摘要
对视频或图像中的行人进行检测使用HOG特征与支持向量机SVM相结合的方法,取得了良好的效果,但是由于HOG特征的计算量非常大,难以满足实时检测的需要,文章针对这一情况提出并实现一种在GPU环境下使用并行加速策略的高效行人检测方法。实验证明这种方法快速有效,大大地提高了行人检测的效率。
The pedestrian detection in video or images has got good effect by combining HOG feature with support vector machine, but due to the computational complexity of calculating HOG feature is too large to meet the needs of real-time detection. In this background, this paper proposes and implements an efficient pedestrian detection method using a parallel acceleration strategy based on GPU environment. Experimental results show that this method is fast and efficient, greatly improved the efficiency of pedestrian detection.
出处
《电子技术（上海）》
2014年第3期1-4,共4页
Electronic Technology
