图像识别中的卷积神经网络应用研究

Research on Application of Convolution Neural Network in Image Recognition

传统图像识别方法存在自适应能力弱的问题,如果待识别对象存在较大残缺或者其他外在噪声干扰,模型则无法获得理想结果。最早在图像处理中成功应用的深度学习是人工智能中非常重要的部分。在图像处理中,带有卷积结构的多层网络的卷积神经网络被加拿大教授及其小组成员提出并优化。在其过程有了突破性发展的情况下,利用卷积神经网络完成了图像识别的设计用以增加模型对图片的识别准确率和在线运算速度,同时减少图像大量特征的提取工作,在识别系统中通过运用随机梯度下降法对系统进行优化,加快模型收敛。根据试验结果,采用卷积神经网络设计的训练模型,对数据集识别的准确率可达到96%,为大规模图像分类更好地发展提供基础支持。

The traditional image recognition method had the problem of weak adaptive ability,if the object to be recognized had large defects or other external noise interference,the model could not obtain the ideal results.The first successful application of deep learning in image processing was a very important part of artificial intelligence.In the case of image processing,convolution neural network with convolution structure was proposed and optimized by Canadian professor and his team members.Under the breakthrough development of the convolutional neural network,the design of image recognition was completed by using the convolution neural network to increase the accuracy of the model recognition of the picture and the speed of online operation,and to reduce the extraction work of a large number of image features.In the identification system,the stochastic gradient descent method was used to optimize the system and accelerate the convergence of the model.According to the experimental results,by using the training model designed by convolution neural network,the accuracy of data set recognition could reach to 96%,which provided basic support for the better development of large-scale image classification.