Fire Source Recognition and Localization Algorithm for Tracked Firefighting Robots Based on Integrated Convolutional Neural Networks

With the extensive use of firefighting robots in complex fire environments, the rapid identification and positioning of fire sources has become an important key for improving robot work efficiency and ensuring fire safety. Innovative to overcome the limitations of existing methods, this paper presents a convolutional neural network based ignition source identification, localization algorithm. The collection of the multiple CNN architectures was conducted in this study to make full use of the feature extraction and classification ability of the ignition source, based on the ensemble learning technology to improve the classification performance and the accuracy of the ignition source localization based on the combination of deep reinforcement learning strategy. Then we use the introduced multi-scale convolution and feature fusion mechanism from the perspective of demand, which makes the model able to cope with diverse ignition source features, and adopt the spatial attention mechanism to improve positioning accuracy. The experimental results demonstrate that the proposed algorithm outperforms traditional methods in ignition source recognition accuracy and positioning speed, and can deliver more effective emergency response in real scenarios.