Road damage detection method and device based on deep learning image classification

Abstract

The invention discloses a road damage detection method and device based on deep learning image classification. The method comprises the following steps: Step one, collecting images of different road scene types and position information to form a sample image set and marking the road scene types; Step two, marking the road scene types; Step three, selecting a classification network model accordingto the need, and training the model; Step four, inputting an image of a road to be detected into the trained deep neural network model to obtain a classification result; and Step five, if a classification result is road damage type, confirming the road segment position information corresponding to the acquired image, and outputting prompt information of the road damage type and the position information. The method improves the accuracy of road damage detection, does not need to set a detection threshold value, and has high real-time performance and diversified mounting position selection.

Description

technical field [0001] The invention relates to the field of intelligent transportation research, in particular to a road damage detection method for road maintenance and management by highway management departments, and in particular to a road damage detection method and device based on deep learning image classification. Background technique [0002] With the continuous improvement of the level of motorization in our country, the construction of urban infrastructure has developed rapidly, the urban area has continued to expand, and the urban road mileage and road density have also increased rapidly. However, due to the impact of vehicle loads, weather conditions, etc., coupled with problems in design methods, construction techniques, construction quality, and lagging maintenance work, road pavement damage is very common. Pavement damage detection plays a vital role in road maintenance and it helps in providing higher quality transportation services. [0003] The tradition...

AI Technical Summary

Problems solved by technology

[0003] The traditional road damage detection work in my country mainly uses manual detection, but manual detection has obvious defects: it relies on the experience of the staff, there is no uniform standard for the measurement results, the original data is incomplete, and the measurement data is difficult to approve

Some existing automatic collection devices have many components and high installation requirements, and the applicable vehicle models are relatively fixed; based on Sobel edge separation and morphological filter road damage detectors, if there are sidewalks, garbage and other non-road damage interference on the road , the detector is prone to failure leading to inaccurate results

Road damage detection based on convolutional neural network has poor real-time performance and requires high equipment hardware requirements

Images