Algorithm Of Small Target Detection In Strong Light Level Background

The technique of small dim target detection in low Signal-to-Noise and low contrast images has been the key technique of the electro-optical detecting system. This master dissertation aims at solving this problem basing on requests of the detecting system.As the characters of the small target in strong light level background are proposed, as well as, the temporal and spatial distribution features of the noise and the clutter, the corresponding algorithms of the images in the filtering, in the enhancement, in the segmentation and in the detection are researched in this paper.1,By utilizing the difference of spatial shape and pertinences between the target and the noise, as well as, the clutter, an improved method basing on gray morphological band-pass filter and basing on three frames correlation filter is applied to filtrate the noise and the clutter. Gray morphological band-pass filter can preserve the target signal within the given scale, and to filtrate lots of insular strong noise and dim noise with high spatial connectivity, at the same time it can filtrate high spatial connectivity clutter; The filter basing on three continuous frames correlation can combine heaping up the energy of the target with filtrating the noise and the clutter. The experimental result shows that the modified algorithm can filtrate effectively the noise and the clutter from the images.2,After analyzing the small target imaging character, the incomplete beta function transformation and piecewise linear transformation are used to enhance the small target image. The advantage of using unitary incomplete beta function is that the transformation curve can be very cragged at the gray segment expected whenα,β's values are chose appropriately. This can expand the gray interval between the target and the background obviously, and catch precisely the several gray grade differences between target and background in the low contrast image. The simple disposal means are explained in order to enhance real time adaptability of the algorithm. When the modified piecewise linear transformation is used to enhance the image, the pattern recognition concept is introduced, and the best threshold got by minimum error means is used to be the subsection point, by this way, the low contrast image is enhanced effectively.3,As the main difficulty in small target image segmentation is analyzed by using routine method, a new method basing on two-time segmentation of the processed images is applied . After the image segmentation by gray threshold, considering the difference of region connectivity between the target and the noise or clutter, region growing is applied to improve the traditional projection detection, and the new method gets over the false alarms greatly.4,Owing to the traditional Detection Before Tracking(DBT) can't accumulate the small target energy, in this paper, a method of Tracking Before Detection (TBD)-dynamic programming algorithm is applied, at the same time, the method is modified by using direction restricted criterion, high correlation algorithm basing on three continuous frames and pipeline algorithm. By this way, the small target in strong light level background is detected successfully.In conclusion, some innovated methods are developed to solve the difficulties for small target detection in strong light level background. The work of this dissertation has some important reference to the future development of the electro-optical detecting system.