Object pose and centroid analysis for automated material handling

Continual advancements of automation made in manufacturing and logistic industries have led to a greater demand of sub systems that provide essential data. An increasing trend in the SOA that reflects this is the development and implementation of 3D vision system. Examples of these 3D vision systems can be found in pick and place stations in automated assembly lines and other structured environments. The limitation of these systems is the dependence on explicit knowledge of the target Object Of Interest (OOI). Typically, these systems depend on CAD files of the OOI to determine the object pose or centre of mass to direct a manipulator to interact with it. In the case of manufacturing processes that have taken a gradual approach to automation, this level of data may not be available. Furthermore, in some applications, such as automated forklifts for logistics, a fixed camera setup for a vision system is not possible which leads to a larger variance in the captured data retrieved for these 3D vision systems.

Current development of 3D vision systems capable of operating in these types of environments, known as unstructured environments, have shown reproducible and repeatable object detection and isolation. This step of isolation allows for further development of these 3D vision systems to encompass the analysis of key properties of the OOI such as the objects volumetric distribution to determine suitability for interaction. In this study, we propose a method of determining these key properties from data retrieved from unstructured environments. This includes the implementation of the SOA in object detection and isolation to retrieve the OOI from an unstructured environment. With the OOI isolated from the unstructured environment, the object is then segmented into detectable planes. This analysis provides the essential data that allows for a close approximation of how the OOIs its mass is distributed. Experimentation is ongoing for this study, with preliminary findings showing promising results.