Research On Infrared Small Target Detection Algorithm Based On Wavelet Transform

Compared with the Remote Sensing Image, Radar Image and the Visible Image, Infrared Image has lots of advantages, such as safety, high resolution and strong infrared radiation penetration and so on. Taking it into consideration, the infrared thermography has become an important technology in precise guidance, and it is widely used in reconnaissance, surveillance, navigation and other important military fields. Besides, infrared target detection and tracking is also widely used in aerospace, intelligent transportation, robot vision and other high-tech fields.Combined with their own characters of imaging, when we consider about the weather, temperature, wind direction and speed, sun radiation and other natural weather conditions' affects, infrared small target detection under complex background is a difficult problem and central issue of signal and information processing technology. Many experts and scholars have invest their concern and done amounts of research, put forward a series productive infrared small target detection algorithms full of innovation and guidance.On this basis, this thesis analyzes the sources of noise and its characters, summarize the typical image processing algorithms like the gray level transformation, spatial filtering, spatial sharpening, frequency domain filtering and the Gabor transform, lists the advantages and disadvantages of these algorithms. Meanwhile, this paper focuses on detection before track (DBT) algorithms under the complex sky and cloud background, the main research work and innovation are as follows:(1) An effective method based on the Wavelet Transform and the improved Top-Hat filter to detect the infrared small target is proposed. Firstly, we use the single level wavelet transform to obtain different sub-bands wavelet coefficients represented as approximate, horizontal, vertical and diagonal respectively. And then we utilize the improved Top-Hat method to filter the approximate sub-band, which is then fusioned with the horizontal sub-band. The coefficients of the sub-bands vertical and diagonal are set to zeros simultaneously. Afterwards, the fusioned coefficients instead of the original sub-bands together with the set-zeros sub-bands to do the inverse wavelet transform. To make the target more prominent, we adopt the grey level transformation based on histogram to enhance the reconstructed image. Then we can figure out the infrared target. The experiments indicate that our method is accurate and effective, and meanwhile with the good robustness.(2) Considering about the noise's distribution characters, we propose an effective method based on wavelet transform and adaptive denoising to detect the infrared small target detection. Firstly, we use two level wavelet transform to obtain different sub-bands wavelet coefficients represented as LL2, HL2, LH2, HH2, HL1, LH1, HH1. And we set all the coefficients LL2 to zeros, we utilize the activity level estimation to compute the each adaptive thresholds of the other six sub-bands to do the soft denoising, then we reconstruct the image. In the end, we set a threshold to segment the reconstructed the image, after the segmentation, we can figure out the target. The experiments indicate that our method is accurate and effective, without the false alarm, successfully overcome the interference of cloud sand, meanwhile with the good robustness.