<abstract xmlns="http://www.w3.org/1999/xhtml"><p>Depth data is an effective tool to locate the intelligent agent in space because it accurately records the 3D geometry information on the surface of the scanned object, and is not affected by factors like shadow and light. However, if there are many planes in the work scene, it is difficult to identify objects and process the resulting huge amount of data. In view of this problem and targeted at object calibration, this paper puts forward a depth data calibration method based on Gauss mixture model. The method converts the depth data to point cloud, filters the noise and collects samples, which effectively reduces the computational load in the following steps. Besides, the authors cluster the point cloud vector with the Gaussian mixture model, and obtain the target and background planes by using the random sampling consensus algorithm to fit the planes. The combination of target Region Of Intelligent agent (ROI) and point cloud significantly reduces the computational load and improves the computing speed. The effect and accuracy of the algorithm is verified by the test of the actual object.</p></abstract>

Depth data is an effective tool to locate the intelligent agent in space because it accurately records the 3D geometry information on the surface of the scanned object, and is not affected by factors like shadow and light. However, if there are many planes in the work scene, it is difficult to identify objects and process the resulting huge amount of data. In view of this problem and targeted at object calibration, this paper puts forward a depth data calibration method based on Gauss mixture model. The method converts the depth data to point cloud, filters the noise and collects samples, which effectively reduces the computational load in the following steps. Besides, the authors cluster the point cloud vector with the Gaussian mixture model, and obtain the target and background planes by using the random sampling consensus algorithm to fit the planes. The combination of target Region Of Intelligent agent (ROI) and point cloud significantly reduces the computational load and improves the computing speed. The effect and accuracy of the algorithm is verified by the test of the actual object.

Depth Data Reconstruction Based on Gaussian Mixture Model

College of Technology, Hubei Engineering University, Xiao Gan 432000, China, Department of Automatic Control, Northwestern Polytechnical University, Xi’an 710072,

University of Victoria, Victoria BC V8P 5C2,

Cybernetics and Information Technologies

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

{"article-title":"Depth Data Reconstruction Based on Gaussian Mixture Model"}

Depth data is an effective tool to locate the intelligent agent in space because it accurately records the 3D geometry information on the surface of the...

Depth Data Reconstruction Based on Gaussian Mixture Model

Data publikacji: 25 sty 2017

Zakres stron: 207 - 219

DOI: https://doi.org/10.1515/cait-2016-0089

Słowa kluczowe
Depth data, point cloud, normal vector clustering, Gaussian mixture model, random sampling consensus algorithm, object calibration, CAMShift

© 2016 Zhe Li et al., published by De Gruyter Open

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Depth Data Reconstruction Based on Gaussian Mixture Model

Data publikacji: 25 sty 2017

Zakres stron: 207 - 219

DOI: https://doi.org/10.1515/cait-2016-0089

Słowa kluczoweDepth data, point cloud, normal vector clustering, Gaussian mixture model, random sampling consensus algorithm, object calibration, CAMShift

© 2016 Zhe Li et al., published by De Gruyter Open

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Słowa kluczowe
Depth data, point cloud, normal vector clustering, Gaussian mixture model, random sampling consensus algorithm, object calibration, CAMShift