A Study on the Application of Image Analysis Techniques to Autocoders in Civil and Hydraulic Engineering

With the continuous updating of computer vision and image analysis technology, image processing, as well as analysis technology has become an important auxiliary means in the analysis of civil engineering and water conservancy projects. In this study, an image analysis model is constructed by defining the autoencoder and its expansion relation, combined with a convolutional neural network. On this basis, an engineering detection model is built by using a sparse-stacked autoencoder. The structural sub-transmission characteristics of civil engineering and water conservancy projects were investigated. The image separation optimization was carried out by using Beer Lambert’s law, and finally the engineering structure extraction and recognition model based on deep learning was formed. Then, the performance of the model is examined. The average value of the repetition rate is higher than 80% in the brightness adjustment, rotation, and scaling operation change test. The Dice and IoU indexes are higher than 90%, and the HD distance is less than 27mm, so the overall performance is excellent. The practical application of civil engineering and water conservancy engineering has a fantastic performance, with a relative error of no more than 2%. The method in this paper has excellent stability and practical effect. It proposes an improvement method for optimizing the image analysis method in civil engineering and water conservancy engineering.