Ping Yang, Shuonan Xiao, Runxi Wu, and Haiyang Lin
Black volute, depalletising system, image recognition algorithm, regional calibration, spatial positioning
In vision-based measurement applications, black light-absorbing objects pose a challenge due to their poor reflection of the structured light emitted by the infrared module of an RGB-D camera. To address this, an algorithm based on reference environment information is proposed. Concurrently, a robotic depalletising system is established for randomly stacked single-layer volutes. During the depalletising process, the algorithm obtains the volutes position information within a depalletising framework from the vision system, which includes an RGB-D camera and an upper computer, and outputs the spatial positioning data and adsorption pose for end adsorption device connect with a six-axis industrial robot. Following this, for the method of positioning each individual volute, the position of the centre in XY-plane is determined by contours at the end of the protrusion and calculated by using the offset algorithm, while the volute’s depth data is inferred from the depth of the cardboard using a linear transformation. Additionally, a regional hand-eye calibration method has been developed to improve two- dimensional calibration precision, which divides the graspable area into quadrants, each with a unique coordinate transformation matrix. By conducting depalletising experiments with three different types of volutes, the maximum error in the image processing procedure is less than 9 mm. The average error is below 4 mm, and the standard deviation is less than 2 mm. After the hand-eye calibration phase, the maximum error observed for the volutes is within 8 mm, with an average deviation slightly less than 4.5 mm and a standard deviation below 2 mm.
Important Links:
Go Back
