Research on Image Recognition Algorithm Based on Deep Convolution Neural Network
Lu Deng · Academic Journal of Computing & Information Science · 2020
In order to improve the performance and robustness of traditional image recognition algorithms based on manually designed features, this paper analyzes and studies convolutional neural network, and constructs a deep convolutional neural network based on AlexNet network. The method of converting single-channel depth data into three-channel data is adopted to solve the problem that the model can only accept three-channel data during fine tuning, thus realizing the object recognition function on the object database. The experimental results show that the deep convolution neural network algorithm has better convergence effect, faster convergence speed, higher recognition rate and stronger generalization ability. At the same time, the application of deep convolution neural network algorithm avoids the local optimal problem in the training process and further improves the training speed.