中文摘要：

为解决人体检测运算效率低的问题,提出一种基于检测窗口估算的快速人体检测方法。假定人体窗口的中心纵坐标服从正态分布,正态分布的均值和方差服从伽玛分布;采用贝叶斯学习方法更新每一帧图像上伽玛分布的4个超参数,计算人体窗口中心纵坐标的候选区域,采用二次多项式计算每一个人体窗口中心纵坐标对应的检测窗口尺寸范围;在约束的检测窗口下提取方向梯度直方图特征,采用支持向量机进行特征分类,检测人体目标。仿真结果表明,该方法的真正率指标高、假正率指标低、检测耗时少。

英文摘要：

For solving the problem of low efficiency of human detection, a fast human detection method based on detection win-dows estimation was proposed. This method assumed that the vertical center of human window belonged to normal distribution, and the mean and variance of the normal distribution belonged to Gamma distribution. Four hyper-parameters of the Gamma dis-tribution on each frame were updated using Bayesian learning method, and candidate regions of vertical were calculated through quadratic polynomial. The histogram of oriented gradients features on the constraint detection windows was extracted, and fea-ture classification was executed using support vector machines to detect human targets. Simulation results show that the pro-posed method has higher true positive rate, lower false positive rate,and less time-consuming.