Automatic acquisition of 3D spatial data in city with air-borne TLS Three Line Scanner
3.Laser Scanner
Data of laser scanner is obtained in a synchronized manner with the TLS. Resolution of data is about 50cm. Fig2 shows data of laser scanner. Location is Roppongi, Tokyo in Japan Ground resolution of the laser scanner data is approximately 50 cm, which is much lower than that of the TLS, though 3D shape
of urban objects can be directly extracted from the laser data. But we can know some objects somewhere from data, then they can be used for auxiliary data.
Fig2 : Data of laser scanner seen with bird's eye view
4.Theory
4.1.Posiion and angle data in each line
Flow chart of this study is shown Fig3. First of all, we have to generate accurate data of position and angle in each line. For that, we conducted relative orientation with some ground control points and GPS Global Positioning System /INS Inertial Navigation System data as auxiliary data. From GPS and INS, position and angle of CCD is obtained directly, but the number of measurement per second are lower than the data acquisition frequency of TLS 500Hz , especially GPS data can be obtained only 5 times per second. And data's reliability is not so high, so we use them as auxiliary data.
Fig3 : Flow chart of this study
4.2.Automatic extraction of buildings and stereomatching
Fig 4 shows TLS image obtained by a test flight without the stabilizer. Distortions coming from attitude changes of the aircraft prevents the stereo viewing and automatic image matching directly using the raw data. So, we have to rectify these images. For the rectification a horizontal plane located at a mean terrain height is defined. And each pixel of the image is projected on the plane to rectify the distortions.
Fig4 : Raw data of TLS
Fig5 is the rectified image. And then, using the rectified image, buildings are extracted. At first, edges of buildings are extracted from image. With application of wavelet transform, image of edge eliminating noise can be obtained. Then, segmentation is conducted by extracting closed areas. At last, correspondence of an identical object among the three different images are determined in stereo matching process. But there are problems when using only images. For example,
