弱光浓雾背景行人探测方法研究

Research on Pedestrian Detection Method in Weak Light and Dense Fog Background

针对弱光浓雾背景下行人信息模糊,难以探测等问题,提出了一种行人探测方法。采用红外面阵成像技术对弱光浓雾背景下行人进行红外和可见光同光轴成像,通过改进蚁群算法对红外图像进行阈值寻优,完成自适应阈值分割,标定目标位置。提出了在YUV色彩空间下对可见光图像采用改进多尺度Retinex图像增强算法,通过对应红外图像目标位置完成对可见光增强图像目标的探测。实验结果表明,以上算法比改进局部OTUS算法在红外图像阈值分割上表现更好,经过上述算法处理的可见光图像在清晰度和质量上比改进多尺度Retinex算法和改进暗通道先验去雾算法得到了极大的提高,且在主观视觉上具有更好的色彩信息,能取得较好的行人探测结果。

A pedestrian detection method is proposed to address the problem of blurred information and difficult detection of pedestrians in low light and dense fog backgrounds.The method adopts the red outside array imaging technique to image the pedestrians in the low light and fog background with the same optical axis in IR and visible light,and the adaptive threshold segmentation is accomplished by thresholding the IR image with an improved ant colony algorithm to mark the target location.An improved multi-scale Retinex image enhancement algorithm is pro-posed for visible images in YUV colour space to complete the detection of visible enhanced image targets by corre-sponding to the target positions of the IR images.The experimental results show that this algorithm performs better than the improved local OTUS algorithm in threshold segmentation of IR images,and the clarity and quality of visible images processed by this algorithm are greatly improved than the improved multi-scale Retinex algorithm and the im-proved dark channel a priori defogging algorithm,and have better colour information in subjective vision,which can achieve better pedestrian detection results.