Small target detection algorithm in optical remote sensing image based on non-local context information

Abstract

Optical remote sensing images are affected by a large amount of noise in the image, which results in inaccurate detection results of small targets in the image. Therefore, an algorithm for detecting small targets in optical remote sensing images based on non- local context information is proposed. Using the optical remote sensing image target template as a reference, scanning targets in different directions through the main direction, radial gradient angle features are extracted, pixel pairs between the detection window and the target template are searched. At the same time, a voting algorithm is introduced to obtain the position and direction of the target, thereby realizing the detection of small targets in optical remote sensing images. The experiment selects some images from the HRSCD optical remote sensing image set as the test object. The test results show that the proposed algorithm has short detection time and high detection efficiency for the same object under different optical remote sensing image sizes.