光照背景下自相关累积泰勒展开的弱亮点检测

Weak Spot Detection Based on Autocorrelation Cumulative Taylor Expansion in Light Background

在光照背景下,弱亮点模型因为受到光照色差的干扰,导致对其检测较为困难。通过对光照背景下的弱亮点模型进行检测,特别是对光照背景下运动人体目标检测,是实现智能视频监控的基础工作。提出一种基于自相关累积泰勒展开的弱亮点目标图像检测算法,首先对光照背景下的弱亮点行人进行目标角点检测与预处理,对每个图层的自相关累积特征进行泰勒展开分离,形成原始的图层自相关累积泰勒展开库,求得小邻域内的亮度变化值,实现对光照背景下的弱亮点运动行人图像的角点检测,并作为前置处理算子,实现基于自相关累积泰勒展开的弱亮点行人检测算法改进。仿真实验表明,该检测算法得到的光照背景下的人体的轮廓特征得到准确凸显,检测性能较优,精度较高,鲁棒性好,在智能视频监控等领域具有重要的应用价值。

In the light of background, weak highlight model because of light color interference, leading to more difficult to detect. Through the detection of a weak spot model to the light conditions, especially for light moving human target detec?tion under, is the basis for the realization of intelligent video surveillance. Proposed an expansion of autocorrelation cumula?tive Taylor algorithm of target image detection based on weak spot, the first corner detection and processing weak spot pe?destrians is obtained in light background. Each layer of the autocorrelation accumulation awakened Taylor separation is tak?en, forming the original layer autocorrelation cumulative Taylor library, the brightness change small neighborhood values, corner weak spot moving pedestrian image on light background detection, it is taken as a pre-processing operator to achieve improved weak spot pedestrian detection algorithm based on Taylor expansion of autocorrelation accumulation. Sim?ulation results show that, the contours of the body characteristics of the detection algorithm of light background obtained prominent, it has a better detection performance, high accuracy, good robustness, it has important application value in the fields of intelligent video surveillance.