基于SVM的复杂背景条码检测算法研究

发布时间：2018-04-26 05:32

本文选题：条形码 + 机器学习　；参考：《浙江大学》2017年硕士论文

【摘要】：随着物联网技术的崛起,条形码技术也在我们的生活中得到了越来越广泛的应用,大到物流运输、仓储管理,小到手机支付、数据下载,条码技术正与我们的生活日益息息相关,而实现复杂背景下的条码的快速检测,将进一步挖掘条码的潜力,为我们的日常生活创造更多的财富和便捷。本文针对现有的条码定位算法,指出了它们共同存在的一些问题:首先,复杂的背景对条码的识别存在一定的干扰,如表格、文字等纹理易与条码混淆;其次,由于不同的条码具有不同的编码格式和纹理,所以不能同时识别图片中多种码制的条码。所以本文提出了一种基于支持向量机的条码定位算法,其过程如下:首先,手工制作条码图像的样本,其中正样本要包含尽可能高比例的条码区域,负样本需要包含各种复杂的背景;其次,使用改进的局部二值模式(LBP)提取条码样本图像的特征,进行训练,输出分类器;再者,采用合理步长的滑动窗口技术对图像进行扫描分块,并且判断这些图像子块所属的分类;最后,聚集所有属于条码区的子块,使用本文提出的算法,将连通的子块进行合并。实验证明,本文提出的条码检测算法在时间复杂度和定位准确率上,都能得到一个较好的效果。
[Abstract]:With the rise of the Internet of things technology, barcode technology has also been more and more widely used in our life. The technology of logistics transportation, storage management, small to mobile payment, data downloading, bar code technology are increasingly closely related to our life, and the rapid detection of barcode in complex background will further excavate the potential of bar code. In order to create more wealth and convenience for our daily life, this paper points out some problems that exist in the existing barcode location algorithm. First, the complex background has some interference to the identification of bar code, such as the form, the text and other textures, which are easily confused with the bar code; secondly, because different bar codes have different codes. Therefore, a bar code location algorithm based on support vector machine is proposed in this paper. The process is as follows: first, the sample of the bar code image is made by hand, in which the positive sample contains as high as possible bar code area, and the negative sample needs to contain a variety of complex background. Secondly, we use the improved local two value model (LBP) to extract the features of the bar code sample image, train and output the classifier. Furthermore, the image is scanned by the sliding window technique with reasonable step length, and the classification of these sub blocks is judged. Finally, all the sub blocks belonging to the bar code area are gathered, and the algorithm proposed in this paper is used. It is proved that the bar code detection algorithm proposed in this paper can get a better effect on the time complexity and the location accuracy.

【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.4

【参考文献】