Abstract: According to the character of fuzzy image edge, noise image, difficult to extract the edge information collected by infrared thermal imager, a new edge detection method was proposed. The method combined LOG operator improved by mathematical morphology with Roberts operator. First the thermal infrared camera images with random noise was filtered using opening and closing operation of mathematical morphological. Then the Laplace algorithm was used for edge detection of thermal infrared images. Second, Roberts operator was introduced to extract edge information. Fusing edge information of images detected by two methods, at the same time, establishing corresponding fusion rule and the threshold condition were applied for this images, the final fused image was obtained. Finally, MATLAB was introduced to perform the experimental simulation for adding salt and pepper noise image. The experimental results show that the method has the advantages of two detection operators and has a good effect in noise suppression and positioning accuracy. Meanwhile the proposed algorithm is insensitive to noise and ideal detection results are obtained.