ViT-DexiNet: a vision transformer-based edge detection operator for small object detection in SAR images

• DOI: 10.1080/01431161.2023.2277167
• Publication Date: 2023-11-21T07:18:15Z
• Authors: M. S. Sivapriya; S. Suresh
• Keywords: Sobel operator; Smoothing

This paper introduces a novel edge detection operator called 'vision transformer-based Dexinet (ViT-DexiNet)' to address the challenges of detecting small objects in synthetic aperture radar (SAR) images. SAR images are typically impacted by strong multiplicative noise, making difficult. Existing traditional methods have limited spectral data preservation capabilities and often result loss clarity integrity salient features The proposed ViT-DexiNet employs series interconnected layers extract refine from It utilizes vision transformer self-attention layer capture pattern structural details image crucial for determining edges. extracted feature maps then processed DexiNet architecture, which consists dense block, transfer upsampling network. architecture helps preserve information at different scales deeper layers. layered blocks generate maps, concatenated averaged through smoothing remove noise enhance images, resulting final high-quality map. To evaluate method, both qualitative quantitative analyses conducted using standard operators such as Canny Sobel. empirical results demonstrate that surpasses baseline operators. achieved values 97.92%, 97.72%, 97.64% 97.41%, respectively metrics accuracy, precision, recall f1-score. offers simplifying interpretation object detection. Overall, method shows promise overcoming limitations approaches improving edges