Pothole Detection Based on Disparity Transformation and Road Surface Modeling

In this project, we developed a robust road pothole detection algorithm [1] based on novel disparity transformation [2] and road surface modeling. Firstly, we utilized our previously published disparity estimation algorithm named PT-SRP to acquire road disparity maps and reconstruct 3D road geometry. The original disparity maps are transformed to better distinguish between damaged and undamaged road areas. Then, the disparities in the undamaged road areas are modeled by a quadratic surface. By comparing the difference between the actual and modeled disparity maps, the potholes can be detected effectively. We created three datasets including 67 pairs of stereo road images using a ZED stereo camera. Please kindly cite our paper when using our datasets in your research. 

[1] Fan, R., Ozgunalp, U., Hosking, B., Liu, M., and Pitas, I., 2020. Pothole Detection Based on Disparity Transformation and Road Surface Modeling. IEEE Transactions on Image Processing, 29(1), pp.897-908. [paper][arxiv]


[2] Fan, R., Bocus, M. J., and Dahnoun, N., 2018. A novel disparity transformation algorithm for road segmentation. Information Processing Letters140, pp.18-24. [paper][arxiv]


Real-Time Dense Stereo Embedded in A UAV for Road Inspection

In this project, we designed an efficient and accurate dense stereo vision system and embedded it in a DJI Matrice 100 drone for road inspection [1]. We mounted a ZED stereo camera on the drone to capture stereo road images. These images were then processed on an NVIDIA Jetson TX2 GPU. We created three datasets including 11368 pairs of stereo images. Please kindly cite our paper when using our datasets in your research.


[1] Fan, R., Jiao, J., Pan, J., Huang, H., Shen, S. and Liu, M., 2019. Real-Time Dense Stereo Embedded in A UAV for Road Inspection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops[paper][arxiv][video]


News: this road surface 3D reconstruction system has been reported by over ten international media agencies, such as VentureBeat, Diamandis, Drobots Company, UAS Vision, Import AI, Impact Lab, US Breaking News, PCNewsBuzz and Engineering 360. (07/07/2019)



Multiple Lane Detection Algorithm Based on Novel Dense Vanishing Point Estimation

In this project, we designed two robust multiple lane detection algorithms [1], [2] based on dense vanishing point estimation. The vanishing points were estimated from disparity maps using dynamic programming [3]. The datasets we used are available on the KITTI website. We will be happy if you cite us:

[1] Ozgunalp, U., Fan, R., Ai, X. and Dahnoun, N., 2016. Multiple lane detection algorithm based on novel dense vanishing point estimation. IEEE Transactions on Intelligent Transportation Systems18(3), pp.621-632. [paper][video]

[2] Fan, R. and Dahnoun, N., 2018. Real-time stereo vision-based lane detection system. Measurement Science and Technology29(7), p.074005. [arxiv][paper][video]

[3] Jiao, J., Fan, R., Ma, H., Liu, M., 2019 Using DP Towards A Shortest Path Problem-Related Application, International Conference on Robotics and Automation (ICRA), May 20-24,  Montreal, Canada. [arxiv][video]


Road Surface 3D Reconstruction Based on Dense Subpixel Disparity Map Estimation

In this project, we proposed a dense subpixel disparity map estimation algorithm for road surface 3D reconstruction [1]. The sensor we used is a ZED stereo camera from StereoLabs. We created three datasets including 91 pairs of stereo road images. The videos recording our experimental results can be found on my YouTube playlists. Please kindly cite our paper when you use our datasets in your research. 

[1] Fan, R., Ai, X. and Dahnoun, N., 2018. Road surface 3D reconstruction based on dense subpixel disparity map estimation. IEEE Transactions on Image Processing27(6), pp.3025-3035. [paper][arxiv][video]