【摘要】 卷积神经网络是一种基于局部权值共享的深度学习网络模型,近些年来被提出并广泛应用于语音识别、图像识别、图像分割、自然语言等领域。文章分析了目前主流的卷积神经网络模型及其实现方法,并在残差神经网络的基础上加以改进,通过增加浅层网络到深层网络的多个通路,将粗糙的背景信息和具有丰富细节的纹理信息加以融合,用以增强深层的卷积网络层的特征信息,从而对具有复杂背景信息和纹理的渣土车图像进行识别与分类。实验结果表明,本方法能进一步提升深度卷积网络对于渣土车顶部覆盖率的分类准确率。更多还原

【Abstract】 Convolutional neural network is a deep learning network model based on local weight sharing. It has been proposed and widely used in speech recognition, image recognition, image segmentation, natural language and other fields in recent years. This paper analyzes the current mainstream convolutional neural network model and its implementation method. Based on the residual neural network, the coarse background information and the texture information with rich details are combined to increase the characteristics of the deep convolution network layer by adding multiple channels from the shallow network to the deep network, which can identify and classify muck car images with complex background information and texture. The experimental results show that the proposed method can further improve the classification accuracy of the depth convolution network for the top coverage of muck.更多还原