目的 基于图像识别和结构改进对嘴棒发射机运行中产生的横滤嘴棒进行智能剔除、识别和预警，以提高生产效率。方法 采集滤嘴棒侧面运输图像，对异常图像进行追踪，从中提取异常滤嘴棒特征点，基于图像生成警示信息，同时对滤嘴棒发射机输送轨道背板进行结构优化，能剔除大部分横滤嘴棒。结果 改进后的模拟样机，经过输送轨道背板的横滤嘴棒的剔除率高达约77%，其余异常横滤嘴棒被及时预警，嘴棒发射机的管道卡堵概率降低约60%，设备运行效率显著提高。结论 图像识别和结构改进能有效识别预警和剔除嘴棒发射机运行中产生的横滤嘴棒。

The work aims to carry out intelligent elimination, identification and early warning of the transverse filter tip produced during the operation of a transmitter to improve the efficiency of the production based on image recognition and structure improvement. The lateral transport images of filter rods were collected, and the abnormal images were tracked, from which the feature points of abnormal filter rods were extracted, and the warning information was generated based on the images. At the same time, the structure optimization of the backplane of the transmission track of the filter rods transmitter was carried out to eliminate most of the transverse filter tips. In the improved simulation prototype, the rejection rate of the transverse filter tip was as high as 77% when the transverse filter tip was transported to the conveying track backplane, and the other abnormal transverse filter tip was immediately warned. The blockage probability of the nozzle transmitter pipeline was reduced by about 60%, and the operating efficiency of the equipment was significantly improved. Therefore, image recognition and structure improvement can effectively identify and eliminate the transverse filter tip generated during the operation of the transmitter.