Research on Pedestrian Target Recognition Algorithm in Low Light Environment

Abstract

With the development of the economy and society, security monitoring issues have received a lot of attention from industry and academia. Among them, invading pedestrian targets in low-light environments have the characteristics of high image noise, low contrast, and lack of color information. How to detect pedestrian targets in a low-light environment is a challenging problem in the current surveillance field. Aiming at the lack of texture and color information of monitoring images in low light environments, a pedestrian targets recognition algorithm in low light environments based on the fusion of visible light and infrared is proposed in this paper. First, use the defogging algorithm to enhance the brightness, clarity, and contrast of the visible light image, improve the signal-to-noise ratio of the image and restore the detailed features of pedestrian targets. Second, convert the RGB image to an HSV image to enhance the sensitivity of the algorithm to color. Then stitch the HSV image and the infrared image to form a four-channel HSV-G image. This paper uses infrared imaging's high sensitivity to light to improve the recognition rate of pedestrians in low light environments. Through experiments, the method proposed in this paper has achieved good results on a self-made low-light environments pedestrian target data set. The mAP reached 82%, FPS reached 67. The research results can provide an important basis for the recognition of pedestrian targets in low-light environments.