基于深度相机的水稻穗层高度检测研究

Research on rice spike layer heights detection based on depth camera

为实现水稻穗层高度检测,基于深度相机对水稻穗层高度计算方法开展研究,以期为水稻割台高度控制、收割机智能化提供技术支持。为选择一种适用于水稻穗层分割的颜色空间,分别在RGB、HSV、LAB和i1i2i3颜色空间中对成熟水稻图像进行3D可视化,通过观察水稻穗层像素分布分析其可分离性;在此基础上,基于多阈值分割法、小联通区域移除、图像腐蚀等方法对水稻穗层进行提取;提出一种基于深度相机的水稻穗层高度(包括穗层最高高度H_(hs)、最低高度H_(ls))计算方法,并分别在实验室及田间进行验证试验。结果表明:水稻穗层在HSV颜色空间中具有较好的可分离性,所采用的图像分割、处理方法可有效提取水稻穗层;水稻穗层高度计算方法的实验室测试结果显示,H_(hs)与H_(ls)的计算误差均小于1.5%,田间测试结果显示H_(hs)与H_(ls)的计算误差分别为1.6%和18.2%。

In order to realize rice spike layer height detection,a research on rice spike layer height calculation method based on depth camera was carried out to provide technical support for the height control of harvester header.In order to select a color space suitable for rice spike layer segmentation,this paper visualized mature rice images in RGB,HSV,LAB and i1i2i3 color spaces in 3D space respectively,and analyzed the separability of rice spike layer by observing pixels distribution.The rice spike layer was extracted based on the multi-threshold segmentation method,small joint region removing and image erosion.A spike layer heights calculation method was proposed based on a depth camera and the method was validated in the laboratory and in the field.The results showed that the rice spike layer had the best separability in HSV color space,and the adopted image segmentation and processing methods could effectively segment the rice spike layer.The laboratory test of the developed calculation method showed that the calculation errors of both H_(hs)and H_(ls)were less than 1.5%,and the field test showed that the calculation errors of H_(hs)and H_(ls)were 1.6%and 18.2%,respectively.