Intelligent 3-D recognition system considering the geometric characteristics of plants and structures in agricultural production systems

到目前为止，植物栽培系统一直被视为一个集中的常数系统，尽管它们被定义为一个分布式常数系统。本研究的目的是将分布式恒方法的思想引入植物栽培系统，并利用三维图像处理系统和人工智能技术实现植物栽培系统的智能机械化一种结合了懒人测距仪和机械手的三维测量系统被开发出来，可以对任何类型的植物和水果的形状进行高度精确的测量利用三维测量系统获取任意类型植物和水果的距离图像，建立植物和水果形状数据库对于实时测量，基于多边形技术开发了一种简单的算法，使我们能够快速测量水果（甜瓜，苹果，番茄，黄瓜等）开发了一种通过局域网或广域网连接的计算机之间的数据通信系统。该系统可以处理大量的数据，如水果的三维图像，并利用图像数据的压缩技术在一台计算机之间快速传输利用虚拟现实技术对植物工厂植物栽培系统进行结构和空间优化设计开发了一种结合神经网络和专家系统的新识别算法，可以正确识别任何水果的形状。在这种识别中，使用了三维距离图像的数据这里新开发的一些技术将有助于工厂生产系统的机械化和电脑化。

Until now, plant cultivating systems have been treated as a concentrated constant system even though these are defined as a distributed constant system. The aim of this study is to introduce the idea of a distributed constant approach to plant cultivating systems and then realize those intelligent mechanization by using a 3-dimensional image processing system and artificial intelligence.1 A 3-dimensional measuring system combining a lazier range finder and a robot hand was developed, which allowed a highly accurate measurement of any types of plant and fruit shapes.2 Range images for any types of plants and fruits were obtained using a 3-dimensional measuring system, and then a data base for the plant and fruit shapes were built.3 For real-time measurement, a simple algorithm that enables us to rapidly measure fruits (muskmelon, apple, tomato, cucumber, etc.) was developed based on a polygon technique.4 A data communication system between computers, which are connected with LAN or WAN,was developed. This system can treat a great deal of data such as three dimensional image for fruits and transmit it rapidly from one computer to another ones using a compression technique of the image data.5 Optimal structural and space design for plant cultivating system in plant factories were implemented using a virtual reality technique.6 A new recognition algorithm combining neural networks and expert systems, which allows the shape of any fruits to be correctly recognized, was developed. For this recognition, the data of a 3-dimensional range image was used.7 Some techniques newly developed here will be useful for the mechanization and computerization of plant production systems.