發(fā)布時間：2018-11-25 12:45

【摘要】：為提高多類支持向量機分類器對多目標(biāo)的分類準(zhǔn)確度,提出一種結(jié)合無向圖模型優(yōu)化的多類支持向量機分類器。首先,利用余弦測度計算訓(xùn)練數(shù)據(jù)之間的相似度,構(gòu)建包含訓(xùn)練數(shù)據(jù)和相似度矩陣的無向圖模型,求解相似度約束矩陣。然后,將相似度約束矩陣引入多類支持向量機求解的目標(biāo)函數(shù),構(gòu)建優(yōu)化的多類支持向量機分類器。最后,將優(yōu)化的多類支持向量機分類器用于智能交通領(lǐng)域,結(jié)合梯度方向直方圖特征檢測行人和車輛目標(biāo)。實驗表明,該方法檢測行人和車輛目標(biāo)的錯誤率低于經(jīng)典的多類支持向量機分類器和目前主流的目標(biāo)檢測方法。
[Abstract]:In order to improve the classification accuracy of multi-class support vector machine classifier, a multi-class support vector machine classifier combined with undirected graph model optimization is proposed. Firstly, using cosine measure to calculate the similarity between training data, an undirected graph model including training data and similarity matrix is constructed, and the similarity constraint matrix is solved. Then, the similarity constraint matrix is introduced into the objective function of multi-class support vector machine, and the optimized multi-class support vector machine classifier is constructed. Finally, the optimized multi-class support vector machine classifier is used to detect pedestrian and vehicle targets in the intelligent transportation field, combining with gradient direction histogram features. Experiments show that the error rate of this method for detecting pedestrian and vehicle targets is lower than that of classical multi-class support vector machine classifiers and the current mainstream target detection methods.
【作者單位】： 常熟理工學(xué)院計算機科學(xué)與工程學(xué)院;鄭州成功財經(jīng)學(xué)院信息工程系;湖北大學(xué)計算機與信息工程學(xué)院;
【基金】：江蘇省高校自然科學(xué)研究項目(12KJB520001)
【分類號】：U495;TP391.41
，

本文編號：2356155