Scholarlink

Purpose: <p> While autonomous rovers have become indispensable to precision farming, achieving consistent operational safety remains a critical challenge. Conventional safety sensors, such as LiDAR, fail to detect obstacles positioned below the plant canopy, posing a significant risk. While camera-based supervised learning methods can detect common objects, they perform poorly when faced with obstacles that were not present in their training data. Current unsupervised anomaly detection offers a solution by learning the normal visual patterns of an environment, but often fails for the dynamic scenes captured by a moving rover. </p> Methods: <p> This paper introduces Video Memory Transformers for Anomaly Detection (VMTAD), a fully unsupervised method designed for real-time obstacle detection in dynamic agricultural scenes. VMTAD utilizes a transformer-driven architecture augmented with a dedicated memory module. This memory module leverages temporal context by processing encoded representations of preceding frames. This approach enables the system to effectively address the dynamic context caused by the robot’s movement. The model is trained using only images that represent normal operation, requiring no data labels. </p> Results: <p> VMTAD was rigorously evaluated on the ’Grillon’ agricultural rover. On a challenging rapeseed dataset, VMTAD achieved competitive performance, reaching a 0.973 detection and 0.997 segmentation Area Under the Receiver Operating Characteristic curve. A lightweight variant provides an optimal balance of high accuracy and real-time inference (14 ms), which is critical for safety, as confirmed by our analysis of the rover’s total stopping distance. </p> Conclusion: <p> VMTAD enhances the detection of agricultural anomalies by combining Transformers with temporal memory. It achieves the highest score against agricultural anomaly detection methods, and a result close to state-of-the-art against industrial methods, with low inference time. Despite cross-domain challenges, this spatio-temporal framework provides a foundation for sustainable, efficient, autonomous farming.</p> <p>Code available in the following repository: <a href="https://github.com/TheoBiardeau/VMTAD">https://github.com/TheoBiardeau/VMTAD</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Unsupervised memory-enhanced video transformers: obstacle detection for autonomous agricultural rover

  • 期刊论文

  • 发表日期：

    2026年5月19日

查看全文
  • Théo Biardeau,
  • Anne-Sophie Capelle-Laizé,
  • Salwan Alwan,
  • David Helbert

摘要

Purpose:

While autonomous rovers have become indispensable to precision farming, achieving consistent operational safety remains a critical challenge. Conventional safety sensors, such as LiDAR, fail to detect obstacles positioned below the plant canopy, posing a significant risk. While camera-based supervised learning methods can detect common objects, they perform poorly when faced with obstacles that were not present in their training data. Current unsupervised anomaly detection offers a solution by learning the normal visual patterns of an environment, but often fails for the dynamic scenes captured by a moving rover.

Methods:

This paper introduces Video Memory Transformers for Anomaly Detection (VMTAD), a fully unsupervised method designed for real-time obstacle detection in dynamic agricultural scenes. VMTAD utilizes a transformer-driven architecture augmented with a dedicated memory module. This memory module leverages temporal context by processing encoded representations of preceding frames. This approach enables the system to effectively address the dynamic context caused by the robot’s movement. The model is trained using only images that represent normal operation, requiring no data labels.

Results:

VMTAD was rigorously evaluated on the ’Grillon’ agricultural rover. On a challenging rapeseed dataset, VMTAD achieved competitive performance, reaching a 0.973 detection and 0.997 segmentation Area Under the Receiver Operating Characteristic curve. A lightweight variant provides an optimal balance of high accuracy and real-time inference (14 ms), which is critical for safety, as confirmed by our analysis of the rover’s total stopping distance.

Conclusion:

VMTAD enhances the detection of agricultural anomalies by combining Transformers with temporal memory. It achieves the highest score against agricultural anomaly detection methods, and a result close to state-of-the-art against industrial methods, with low inference time. Despite cross-domain challenges, this spatio-temporal framework provides a foundation for sustainable, efficient, autonomous farming.

Code available in the following repository: https://github.com/TheoBiardeau/VMTAD.