Research On Aerial Infrared Small Target Detection Algorithm Under Complex Cloud Background

With the rapid development of infrared imaging technology and computer vision algorithms,along with the combat environment is becoming increasingly complex,infrared guidance technology is gradually replacing radar,laser and other active detection guidance technology.As an important part of infrared guidance technology,infrared small target detection technology has always been a hot topic researched by scholars at home and abroad.Due to the long distance between the target and the imager,the target usually appears as a small spot in the image,resulting in the target lacking in details characteristics such as shape,texture,etc.In addition,complicated scenes such as clouds and waves in the infrared image may affect the detection result.Therefore,the research on the detection method of small infrared targets is still a very challenging task.The research topic of this paper is the infrared small target detection method under complex cloud background.The aim of this paper was to design an infrared small target detection algorithm with good detection performance,and in order to meet the real-time requirements of practical applications,the algorithm should be able to adapt to the acceleration of the embedded FPGA processor.Based on the in-depth study of traditional algorithms,this paper selects the method based on the human visual system as the research direction.It has the advantages of better detection effect and simple computing structure.The main work of this paper is as follows:The LCM algorithm based on local contrast measurement has the problem of insufficient background suppression and target enhancement in complex cloud background.This paper deeply analyzed the contrast characteristics of the local area in infrared images,and proposed an enhanced LCM algorithm based on local direction gradient.Comparison experiments show that the improved method significantly improves the ability of the algorithm to suppress the background and enhance the target,and the algorithm has good detection performance in complex cloud scenarios.In order to solve the problem that the LCM algorithm and its improved GELCM algorithms have low real-time performance due to repeated calculations at multiple scales,a fast small target detection algorithm based on local contrast measurement and scale space theory was proposed.In the proposed algorithm,the hypothesis model of the small target point source diffusion is introduced,and the candidate targets are quickly extracted by the comparison mechanism and the adaptive scale selection theory in the human visual system,thereby avoiding the repeated calculation of invalid pixels by the algorithm.The comparison experiments show that the proposed method has good detection performance and real-time performance at the same time,and has a high practical significance.For the contradiction between the low tolerance to false alarms in practical applications and the traditional algorithms always generate false alarms in complex scenarios.In this paper,through the in-depth study of the target and the characteristics of the convolutional neural network,an infrared small target classification network is designed to suppress false alarms in the detection results of existing algorithms,and the detection results of existing algorithms are used as the training set.Improve the relevance of the classification network.The experimental results show that the designed network has high classification accuracy,and it has the advantages of small parameter and small computation.In addition,this paper combined the fast detection algorithm proposed in this paper with the classification network,and proposed a small infrared target detection algorithm based on CNN.The experimental results show that the detection algorithm has the advantages of high detection rate and low false alarm rate in the complex cloud background.Moreover,the algorithm has a simple calculation structure.Therefore,the algorithm has extremely high practical significance.