基于改进SURF算法的大规模群体人数统计

Large crowd count based on improved SURF algorithm

为了在高密度大规模群体人数统计的问题上有效的克服遮挡与摄像机透视畸形带来的影响,文中采用了一种基于线性内插透视矫正的SURF(Speeded Up Robust Feature)算法。首先,采用背景差与滑动平均相结合的方式得到人群前景,并通过对二值前景图像的形态学处理进行去噪。其次,对获取到的前景图像进行多特征提取,将传统的灰度共生矩阵特征与SURF算法特征相结合,并通过线性内插权值的透视矫正方法进行摄像畸形矫正,将矫正后的特征值组成了表征人群数目特征的特征向量。从而减少了深度信息丢失而引起的误差,得到了优化的人群特征向量;最后,通过支持向量回归的方式拟合出人群人数统计模板,以此预测监控区域的人数。实验表明文中方法具有较高的准确性,较传统SURF算法准确率有了很高的提升。

The SURF based on the method of Linear Interpolation for camera distortion calibration is a- dopted for high-density crowd counting. The eigenvalues are built on the Gray Level Co-occurrence Ma- trix（GLCM） features and the SURF features. To get the foreground image ,firstly, gray and smooth the in- put image. Then getting foreground image by background subtraction operation and moving average meth- od. And also morphology processing was performed on the binary image to eliminate noise. And then, ex- tracting feature parameters of foreground image. Though the method of linear interpolation, weight values are interpolated to reducethe error, which is caused bycamera distortion calibration. Linear interpolation weights perspective correction method is considered for camera deformity correction. The optimized crowd feature vector can be obtained then. Through the method of support vector regression （SVR）, the crowd number can be forecasted by the training model. The experiment result shows that the method of this pa- per has a higher accuracy than the previous methods.