摘要
在目前人工智能技术的发展与需求加大的背景下,行人检测作为计算机视觉领域中的重要问题,也成为重点的研究方向。结合传统算法支持向量机,提出了一种将改进卷积神经网络的特征提取与传统的支持向量机的分类器相结合的方法。首先对图像构建多尺度图像子块,对子块进行CNN特征提取分类,再对输出特征图用支持向量机进行二次分类。实验表明,该方法对于行人检测具有较高的准确率和召回率,高于传统方法。
In the context of the current development of artificial intelligence technology and increasing demand, pedestrian detec- tion as an important issue in the field of computer vision has also become a key research direction. Combining with the traditional algorithm support vector machine, this paper proposes a method that combines the feature extraction of the improved eonvolutional neural network with the classifier of the traditional support vector machine. Firstly, the multi - scale image sub - blocks are con- structed for the images, and the sub - blocks are extracted and classified by the CNN feature. Then the output feature maps are sub - categorized by the support vector machine. Experiments show that this method has higher accuracy and recall rate for pedestrian detection than traditional methods.
作者
焦佳丽
李雷
JIAO Jiali;LI Lei(College of Science,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)
出处
《电视技术》
2018年第9期1-4,10,共5页
Video Engineering
基金
国家自然科学基金(61070234
61071167
61373137)
关键词
行人检测
支持向量机
HOG特征
卷积神经网络
Pedestria detection
Support vector machine
Histogram of oriented gradient
Convolution neural network
