Road disaster remote sensing intelligent detection method based on deep learning

Abstract

The invention discloses a deep learning-based road disaster remote sensing intelligent detection method. The method comprises the following steps of A, image preprocessing: performing radiometric calibration, atmospheric correction, geometric correction, ortho-rectification and image registration on a remote sensing image; B, establishing a road buffer area: according to the road vector data and the image resolution, establishing an area 3-4 times based on the road width as the road buffer area; C, image masking: performing masking processing on the remote sensing image according to the established road buffer area; D, image block cutting: cutting the masked remote sensing image into a plurality of image blocks; E, image block prediction: inputting the cut image blocks, namely effective image blocks, into the trained YOLOV3 network, and predicting the probability that the image blocks belong to different damage categories of the road; and F, merging the image blocks: merging the predicted image blocks to generate an image result image with the original size. The method is convenient and easy to operate, and an accurate road disaster detection distribution result is obtained.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image processing, and more specifically relates to an intelligent detection method for road disasters based on deep learning, which is suitable for optical remote sensing images with a resolution of 0.5m-2m. Background technique [0002] Transportation plays an indispensable and important role in post-earthquake rescue. However, due to the impact of the earthquake, transportation elements are often damaged to varying degrees and cannot pass normally, which brings many challenges to the entire rescue operation. Rescue has the biggest impact. Therefore, how to quickly and accurately obtain the road disaster situation after the earthquake is very important for the smooth progress of the entire post-disaster rescue activities. Relying on traditional methods, it is difficult to quickly understand the disaster situation of the roads in the disaster-stricken area from the ground, so that it is im...

AI Technical Summary

Problems solved by technology

First, use the constructed road disaster sample to lock the detection object to the disaster itself, and solve the error product effect caused by the multi-step operation of the general disaster extraction technology; then input the sample into the YOLOv3 network for training, use the model to learn the semantic features of the road disaster, and the training is completed Finally, the detection of road disasters solves the problem that the conventional method is not highly automated, the detection accuracy of road disasters depends on the accuracy of road extraction and the detection time is long; in addition, before the data to be detected is input into the network, the road vector data is used to construct the relevant The road buffer zone of the road, and the road and its surrounding areas are extracted through the image mask technology, which solves the problem of false detection caused by the excessive image range in the detection process, and retains the necessary semantic information for detection, further improving the detection efficiency

Images