模拟复杂地形的喷雾靶标激光检测与三维重构

Spray target laser scanning detection and three-dimensional reconstruction under simulated complex terrain

喷雾靶标的检测是精密变量喷雾的重要环节,激光传感器以其精度高、速度快、不受光照干扰等特点被广泛应用于精密变量喷雾研究中。为了消除激光传感器姿态角的偏移对喷雾靶标检测的影响,获取精确的喷雾靶标外形尺寸信息以及三维重构图像,进行了室内模拟复杂地形激光检测矫正研究。该文基于UTM-30LX型激光传感器搭建了室内靶标检测试验平台,模拟复杂路况设计了滚转角、俯仰角和偏航角等姿态角偏移检测试验,提出了采用极坐标值与三角函数重新匹配、检测帧与检测点重新组合和深度值系数矫正等3种姿态角偏移矫正方法。首先对树形雕花板进行单一姿态角偏移检测试验,选取适合的检测距离与行进速度,对每种姿态角的多个偏移角度进行多次重复试验,然后矫正所获取的数据信息,对矫正后的目标尺寸进行误差分析并重构目标三维图像。再以仿真树为试验对象,验证在3种姿态角度同时改变时矫正方法的有效性。试验结果显示树形雕花板高度、宽度、树冠高度等尺寸相对误差均小于5%,仿真树相应参数尺寸相对误差均小于10%,满足变量喷雾检测的精度要求。

Spray target detection is one of critical steps in variable-rate spray applications. Laser sensor scanning technologies have been used to detect spray targets for precision variable-rate sprayers due to its high accuracy, rapid scan speed, and insensitivity to light sources. However, potential uneven road conditions could reduce detection accuracy of object scanning with laser sensor. In order to obtain accurate shapes of laser scanning targets and acquire three-dimensional（3-D） reconstruction images under complex terrain conditions, an indoor target detection platform with a laser sensor scanner was built to conduct spray target detection experiments under simulated uneven road conditions, and 3 correction methods were proposed to correct laser scanning data distortion due to the influence of complex field conditions. The target laser scanning detection platform consisted of a slide motion control unit and a laser sensor data collection unit. The slide motion control unit was able to control laser travel speed and laser moving distance on the sliding table. The laser sensor data collection unit was capable of detecting spray targets with the laser scanner and saving laser data for post process. The changes of attitude angles of laser sensor in 3 directions including pitch angle, roll angle and yaw angle were used to simulate complex terrain during field operations. Three data correction methods were proposed to diminish the influence caused by attitude angle deviations. They were the methods of re-matching polar coordinate value and trigonometric function for pitch angle deviation correction, re-combination of detection frames and detection points for roll angle deviation correction, and correcting the coefficient value of depth data for yaw angle deviation correction, respectively. The verification experiments for the proposed 3 correction methods to overcome complex field road conditions were divided into 3 test steps. Firstly, an artificial tree and a tree-shape carved board were detected with zero attitude angle deviation of laser sensor by specifying constant detection distances and laser travel speeds. Secondly, the tree-shape carved board was selected as the laser scanning target to test each deviation correction method of single attitude angle. Six angle values including-30 o,-20 o,-10 o, 10 o, 20 o, and 30 o were chosen for roll angle to simulate single uneven road conditions. So were angle values setting for pitch angle and yam angle. Finally, the artificial tree was selected as the laser scanning test target to verify the effectiveness of the combined attitude angle deviation correction. Three groups of combined attitude angle deviations were selected to simulate complex road conditions and to test the correction effects under the combinations of roll angle, pitch angle and yaw angle. All of acquired laser scanning object data were analyzed and corrected by the 3 proposed data correction algorithms. The data correction process and 3-D image reconstruction were conducted using Matlab software. The experiment results showed the spray object scanning with laser sensor could achieve precise outline shape detection of targets and obtain accurate 3-D reconstruction images when the laser sensor had not attitude angle deviations and the target detection test platform was under ideal and flat road environments. The relative errors of the height, width and canopy height of tree-shape carved board were all less than 5.0% by using the deviation correction method of single attitude angle to overcome simulated uneven road conditions. The relative errors of the corresponding parameters of the artificial tree were all less than 10.0% by using the deviation correction method of combined attitude angle to decrease the influence of complex uneven road conditions. The 3-D reconstruction images also had significant improvements after the correction algorithms. The test results verified the effectiveness of the 3 data correction methods for attitude angle deviation correction under simulated complex road conditions. The proposed methods have the potential to be integrated into variable-rate sprayers for precise spray field applications.