一种基于视觉定位纠偏的面单特征识别优化

An Optimization of Express Delivery Sheet Feature Recognition Based on Visual Positioning and Rectification

利用视觉识别快递分拣成为主流,然而在识别快递主要特征信息时,因其所处背景的复杂多变,快递的摆放位置随意,导致识别效率较为低下。针对能够在复杂环境下也能较好快捷地识别快递的主要特征信息,提出了几何矫正优化定位目标区域以及图像处理参数的优化来抑制外界的干扰因素。通过图像预处理阶段的各种分析以及形状匹配,寻找目标区域的大致的位置,利用几何向量求取更为精确的位置关系后,再将优化定位后的目标区域图像以动态参数调整,以达到对环境的自适应调整。结果表明相较于没有优化时的结果,同一快递单号在任意角度摆放下,识别准确率提高了17%,不同快递单号任意摆放下的识别准确率提高了28.7%。数据表明,该方法能快速准确地定位识别三段码,具有更高的分拣准确度和稳定性,实现了预期目标。

Visual recognition of express delivery sorting has become the mainstream,but the efficiency is low in the identification of the main feature information of express delivery,due to the complex and changeable background and its arbitrary location.In order to solve this problem,the paper proposes the optimization of geometric correction to locate the target area and the optimization of image processing parameters to suppress external interference factors.The general position of the target region is found through various analysis and shape matching in the pre-processing stage of the image,and the more accurate position relationship obtained with geometric vectors used.The image of the target region after optimized positioning is adjusted with dynamic parameters to achieve adaptive adjustment to the environment.The results show that,compared with the results without optimization,the recognition accuracy of the same tracking number placed at any angle increases by 17%,and the recognition accuracy of different tracking number placed at any angle by 28.7%.The data show that the method is able to locate and identify the three-segment code quickly and accurately,and has higher sorting accuracy and stability.