Abstract: Tobacco silos are important equipment used to store tobacco leaf materials in the primary department of a cigarette factory. In order to solve the problem of frequent start and stop of buffering equipment such as tobacco feeder due to large leaf throughput fluctuations under traditional leaf feeding mode, a control method for tobacco throughput from tobacco silos based on the point cloud reconstruction and fuzzy neural network was proposed. Firstly, a 2D laser radar installed on the spreading vehicle in the silo was used to scan the tobacco leaf materials stored in the silo to obtain their 2D point clouds. The acquired 2D point cloud was processed by converting into a coordinate system, followed by filtering, registration and splicing to reconstruct a corresponding 3D point cloud of the material. The distribution information of the material volume was obtained by the slicing method. Finally, the fuzzy neural network control algorithm was applied dynamically by adjusting the operating frequency of the bottom belt motor in the silo. The proposed method was verified on the primary production line in Yanji Cigarette Factory. The results showed that the volume measurement error of the method was less than 2.0% for the standard boxes placed in the silo and less than 3.0% for the leaf material stored in the silo. By using this method, the tobacco throughput from the silo was more uniform and the average downtime of the feeder and other equipment was reduced from 13 to 9 times per production batch. This method provides technical support for extending the life of cigarette makers.