Development of PF Systems Using Field Environmental Sensors based on Multi-spectrum Machine Vision

这项研究的目标是开发一种机器视觉，它可以检测作物状况。作物状况，包括作物健康和生长状况，是促进精准农业的最重要信息之一。该视觉系统能够通过计算叶色指数(LCI)和植被覆盖率(VCR)来估计由作物高度和叶面积组成的作物生长参数和叶绿素含量。最后，将所开发的视觉系统安装在无人直升机上。与一般的地面系统相比，直升机基地传感系统能够有效地收集田间信息。为了实时感知作物状况，并通过及时的信息高效地进行田间作业，研制了实时视觉传感器。R-G-NIR机器视觉(MS2100，Duncan Tec.)并将其应用于视觉传感器硬件。在这项研究中，一个估计叶绿素含量和作物生长的因子使用了L…更多CI和VCR作为输入参数。被定义为叶面积与整个图像面积之比的VCR与作物高度有显著的相关性。此外，还研究了视觉信息与产量的关系。定义了LCI与VCR乘积的实际生长指数(AGI)来表示叶绿素含量的绝对体积。田间试验表明，产量与产量的相关系数R^2为0.684，达到极显著水平。因此，视觉系统可以在收获前估计产量。基于无人直升机的传感平台&本研究开发了一种系统，可以生成安装在无人直升机上的视觉传感器(MS2100)获取的作物状况地理信息系统地图。对于本研究中使用的无人直升机，采用了RTK-GPS作为定位传感器，并在直升机上安装了提供姿态(横倾角和俯仰角)的惯性传感器。当直升机上的视觉传感器获取图像时，图像中会出现因直升机姿态变化而引起的声学失真。为了消除这种失真，迫切需要通过将图像坐标转换到全局坐标来进行几何校正。通过对图像进行几何校正，可以生成最大误差为29厘米的地图。这一精度足够高，足以利用这种直升机基地传感系统来进行精准农业。较少

The objective of the research was to develop a machine vision, which could detect crop status. Crop status including crop health and growth was one of the most important information for promoting precision agriculture. The vision system was able to estimate both chlorophyll content and crop growth parameter composed of crop height and leaf area by calculating Leaf Color Index (LCI) and Vegetation Cover Ratio (VCR) from the image. Finally, the developed vision system was installed on an unmanned helicopter. The helicopter-base sensing system was able to effectively gather field information compared to a general ground-base system.<Development of a real-time vision sensor>In order to sense the crop status and efficiently conduct field operation through the timely information, the real-time vision sensor was developed. The R-G-NIR machine vision (MS2100, Duncan Tec.) was adopted to the vision sensor hardware. In this research, an estimator of the chlorophyll content and crop growth used L … More CI and VCR as input parameter. The VCR defined as ratio of leaf area to whole image area had significantly high correlation with crop height. In addition, the relationship between vision information and yield was also investigated. Actual Growth Index (AGI) which was product of LCI with VCR was defined to express the absolute volume of chlorophyll content. High correlation between AGI and yield, which was R^2 of 0.684, was obtained in the filed experiment. Therefore, it was concluded that the vision system could estimate yield before harvest.<Sensing platform based on an unmanned helicopter>This study developed a system that can generate a GIS map regarding crop status obtained by the vision sensor (MS2100) mounted on an unmanned helicopter. As for the unmanned helicopter used in this research, an RTK-GPS was adopted as a positioning sensor, and an inertial sensor that provides posture (roll and pitch angles) was installed in the helicopter. When obtaining the image by the vision sensor on the helicopter, sonic distortions caused by change of helicopter's posture arise in the image. In order to remove this distortion, geometric correction by converting from image coordinate to global coordinate was badly needed. By applying geometric correction to the image, it was possible to generate a map including maximum error of 29 cm. This accuracy was high enough to utilize this helicopter-base sensing system to Precision Farming. Less