| Through the fusion of infrared and low-light-level images, the color fusion images which contain the information of them can be obtained. The fusion images can help observers percept the scene comprehensively. However, because the targets in the fusion images which are fused by long-distance infrared and low-light-level images are not obvious, some important information will be lost when targets extraction is adopted, and the perception of the scene will be affected seriously. To solve this problem, a new fusion method based on visual perception is proposed in this paper. The extraction of the visual targets ("what") information and parallel processing mechanism are applied in traditional color fusion methods. The infrared and low-light-level color fusion images based on efficient typical targets learning are achieved.In this paper a new method of typical targets recognition based on the watershed segmentation of regions of interest and BP neural network is studied. It is used to percept "what" information. In the training period, the features of targets and background are described in four aspects:color, gradient, Hu and texture, and BP neural network is applied to train the features of sample points by supervision. So a feature collection of targets and background can be obtained. During the targets recognition period, an image preprocessing based on watershed segmentation of regions of interest is adopted. Decreasing test points selected from the intersected blocks are detected by BP neural network and the category of every block can be acquired. This preprocessing not only maintains the accuracy of recognition, but also reduces the recognition time. A color night-vision method based on visual parallel processing mechanism is also studied in this paper. The targets recognition and image fusion are conducted at the same time, and then the re-fusion of targets and background is conducted. The result is that natural color fusion images with obvious targets are obtained.In order to test the effectiveness of the color night-vision method based on visual perception, a night-vision system based on DM642which is used to fuse double-spectral images is established. The acquisition and fusion of images are completed on this system. Experimental results show that the fusion images obtained by this algorithm not only can improve the detection rate of targets, but also can get rich natural information of background. |
