基于卷积神经网络的目标检测与识别

Target Detection and Recognition Based on Convolutional Neural Network

针对计算机对图像或者视频中目标的识别和检测,提出了基于区域建议网络和卷积神经网络的目标检测识别算法。这种算法由生成建议框的卷积神经网络、用于目标检测的Fast R-CNN网络和使边界框回归更精确的LocNet网络构成。利用该算法对VOC2007数据集进行训练和测试,并与其他的卷积神经网络进行比较,实验数据显示,以Fast R-CNN网络为基础,结合RPN网络和LocNet网络可以极大提高目标检测识别的速率和准确率。

Aiming at the recognition and detection of objects in images or video by computer,a target detection and recognition algorithm based on regional recommendation network and convolution neural network is proposed. This algorithm consists of a convolution neural network(region proposal network,RPN),a Fast R-CNN network for target detection,and a more precise LocNet network for the boundary box regression. Using the algorithm to train and test the VOC2007 data sets,and compared with other convolution neural networks,experimental data show that based on Fast R-CNN network,combining RPN network and LocNet network can greatly improve the speed and accuracy of target detection and recognition.