FieldNet: Efficient real-time shadow removal for enhanced vision in field robotics

📅 2024-03-13
🏛️ Expert systems with applications
📈 Citations: 1
Influential: 0
📄 PDF
🤖 AI Summary
To address degraded detection and localization performance in field robotics vision caused by dynamic shadows, this paper proposes a lightweight, end-to-end differentiable real-time shadow removal method. We introduce the first physics-inspired shadow modeling module explicitly designed for dynamic outdoor scenes, integrated with a dual-branch U-Net variant to enable illumination-invariant feature extraction, adaptive shadow mask prediction, and edge-aware reconstruction. Evaluated on the FieldShadow benchmark, our method achieves 47 FPS on Jetson AGX Orin with a PSNR of 32.7 dB, significantly outperforming existing state-of-the-art approaches. Key contributions include: (i) the first efficient, differentiable shadow modeling framework tailored to dynamic outdoor environments; (ii) a lightweight architecture balancing physical plausibility and millisecond-scale inference; and (iii) substantially improved robustness of visual perception under complex, varying illumination conditions.

Technology Category

Application Category

Problem

Research questions and friction points this paper is trying to address.

Real-time shadow removal for outdoor field robotics vision
Addressing inconsistent shadow boundaries and artifact generation
Enhancing weed detection accuracy in precision agriculture
Innovation

Methods, ideas, or system contributions that make the work stand out.

Deep learning framework for real-time shadow removal
Probabilistic enhancement module and novel loss function
No shadow masks required during inference
A
Alzayat Saleh
College of Science and Engineering, James Cook University, Townsville, 4814, QLD, Australia
A
A. Olsen
AutoWeed Pty Ltd, Townsville, 4814, QLD, Australia
J
Jake C. Wood
AutoWeed Pty Ltd, Townsville, 4814, QLD, Australia
B
B. Philippa
College of Science and Engineering, James Cook University, Townsville, 4814, QLD, Australia
M
M. Azghadi
College of Science and Engineering, James Cook University, Townsville, 4814, QLD, Australia