基于机器视觉的玉米定位施肥控制系统设计与试验

Design and Experiment of Maize Fertilization Control System Based on Machine Vision

为了实现农田环境下的苗期玉米定位定量施肥作业,提高肥料利用率,本文设计了基于机器视觉的苗期玉米定位施肥控制系统,控制自驱动履带式田间玉米定位施肥机,使用切片式排肥器结构实现间歇式施肥。采集苗期玉米冠层图像后,利用颜色特征区分连续拍摄过程中的大量背景图像和识别错误的株心,改进了株心识别定位方法;建立施肥滞后误差补偿模型,实时计算施肥相对距离,准确计算排肥器的落肥时刻,实现了按株间歇式自动定位施肥。经过试验与验证分析,改进的株心定位识别方法明显缩减了图像处理时间,提高了算法鲁棒性;3个挡位施肥量(7.25、14.5、21.75 g)的施肥稳定性变异系数分别为1.93%、1.87%、1.93%,施肥量控制精度大于95%,施肥位置准确性平均误差为3.2 cm,可以实现精准定位施肥。本文设计的施肥控制系统可以实现苗期玉米按株定位定量施肥,达到了减施提质、科学施肥的目的。

In order to realize the positioning and quantitative fertilization of maize in seedling stage under farmland environment,a control system was designed,which was based on machine vision to position and fertilize maize in seedling stage and improve fertilizer utilization rate.The control system controlled the maize positioning fertilizer applicator to travel in the field through the self-driving crawler.The fertilizer applicator realized intermittent fertilization according to plant location by using the structure of slice fertilizer discharger.After collecting the images of maize canopy at seedling stage,the area of maize plants at seedling stage was small,and there were many collected images without maize.In order to solve the influence of the above problems on operation efficiency and accuracy,the system used color features to distinguish a large number of background images and misrecognized plant centers during continuous shooting.By adding image preprocessing and algorithm post-processing,the method of plant center recognition and location was improved;in order to solve the problem of fertilizer lag caused by fertilizer falling time,system response time and positioning error,the compensation model of fertilization lag error was established,the relative distance of fertilization was calculated in real time,and the time of fertilizer falling was calculated accurately,which improved the control precision of intermittent automatic positioning fertilization according to plant.After experiment and verification analysis,the improved plant center location recognition method obviously reduced the image processing time and improved the robustness of the algorithm;according to the amount of single fertilization,the fertilizer applicator was set three gears(7.25 g,14.5 g and 21.75 g),and the coefficient of variation of fertilization stability,which represented the stability of fertilizer application per plant,was 1.93%,1.87%and 1.93%,respectively.The control precision of the actual fertilization amount and the target fertilization amount of the fertilizer discharger was more than 95%.The average error of the distance between the actual fertilizer drop point and the target drop point can reach 3.2 cm,which met the basic operation requirements of location fertilization.The fertilization control system designed can realize the quantitative fertilization according to the location of maize plants at seedling stage,and achieve the purpose of reducing application,improving quality and scientific fertilization.