Image analysis of topology changes for laser-based soft material processing

基于激光的材料加工是一个新兴领域，主要是切割，但也有连接技术，如焊接，焊接，烧结和缝纫或去除程序，如烧蚀和雕刻。基于激光的软材料加工的应用领域包括纺织业的织物处理，农业除草等植物处理，以及生物学和医学中的活组织处理。这些领域的进一步自动化将节省成本，优化流程，并可能减轻饥饿和疾病。本发现基金的研究将集中在基于激光的软材料加工自动化的未解决的方面。激光切割的创造产生了柔软材料的拓扑变化。对于自主切割，需要补偿材料的潜在变形和变化，包括考虑实际拓扑及其进展的算法。一般来说，对软质材料的操作往往会导致形貌和拓扑结构的变化。这些形态变化等同于变形，而拓扑变化可以理解为材料内部孔洞的发展。本研究将利用三维视觉方法检测和补偿激光加工引起的拓扑变化。在实用性方面，激光加工具有优势，即发生修饰的区域不会被较大的物体（例如刀片）覆盖。总体假设是，实时结合拓扑变化将允许对软材料变形进行强大的跟踪，并在自主激光加工过程中获得先进的结果。为了证明这一假设，该发现基金寻求回答以下科学关键问题（SKQ）： A -可以自动识别不同类型的拓扑变化吗？B -如何通过计算区分拓扑和形态变化？C -哪些计算机视觉方法可以用来检测软材料上/内部拓扑结构及其空间位置的变化？D -如何在闭合计算机视觉和激光加工方法之间的回路的同时考虑形态和拓扑的变化？E -在周围软质材料进一步变形或不进一步变形的情况下，检测拓扑变化位置的准确性和延迟是多少？该项目分为四个目标来回答SKQ：目标1关注软材料激光加工的拓扑和形态变化数据集的生成。目标2包括本提案的核心工作，并产生基于视觉的拓扑和形态变化及其特定位置识别的算法原型。目标3是将这些算法转化为闭环机器人系统，目标4是根据激光加工结果量化系统特性和由此产生的精度。

Laser-based material processing is an emerging field, predominantly cutting, but also joining technologies such as welding, soldering, sintering and sewing or removal procedures like ablation and engraving. Application areas for laser-based soft material processing include fabric handling for textile industry, treatment of plants such as weeding in agriculture, as well as living tissue manipulation in biology and medicine. Further automation in these areas will save costs, optimize processes and may alleviate hunger and disease. The research in this Discovery Grant will focus on unsolved aspects for automation of laser-based soft material processing. Creation of a laser cut produces a topological change in soft material. For autonomous cuts, potential deformation and changes of the material need to be compensated for, including algorithmic consideration of the actual topology and its progression. In general, manipulation of soft material often results in changes of morphology and topology. While those morphological changes are equivalent with deformations, changes of topology can be understood as the development of holes inside the material. This research will detect and compensate topological changes induced by laser processing with 3D vision methods. In terms of practicability, laser processing has the advantage, that the area where the modification happens is not covered by larger objects (e.g. a knife blade). The overall hypothesis is that incorporating the change of topology in real-time will allow for a robust tracking of soft material deformation and lead to advanced results during autonomous laser processing. To prove the hypothesis, this Discovery Grant seeks in answering the following scientific key questions (SKQ): A -Can different types of topological changes be identified automatically? B -How can topological and morphological changes be differentiated computationally? C -Which computer vision methods can be leveraged to detect the change of topology and its spatial location on/inside the soft material? D -How to take morphological and topological changes into account while closing the loop between computer vision and laser processing methods? E -What is the accuracy and latency of detection of a topological change location with or without further deformation of the surrounding soft material? This project is sub-divided into four aims to answer the SKQ: Aim 1 cares about generation of a dataset on topological and morphological changes of soft material laser processing. Aim 2 include the core work of this proposal and result in algorithmic prototypes of the vision-based recognition of the topological and morphological changes and their specific locations. Aim 3 is translating theses algorithms into a closed-loop robotic system and Aim 4 quantifies the system characteristics and resulting accuracy in terms of the laser processing results.