like.At the same time,we can better understand the influencing factors that may lead to recognition failure,design and debug the infrared target recognition algorithm in a targeted manner,and finally analyze and verify the validity and feasibility of the recognition results by identifying the indicators.(3)Establish the target feature database and analyze and extract the infrared image target features,including global color features,local motion features and edge features,and selectively perform feature data fusion processing.It turns out that the recognition algorithm based on fusion features can greatly simplify the operation process and solve the major problems left by using a single feature quantity.(4)Design an infrared target recognition algorithm based on feature fusion.Based on the steady-state motion of the target and the target shape approximation does not change,the sequence Monte Carlo(particle filter)recognition algorithm is established,and the application performance is verified by experiments,which achieves shorter recognition time and good robustness effect The target recognition algorithm of particle swarm optimization self-organizing feature map network is designed.By establishing a more comprehensive simulation experiment,the problem of inaccurate target recognition and loss of target Monte Carlo algorithm is solved when the target sharp mutation and the target outer contour are obviously changed.It achieves the purpose of detecting and recognizing the center position of the target in a stable and real-time manner,so it has an advantage in the field of infrared target recognition applications.