Drone-based deep learning system for accurate camel counting in desert rangelands with integration into Sarebanyar platform

Articles in Press

Document Type : Research Article (Regular Paper)

Authors

1 Animal Science Research Department, Yazd Agricultural and Natural Resources Research and Education Center, Agricultural Research Education Extension Organization (AREEO), Yazd, Iran

2 Animal Science Research Institute of Iran, Agricultural Research, Education and Extension Organization (AREEO), Karaj Iran

3 Plant and Seed Improvement Department, Yazd Agricultural and Natural Resources Research and Education Center, Agricultural Research Education Extension Organization (AREEO), Yazd, Iran

4 Tashk Kavir Company, Yazd, Iran

5 Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran

Abstract

Data-driven herd management is essential for scientific herd management in arid and semi-arid regions; however, traditional counting methods are labor-intensive, time-consuming, and prone to significant human error. This study proposes a drone-based deep learning framework for automated counting and monitoring of camel herds under prevailing desert conditions. Aerial videos were collected from nine camel herds across multiple regions in Iran using unmanned aerial vehicles (UAVs). A dedicated dataset was constructed from extracted video frames and annotated using polygon-based labeling. Two detection strategies were comparatively evaluated, including body-without-head and head-only detection. Object detection models were trained using the YOLOv8 architecture and tested under different training data scales to analyze the effect of dataset size on counting performance. Experimental results showed that head-based detection showed promising robustness under the tested conditions, although a fully controlled comparison with body-based detection would require matched training data. Expanding the training dataset from 100 images (≈900 frames) to 220 images (≈1800 frames) led to substantial improvements in detection stability and counting accuracy. Using the expanded dataset, the proposed system achieved > 98% counting accuracy in multiple real herds, including large groups containing up to 271 camels, with 1 counting error. The trained model was successfully integrated into the Sarebanyar App, enabling automated processing of newly acquired UAV videos and direct storage of results within an operational management workflow. The findings demonstrated that combining the UAV imaging, optimized region-of-interest selection, and diverse real-world training data enables a robust, scalable, and economically efficient solution for intelligent livestock monitoring in harsh desert environments, providing a practical pathway toward AI-driven smart herd management systems.

Keywords

Main Subjects

References

Ampatzidis, Y., Partel, V., Costa, L., 2019. UAV-based high throughput phenotyping in citrus utilizing multispectral imaging and artificial intelligence. Remote Sensing 11,410.
Barbedo, J.G.A., 2019. A study on the detection of cattle in UAV images using deep learning. IEEE Access 7, 158824_158835.
Bitaraf Sani, M., Hasheminejhad, Y., Asadzadeh, N., Karimi, O., Khojastehkey, M., Abolghasemi, M, Banabazi, M. H., 2025. A novel ICT-based system for camel husbandry for online monitoring, phenotyping and rangeland assessment. Discover Animals 2, 68. ‏‏
Guo, X., Shao, Q., Li, Y., Wang, Y., Liu, Q., 2022. Surveying wild animals from satellites, manned aircraft and unmanned aerial systems (UASs): A review. Remote Sensing 14, 834.
Huang, B., Fan, C., Chen, K., Rao, J., Ou, P., Tian, C., Zhao, H., 2024. VCAT: an integrated variant function annotation tools. Human Genetics 143, 1311_1322.
Iskandar, R. J., Fatichah, C., Yuniarti, A., 2024. Object detection in low-light conditions: a comparison using YOLOv5 and YOLOv8. The 4th International Conference on Science and Information Technology in Smart Administration (ICSINTESA) (pp. 558-563). IEEE. ‏
Khoramshahi, E., Oliveira, R. A., Koivumäki, N., Honkavaara, E., 2020. An image-based real-time georeferencing scheme for a UAV based on a new angular parametrization. Remote Sensing 12, 3185. ‏
Liu, S., Yin, D., Feng, H., Li, Z., Xu, X., Shi, L. and Jin, X., 2022. Estimating maize seedling number with UAV RGB images and advanced image processing methods. Precision Agriculture 23, 1604_1632.
Liu, S., Yin, D., Feng, H., Li, Z., Xu, X., Shi, L., Jin, X., 2022. Estimating maize seedling number with UAV RGB images and advanced image processing methods. Precision Agriculture 23, 1604_1632. ‏
Maes, W., Steppe, K., 2019. Perspectives for remote sensing with unmanned aerial vehicles in precision agriculture. Trends in Plant Science 24,152_164.
Redmon, J., Divvala, S., Girshick, R., Farhadi, A., 2016. You only look once: Unified, real-time object detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, pp.779_788.
Sohan, M., Sai Ram, T., Rami Reddy, C. V., 2024. A review on YOLOV8 and its advancements. International Conference on Data Intelligence and Cognitive Informatics, Springer, Singapore, pp. 529_545.
Vlaminck, M., Diels, L., Philips, W., Maes, W., Heim, R., Wit, B. D., Luong, H., 2023. A multisensor UAV payload and processing pipeline for generating multispectral point clouds. Remote Sensing 15, 1524. ‏
Zhang, C., Kovacs, J. M., 2012. The application of small unmanned aerial systems for precision agriculture: a review. Precision Agriculture 13, 693_712. ‏
Zhang, Y., Wu, K., Wang, J., 2021. Livestock classification and counting in quadcopter aerial images using Mask R-CNN. International Journal of Remote Sensing 42, 2487_2506.
Zhou, X., Ren, H., Zhang, T., Mou, X., He, Y., Chan, C. Y., 2022. Prediction of pedestrian crossing behavior based on surveillance video. Sensors 22, 1467. 
Journal of Livestock Science and Technologies

Articles in Press, Accepted Manuscript
Available Online from 10 June 2026

Files

Share

How to cite

Statistics