Corrosion Residual Strength Evaluation of Comprehensive Pipe Gallery Concrete Based on Improved Faster-RCNN

Aiming at the concrete corrosion problem in comprehensive pipeline corridors and other projects, the study proposes a residual strength assessment method for concrete corrosion based on Faster Region-based Convolutional Neural Networks (Faster-RCNN). This method indirectly evaluates the residual strength of concrete corrosion by identifying the corrosion characteristics of the concrete sample surface through Faster RCNN. Faster RCNN consists of Region Proposal Network (RPN) and Fast RCNN networks. The study improved the original Visual Geometry Group 16 (VGG16) network in both RPN and Fast RCNN networks to enhance the accuracy and efficiency of the Faster RCNN model in evaluating the residual strength of concrete corrosion. The results show that the Faster-RCNN model constructed in the study is more accurate and efficient for the assessment of concrete corrosion residual strength. The results show that the Faster-RCNN model has very high recall and precision, with a precision of 0.841 when the recall is 0.900, and a precision of 0.761 when the precision is 0.900. It takes about 2500-2600s to process 1000 concrete images with the evaluation model, and the accuracy of the evaluation is maintained at more than 96%. . It shows that the Faster-RCNN model can effectively evaluate the residual strength of concrete corrosion and provide technical support for the evaluation of the quality of the same type of concrete.