The Research Of Artificial Shallow Blindage Target Detection Based On Infrared Image

The exposed ability to the camouflage targets, which is very important for the application of underground nuclear silo, border surveillance, jungle warfare military targets, and the armed forces responsible for the rescue and relief missions and other paramilitary application is essential for the development of the army under the condition of information. The paper focus on how to detect the artificial shallow blindage, which is a typical hidden target, and make use of the information of natural background and the texture of single frame image to detect the target, and put forward a hidden target detection method based on the fusion of fractal measurement results and saliency detection result, the main contents are summarized as follows:A detection method for man-made object is proposed based on the combination of weighted fractal characteristic. The fractal feature of infrared image is extracted to distinguish artificial target and natural background, and two kinds of fractal feature, that are texture roughness and fractal fitting error are weighted and normalized to enlarge the regional differences of artificial and natural areas.In order to retain the highlighted area, a saliency detection method based on double error of reconstruction is proposed through extracting the salient region of infrared image. The super pixel of image edge is extracted as a background template, which is used to construct sparse appearance model and dense appearance model. For each image region, the dense reconstruction error and sparse reconstruction error is calculated firstly, and then the reconstruction error is propagated by context clustered by K-means clustering method, and then the detection results based on the sparse and dense are fused according to the Bayes criterion, and achieve the pixel saliency detection by the integration of multi-scale reconstruction error information and the modified Gauss-basis informationFinally, the shallow blindage target is extracted by the fusion of these two kinds of detecting results and further morphological process.