目的 为适应快递包裹外形的多样性，分拣标签的张贴无规律性，提高物流分拣环节无人化和绿色化，降低分拣成本，研制一种异形包裹全方位自动扫描分拣系统。方法 该系统对现有物流传送装置进行结构改进，设计旋转式分拣平台及玻璃板内嵌传送带，通过Solidworks仿真软件构建整体模型，并进行场景仿真，开发基于OpenCv4.5.4-Python、Zbar的多线程ROI模板追踪算法与Blob检测算法结合高斯滤波、模板匹配及透视变换获取条码信息，并搭建基于STM32F407ZGT6输出PWM进行运动控制的异形包裹全方位自动扫描分拣物理样机。结果 实验结果表明，物理样机对异形包裹平均识别时间为0.91 s，平均分拣时间则可稳定在1.78 s左右。结论 该系统可以全方位无死角扫描包裹条码，准确率可达100%，具有良好的稳定性和准确性。

The work aims to develop an all-directional automatic scanning and sorting system for irregular packages, in order to adapt to the diversified shapes of express packages and the irregular posting of sorting labels, improve the unmanned operation and environment-friendly performance of logistics sorting links, and reduce the sorting cost. In the system, the structure of the existing logistics transmission device was improved and the rotary sorting platform and the glass plate embedded conveyor belt were designed. Through Solidworks simulation software, the overall model was constructed and scene simulation was carried out. Multi-thread ROI template tracking algorithm and Blob detection algorithm based on OpenCv4.5.4-Python and Zbar was developed, which was combined with Gaussian filtering, template matching and perspective transformation to obtain bar code information. A physical prototype of the all-directional automatic scanning and sorting system for irregular packages was built based on the output PWM of STM32F407ZGT6. The experimental results showed that the average recognition time for irregular packages was 0.91 s, and the average sorting time could be stable at about 1.78 s. The system can scan the package bar code in an all-round way without dead angle, and the accuracy can reach 100%, which has good stability and accuracy.