Untersuchungen zur Laserlichtwirkung auf der Basis von Bildanalysen auf juvenile Pflanzen zur Unkrautregulierung

非化学杂草控制是植物生态和可持续生产的重要目标。未来的一种替代方法可能是利用激光技术对植物早期生长阶段进行特定的杂草控制。应用研究项目的目标是在生物系统和激光工程师的跨学科研究方法中扩大基于激光的杂草管理。因此，第一阶段项目的良好结果提供了依据。在建立了基于激光技术和植物栽培参数的损伤模型后，可以设计图像采集和处理系统。确定最佳瞄准点后，对激光束进行定位。目前正在进行的实验确定了有关杂草密度和操作速度的激光应用的极限值。更新建议的重点在于将结果转移到复杂的条件。关键的挑战是从实验室环境到现场应用的步骤。在考虑倾斜和振动的情况下，利用环境动力学对图像采集和处理以及激光束定位进行优化。这项改进旨在可靠、快速地检测植物行间和行内的杂草。因此，在不同植物种类和生长阶段的重叠情况下，对分离质量进行了研究。要做到这一点，有必要对真实的杂草丰度进行调查，从而获得有关生长早期形态特征的单个杂草物种的训练数据集。设计特定的主动形状模型，根据单叶形状和分生性确定接近分生组织的理想目标。通过3d相机包含高度信息是有利的，但需要调查相机细节的影响。检测到的植物分类将根据快速决策进行优化，以最大限度地减少衍生激光束定位的延迟。瞄准以全局坐标为基础，相机和振镜都根据全局坐标进行校准。与受控的实验室条件相反，在粗糙的环境条件下，激光加工头将被放置在远离土层的地方。考虑致动器的角分辨率，创建了空间校正矩阵。沿着移动方向跟踪激光束位置的发展允许动态辐照与适应的曝光时间。将系统安装到移动平台后，会发生更复杂的驾驶事件。这些振动还包括倾斜和漂移。在击球过程中，为了避免误击，应考虑对目标的修正。补偿应由适当的轴承以及倾斜和加速度传感器支持。

Non-chemical weed control is an essential objective of an ecological and sustainable plant production. A future alternative might be the specific weed control of plants in early growth stages using laser technology. The objective of the applied research project is to enlarge upon a laser based weed management in an interdisciplinary research approach of bio systems and laser engineers. Thereby, the promising results of the first project phase provide the basis. After the development of a damage model depending on laser technical and plant cultivation parameters, an image acquisition and processing system could be designed. As the optimal targeting points could be determined, the laser beam was accordingly positioned. Ongoing experiments currently determine the limit values of the laser application concerning the weed density and the operating speed. The focus of the renewal proposal lies on the transferring of the results to complex conditions. The key challenges are the steps from a laboratory environment to applications in the field. The image acquisition and processing as well as the laser beam positioning will be optimized with the environmental dynamisation, as tilting and vibration will be considered. This improvement aims at a reliable and fast detection of weed plants between and in plant rows. Thereby, the separation quality is investigated under overlapping situations as well as with diverse plant species and growth stages. To do this, surveys of real weed abundances are necessary, allowing the acquisition of training datasets for individual weed species regarding morphological characteristics of early growth stages. Specific active-shape-models will be designed which are able to determine an ideal targeting close to the meristem based on single leaf shape and phyllotaxis. The inclusion of height information via 3D-cameras is advantageous, while the influence of the camera specifics shall be investigated. The classification of detected plants will be optimized respecting fast decisions to minimize delays for the derived laser beam positioning.The targeting bases on global coordinates, to which the camera as well as the galvanometer is calibrated. In contrast to the controlled laboratory conditions the laser machining head will be placed far away from the soil layer under rough environment conditions. Respecting the angular resolution of the actuators, a spatial correction matrix is created. The development of a tracking of the laser beam position along the moving direction allows on-the-fly irradiations with adapted exposure times.After the installation of the system on to a mobile platform, even more complex driving events occur. These include beside vibrations also tilting and drifting. During the driving, corrections of the targeting shall be considered in order to avoid false hits. The compensations shall be supported by appropriate bearing as well as by tilt and acceleration sensors.