Humanoid Robot Portrait Drawing Based on Deep Learning Techniques and Efficient Path Planning

This paper presents a complete robotic portrait drawing system. First, facial features are extracted robustly using deep learning-based approaches. Then an efficient path planning algorithm is proposed to draw the portrait. To extract facial features, it goes through the following steps: face detection, facial edge extraction, image synthesis and image binarization. Given the processed image, the robot maps the coordinates of the image pixels to the coordinates of the drawing board after calibration. Then drawing paths are planned. The path planning principle is that the robot tries to travel as minimum distance as possible. To evaluate the proposed system, the facial landmarks are extracted from both the reference image and the binary image output of our system. Point-to-point distances are calculated to measure the binary image quality. On the other hand, the drawing time of each individual portrait is saved for efficiency comparison.