FuseDepth

Implementation of our paper entitiled LiDAR-ToF-Binocular depth fusion using gradient priors published in CCDC.

1. Get Started

a. Test environment:

Ubuntu 16.04
OpenCV 3.4.3

b. Install:

mkdir build
cd build
cmake ..
make

2. Test

To run a demo, do:

./Fusion image_folder_name

image_folder_name the name of an image, or a video frame from a sequence. The folder should include the corresponding RGB image, depth from a 16-wire lidar, depth from a stereo/binocular, and depth from a ToF. The depth data are all saved as 16-bit images.

A few samples are shown below. They are RGB image, lidar depth, stereo depth, and ToF depth from left to right.

      

3. Results

The algorithm fuses lidar, stereo, and ToF depth via gradient priors. It takes about 10-20ms to compute a video frame with a resolution of 640x480. The result is shown below.