Abstract: For the shadow of moving objects causes the physical deformation of moving objects in the visual video sequence, affects moving objects detection and tracking; meanwhile, the currently used shadow detection algorithm that always firstly detects the shadow, then splits the shadow, which takes two steps to achieve the purpose of eliminating the shadow, and it is difficult to achieve real-time detection requirements, therefore, this paper presents a fast video-based method to eliminate shadows. The method analyzes the shadow and the HSV color space characteristics of its background, and uses H, S, V, components different characteristics between shadows and moving object to calculate the corresponding threshold. By using this threshold to segment and eliminate shadows, it detects moving targets at one-time in video image sequences, at the same time, eliminates the shadows. Experiments show that in outdoor conditions, the algorithm is accurate and effective.

Key words: shadows elimination, threshold, video image

摘要： 针对在视频序列图像中运动目标的阴影造成运动目标的物理变形，影响运动目标的检测与跟踪；同时，由于目前常用的阴影检测算法，总是要先检测到阴影，再分割阴影，需两步才能达到消除阴影目的，难以达到实时检测要求。为此，提出了一种基于视频图像快速阴影消除方法。该方法通过分析阴影与背景的HSV彩色空间中的特性，利用阴影与运动目标在H，S，V三个分量中的不同特点，计算其相应的阈值，运用该阈值进行分割并消除阴影，能在视频序列图像中一次性检测到运动目标的同时就消除阴影。实验表明，在户外条件下，该算法是准确有效的。

关键词: 阴影消除, 阈值, 视频图像