A novel hybrid statistical model for camouflaged target detection

Abstract

Camouflage is a defense mechanism employed as a concealing technique to reduce the possibility of the target being identified. In order to complete military tasks accurately, the detection of camouflaged targets in different scenarios plays a crucial role. Traditional target detection systems have the problem of detecting camouflaged targets in complex environmental conditions. There are various challenges in camouflage target detection system development, such as size, random texture, and recognition accuracy. To resolve these issues, a hybrid statistical model is proposed to identify the target from camouflage images. In this work, hybrid statistical model consists of two modules. A feature extraction module is utilized to extract the low-level features of the image based on texture and scale. A segmentation module is utilized to extract the target region and enhance the boundaries based on seed point selection and detection of edges. Further, the use of morphological processes to highlight the target region. Experiments demonstrate that the proposed hybrid statistical model performs well in camouflage detection in different environments.