[Paper Review] Systematic Literature Review of Vision-Based Approaches to Outdoor Livestock Monitoring with Lessons from Wildlife Studies

Stacey D. Scott, Zayn J. Abbas+7 more|arXiv (Cornell University)|Oct 7, 2024

Wildlife Ecology and ConservationEnvironmental Science3 citations

TL;DR

This systematic literature review analyzes vision-based computer vision approaches for outdoor livestock monitoring, drawing parallels with wildlife studies to address challenges in large, unstructured environments. It identifies deep learning as the dominant approach for animal detection, counting, and multi-species classification, while highlighting open technical challenges and future research directions in precision livestock farming.

ABSTRACT

Precision livestock farming (PLF) aims to improve the health and welfare of livestock animals and farming outcomes through the use of advanced technologies. Computer vision, combined with recent advances in machine learning and deep learning artificial intelligence approaches, offers a possible solution to the PLF ideal of 24/7 livestock monitoring that helps facilitate early detection of animal health and welfare issues. However, a significant number of livestock species are raised in large outdoor habitats that pose technological challenges for computer vision approaches. This review provides a comprehensive overview of computer vision methods and open challenges in outdoor animal monitoring. We include research from both the livestock and wildlife fields in the review because of the similarities in appearance, behaviour, and habitat for many livestock and wildlife. We focus on large terrestrial mammals, such as cattle, horses, deer, goats, sheep, koalas, giraffes, and elephants. We use an image processing pipeline to frame our discussion and highlight the current capabilities and open technical challenges at each stage of the pipeline. The review found a clear trend towards the use of deep learning approaches for animal detection, counting, and multi-species classification. We discuss in detail the applicability of current vision-based methods to PLF contexts and promising directions for future research.

Motivation & Objective

• To identify and analyze computer vision methods used for outdoor livestock monitoring in large, unstructured environments.
• To explore the applicability of wildlife monitoring techniques to livestock contexts due to shared challenges in appearance, behavior, and habitat.
• To map the current state of the art across the image processing pipeline stages: detection, counting, classification, and tracking.
• To identify open technical challenges and research gaps in vision-based livestock monitoring systems.
• To provide actionable insights and future research directions for precision livestock farming (PLF) using vision technologies.

Proposed method

• Conducted a systematic literature review focusing on computer vision methods applied to outdoor monitoring of large terrestrial mammals in livestock and wildlife contexts.
• Used an image processing pipeline framework to categorize and analyze methods across key stages: detection, counting, classification, and tracking.
• Reviewed 281 papers from both livestock and wildlife research to identify trends, techniques, and limitations in vision-based monitoring.
• Focused on deep learning-based models, particularly convolutional neural networks (CNNs) and transformer-based architectures, for object detection and classification.
• Evaluated methodological choices, dataset characteristics, and performance metrics across studies to assess robustness and generalizability.
• Synthesized findings into a structured analysis of technical challenges and future research opportunities in outdoor animal monitoring.

Experimental results

Research questions

• RQ1 What are the dominant computer vision techniques used for outdoor livestock monitoring, and how do they compare to those used in wildlife studies?
• RQ2 To what extent can wildlife monitoring methodologies be adapted for precision livestock farming in outdoor settings?
• RQ3 What are the key technical challenges in animal detection, counting, and classification under real-world outdoor conditions?
• RQ4 How do current vision-based systems perform across different species and environmental conditions in large, unstructured habitats?
• RQ5 What are the most promising research directions for improving the robustness and scalability of vision-based livestock monitoring systems?

Key findings

• Deep learning models, particularly CNNs and transformer-based architectures, are the dominant approach for animal detection, counting, and multi-species classification in outdoor settings.
• A clear trend is observed toward end-to-end learning frameworks that integrate detection and segmentation, improving accuracy in complex outdoor scenes.
• Challenges related to occlusion, varying lighting, and seasonal changes in vegetation significantly impact model performance, especially in dense or dynamic environments.
• Multi-species classification remains a challenge due to visual similarities between species and limited annotated datasets for diverse outdoor livestock species.
• Wildlife monitoring research offers valuable methodological insights—such as transfer learning and domain adaptation—that can be leveraged to improve livestock monitoring systems.
• Despite progress, few studies validate systems under long-term, real-world outdoor conditions, highlighting a gap in robustness and scalability testing.