Dimensional Arrow Detection from CAD Drawings

Abstract

Objectives: This paper proposes an approach to effectively identify dimensional arrow heads from the CAD drawing images. Methods/Analysis: The proposed approach implements a multi-level thresholding and morphological operations in order to detect arrow like entity and some additional morphological steps in order to filter out non-arrow like entities from CAD drawing images. A histogram based multi-level thresholding is implemented. Morphological Black hat is used to detect solid arrow heads while morphological white hat is used to detect line arrow heads. Properties of arrow heads are used to filter out unwanted entities. Findings: CAD drawings contain arrows to depict dimensions of the models. In many situations, users do not have access to the CAD file for the drawing or do not have suitable software to visualize the CAD file. In such cases, an image of 2D drawing can be processed in automated or semi-automated way to extract dimensional relationships between entities. One step in such extraction is to detect dimensional arrow heads. The approach was tested on variety of images of CAD drawings with increasing difficulties with 96.655% success rate. The F1 score for each image was calculated and the cumulative average F1 score was found out to be 0.9596, precision rate is 95.95%, recall rate is 96.83% which suggests acceptable accuracy. Applications/Improvement: The proposed approach helps to detect dimensional arrow heads from CAD drawing images using image processing.