With the advancement of industrial automation and intelligent technology, detecting text in industrial design drawings has become critical for improving efficiency in industrial pipelines. These drawings present unique challenges, including intricate backgrounds, linear interference, and diverse text styles, such as curved, inclined, and irregular shapes. While segmentation-based methods have demonstrated potential in detecting arbitrarily shaped text, existing approaches often suffer from inaccuracies and limited robustness when applied to industrial design drawings. To address these limitations, this paper proposes a novel text detection algorithm that employs multi-dimensional feature fusion to enhance feature extraction and robustness. By integrating this strategy into a segmentation-based model, the proposed method effectively handles the complexities of text detection in industrial scenarios. Experimental results on industrial design drawing datasets demonstrate significant improvements in detection accuracy and robustness compared to state-of-the-art methods, making the algorithm well-suited for practical industrial applications.