Generating Vectors from Images using Multi-Stage Edge Detection for Robotic Artwork

摘要

Edge detection is a vital aspect in image processing, especially in vectorisation of images which is an important component fed to computer numerical controlled vector devices, to obtain desired output. The problem addressed in the paper deals with vectorisation of edge detected artworks for converting into computer numerical controlled vector system which can be used as spatial robotic artwork. Here as an experimental benchmark, we have used XY plotter. The XY computer numerical controlled plotters can process these vectors as tangible artwork on larger surfaces. These works are very tedious to convert from a regular image to drawing object. We try to apply multi stage edge detection so that the proper edges can be converted conveniently for vectorisation. The most extensively used algorithms applicable for colour and intensity prime images come with certain limitations. The innovation proposed here is to combine the Canny operator and Deep Learning models together for effective and enhanced edge detection, removing noise, unwanted objects and complete the artwork component in the most meaningful and inferable condition. The goal of this algorithm is to produce flexible and robust results with scalable hardware configuration at an improved speed. This problem is fundamental in producing accurate vector images. In this paper, we try to incorporate the robustness of various existing algorithms but with a challenge of multi-stage computation for a better output.