Technology

The company independently developed a both effect and performance optimized semi-global matching (SGM) algorithm, a multilevel triangle-based ground filtering algorithm, a block-regression (BR) based pansharpening algorithm, and a series of parallel techniques with different processing levels. Cutting-edge techniques adopted in MASI software:

(1) Dense stereo matching

VisionOnSky Co., Ltd. independently developed a both effect and performance optimized semi-global matching (SGM) algorithm which is applied to dense stereo matching of aerial image (including oblique aerial image), ADS (ADS40, ADS80, ADS100) images, satellite image and UAV image. It generates highly dense point clouds and digital surface model (DSM) through finding the point corresponding for each pixel.

(2) Ground filtering

A multilevel triangle-based ground filtering algorithm is presented, which can transform DSM to DTM (Digital Terrain Model) automatically, i.e., removing the height of building and tree, only remaining the height of pure ground. The algorithm can select different level of terrain details in the light of the requirements. The DSM transformed to DTM can be extracted from satellite image, aerial image, ADS image, UAV image as well as LiDAR points.

(3) Pansharpening

We presented a block-regression (BR) based pansharpening algorithm. The algorithm can combine a lower resolution multispectral image and a higher resolution panchromatic image both covering a same area, and produce a new higher resolution image, in which not only the color content are remained but also the spatial details are enhanced by adopting the higher resolution panchromatic image. BR algorithm can achieve a maximum enhancement of its spatial details at the expense of a minimum spectral distortion for the multispectral images. In the algorithm, the bands of the multispectral image used can not be limited and the algorithm is specially optimized for the pansharpening of very high resolution satellite image.

(4) Parallel processing techniques

We also developed a series of parallel techniques with different processing levels, which are suitable for different steps of the whole flowchart. Not only concurrently processing of multiple tasks but also parallel processing in a single task is supported. The series of parallel techniques can be used either on multi-core computer or on cluster computer.

Technical articles (the following items can be accessed by clicking):

The usage of commands provided by Windows OS in the command operating manner for MASI software, 03/2018