Robust Object Detection under Extreme Weather Using Physics-Aware Deep Learning
DOI:
https://doi.org/10.63282/3117-5481/AIJCST-V6I2P103Keywords:
Strong Object Detection, Bad Weather, Physics-Conscious Deep Learning, Multi-Modal Fusion, Autonomous Driving, Surveillance, Low-Light Enhancement, Cybersecurity StructuresAbstract
Weaknesses in object detection under extreme weather conditions are an important problem in the current computer vision, especially in applications in autonomous driving, surveillance, and remote sensing. Traditional deep-learning methods like CNNs, YOLO, SSD and Faster R-CNN have been shown to drastically decrease the performance under low-light, fog, rain, snow, and low-contrast environments as a result of occlusion, noise and low-contrast. Physics-aware deep learning combines environmental and physical priors such as atmospheric scattering, rain streak formation and low-light noise models into a neural network to increase feature visibility and enhance detection robustness. In this review, the author thoroughly examines the challenges related to extreme weather, compares the traditional and hybrid physics-informed approaches, and also addresses such architectures as CNN-based, Transformer-based, and multi-modal networks. Surveys are made on datasets, metrics of evaluation, and real-world applications, and the advantages of physics-aware to safety-critical systems are discussed. Lastly, the future research directions and open issues such as multi-modal fusion, self-supervised learning, edge deployment, and secure AI integration are given to inform the creation of dependable object detection systems in unfavorable environmental factors.
References
[1] Wang, X., “Deep learning in object recognition, detection, and segmentation,” Foundations and Trends® in Signal Processing, vol. 8, no. 4, pp. 217–382, 2016.
[2] Anastasiu, D. C., A. Sharma, R. M. de Sousa, and S. A. Khan, “Efficient city-wide multi-class multi-movement vehicle counting: A survey,” Journal of Big Data Analytics in Transportation, vol. 2, no. 3, pp. 235–250, 2020.
[3] Lahoz-Monfort, J. J., G. Guillera-Arroita, and B. A. Wintle, “Imperfect detection impacts the performance of species distribution models,” Global Ecology and Biogeography, vol. 23, no. 4, pp. 504–515, 2014.
[4] Shenoy, S. A. N. A. T. H., “Object detection under poor lighting conditions using deep learning,” 2021.
[5] Meng, Y., Physics-Informed Machine Learning for Smart Decision-Making in Ultrasonic Metal Welding, Ph.D. dissertation, Univ. of Illinois at Urbana–Champaign, Urbana, IL, USA, 2023.
[6] Zhao, Z.-Q., P. Zheng, S.-T. Xu, and X. Wu, “Object detection with deep learning: A review,” IEEE Transactions on Neural Networks and Learning Systems, vol. 30, no. 11, pp. 3212–3232, 2019.
[7] Bennamoun, M., and G. J. Mamic, Object Recognition: Fundamentals and Case Studies. Berlin, Germany: Springer, 2012.
[8] Kallie, C. S., G. E. Legge, and D. Yu, “Identification and detection of simple 3D objects with severely blurred vision,” Investigative Ophthalmology & Visual Science, vol. 53, no. 13, pp. 7997–8005, 2012.
[9] Singh, D., and V. Kumar, “A comprehensive review of computational dehazing techniques,” Archives of Computational Methods in Engineering, pp. 1–19, 2018.
[10] Cao, X., Y. Huang, J. Chen, and L. Huang, “Region-based CNN for foreign object debris detection on airfield pavement,” Sensors, vol. 18, no. 3, p. 737, 2018.
[11] Bossu, J., N. Hautière, and J.-P. Tarel, “Rain or snow detection in image sequences through use of a histogram of orientation of streaks,” International Journal of Computer Vision, vol. 93, no. 3, pp. 348–367, 2011.
[12] Xu, X., J. Lu, Y. Zhao, and H. Liu, “Exploring image enhancement for salient object detection in low light images,” ACM Transactions on Multimedia Computing, Communications, and Applications, vol. 17, no. 1s, pp. 1–19, 2021.
[13] Mușat, V., M. Cordts, and A. Geiger, “Multi-weather city: Adverse weather stacking for autonomous driving,” in Proc. IEEE/CVF Int. Conf. Computer Vision (ICCV), 2021.
[14] Chen, L., Z. Zhang, and Y. Li, “Machine learning methods in weather and climate applications: A survey,” Applied Sciences, vol. 13, no. 21, p. 12019, 2023.
[15] Tian, Y., “Effective image enhancement and fast object detection for improved UAV applications,” 2023.
[16] Olorunshola, O., P. Jemitola, and A. Ademuwagun, “Comparative study of some deep learning object detection algorithms: R-CNN, Fast R-CNN, Faster R-CNN, SSD, and YOLO,” Nile Journal of Engineering and Applied Sciences, vol. 1, no. 1, pp. 70–80, 2023.
[17] Yang, J., C. Li, and Z. Wang, “Focal attention for long-range interactions in vision transformers,” in Advances in Neural Information Processing Systems, vol. 34, pp. 30008–30022, 2021.
[18] Ozcan, A., and O. Cetin, “A novel fusion method with thermal and RGB-D sensor data for human detection,” IEEE Access, vol. 10, pp. 66831–66843, 2022.
[19] Henderson, P., and V. Ferrari, “End-to-end training of object class detectors for mean average precision,” in Proc. Asian Conf. Computer Vision (ACCV). Cham, Switzerland: Springer, 2016.
[20] Fazlali, H., M. Ebrahimi, and A. A. Aghagolzadeh, “Single image rain/snow removal using distortion type information,” Multimedia Tools and Applications, vol. 81, no. 10, pp. 14105–14131, 2022.
[21] Lin, X., Y. Zhou, and H. Wang, “Multi-sensor data fusion method based on self-attention mechanism,” Applied Sciences, vol. 13, no. 21, p. 11992, 2023.
[22] Ahmed, F., “Cloud security posture management (CSPM): Automating security policy enforcement in cloud environments,” 2023.
[23] Schumann, G. J. P., P. D. Bates, M. S. Neal, and J. C. Trigg, “Assisting flood disaster response with earth observation data and products: A critical assessment,” Remote Sensing, vol. 10, no. 8, p. 1230, 2018.
