A method for automatic extraction and change analysis of low-level roads

Abstract

The invention discloses a method for automatic extraction and change analysis of low-level roads. The method can realize automatic extraction and change analysis of road network, and conduct sample analysis of geometric, texture and spectral characteristics of road materials through high-resolution remote sensing images. Select, generate a data set for model training, and use the generated model to automatically extract roads. The road is extracted based on the data model, and the automatic registration of the image extraction result and the network data is realized. Set thresholds, confidence intervals, etc. for comparative change analysis, and generate fusion and reduction data sets. It can greatly reduce the work of generating training data sets. The image-based road extraction results are fused with network road data. The road data of the fusion results are more accurate. A new data set can be automatically generated according to the fusion results, which greatly reduces the cost of training data sets. Generate jobs.

Description

technical field [0001] The invention relates to the technical field of road image extraction and analysis, in particular to a method for automatic extraction and change analysis of low-level roads. Background technique [0002] Highways are an important part of national infrastructure, and as the "main artery" of transportation, they are closely related to local economic development. With the rapid development of my country's economy, the speed of urban construction has greatly increased, and the road network has been updated rapidly. The delay in updating road network data is a common problem faced by developed and developing countries. Fast and accurate road network updates can provide assistance to regional economic development, and also provide services for travel route planning, urban construction, disaster warning and even military strikes, laying the foundation for the development of smart cities and driverless technology. [0003] The traditional road network update...

AI Technical Summary

Problems solved by technology

[0004] At this stage, major map manufacturers use remote sensing data and mobile GPS data to update the traffic road network of large and medium-sized cities, but the update frequency of the road network at the county and town levels is slow, and the update frequency even reaches 3-5 years

Obtaining road trajectory data through GPS data requires massive basic data as support, but obtaining massive trajectory data in areas with low traffic density is usually time-consuming and expensive. High-level trajectory data often leads to large errors in trajectory data extraction, which cannot meet the needs of practical applications.

[0005] In the existing technology, although many adjustments have been made in the use characteristics and methods of the road network update technology, in essence, the road is extracted by manually giving the computer some knowledge of road color features, texture features, and shape features in advance. The extraction method is manual. The subjective initiative is relatively large, the road feature information is easy to understand, and the extraction effect can meet the needs of practical applications, but the extraction accuracy has reached the bottleneck period and it is difficult to improve it to a large extent

Images