基于改进高斯混合模型的红外人体目标检测方法

Detection method for human object based on improved Gaussian mixture model

针对高斯混合模型在阴影不显著情况下,容易把随光线突变而变化的背景像素点当作前景目标从而造成目标误检的缺点,提出了一种基于改进的高斯混合模型的红外人体目标检测方法。该方法引入边缘检测信息增强红外人体目标检测效果。首先,该算法利用Canny边缘检测来提取人体目标的边缘信息。然后,以此对每个像素建立高斯混合模型来完成人体目标的检测。实验结果表明,该方法能够有效消除光照突变所产生的阴影影响,提高了检测的准确性。

If shadow was not significant the Gaussian mixture model is easy to regard background pixels which follow the change of the light mutation as foreground objects In order to solve the shortcomings of the false target detection, this papers present a detection method for infrared human based on improved Gaussian mixture model. The method make use of the edge information to enhance the human detection effect.First of all the algorithm uses canny edge detection to extract the edge information of the human target, and then Gaussian mixture model complete the detection of human goals.with it. The experimental results show that this method can effectively eliminate the impact of the shadow of light mutation to improve the detection accuracy.