基于无人机可见光谱遥感的玉米长势监测

Research on Maize Growth Monitoring Based on Visible Spectrum of UAV Remote Sensing

玉米是我国重要的粮食作物之一,在我国种植规模最大、发展最快。玉米的长势会直接影响到其产量和品质,因此通过对玉米的长势进行有效监测,可以为田间管理、早期产量估算提供宏观的参考信息,为国家和相关部门决策提供重要的参考依据。以无人机为遥感平台,搭载影像传感器构建遥感系统,获取玉米可见光谱遥感影像。利用ENVI软件对获取的玉米冠层可见光谱彩色图像进行几何校正和辐射校正,然后对图像进行彩色图像灰度化和增强处理。利用对农田复杂背景适应能力较好以及具有较强光照适应性的AP-HI算法完成作物分割来提取玉米覆盖度信息。在计算玉米覆盖度时,首先利用AP-HI算法将图像进行分割,并转换为二值图,来去除图像中的土地、水管、道路、作物残渣等背景,以保留玉米的二值图像。图像中的农田存在道路区域,计算实际作物覆盖度时需将其排除。道路区域出现在图像的四个边界以及相对正中的位置处,对这些位置分别进行处理,统计其中黑色像素点的个数,根据像素点个数确定道路宽度,并将道路部分从二值图中去除。去除后的二值图中,白色像素为无作物区域,黑色像素为玉米种植区域,统计黑色像素占总像素的比例,以此确定作物的多少。选取80×80像素值作为单位面积,对处理图像进行分块标记,得到区块数为720,对单位面积的分块进行全区域扫描,每当扫描到一个黑色像素值就将总的统计面积加1,直至扫描到6400个像素点,计算其中含有的总的黑色像素值数目与6400的比值,直至将720个区块黑色像素点占总像素比例统计完全,即可计算图像中黑色像素数与总像素数之比,即为玉米覆盖度。在此基础上,根据实际情况计算玉米冠层孔隙率,并建立覆盖度与叶面积指数模型,完成玉米叶面积指数反演,为玉米长势监测提供理论依据。

Maize is one of the most important food crops in China,which has the largest planting scale and the fastest growing trend.The growth of maize will directly affect its yield and quality.Therefore,through effective monitoring of the growth of maize can provide macro information for field management and yield estimation,and provide an important basis for decision-making by the relevant national departments.In this study,Unmanned Aerial Vehicle(UAV)equipped with an image sensor was used as a low-altitude remote sensing platform to obtain visible spectral remote sensing images of maize.First we made geometric and radiometric correction of maize canopy visible spectrum image by ENVI software,and then we made gray and enhancement processing of the color image.AP-HI algorithm was used to obtain the maize coverage information,for it has strong light adaptability to the complex background of farmland.The image was segmented by AP-HI algorithm and converted it into a binary image to remove the background of the land,water pipes,roads and residues in the image,so as to retain the binary image of maize.The road existed in the farmland of the image,which needed to be excluded when calculating the actual crop coverage.The road area appeared in the four boundaries and center of the image.The number of black pixels in the road area was counted and the road width was calculated according to the number of pixels,and then the road part was removed from the binary image.In the processed image,the white pixels are the non-crop area,and the black pixels are the maize planting area.In order to calculate the size of maize crops,the proportion of black pixels to total pixels in the binary image needed to be counted.The unit area was selected as 80×80 pixels,and the image was marked by blocks from top to bottom and left to right and got the number of blocks was 720.The unit area was scanned,and the proportion of the black pixels per unit area to the total number of pixels(6400 pixels)was calculated.Until the 720 blocks were completely counted,the proportion of the number of black pixels to total pixels in the image could be calculated,which is the maize coverage.The relational model between coverage and Leaf Area Index(LAI)through canopy porosity was established to complete maize LAI inversion,so as to provide theoretical basis for monitoring maize growth.The results show that low altitude UAV visible spectrum remote sensing can be used as an effective method to extract crop coverage,which has a good prospect.