Monocular Image Road Sign Update via Semantic Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for updating road signs and markings are inefficient and labor-intensive, requiring human intervention and struggling to provide accurate, real-time data for urban road maintenance and navigation, especially with the limitations of traditional field surveys and image reconstruction technologies.
Innovation Solution
A method utilizing monocular images and deep learning semantic segmentation to preprocess and correct street images, constructing a sparse 3D model, and calculating spatial positions of road signs and markings through multi-view geometry and depth values, integrating GPS/IMU data for accurate data acquisition and updating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional field acquisition or manual acquisition on remote sensing images is used to obtain road signs and markings information, then data can be acquired, but the acquisition workload is large and the update cycle is long
Solution Approach 1:
The patent replaces traditional mechanical field surveying methods with automated image processing technology. By using monocular images captured from mobile platforms and applying deep learning semantic segmentation, the system automatically extracts road signs and markings information without manual field acquisition, dramatically improving efficiency and reducing update cycles
Solution Approach 2:
The system enables self-service data acquisition by automatically processing monocular images to extract and update road sign and marking information. The automated workflow includes image capture, semantic segmentation, feature extraction, and data updating without requiring manual intervention at each step, allowing the system to continuously update itself
2Loss of information
If reconstructed three-dimensional model is generated using images or laser point clouds, then spatial information can be obtained, but the model does not have monolithic semantic information and is difficult to be used for urban road planning and management
Solution Approach 1:
The patent applies semantic segmentation to divide the monocular image into distinct semantic regions corresponding to different road elements (signs, markings, etc.). This segmentation process extracts meaningful semantic information from the image, transforming raw visual data into structured information that can be directly used for urban road planning and management
Solution Approach 2:
The patent transitions from traditional 3D spatial modeling to a 2D image-based semantic extraction approach. By working directly with monocular images and applying semantic segmentation in the image domain, the system obtains both spatial and semantic information simultaneously, avoiding the complexity of generating and annotating 3D models
3Measurement precision
If deep learning semantic segmentation is applied to preprocess and correct streetscape images, then extraction accuracy of road signs and markings is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary preprocessing and distortion correction on streetscape images before semantic segmentation. By pre-processing the images to correct geometric distortions and enhance features beforehand, the system improves the accuracy of subsequent semantic segmentation while managing computational complexity through efficient preprocessing algorithms
4Productivity
If monocular images are used for data acquisition instead of traditional field surveying, then acquisition costs and time are reduced, but the method requires sophisticated image processing algorithms
Solution Approach 1:
The patent replaces complex field surveying operations with automated monocular image processing. By capturing images from mobile platforms and using deep learning algorithms for automatic extraction, the system eliminates the need for manual field work while managing algorithmic complexity through standardized processing pipelines
Data Source
AI summary
A method for updating road signs and markings on the basis of monocular images comprises the following steps: acquiring street images of urban roads and GPS phase center coordinates and spatial attitude data corresponding to the street images; extracting coordinates of the road sign marking images; constructing a sparse three-dimensional model, and then generating a streetscape image depth map; calculating the space position of the road sign and marking according to the semantic and depth values of the image, the collinear equation and the space distance relation; if the same road sign and marking is visible in multiple views, solving the position information of the road sign; and vectorizing the obtained road sign position information, and fusing the information into the original data to realize the updating of the road sign data.