• Shubham Shinde
• Triveni Dhamdhere
• Poonam Bhawake

Detecting building shadows in aerial imagery is a complex task due to varying lighting conditions, occlusions, and complex building geometries. Shadows can significantly affect the performance of applications such as building detection, segmentation, and classification. In this paper, we propose a novel method for building shadow detection using aerial imagery based on the YOLOv8 object detection algorithm, which incorporates annotations generated using the LabelMe tool. We first preprocess the aerial images using the HSV color space to extract shadow regions, followed by morphological operations to refine shadow regions and obtain a shadow mask. We then use the YOLOv8 object detection algorithm to detect buildings and their shadows simultaneously, using annotations generated through the LabelMe tool to improve the algorithm's accuracy. Finally, we perform post-processing on the detected shadows to eliminate false positives. We evaluate our method on a dataset of aerial images and compare it to other state-of-the-art methods. Our results show that our approach achieves an F1-score of 0.91 and a recall of 0.93, outperforming other methods in terms of accuracy, speed, and robustness. The processing time of our method is 0.025 seconds per image, demonstrating its computational efficiency. Our proposed method combines color-based and edge-based features with the YOLOv8 object detection algorithm and LabelMe annotations to detect building shadows in aerial imagery accurately. Our method is suitable for various applications that require building shadow detection, such as urban planning and environmental monitoring. Future work includes extending the method to detect other types of shadows, improving building edge detection in the presence of shadows, and exploring the use of deep learning techniques to improve accuracy further. In conclusion, our proposed method demonstrates that the integration of LabelMe annotations can significantly improve the accuracy of building shadow detection in aerial imagery. Our method's high accuracy, speed, and robustness make it a promising approach for building shadow detection in various applications.

Deep Learning , Machine learning , YOLO , SSD

" Building Shadow Detection Using Aerial Imagery", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.10, Issue 9, page no.c800-c805, September-2023, Available :http://www.jetir.org/papers/JETIR2309292.pdf

" Building Shadow Detection Using Aerial Imagery", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.10, Issue 9, page no. ppc800-c805, September-2023, Available at : http://www.jetir.org/papers/JETIR2309292.pdf