针对数字图像的高精度分类问题, 提出了一种新型数字图像分类算法。在该算法中, 局部二值模式(Local Binary Patterns, LBP)算子被用于数字图像的LBP图构建LBP图的直方图被用于构建 图像样本的特征向量; 大量样本的特征向量构建的训练数据集被送入最小二乘支持向量机(Least Squares Support Vector Machines, LSSVM)进行最优分类模型的构建。在测试数据集的分类测试中 , 对本文所提出算法与传统支持向量机算法、极限学习机算法和Hopfield神经网络方法进行了比较, 在宏查准率、宏查全率和分类时间几个典型性能指标的测试方面, 本文所提出的LBP-LSSVM算法 均表现出了优异的性能。

Aiming at the high-precision classification problem of digital images, a new digital image classification algorithm is proposed. In this algorithm, the local binary patterns (LBP) operator is used to construct the LBP image of the digital image, and then the histogram of the LBP graph is used to construct the feature vector of the image sample. Finally, the training data set constructed by the eigenvectors of a large number of samples is sent to the least squares support vector machines (LSSVM) for the construction of the classification model. In the classification test of test data sets, the proposed algorithm is compared with traditional support vector machine algorithm, extreme learning machine algorithm and Hopfield neural network method. The LBP-LSSVM algorithm presented shows excellent performance in the testing of typical performance indicators such as macro-precision rate, macro-recall rate and classification time.