Estimating Weight of Unknown Objects Using Active Thermography

Abstract

Successful manipulation of unknown objects requires an understanding of their physical properties. Infrared thermography has the potential to provide real-time, contactless material characterization for unknown objects. In this paper, we propose an approach that utilizes active thermography and custom multi-channel neural networks to perform classification between samples and regression towards the density property. With the help of an off-the-shelf technology to estimate the volume of the object, the proposed approach is capable of estimating the weight of the unknown object. We show the efficacy of the infrared thermography approach to a set of ten commonly used materials to achieve a 99.1% R 2 -fit for predicted versus actual density values. The system can be used with tele-operated or autonomous robots to optimize grasping techniques for unknown objects without touching them.

Development of the Biomechanical Technologies for the Modeling of Major Segments of the Human Body: Linking the Past with the Present

“…An important assumption to determine the COM optically is to know the body density in advance. Optical methods, specifically the ones that provide information on the temperature of an object (tele-thermography), can be used to estimate density by mathematical procedures of neural networking, and used in parallel with shape estimation and limbs centroid calculations, to allow the determination of the center of mass [ 60 ].…”

Learning About Objects by Learning to Interact with Them

“…[54] infer the mass in a supervised setting. There are also approaches for mass estimation for specific object categories (e.g., [47,3]). Our work differs from these works since we infer the relative mass of objects by interacting with them.…”

“…The agent obtains a sequence of raw visual feedback by applying forces to the chosen interaction points in the order determined by the model. Let F = {f 0 , f 1 , f 2 } be a set of force magnitudes that are chosen a priori and correspond to magnitudes typically required to move light, medium and heavy objects 3 . Let D be a set of 8 force directions corresponding to a quantized set of directions in space, but avoiding those that point into the ground.…”

“…Threshold 1 is determined by the scale of the Embedding Loss. See Appendix B 3. We have selected typical minimal forces that move objects which correlate strongly, but not perfectly, with true physical masses.…”

Manipulation Tasks in Hazardous Environments Using a Teleoperated Robot: A Case Study at CERN

“…Different approaches have been developed to estimate the object’s weight. In [ 20 ], active thermography and custom multi-channel neural networks are used to classify the density property. The approach is capable to estimate the weight of the unknown object by evaluating its volume.…”