Abstract:

Sample preparation is the basis of machine learning，which is directly related to the final performance of the algorithm for the object recognition.It is a resource-consuming task with trivialities.Thus，an automatic method of sample preparation is proposed，which includes two stages：coarse-sample preparation stage and fine-sample preparation stage.Coarse-sample preparation stage aims to collect all samples according to the current standard based on image segmentation algorithm；subsequently，fine-sample preparation selects the boundary samples from the coarse samples based on SVM to reduce the sample scale and ensure the high learning efficiency of machine learning.Applying this method to vehicle recognition，the experiment results demonstrate that the proposed method is effective and efficient and it has great value to be applied and popularized.

Key words: training sample, sample classification, boundary sample, Support Vector Machine（SVM）

摘要：

样本的准备是机器学习的基础，直接关系到算法对图像目标物的最终识别性能，也是一项非常繁琐和耗资源的任务，为此，文中提出一种样本自动化准备方法，分两个阶段：粗精度样本准备阶段和细精度样本准备阶段。粗精度样本准备阶段基于图像分割算法收集符合标准的样本，细精度样本准备阶段基于SVM方法选择边界样本，以减少样本规模，确保机器学习过程中对训练样本学习的高效性。提出的方法应用于车辆识别中，实验数据表明了该方法的有效性和高效性，具有良好的应用和推广价值。

关键词: 训练样本, 样本分类, 边界样本, 支持向量机