Road surface crack defect detection method based on combination of image processing and convolutional neural network

Abstract

The invention discloses a road surface crack defect identification method based on combination of image processing and a convolutional neural network. The method comprises the following steps: step 1,data preparation: acquiring 2000 1024 * 512 road grayscale images; step 2, image calibration: classifying the collected road images, calibrating the images without cracks as 0 folder, and calibratingthe images with cracks as 1 folder; step 3, carrying out picture preprocessing on the calibrated picture, and the unified size being 1024*512; step 4, establishing a core convolutional neural networkalgorithm frame AsphaltCackNet; step 5, generating a principle of a feature extractor for feature mapping; step 6, dividing an image data set for training and testing; step 7, training the data set and calculating a cost function; step 8, a learning method of the neural network; step 9, randomly initializing and losing; step 10, performing parallel computing; step 11, obtaining a training result,and performing testing and evaluation. Under the condition that the width and the height of an original image are not changed, perfect pixel accuracy can be ensured.

Description

technical field [0001] The invention relates to a method for detecting and classifying crack defects on the road surface. Background technique [0002] In the 21st century, my country's comprehensive national strength has developed rapidly, and the vigorous development of expressways has become an important symbol, and the protection of roads has also received equal attention. However, the efficiency of manual work is low, especially there are many unpredictable factors in the on-site construction, which makes the detection of road defects particularly difficult, and the accuracy cannot be guaranteed, resulting in many road defects not being detected. In the past, there have been many algorithms to detect road surface crack defects, but all have their limitations. Algorithms such as edge detection and image segmentation and morphological transformation based on image preprocessing are widely used to detect the edges of pavement cracks, but cannot detect the complete crack p...

AI Technical Summary

Problems solved by technology

However, due to the existence of the pooling layer, the original data gradually shrinks by increasing the level of abstraction, resulting in

Images