Method for constructing recognition model of moisture damage of asphalt pavement and method and system for recognizing moisture damage of asphalt pavement

A method for constructing a recognition model of a moisture damage of an asphalt pavement, comprising the following steps:

• S1, pre-processing a Ground Penetrating Radar (GPR) pavement investigation data set to obtain an initial data set with the moisture damage, a bridge joint and a normal pavement;
• S2, performing a continuous wavelet transform on the initial data set by using the continuous wavelet transform, and taking an amplitude of the continuous wavelet transform to construct a first time-frequency image data set;
• S3, filtering a first time-frequency image in the first time-frequency image set to obtain a second time-frequency image to construct a second first time-frequency image data set;
• S4, performing a normalization processing on the second time-frequency image in the second time-frequency image set to obtain a third time-frequency image to construct a third time-frequency image set, and tagging a moisture damage classification label for the third time-frequency image of the third time-frequency image set;
• and
• S5, constructing the recognition model:
• training a preconstructed GPRMCNN deep learning model to obtain a trained recognition model by taking the third time-frequency image set as input data and the moisture damage classification label as output data,
• wherein the preconstructed GPRMCNN deep learning model adopts 16 layers of convolutional neural networks, comprising three convolutional layers and one full connecting layer, a size of a convolution kernel used by a convolutional operation is 3*3, a down sampling is arranged behind previous two convolutional layers and a last convolutional layer does not have a pool layer of the down sampling and is connected to the one full connecting layer through a Drop network.