ISSN 1000-1239 CN 11-1777/TP

• 论文 • 上一篇    下一篇


徐 晶 刘 鹏 刘家锋 唐降龙   

  1. (哈尔滨工业大学计算机科学与技术学院 哈尔滨 150001) (
  • 出版日期: 2009-11-15

A Detection Algorithm for Rain-Affected Moving Objects

Xu Jing, Liu Peng, Liu Jiafeng, and Tang Xianglong   

  1. (School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001)
  • Online: 2009-11-15

摘要: 提出一种在户外受雨滴影响的视频场景中检测运动目标的方法.在R,G,B空间构建雨滴在视频中的成像模型,该模型可以计算受雨滴影响像素的亮度变化值.能够有效克服现有模型只能针对某些特定类型雨滴进行辨识的局限性.在使用基于颜色信息的雨滴成像模型基础上,提出运动目标检测函数,此函数可以有效抑制雨滴产生的干扰.实验结果表明,提出的雨滴成像模型和相应的检测函数与现有模型比较,能够适用于多种不同受雨滴影响的图像序列采样环境,对于运动目标具有更好的分辨能力,并有更强的鲁棒性.

关键词: 成像模型, 运动目标检测, 复杂场景, 视频处理, 户外视觉与天气

Abstract: The degradation effects caused by raindrops are complex, which brings difficulties to image processing algorithms and objects detection. It is necessary to distinguish objects from the various fast moving raindrops or the blur effects caused by rain streaks when capturing videos in the rain. An algorithm is proposed to detect the moving objects in rain-affected videos which are obtained by the outdoor vision system. The chromatic property is used to structure the imaging model of raindrops and an imaging equation that includes the chromatic information of R, G and B channels is presented. The proposed imaging model is a measurement to describe the intensity changes of the pixels which are affected by rain streaks. This imaging equation has capability to explain the limitation of the existing models which can only process some special kind of raindrops, and also the imaging model is able to overcome those limitations. Thereafter, a method based on this imaging model is proposed to detect the rain-affected moving objects. Finally, the experimental results show that the imaging model has capability to describe the imaging properties of raindrops, and it has stronger robustness and better performance for outdoor vision systems than the traditional methods.

Key words: imaging model, moving object detection, complex scene, video processing, outdoor vision and weather