CAJ | 학술논문

视频电子交通警察系统中，常见的人行横道灯识别方法易受光线变化等干扰影响其正确率。提出一种基于分类器的识别方法，首先利用颜色信息和几何约束粗过滤，再提取HOG特征，对PCA降维后的HOG特征训练SVM分类器进行分类。实验表明，该方法检测效果较好。
시빈전자교통경찰계통중，상견적인행횡도등식별방법역수광선변화등간우영향기정학솔。제출일충기우분류기적식별방법，수선이용안색신식화궤하약속조과려，재제취HOG특정，대PCA강유후적HOG특정훈련SVM분류기진행분류。실험표명，해방법검측효과교호。
In an eIectronic poIice system based on video capture,the common Cross WaIk signaIs recognition aIgorithm are sus-ceptibIe to Iight changes such as interference affect its accuracy.In this paper,a cIassifier based Video Cross WaIk signaIs detect aIgorithm is proposed.FirstIy,using coIor information and geometric constraints fiItering,then extracting HOG features and its high dimensionaI feature vector can be reduced by projecting into Iower dimensionaI space using PrincipaI compo-nent anaIysis.FinaI y,a SVM cIassifier trained to judge the Candidate region.