基于深度卷积神经网络的图像目标检测

Image Object Detection Based on Deep Convolutional Neural Network

针对从动态背景提取前景目标较差的问题,提出了一种基于卷积神经网络的图像前景目标检测方法。首先,基于传统的卷积神经网络构建了特征提取的网络模型,然后利用反卷积和金字塔池化,解决了传统卷积神经网络VGG-Net只能对整张图片分类以及只能输入固定尺寸图象的缺陷。针对R-CNN和SPP-Net网络模型提出了一种优化的boundingbox选择方法,使得对检测目标的定位更快更准确。在实际应用中,能够获得更好的前景目标检测效果,为后续的视频分析任务的研究提供了更好的条件。

To address the poor results of foreground extraction from dynamic background,a method of image foreground object detection based on deep convolution neural networks is proposed in this paper.Firstly,it constructs a network model based on traditional convolutional neural network to extract feature map,Then,using the deconvolution method and pyramid pooling solved the problem that the traditional convolution neural network VGG-Net can only classify the entire picture and the model can only receive the fixed size image.