基于多特征融合和深度学习的图像分类算法

Image classification algorithm based on multiple feature fusion and deep learning

图像特征提取方法以及分类器的选择是影响图像分类精确度的关键因素.传统算法利用单一的图像特征和浅层结构对图像进行分类,算法实现简单但结果精确度不高.针对这一情况,提出基于多特征融合和深度学习的图像分类算法.算法利用颜色矩、LBP和梯度直方图等算法提取图像的颜色、纹理以及形状特征,继而通过融合算法将这些不同属性的特征进行融合,作为深度学习网络的输入层.实验结果表明,相对于单特征浅层分类,算法在保证时效性的同时,图像分类精确度得到了提高,分类效果更加可靠.

The image feature extraction method and the choice of classifier are the key factors affecting the accuracy of image classification.The single feature extraction and shallow structure analysis of the image are simple but the accuracy of the results is not high.In response to this situation,an image classification algorithm based on multifeature fusion and deep learning was proposed,the algorithm extracts the color,texture,and shape features of the image using color moments,gray-level co-occurrence matrix,etc.,and then merges the features of these different attributes through the fusion algorithm as the input layer of the deep learning network.The experimental results showed that compared with the single feature shallow classification,the algorithm of this paper ensures the timeliness,and the accuracy of image classification was improved and the effect was more reliable.