Research on Flame Detection of Firefighting Robot Based on YOLOv5
Firefighting robots strongly assist firefighters in obtaining information about fire scenes and making timely decisions.However,the detection accuracy and high computational complexity of the model in firefighting operations needs to be further improved.An improved YOLOv5 flame detection algorithm for fire-fighting robots is proposed.This method adopts the Involution operator to the feature extraction part of the original YOLOv5 model,which expands the receptive field and makes the network lighter.CBAM attention mechanism is introduced to the feature extraction and feature fusion part to further reform the feature extraction in the network.Furthermore,to ensure the feature information of the underlying feature map,a new bounding box loss function α-CIoU is added in the loss function calculation to improve the convergence speed of the model and the robustness to the data set.The experimental results demonstrate that the improved model achieves an average detection accuracy of 93.6%and reduces the computational workload of the model by 58%.This method effectively enhances the accuracy of flame detection while reducing the model's computational overhead.
flame detectionattention mechanismsboundary box loss function
NETL
NSTL
万方数据
