基于激光与机器视觉技术的立木胸径检测方法研究

The diameter inspection of standing trees is one of the major tasks in forest resource investigation. The limitations of traditional detection methods are obvious: low efficiency and heavy workload. Machine vision inspection technique is the development trend of tree inspection. The primary task of the machine vision inspection technique is system calibration. Now the system calibration of tree inspection is carried out using benchmarking. The efficiency of this method is very low and can not match with the vision inspection. A mathod of the diameter inspection of standing trees based on laser and machine vision was put forward in the project. The spot size of laser beam on trunk was used as the sysem calibration, so the detection precision and the inspection efficiency are improved. The distance between the upside edge and the downside edge of the laser beam spot and the laser emitter is detected by the laser range finders, and the actual size of the laser beam spot is computed with the distance. The cluster and the shape parameters extraction techniques are used to detect the trunk region in image. Models for mathematical ecology are used cutting separation of superposed trunk. The edge shape detectiong method based on saltation gray scale is used to detect the edge of trunk. All the methods above are used to ensure the inspection efficiency and detection precision of the system.

在森林资源调查中，立木的胸径测量是最重要的任务之一，而传统的检测方法记录、统计功效慢，作业工作量较大，采用机器视觉技术进行检测已经成为树木检测发展的一个趋势。采用机器视觉技术进行立木胸径检测首先要解决视觉系统的标定问题。现有的立木视觉检测方法采用放置标杆对视觉系统进行标定，精度差效率低，与视觉检测的高速自动化的特征不相匹配。本项目提出一种基于激光与机器视觉技术的立木胸径检测方法，选用激光光束照射到树干上的光斑作为标尺对视觉系统进行标定，提高了检测精度和检测效率。本检测方案用激光测距仪测量出树干上激光光斑上下侧与激光发射器之间的距离，进而计算出激光光斑的实际尺寸。运用聚类和提取形状参数的方法识别出立木树干区域，采用数学生态学的方法对重叠的树干进行切割分离，采用基于灰度突变的区域边缘形态检测算法检测树干的边缘，通过这些方案保证了系统的检测精度和检测效率。

在森林资源调查中，立木的胸径测量是最重要的任务之一，而传统的检测方法记录、统计功效慢，作业工作量较大，采用机器视觉技术进行检测已经成为树木检测发展的一个趋势。采用机器视觉技术进行立木胸径检测首先要解决视觉系统的标定问题。现有的立木视觉检测方法采用放置标杆对视觉系统进行标定，精度差效率低，与视觉检测的高速自动化的特征不相匹配。本项目提出一种基于激光与机器视觉技术的立木胸径检测方法，选用激光光束照射到树干上的光斑作为标尺对视觉系统进行标定，提高了检测精度和检测效率。本检测方案用激光测距仪测量出树干上激光光斑上下侧与激光发射器之间的距离，进而计算出激光光斑的实际尺寸。运用聚类和提取形状参数的方法识别出立木树干区域，采用数学生态学的方法对重叠的树干进行切割分离，采用基于灰度突变的区域边缘形态检测算法检测树干的边缘，通过这些方案保证了系统的检测精度和检测效率。通过Otsu法对激光光斑进行分割，根据激光光斑的实际尺寸和像素数确定图像上立木胸径的位置，根据立木胸径的像素数计算出立木胸径值。实验结果表明，本方法的测量结果与轮尺测量的胸径值最大误差1.22%，符合国家对森林资源清查操作的误差要求。