Hybrid CNN-Based Border Pixel Extraction with Attention-Enhanced Feature Fusion and Explainable AI for Real-Time Traffic Object Detection
Received: 14 December 2025 | Revised: 18 January 2026 | Accepted: 4 February 2026 | Online: 4 April 2026
Corresponding author: D. Manju
Abstract
This study presents a novel object recognition model, called Object Border Pixel Extraction with Precise Shape Detection using CNN (OBPE-PSD-CNN), to improve fine-grained object recognition in complicated traffic scenarios. In contrast to traditional deep learning-based object detection techniques that mainly use bounding boxes or coarse segmentation masks, the suggested one incorporates border pixel extraction and skeletal-contour-based shape modeling into the detection pipeline and allows the object to be represented precisely in terms of structure and geometry. In addition to enhancing discriminative ability in groups of visually similar objects, a hybrid deep feature fusion architecture based on ResNet-50 and AlexNet is proposed, and then an attention-enhanced classifier based on a Convolutional Block Attention Module (CBAM) is used to refine features adaptively, spatially, and channel-wise. Unlike the current literature where explainability is seen as a post-hoc analysis, the suggested framework entails Explainable AI (XAI) mechanisms in the architecture, offering visual pieces of evidence on model decisions in real-time. The suggested OBPE-PSD-CNN is an end-to-end solution to object detection, accurate shape extraction, segmentation, classification, and temporal tracking. Large-scale testing on benchmark traffic data such as CityFlow and UA-DETRAC demonstrates that the suggested algorithm is more effective than the state-of-the-art strategies, especially in dense, heterogeneous, and visually demanding traffic cases.
Keywords:
object detection, object border, object shape, human-computer interface, deep learning, convolutional neural networks, traffic monitoringDownloads
References
X. Zou, H. Liu, and Y. J. Lee, "End-to-End Instance Edge Detection." arXiv, 2022.
L. C. Chen, J. T. Barron, G. Papandreou, K. Murphy, and A. L. Yuille, "Semantic Image Segmentation with Task-Specific Edge Detection Using CNNs and a Discriminatively Trained Domain Transform." arXiv, 2015. DOI: https://doi.org/10.1109/CVPR.2016.492
K. Dong et al., "Attention‐enhanced multiscale feature fusion network for pancreas and tumor segmentation," Medical Physics, vol. 51, no. 12, pp. 8999–9016, Dec. 2024. DOI: https://doi.org/10.1002/mp.17385
M. Xiao, B. Yang, S. Wang, Z. Zhang, X. Tang, and L. Kang, "A feature fusion enhanced multiscale CNN with attention mechanism for spot-welding surface appearance recognition," Computers in Industry, vol. 135, Feb. 2022, Art. no. 103583. DOI: https://doi.org/10.1016/j.compind.2021.103583
M. Flores-Calero et al., "Traffic Sign Detection and Recognition Using YOLO Object Detection Algorithm: A Systematic Review," Mathematics, vol. 12, no. 2, Jan. 2024. DOI: https://doi.org/10.3390/math12020297
Y. Luo, Y. Ci, H. Zhang, and L. Wu, "A YOLOv8-CE-based real-time traffic sign detection and identification method for autonomous vehicles," Digital Transportation and Safety, vol. 3, no. 3, pp. 82–91, 2024. DOI: https://doi.org/10.48130/dts-0024-0009
A. Andres, A. Martinez-Seras, I. Laña, and J. Del Ser, "On the black-box explainability of object detection models for safe and trustworthy industrial applications," Results in Engineering, vol. 24, Dec. 2024, Art. no. 103498. DOI: https://doi.org/10.1016/j.rineng.2024.103498
Y. Wang, Y. Liu, R. Yi, and Y. Jiang, "Real-time traffic object detection algorithm with deep stochastic configuration networks," Information Sciences, vol. 700, May 2025, Art. no. 121848. DOI: https://doi.org/10.1016/j.ins.2024.121848
Y. Zheng, S. Liu, and L. Bruzzone, "An Attention-Enhanced Feature Fusion Network (AeF2 N) for Hyperspectral Image Classification," IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1–5, 2023. DOI: https://doi.org/10.1109/LGRS.2023.3320193
M. K. Rao and P. M. A. Kumar, "Advanced Object Tracking in Video Surveillance Systems with Adaptive Deep SORT Enhancement," Engineering, Technology & Applied Science Research, vol. 15, no. 2, pp. 20871–20877, Apr. 2025. DOI: https://doi.org/10.48084/etasr.9529
L. Jia et al., "MobileNet-CA-YOLO: An Improved YOLOv7 Based on the MobileNetV3 and Attention Mechanism for Rice Pests and Diseases Detection," Agriculture, vol. 13, no. 7, Jun. 2023. DOI: https://doi.org/10.3390/agriculture13071285
"2022 Data and Evaluation – AI City Challenge." https://www.aicitychallenge.org/2022-data-and-evaluation/.
"UA-DETRAC_dataset." Kaggle, [Online]. Available: https://www.kaggle.com/datasets/dtrnngc/ua-detrac-dataset.
Downloads
How to Cite
License
Copyright (c) 2026 M. Koteswara Rao, D. Manju, K. Kishore Kumar, Rajesh Kumar Verma, Padmini Debbarma, Boda Sindhuja
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain the copyright and grant the journal the right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) after its publication in ETASR with an acknowledgement of its initial publication in this journal.
