Abstract:Early wildfire detection in remote and forest areas is crucial for minimizing devastation and preserving ecosystems. Autonomous drones offer agile access to remote, challenging terrains, equipped with advanced imaging technology that delivers both high-temporal and detailed spatial resolution, making them valuable assets in the early detection and monitoring of wildfires. However, the limited computation and battery resources of Unmanned Aerial Vehicles (UAVs) pose significant challenges in implementing robust and efficient image classification models. Current works in this domain often operate offline, emphasizing the need for solutions that can perform inference in real time, given the constraints of UAVs. To address these challenges, this paper aims to develop a real-time image classification and fire segmentation model. It presents a comprehensive investigation into hardware acceleration using the Jetson Nano P3450 and the implications of TensorRT, NVIDIA's high-performance deep-learning inference library, on fire classification accuracy and speed. The study includes implementations of Quantization Aware Training (QAT), Automatic Mixed Precision (AMP), and post-training mechanisms, comparing them against the latest baselines for fire segmentation and classification. All experiments utilize the FLAME dataset - an image dataset collected by low-altitude drones during a prescribed forest fire. This work contributes to the ongoing efforts to enable real-time, on-board wildfire detection capabilities for UAVs, addressing speed and the computational and energy constraints of these crucial monitoring systems. The results show a 13% increase in classification speed compared to similar models without hardware optimization. Comparatively, loss and accuracy are within 1.225% of the original values.

What problem does this paper attempt to address?

### Problems Addressed by the Paper The paper aims to address the issue of early wildfire detection in remote and forested areas. Specifically, it focuses on how to utilize autonomous Unmanned Aerial Vehicles (UAVs) for real-time wildfire detection. Although autonomous UAVs are equipped with advanced imaging technologies that can provide high temporal and spatial resolution images, their limited computational power and battery resources restrict the deployment of efficient image classification models on these platforms. Current research mostly relies on offline processing, lacking real-time inference capabilities. Therefore, this paper proposes a real-time image classification and fire segmentation model and explores the impact of hardware acceleration technologies (such as Jetson Nano P3450 and TensorRT) on fire classification accuracy and speed. ### Main Contributions 1. **Model Optimization and Training**: - Proposed a combined classification and segmentation model architecture based on MobilenetV3 and DeepLabV3+, optimized through Quantization Aware Training (QAT) and Automatic Mixed Precision (AMP). - Developed a custom CUDA kernel to optimize classification speed, reduce power consumption, and minimize memory management overhead. 2. **Hardware Acceleration**: - Utilized Jetson Nano P3450 and CUDA 12.3 to improve model inference speed and efficiency through hardware acceleration technologies. - Explored the impact of different activation functions (ELU, ReLU, PReLU) on model training and memory efficiency. 3. **Performance Evaluation**: - Conducted experiments using the FLAME dataset to evaluate the model's performance metrics under different configurations, including latency, Mean Pixel Accuracy (MPA), Mean Intersection over Union (MIOU), Frames Per Second (FPS), batch size, throughput, and memory utilization. - Experimental results showed a 13% increase in classification speed compared to similar models without hardware optimization, with loss and accuracy within 1.225%. ### Experimental Design and Results - **Dataset**: Used the FLAME dataset, containing 2,003 high-resolution fire images, expanded to 4,006 images through data augmentation. - **Experimental Setup**: Implemented the model using PyTorch and trained it with a custom CUDA optimizer. Evaluated the performance of different activation functions (ReLU, ELU, PReLU) in the model. - **Performance Metrics**: Results indicated that ReLU exhibited the best performance in classification tasks, with higher MPA and MIOU, and faster FPS. - **Quantization and Optimization**: FP16 quantization significantly improved throughput and reduced latency, providing an effective method for real-time fire detection on edge devices. ### Conclusion Through hardware acceleration and model optimization, this study successfully achieved efficient real-time fire detection on resource-constrained UAV platforms. This achievement not only enhances the speed and accuracy of fire detection but also provides valuable experience and technical support for deploying similar real-time detection systems on edge devices in the future. Faster inference speeds enable quicker responses to fires, thereby saving lives and protecting ecosystems.

Hardware Acceleration for Real-Time Wildfire Detection Onboard Drone Networks

Real-Time Wildfire Monitoring Using Low-Altitude Remote Sensing Imagery

Detecting Wildfires on UAVs with Real-time Segmentation Trained by Larger Teacher Models

AI Enabled Fire-Detecting Drone Using Deep Learning Model

Papaya Tree Detection with UAV Images Using a GPU-Accelerated Scale-Space Filtering Method

An Efficient Wildfire Detection System for AI-Embedded Applications Using Satellite Imagery

Streamlining Forest Wildfire Surveillance: AI-Enhanced UAVs Utilizing the FLAME Aerial Video Dataset for Lightweight and Efficient Monitoring

An FPGA smart camera implementation of segmentation models for drone wildfire imagery

Implementation of Lightweight Convolutional Neural Networks with an Early Exit Mechanism Utilizing 40 nm CMOS Process for Fire Detection in Unmanned Aerial Vehicles

SegNet: A Segmented Deep Learning based Convolutional Neural Network Approach for Drones Wildfire Detection

Accelerated Fire Detection and Localization at Edge

Autonomous Satellite Wildfire Detection Using Hyperspectral Imagery and Neural Networks: A Case Study on Australian Wildfire

Development of a deep learning-based surveillance system for forest fire detection and monitoring using UAV

Wildland Fire Detection and Monitoring Using a Drone-Collected RGB/IR Image Dataset

Deep Learning and Transformer Approaches for UAV-Based Wildfire Detection and Segmentation

Obscured Wildfire Flame Detection By Temporal Analysis of Smoke Patterns Captured by Unmanned Aerial Systems

Applying deep learning to real-time UAV-based forest monitoring: Leveraging multi-sensor imagery for improved results

Fast Forest Fire Detection and Segmentation Application for UAV-Assisted Mobile Edge Computing System

A UAV-Assisted Edge Framework for Real-Time Disaster Management

AERO: AI-Enabled Remote Sensing Observation with Onboard Edge Computing in UAVs

Lightweight and attention-based CNN architecture for wildfire detection using UAV vision data