基于小波包分解和分形法的人群密度等级分类算法研究

Population density of classification algorithm based on the wavelet packet decomposition and fractal method

为了更好地解决中高密度环境下人群密度估计问题,提出了基于小波包和分形法的人群密度等级分类算法。区别于单一的分形维数表征方法,利用小波包分解后子图的分形维数和原图分形维数共同表征目标人群纹理,最后通过支持向量机实现分类。以查全率和查准率2个指标对密度估计结果进行评价,该算法的查全率和查准率最高,且运行时间最短。

In order to better solve the problem of the crowd density estimation in high density environment,the method based on the wavelet packet and fractal is proposed.Different from the single fractal dimension characterization methods,the fractal dimension and the fractal dimension of the subgraph are used to represent the texture of the target population by using wavelet packet decomposition,and finally the classification by support vector machine（SVM）is realized.Two indicators of recall ratio and precision ratio are selected to evaluate the result of the density estimation,the call ratio and the precision ratio of the algorithm are the highest,and the run time is the shortest.