Research On The Key Algorithms Of Power Station Fire Extinguishing Robot

In recent years,there have been dangerous accidents in the power plant,especially in the high-voltage power distribution room.In the emergency repair operation,facing the dangerous environment,it is easy to cause casualties,and huge property losses.In recent years,artificial intelligence technology has continued to grow.The fire fighting robots has replaced the field operations of people,making fire extinguishing operations safer and more efficient,and greatly reducing unnecessary loss.This paper studies several key technologies of fire fighting robots for fire scenarios.The main contents of this paper are as follows:First,this paper studies the flame detection algorithm of fire fighting robots.This paper proposes a flame detection algorithm based on the integration of multiple classifiers.First,an adaptive parameter updating algorithm is proposed.The background complexity is used to adaptively adjust the discrimination threshold and update rate of the background.Then this paper proposed a fire detection algorithm based on multiple classifiers.The foreground regions are sent to the color,shape and motion classifiers,then the total values of the three classifiers are obtained as the final result.The algorithm comprehensively considers the various characteristics of the fire.Then,based on the detection of the flame,an efficient and accurate target tracking technology is proposed for the specific scene,which is improved on the basis of SiamRPN.The fusion method is used to enable the network to simultaneously utilize the structural and semantic features of the image.The anchorless network is designed.And the backbone network is equipped with an online update module,so that the network can further adapt to changes in the target.The improved algorithm has robustness to target changes,and solves the drift phenomenon caused by traditional tracking algorithms.Then on the basis that the emergency robot can detect and track the fire in real time,this paper studies the target positioning algorithm,and proposes an image alignment algorithm and a point cloud generation algorithm.And the robot follow-up control system is designed,so that the robot can follow the flame movement.Finally,based on the content of the previous chapters,this article completes the fire extinguishing operation experiment.First,the emergency robot follow-up control algorithm was verified,and its follow-up result in three different flame motion types was verified;then the robot fire extinguishing operation was verified in the case of interference with suspected flames;finally,the experiment was performed on under the different flame moving speed,when equipped with different tracking algorithms.