Road surface 3d reconstruction based on dense subpixel disparity map estimation (1807.01874v2)

Abstract: Various 3D reconstruction methods have enabled civil engineers to detect damage on a road surface. To achieve the millimetre accuracy required for road condition assessment, a disparity map with subpixel resolution needs to be used. However, none of the existing stereo matching algorithms are specially suitable for the reconstruction of the road surface. Hence in this paper, we propose a novel dense subpixel disparity estimation algorithm with high computational efficiency and robustness. This is achieved by first transforming the perspective view of the target frame into the reference view, which not only increases the accuracy of the block matching for the road surface but also improves the processing speed. The disparities are then estimated iteratively using our previously published algorithm where the search range is propagated from three estimated neighbouring disparities. Since the search range is obtained from the previous iteration, errors may occur when the propagated search range is not sufficient. Therefore, a correlation maxima verification is performed to rectify this issue, and the subpixel resolution is achieved by conducting a parabola interpolation enhancement. Furthermore, a novel disparity global refinement approach developed from the Markov Random Fields and Fast Bilateral Stereo is introduced to further improve the accuracy of the estimated disparity map, where disparities are updated iteratively by minimising the energy function that is related to their interpolated correlation polynomials. The algorithm is implemented in C language with a near real-time performance. The experimental results illustrate that the absolute error of the reconstruction varies from 0.1 mm to 3 mm.

• The paper presents a novel method for 3D road surface reconstruction using dense subpixel disparity estimation, enhancing accuracy through perspective transformation and iterative refinement.
• Achieving millimetre accuracy (0.1-3 mm errors), this approach is viable for detecting road damage like potholes and has potential for broader infrastructure monitoring.
• The methodology demonstrates near real-time capability with future optimization potential for dynamic systems like autonomous vehicles and improved road surface SLAM applications.

Road Surface 3D Reconstruction via Dense Subpixel Disparity Estimation

The paper presents a novel approach to 3D reconstruction of road surfaces which is crucial for detecting and assessing road damage such as potholes. The paper focuses on achieving millimetre accuracy using a dense subpixel disparity estimation algorithm that exhibits both computational efficiency and robustness. This method addresses the limitations in existing stereo matching algorithms which are not particularly tailored for road surface reconstruction.

Methodology

The primary novelty of the algorithm lies in the dense subpixel disparity estimation method. The process begins with a perspective transformation that aligns the target frame’s view to match the reference frame. This step increases the block matching accuracy and enhances processing speed. Disparities are iteratively estimated using a previous proprietary algorithm that propagates the search range from three neighboring disparities. Correlation maxima verification is employed to rectify errors caused by an insufficient propagated search range, ensuring that each disparity achieves subpixel resolution through parabola interpolation enhancement.

To further refine the accuracy, a novel disparity refinement approach integrates Markov Random Fields (MRFs) and Fast Bilateral Stereo (FBS). Disparities are updated iteratively by minimizing an energy function related to their interpolated correlation polynomials. This iterative local stereo matching substantially enhances the trade-off between speed and precision, providing a feasible methodology for real-time road surface assessment.

Results and Implications

The experimental results demonstrate the method's efficacy with absolute errors in the reconstruction process ranging from 0.1 mm to 3 mm, thus meeting the millimetre accuracy target required for road condition assessment. Such precision is beneficial not only for immediate road damage detection but also holds potential for broader infrastructure monitoring applications, including bridges and tunnels.

The implementation was executed in C language, showing promising near real-time capabilities. Despite not achieving actual real-time processing, there is potential for optimization using parallel computing architectures, which could extend the algorithm’s applicability to dynamic monitoring systems such as intelligent transportation systems and autonomous vehicle navigation.

Future Directions

The paper opens avenues for improving road surface SLAM (Simultaneous Localization and Mapping) applications, potentially transforming approaches in smart city data assessments. Future work could focus on enhancing parallel computation strategies for the disparity estimation process to achieve true real-time performance. Additionally, there is potential scope for integrating a self-calibration algorithm to counter stereo calibration errors, further enhancing the system's robustness.

Overall, the paper presents a significant advancement in stereo vision-based road surface reconstruction with wide-reaching implications across civil engineering and smart city applications.