Study on the Development of Neuro-Chip for the Quality Evaluation of Brown Rice

利用神经网络对糙米的品质评价进行了模拟，作为开发用于品质评价装置的神经处理器的第一步。该系统根据稻米的光学特性，将糙米按外观分为五类。通过改变BP网络的隐含单元数和学习周期进行了仿真，在仿真的基础上，明确了神经网络的判断水平与有经验的大米品质评价员的信息处理水平相当，神经网络用于基于糙米外观的品质评价是有效的。此外，开发了BP网络程序作为神经处理器的程序，并进行了C语言编写的程序向机器语言的转换。

A simulation of the quality evaluation of brown rice using neural network was carried out as a first step into the development of a neuro-processor for the quality evaluation apparatus. The simulation classifies brown rice to five quality categories based on external appearance according to optical characteristics. The simulation was conducted by changing the number of hidden units and learning cycles of the back-propagation network.Based on the simulation, it was clarified that the judgment level of the neural network was nearly equal to that of information processing of an experienced rice quality evaluator, and the use of neural network was effctive for quality evaluation based on external appearance of brown rice.Furthermore, the BP network program was developed as a program for the neuro-processor and the preparation to converse the program written by C language to the machine language was also conducted.