Scanning trajectory optimization for industrial robot based on the scanned objects surface optical properties

Abstract

This work presents a workflow to automatically calculate optimized scanning trajectories for industrial robot system using the estimation of a surface reflectance model of 3-D shape and parameters from multiple views. To solve the problem of determining the views without Lambert's surface, a 3-D reconstruction algorithm based on a convolutional neural network is proposed. In the first step, the encoder is trained for the descriptor description of the input image. In the second step, a fully connected neural network is added to the encoder for regression for choosing the best views. The coder is trained using the generative adversarial methodology to construct a descriptor description that stores spatial information and information about the optical properties of surfaces in different areas of the image. The codec network is trained to recover the defect map (depends directly on the sensor and scene properties) from RGB image. As a result, this method uses nonLambertian properties, and it can compensate for triangulation reconstruction errors caused by view-dependent reflections. Experimental results on both synthetic and real objects show that the proposed method automatically finds trajectories that enable 3-D reconstructions, with a significant reduction of scanning time.