Study On Infrared Specific Target Detection Algorithm Under Complex Backgrounds

In recent years, along with the rapid development of computer and digital imageprocessing technology, infrared imaging technology is widely used in various civil andmilitary fields. A large number of background clutter and noise in Infrared imageincreases the difficulty of detection of small targets, so the study of real-time detectionalgorithm for low SNR images of small targets, is of great significance.From the perspective of the infrared image processing method, this paper analysisand research the two main aspects to deal with the problem of the specific target incomplex background.The first problem is the infrared image preprocessing. The original image oftencontains a variety of noise, how to use the imaging characteristics of infrared imagesto remove the noise that is caused by imaging system and other external environmentis an important aspect of this study. The spatial filtering and frequency domainfiltering method is discussed and the results were compared and analyzed. After that,this paper presents an improved adaptive Gaussian filtering method. Blurred edges andother issues exist for traditional Gaussian filter, design a direction and scale adaptiveGaussian filter. The direction and scale of the filter is determined by the local gradientand local variance. The experimental results show that the method to protect imagedetails while effectively filter out the noise.The infrared target detection problem is discussed. Two major categories ofalgorithms which are detect before track and track before detect are analyzed. On thisbasis, an infrared target detection method based on local entropy of the image isproposed; this method compromises the local variance information of image to detectthe infrared target. Experimental analysis and comparison show that this methodcompared with the previous method, has more accurate detection rate and low falsealarm rate.