SYSTEMATIZATION OF CULTIVATING PROCESSES BY ARTIFICIAL INTELLIGENCE

近年来，工厂生产引起了人们的极大兴趣，这使得我们能够改进复杂的传统工厂生产控制系统。对于工厂生产系统等复杂系统的控制，基于人工智能的计算机控制系统比传统的数学控制系统更适合。对于多任务耕作，还需要一个由多台计算机组成的计算机网络系统。为实现农业上植物生产系统的系统化及其高效控制，本研究开发了由不同类型的计算机组成的计算机集成系统(CIS)。(1)开发了由工作站、人工智能型计算机和多台个人计算机组成的计算机集成系统(CIS)，并与局域网(LAN)相连。(2)利用图像处理技术，可以有效地实现番茄果实形状的模式识别、花腐病的诊断和品质判断。研究了一种传输数据量非常大的植物图像数据的有效方法。将神经网络和遗传算法相结合，开发了一种新型的智能控制系统，该方法首先利用神经网络对植物受环境因素影响的生理过程进行辨识，然后利用遗传算法对辨识出的模型进行仿真，寻找环境因素的最优设定值。将该控制系统应用于实际的番茄栽培系统。控制结果表明，该系统具有良好的营养生长和生殖生长。(4)因此，本研究开发的植物生产计算机辅助系统对植物栽培这样一个复杂、不确定的系统的最优控制是非常有效的。

In recent years, there has been great interest in plant factory which enables us to improve the complicated traditional control system for plant production. For the control of such complex systems as plant production systems, a computer-based control system based on artificial intellgence is more suitable than traditional mathematical control system. For multi tasks in cultivation, a computer network system composed of several computers is also necessary. In this study, a computer integrated system (CIS) composed of different several types of computers was developed in order to realize the systematization of plant production system and its efficient control in agriculture.(1) A computer integrated system (CIS) composed of workstation, artificial intelligence-typed computer and several personal computers, which are connected with local area network (LAN), was developed.(2) The use of image processing allows a pattern recognition of fruit shape, diagnosis of blossom-end rot and of quality judgment of tomatoes to be efficiently implemented. An effective way for transmitting the image data of plants, which have a very large data, was examined. The completion of image data shows a good transmission performance.(3) A new intelligent control system combining with neural networks and genetic algorithms was developed.In the method, the physiological processes of a plant, as affected by environmental factors, are at first identified using neural networks and then optimal setpoints of environmental factors are searched for through simulation of the identified model using genetic algorithms. This control system was applied to an actual tomato cultivating system. The control results showed a good vegetative growth and reproductive growth.(4) Thus, a CIS for plant production developed in this study was shown to be very effective on the optimal control of such a complex and uncertain system as plant cultivating system.