Water leakage detection and its application of turbine floor equipment based on MaskRCNN

Abstract

Water leakage detection is of great significance to reduce economic losses and safety hazards in industrial sites. Aiming at the problem of water leakage detection caused by weak light condition and uneven light on turbine layer, this paper proposes a water leakage detection method based on weak light compensation for MaskRCNN turbine layer equipment. Firstly, due to the poor light of the turbine layer, the boundary of the seepage area photographed by the inspection robot is not clear, and the segmentation method is difficult to determine the target boundary. In this paper, low light enhancement algorithm is used to process the image of water seepage area taken by inspection robot to strengthen the boundary of water seepage image. Then, software Labelme was used to label the water seepage area of the enhanced inspection image and generate the configuration file corresponding to the inspection image. Thirdly, MaskRCNN instance segmentation network is introduced to train the image of the seepage area, and the network model of the detection task of the seepage area is generated. Finally, the test image is used to detect the water seepage area. After weak light enhancement, the trained MaskRCNN model is imported to obtain the detection result of the water seepage area. The experimental results show that the proposed method is effective and accurate, and it has a good application effect in detecting water dripping and leakage of turbine floor equipment by the wheel inspection machine in Qingyuan Energy Storage Power Generation Station.