An asphalt pavement structure depth calculation method based on a generalized regression neural network

Abstract

The invention discloses an asphalt pavement structure depth calculation method based on a generalized regression neural network. The method comprises steps of collecting two-dimensional images of thesurfaces of the prepared concrete test piece before and after sand laying; reconstructing a pavement texture three-dimensional model according to the gray value of the two-dimensional image; calculating a sand paving plane of the sand paving image by using a digital image processing technology; meanwhile, the sand laying plane is also a reference surface of the pavement texture three-dimensional model; fitting the pavement texture three-dimensional model above the reference surface by adopting a least square method, and generating a fitting curved surface; calculating the volume between the two sides of the reference surface and the three-dimensional model according to an integration method; wherein the ratio of the volume to the projection area of the pavement texture three-dimensional model on the horizontal plane is the average elevation Ha of the asphalt concrete test piece image; and taking the average elevation Ha, the extreme value and the gray average value of the test piece image as input samples of a generalized regression neural network, taking the actual pavement construction depth as output samples, carrying out neural network model training, and predicting the pavement construction depth by using the trained model, with the prediction precision reaching 90% or more.

Description

[0001] Field [0002] The invention belongs to the technical field of road engineering subject detection, and mainly relates to road non-destructive detection technology, in particular to a method for calculating depth of asphalt pavement structure based on generalized regression neural network. Background technique [0003] The wide application of asphalt pavement highlights the limitations of traditional pavement detection methods. Digital image technology not only enriches the pavement detection methods, but also changes the detection method from manual to semi-automatic, which greatly improves the detection efficiency and has high measurement accuracy and no detection. Advantages of injury. [0004] The depth of pavement structure is an important index to characterize the macroscopic structure of pavement. The current measurement methods of pavement structure depth mainly include volumetric method, digital image technology method, laser method and so on. The human factors...

AI Technical Summary

Problems solved by technology

However, the rationality of the single digital image method to calculate the pavement structure depth needs to be improved. The integral idea based on the volume method makes the calculated structure depth inconsistent with the actual situation. Therefore, a new and more reasonable pavement structure depth calculation method is needed.

Images