本文介绍了笔者开发的面向航空遥感领域的计算机视觉系统π中采用的黑板控制策略, 着重论述了为解决航空遥感影像理解问题而采用的基于神经网络的黑板模型的控制机制。

This paper addresses the problem of neural-network-based control mechanism of Blackboard of image understanding system.Blackboard architecture has been used as a model for intelligent information fusion in π, A feedforward neural network model was proposed as the Backboard control mechanism of π. The blackboard architecture was developed to deal with the difficult characteristics of the speech understanding problem: a vary large search space: erroneous or incomplete input data, and imprecise and/ or incomplete problem-solving knowledge and it has proven to be popular for AI problems. The image understanding system requires a problem-solving model that supports the incremental development of solutions, can apply diverse types of knowledge, and can adapt its strategy to the particular problem situation. The neural-network-based control mechanism of the blackboard can offer efficient control for information extraction by image understanding system π.