To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Dense Feature -based Landmark Identification for Mobile Platform Localization |
|
Author |
Ba-Viet Ngo, Thanh-Hai Nguyen |
| Citation |
Vol. 18 No. 12 pp. 186-200 |
|
Abstract |
Recognition of natural landmarks for mobile platform localization has attracted researchers in many recent years. Feature extraction plays a key role in the recognition of these natural landmarks based on feature density. This paper shows the recognition of natural landmarks in indoor environments with different objects using an RGB-D camera system installed with a mobile platform. A FAST detector is applied to extract feature information of all objects in an RGB image to produce many binary feature frames corresponding to the objects in the RGB image. The proposed method in this paper is that one of the feature frames containing the largest feature density is identified to be a natural landmark. For identifying one feature frame of one object to be one natural landmark, dilation and contour will be employed for determining all feature frames of objects in an image. After finding the feature frames, a maximum density of key points in each feature frame is calculated in order to identify the best landmark. For evaluating the accuracy of landmark identification, a SURF method is employed to recognize the identified landmark and the landmark recognition is evaluated using a proposed method based on the matching rate of feature points between the accrual landmark and the identified landmark. The experimental results show that the proposed method for identifying reliable natural landmarks without a priori information is effective for mobile platform localization in indoor natural environments. |
|
Keywords |
Natural landmark identification, Feature detection using FAST, dilation for feature connection, Maximum feature density. |
|
URL |
