Research On Infrared Small Target Image Preprocessing And Detection Algorithms

In the modern high-tech war, infrared small target detection and tracking, which is the core technology of infrared pre-warning, infrared guidance areas, plays an increasingly important role. Affected by the detector's noise, atmospheric transmission as well as long detection distance, infrared small target image shows a low SCR, and partly missing edge information. Hence, it is relatively difficult to detect infrared small target, especially under the complex background. In this paper, the infrared image preprocessing and small target detection algorithms are studied right under the complex background.Firstly, this paper analyzes the characteristics of infrared image and carries on a detailed analysis and research to its three aspects. Then on the basis of these research,a mathematical model of infrared image is constructed.Secondly, based on the research of the classical median filter algorithm and its various improved algorithms,this paper proposes an average median filter algorithm,which is suitable for simple background infrared small target image. This algorithm selects threshold according to the average grey value of the image, and effectively distinguishes background and target. Better effect was achieved through separated handling of background and signals.Then the algorithm combines the morphology opening Top-Hat transformation. Taken advantage of its background suppression capability, this algorithm was successfully introduced into the condition of complicated background, and the SCR of the infrared images was significantly improved after pretreatment.Finally, in small target detection, the algorithm of wavelet transform for image singularity detection was emphatically analyzed in this paper. By means of the excellent direction resolving capacity and time-frequency localization analysis capacity of directional wavelet, it reserves target details in the realization of small target detection. This paper makes improvement on the judgment of small target. By marking the large-scale transition points among the results of several directions' wavelet transform and referencing these points'distribution areas in the image,the small target is confirmed. This method not only recognizes the small target effectively, but also the false alarm caused by noise is limited.