基于机器视觉的玉米苗期土壤湿度检测试验研究  

Research on soil moisture detection of corn seedling based on machine vision

作　　者：卜小东[1] 张馨月[1] 黄可京[1] 郭辉[1] Bu Xiaodong;Zhang Xinyue;Huang Kejing;Guo Hui(Institute of Mechanical and Electrical Engineering,Beijing Vocational College of Agriculture,Beijing,102208,China)

机构地区：[1]北京农业职业学院机电工程学院,北京市102208􀇙

出　　处：《中国农机化学报》2020年第7期150-156,共7页Journal of Chinese Agricultural Mechanization

基　　金：北京市教委科技一般项目(KM201812448006)。

摘　　要：在玉米苗期进行土壤湿度动态监测是提供精准灌溉的重要依据,对于玉米在此阶段快速健康生长具有重要意义。本文模拟超低空图像采集设备的试验方式,通过试验平台采集玉米苗期土壤水分的变化情况,以期建立图像与土壤水分数据的联系。利用超绿特征(2G-R-B)对采集到的玉米苗期土壤图像进行分割以排除植株本身对图像的影响。对试验中分割处理后的土壤图像的均值、归一化方差特征参数与试验平台测得土壤水分数据进行分析比较,分析后对所处理图像采用4G-R-B颜色特征修正,通过计算归一化方差σ4G-R-B作为特征参数与实测土壤湿度进行线性回归分析,二者相关性验证结果为:R2=0.73,RMSE=3.2%。表明修正处理后图像归一化方差σ4G-R-B图像特征参数能够较好的表征土壤的水分变化。Soil moisture dynamic monitoring was an important basis for providing precise irrigation in the corn seedling stage.It was of great significance for the rapid and healthy growth of corn at this stage.This paper simulates the experimental method of ultra-low-altitude image acquisition equipment.This experiment used the equipment to collect information about the changes in soil moisture,and established the connection between the image and the soil moisture data.We used super green features(2 G-R-B)to segment the collected images,in order to eliminate the influence of the plant itself on the image.We analyzed and compared the average and normalized variance of the processed images R,G,B and the measured soil moisture data.This paper analyzes and compares the mean and normalized variance characteristic parameters of the soil image after the segmentation process in the experiment with the soil moisture data measured by the experimental platform.After the modified 4 G-R-B characteristic parameter was applied to the processed image,linear regression analysis was carried out through this characteristic parameter and the measured soil moisture.The correlation verification results of R2 was 0.73,RMSE was 3.2%.It is found that the normalized varianceσ4 G-R-B of the image after the correction process can better characterize the soil moisture changes.

关 键 词：玉米 苗期 机器视觉 土壤湿度 图像分割 回归分析 

分 类 号：S24[农业科学—农业电气化与自动化] TP391.4[农业科学—农业工程]