Research On Real-time Detection And Localization Technology Of Fire And Smoke For Firefighting Robots

Fire is a serious natural disaster that causes significant human casualties and property damage.In recent years,fire-fighting technology based on fire-fighting robots has developed rapidly.However,due to their poor autonomous perception capabilities in complex fire environments,the intelligence level of fire-fighting robots is low,making it difficult to achieve autonomous fire-fighting.This article focuses on the core issue of firefighting robot perception in fire environments: fire detection and fire center point localization,and the main content is as follows:Firstly,a fire and smoke detection algorithm,FSD-YOLO,based on deep learning was designed and implemented.A fire and smoke detection dataset that reflects the distribution of real fire scenes was created.An efficient sampling augmentation module was designed to enhance the feature extraction of fires and smoke.A class-balanced loss function was designed to address the problem of uneven learning effects between the two categories of fires and smoke.The designed FSD-YOLO algorithm for fire and smoke detection improved the detection accuracy by 8%.Secondly,a fire center localization algorithm was proposed by combining a semantic segmentation algorithm with stereo vision to locate the fire point,and an unsupervised clustering algorithm was used to obtain the fire center and filter out outliers.A validation dataset was created to test the accuracy of the localization algorithm,and the experimental results effectively demonstrated the accuracy of the fire center localization algorithm.Finally,a hardware and software system for fire detection and localization was designed.Under this system,mobile firefighting robots can automatically map and locate themselves and conduct fire inspections.When a fire is detected,the fire center localization algorithm can be used to broadcast the fire point location and automatically extinguish the fire.The validation experiments demonstrated the feasibility and effectiveness of the system.In summary,this article presents a comprehensive approach to the detection and localization of fires by mobile firefighting robots.The proposed FSD-YOLO algorithm for fire and smoke detection and the fire center localization algorithm show promising results,and the designed hardware and software system provides a feasible solution for fire detection and extinguishing in real-world scenarios.