基于机器视觉的菊花生长自动无损监测技术

Non-destructive measurement system for plant growth information based on machine vision

本研究采用机器视觉及图像处理技术对多株植物生长信息进行了监测研究。该监测系统使用的CCD相机被安装在四周均布10株被测植物的精密旋转云台上,相机的镜头前安装了近红外滤片,在相机顶部安装了近红外光源。旋转云台按一定时间间隔自动定角度旋转,旋转到设定角度后,CCD相机对被测植株进行取像。利用该系统,可以长时间自动连续对多株植物进行监测。为了提高测量精度,实验过程中在被检测对象的茎节叶腋处粘贴反射亮片,在图像分析处理过程中,采用一种基于亚像素和区域匹配的误差消除估计算法EEC图像分析的处理方法,有效地提高了检测精度。并且,通过研究发现在图像匹配过程中,参考模板的最小尺寸应大于15×15pixel。利用该系统对菊花的茎长生长进行了监测分析,发现白天菊花的茎长生长率要小于夜间的生长率。该研究为利用机器视觉技术连续非接触监测植物的生长信息提供了技术理论基础。

A non-destructive measurement system for plural plants morphological characteristics of was developed and constructed.A CCD camera （Sony,XC-77） with an infrared filter and an 850 nm infrared LED lighting source was installed on a stepping motor （Oriental motor,DG130R-ASAA）.The all plants images was captured and stored by rotated the stepping motor 20° according to the time interval that can be preset to any interval 10 min was used in this study.The images that captured by infrared irradiation take the same level of intensity value during day and night.The reflective marker was attached at the node position of a plant;stem elongation up to the node position could be measured by tracking the reflective marker.The image analysis and measurements were performed using an image processing of area-based pattern matching technique.Plant stem node coordinates were computed by the image processing algorithms―the sub-pixel estimation on area-based matching technique.One of the problems in the pattern matching technique is determination of an appropriate reference image size.The result of the research shows that the measurement accuracy became raise if the reference window size is more than 15×15 pixel in this research.The system was used to measure the stem elongation of Chrysanthemum （Dendranthema×grandiflorum ＂Reangan＂）.In the research,the daily pattern of chrysanthemum stem elongation was constructed,and it was observed that the daily stem elongation was greater than that during the night.