电子设计工程
電子設計工程
전자설계공정
ELECTRONIC DESIGN ENGINEERING
2014年
23期
51-53
,共3页
v-FDF%SVM%级联%行人检测
v-FDF%SVM%級聯%行人檢測
v-FDF%SVM%급련%행인검측
v-FDF%SVM%cascade%pedestrian detection
为了提高行人检测性能和速度，本文提出了一种新颖的基于v-FDF和SVM级联的行人检测方法。该方法既融合了v-FDF计算复杂度低和特征提取准确的优势，同时又采用支持向量机的SVM模式，提高矩阵运算速度。本文运用Matlab进行了多组仿真实验，实验结果表明本文的方法性能比基于Adaboost和SVM的方法好，在相同的虚警率下漏检率较低，并且检测速度能达到10帧每秒，基本上能达到实时性要求，可见本文算法检测率高，方法的性能好。
為瞭提高行人檢測性能和速度，本文提齣瞭一種新穎的基于v-FDF和SVM級聯的行人檢測方法。該方法既融閤瞭v-FDF計算複雜度低和特徵提取準確的優勢，同時又採用支持嚮量機的SVM模式，提高矩陣運算速度。本文運用Matlab進行瞭多組倣真實驗，實驗結果錶明本文的方法性能比基于Adaboost和SVM的方法好，在相同的虛警率下漏檢率較低，併且檢測速度能達到10幀每秒，基本上能達到實時性要求，可見本文算法檢測率高，方法的性能好。
위료제고행인검측성능화속도，본문제출료일충신영적기우v-FDF화SVM급련적행인검측방법。해방법기융합료v-FDF계산복잡도저화특정제취준학적우세，동시우채용지지향량궤적SVM모식，제고구진운산속도。본문운용Matlab진행료다조방진실험，실험결과표명본문적방법성능비기우Adaboost화SVM적방법호，재상동적허경솔하루검솔교저，병차검측속도능체도10정매초，기본상능체도실시성요구，가견본문산법검측솔고，방법적성능호。
In order to improve the detection performance and speed, this paper proposes a novel based on v-FDF and SVM cascade for pedestrian detection, and a lot of experiments are conducted with Matlab. This method combines both the advantages of v-FDF that has low computing complexity and accurate feature extraction, and the SVM model that is based on support vector machine, increasing the speed of matrix operations. The experiment results show that the performance of the method in this paper is better than the methods based on SVM and Adaboost, which has lower miss rate under the same false alarm rate, and the detection speed can reach 10 frames per second, can achieve real-time demand basically. So the algorithm in this paper has the high detection rate, and method of performance is good.
