Research On Fire Sensing System Of Fire Robot Based On Multi-source Information Fusion

With the rapid development of all walks of life in our country,it is also accompanied by the occurrence of many major fire accidents.Human beings’ ability to control and eliminate fires is one of the important manifestations of social progress.Firefighting robot is a new type of firefighting tool.Its adaptability is stronger than that of firefighters,and it can work in high-risk environments such as low oxygen concentration,dense smoke and toxic gases.Therefore,it is particularly important to study the fire-fighting robot’s ability to perceive fire.The fire-sensing ability of a firefighting robot is the basis for the robot to perform tasks.By carrying a variety of sensors and cameras,and using multi-source information fusion technology to make decisions,robots can better perceive the environment,so as to formulate the best mission execution plan,and provide more efficient and safe technical support for disaster rescue.This thesis focuses on the on-site work process of fire-fighting robots,and conducts research on the fire perception system in complex environments.The specific research contents are as follows:(1)This article proposes a fire perception scheme that comprehensively analyzes the changes in fire characteristics in the environment surrounding a firefighting robot by integrating the detection data from various sensors and video image data.The scheme adopts a complementary strategy between the camera and sensor devices,which can not only reduce the demand for sensor devices in the fire perception system to a certain extent,but also make the fire detection more comprehensive and accurate.(2)This thesis proposes a fire perception method based on multi-sensor data fusion,aiming to address the issues caused by a single sensor in fire perception.By analyzing the limitations of traditional data detection methods,the proposed method utilizes the Kalman filter algorithm to fuse and update the collected raw data.Then,fuzzy logic reasoning is used to construct the basic probability assignment function of fire-related evidence.Finally,the D-S evidence theory algorithm is employed for global fusion of multi-sensor data to determine the fire situation and make decisions.(3)This thesis proposes a fire recognition method based on video image detection to address the issue of reduced detection effectiveness or even failure when the distance between the fire source and multiple sensors is too far.The method uses the YOLOv5 object detection algorithm as its foundation and improves the model by introducing attention mechanism and permutation bounding box regression loss function.Experimental results show that the improved object detection network in this article has a good performance in fire recognition.In summary,this thesis proposes the above two methods to perceive fire according to the needs of fire-fighting robots.The experimental results show that the multi-sensor perception method can effectively obtain the characteristic information of the fire,and the video image-based perception method can solve the problem of poor effect of multisensor long-distance perception of fire.This thesis effectively combines two fire perception methods to further improve the comprehensiveness and accuracy of fire perception by fire-fighting robots.