Search for Projects

1 Apr 2018

Pothole Detection, Classification and Tracking

This project mainly aims at detecting potholes from a road surface using the disparity information.

The sensor used for acquiring disparity information is a ZED stereo camera. We use our previously published algorithm to estimate dense subpixel disparity maps. For more details on the algorithm, please kindly refer to our paper published on IEEE Transactions on Image Processing. The paper is also publicly available on ResearchGate.

The datasets for pothole detection include 67 pairs of stereo images. The subpixel disparity map and 3D point cloud of each stereo pair are also provided. Some examples of pothole detection and classification results are shown below: 

The demo video of pothole tracking using DSST is shown below: 
If you have any problems with the datasets, please contact me without any hesitation.

24 Feb 2018

Multiple Lane Detection Algorithm Based on Novel Dense Vanishing Point Estimation

In this project, the main purpose of the proposed algorithm is to detect the painted lane markings on the road surface. The proposed algorithm can also be used to detect the road boundaries. 

The datasets used in this project are from KITTI, and the demo videos can be accessed on Youtube.

Please kindly refer to our papers, published on IEEE Transactions on Intelligent Transportation Systems (TITS) and Measurement Science and Technology (MST), for more details:

The TITS paper and the MST paper.

16 Jul 2017

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

This project mainly aims at reconstructing the 3D model of a road surface for pothole detection. Please kindly refer to our paper published on IEEE Transactions on Image Processing, for more details. 

The sensor we use is a ZED stereo camera from STEREOLABS. More details about the sensor are available at:

We created three datasets including 91 stereo image pairs, they are available at: datasets

The datasets include uncalibrated images, calibrated images and the calibration parameters. The datasets are used only for research, if you have any problems with the datasets, please contact me without any hesitation. 

The videos of the experimental results are also available on YouTube. You can simply access them at: videos.