中文摘要：

目前二维条码定位一般使用几何方法或纹理分析方法，其鲁棒性或实时性较差，尤其是在金属材质表面．针对传统二维条码定位方法的不足，提出了基于机器学习和级联过滤器联立的方法滤除背景区域，结合二维条码的几何性质检测候选区域，然后利用聚类生长法包络二维条码区域．实验结果表明：与传统算法相比，本文的算法对于各种复杂金属背景上的二维条码定位具有很高的鲁棒性与实时性．利用训练后的级联分类器和连通区域判决器，平均定位准确率可达到97％，并且处理时间控制在700ms以内，对金属零件上二维码信息的可靠获取具有重要价值．

英文摘要：

Current detection algorithms include geometric method and texture analysis method. In locating 2D code under various material backgrounds, especially under metal background, geometric method is characterized by poor robustness and texture analysis method by slow processing speed. To solve the drawbacks mentioned above, the inte- gration of machine learning method into cascade filter method is proposed in this paper to filter background areas, then the geometric properties of 2D barcode are used to detect candidate target area, and finally clustering growth method is employed to envelope 2D barcode region. The experiments reveal that, compared with traditional methods, the method proposed in this paper has achieved higher detection rate with better robustness. With the trained cascade classifier and the connected region classifier, the average positioning accuracy of 97% can be achieved and the processing time can be controlled within 700 ms, which has a great value in obtaining reliable information of 2D barcode on metal parts.