基于轮廓查找与线扫描的作物行检测算法

Detection Algorithm of Crop Row Based on Contour Searching and Line Scanning

针对传统作物行直线检测算法存在的检测时间长、检测误差大的问题,提出一种基于轮廓查找与线扫描的作物行检测算法。首先,在农田图像下端一定宽度条形图上进行轮廓查找,求取轮廓的外接矩形,得到作物直线下端定位点;其次,对每一个下端定位点在图像顶端一定区域内进行线扫描,上下两端点形成一条候选作物行直线,记录直线上的绿色像素数量最大时上端顶点的位置,此直线即为最佳作物行直线。仿真实验表明,研究结果可以准确地检测出主要作物行直线,准确率为90.0%,平均检测时间为30 ms。

Aiming at the problem of long detection time and large detection error of the traditional crop rows detection algorithms,a kind of crop rows detection algorithm based on contour searching and line scanning was proposed.Firstly,in farmland image bottom outline for a certain width on the bar image,the bounding rectangle was calculated,and the bottom anchor point of crop row was got.Secondly,sanning in a range for each bottom anchor point at the top of image,a line was constructed with top point and bottom point,the location of up point was recorded when the number of the green pixels in the crop row attained maximum,and the crop with maximum pixels was the best crop row line.Simulation results show that the algorithm can accurately detect the main crop lines,and the accuracy is 90.0%,also the average detection time is 30 ms.