基于卷积神经网络的图像识别方法

Image Recognition Method Based on Convolution Neural Network

为了提高图像识别的全面性及准确性,研究了一种基于卷积神经网络(Convolutional Neural Network,CNN)的图像识别方法。该方法利用萤火虫算法获取分割阈值,实现图像目标和背景的分割;利用灰度共生矩阵和基于加速分割测试的特征(Features From Accelerated Segment Test,FAST)算法提取图像纹理和角点特征;以特征为输入,利用卷积神经网络实现目标类别识别。测试结果表明,设计的基于CNN的识别方法的F1分数为最大值,均在0.8以上,能够更全面、更准确地识别图像中的目标类型。

In order to improve the comprehensiveness and accuracy of image recognition, an image recognition method based on Convolutional Neural Network(CNN) is studied. The firefly algorithm is used to obtain the segmentation threshold to segment the image object and background. The gray level co-occurrence matrix and Features From Accelerated Segment Test(FAST) algorithm are used to extract image texture and corner features. Taking the feature as the input, the target category recognition is realized by using convolutional neural network. The results show that the F1 scores of CNN based recognition methods are the maximum, all of which are above 0.8, indicating that the method studied can recognize the object type in the image more comprehensively and accurately.