基于边缘信息增强的CNN行人检测算法

Pedestrian detection CNN algorithm based on edge enhancement

行人检测是模式识别与分类的经典问题,一直受到研究人员广泛的关注。卷积神经网络作为深度学习的重要模型,在解决行人检测问题上有着良好的效果。基于以上背景,文章提出一种利用双输入卷积神经网络结合图像边缘强化特征进行行人检测的方法。该方法使用Canny边缘检测算子对图像进行边缘检测,得到完整的行人边缘信息,并将其与完整的行人图像分别作为卷积神经网络输入,达到边缘增强的目的,最后通过对卷积神经网络结构参数的优化,实现对行人的检测分类,经在标准行人样本集进行测试,证明了文章算法的有效性。

Pedestrian detection is a classic problem of pattern recognition and classification, which has been widely attracting the attention of the researchers. Convolution neural network as an important model of deep learning, has a good result to deal with the problem of pedestrian detection. Based on this, this paper puts forward a kind of double input convolution neural network combined with the feature of image edge reinforcement method for pedestrian detection. Using the improved Canny edge detection of image edge detection operator, we get complete pedestrian edge information, with the complete pedestrian images as two input convolutional neural network respectively, and achieve the goal of edge enhancement. Finally through optimizing the parameters of convolutional neural network structure implementation of pedestrian detection classification is achieved, and finally the standard pedestrian sample is set to test to prove the effectiveness of the algorithm in this paper.