3-D Laser Scanner and 3-D Mapping

We have developed a method of 3-D mapping by using a tracked vehicle with 3-D laser scanner. We proposed a reticulate scan method for measuring dense 3-D shape. 3-D shapes are measured at several locations as point cloud data by using the reticulate scan method during exploration and 3-D map is built by matching these 3-D point cloud data with each other. We have evaluated and improved the 3-D mapping method with Japanese firefighters.

Small-size 3-D Laser Scanner

We had developed a small-size and light-weight 3-D laser scanner named TK-Scanner. We proposed a reticulate scan method which can measure dense and wide view angled shape around the scanner. TK-Scanner consists of a 2-D LIDAR and a pan-tilt mechanical base.

3-D Mapping in Dynamic Environment

Reticulate scan method measures distance two times at same direction during one scan. Moving objects can be detected by comparing measured distances with each other. Therefore, 3-D map without moving objects can be built in dynamic environment.

Scan Matching with Gravity Constrain

6DOF scan matching sometime fall into local minima. We want to avoid falling local minima and to accomplish robust matching. Gravity constraint can increase the robustness of the scan matching. We confirmed the validity of gravity constrain by using point cloud data measured at Sendai subway station.

Dense 3D Shape Reconstruction by Fusion of LIDAR and Camera Data

3-D laser scanner measures 3-D shape as point cloud data. The resolution of point cloud data decreases and becomes insufficient as distance become far apart. We proposed a method of dense 3-D shape reconstruction by fusion of point cloud data and camera data. It is similar to voxel curving technique. Figure a) and b) on the left shows the target environment and a result of dense 3-D shape reconstruction respectively. Detail of orange bridge surface was recovered by using our proposed method.

Real-time 3-D Mapping Using 3-D Range Camera

Real-time 3-D mapping is one of the hot research topics because range cameras (like KINECT) can measure 3-D shape in real-time. We developed a fast and robust 3-D shape matching algorithm. Figure shows a 3-D mapping result. Camera’s trajectory and 3-D map were simultaneously built in real-time.