Stability of cotton flow velocity determines the accuracy of foreign fiber detection system. There are many factors affecting fiber flow flux. In the process of operational control, recent research has been directed towards synthetically and effectively adjusting the relative parameters, and thus achieving a stable and economic foreign fiber detection system in textile sector. In the foreign fiber detection system, the parameters of flow velocity are affected by the temperature, pressure, density, which are also interrelated and redundant information. Based on Clustering Fusion, the design of the flow velocity used in the detection device is provided in this paper. Using captured parameter characteristic information, clustering fusion control, conducts the second fusion for ART-2 and BP neural network, and sends to fusion center, then fusion center process the data using neural network method. Linking with synthesis repository and global data base, different control strategies can be utilized to adjust velocity of flow parameter. The cluster control strategy that keeps output velocity of flow stable are proposed to improve the fiber measure precision in foreign fiber detection system. This system can be used in indigenous foreign fiber detection system, and significantly improve the performance of the entire system.