A Vision-Based System for Rapid DXF Generation in Metalworking: Image Processing Pipeline and Economic Feasibility Case Study

Abstract

Low-volume and custom production jobs in small and medium-sized metalworking enterprises are often declined due to the high fixed costs and time requirements of manual CAD modeling. This paper presents a practical, production-ready desktop application for the automated generation of DXF files from photographs of physical parts, designed to support rapid laser cutting and prototyping workflows. The system combines a user-friendly GUI with a robust computer vision pipeline including adaptive thresholding, contour extraction, geometric scaling, and vector simplification. Unlike generic raster-to-vector tools, it targets the specific needs of SMEs: offline use, closed contours, minimal training, and compatibility with CAM systems. A real-world case study demonstrates the successful replication of an agricultural machine parts, showing ~90% reduction in preparation time and ~45% profit margin on the order. Interviews with industry stakeholders confirm the potential of the proposed solution to recover previously unviable jobs and increase monthly revenues by an estimated 5-7%. This interdisciplinary contribution integrates computer vision and applied economics, offering a scalable solution to improve productivity and digital inclusion in SMEs.