| MONOCULAR VISION SLAM RESEARCH FOR PARKING
ENVIRONMENT WITH LOW LIGHT |
| Sumin Zhang , Yongshuai Zhi , Shouyi Lu , Ze Lin , Rui He |
| Jilin University |
|
|
|
|
|
|
|
| ABSTRACT |
|
The low light condition poses a challenge when utilizing direct monocular visual simultaneous localization and mapping (vSLAM) in a parking environment. To make the monocular vSLAM perform well under parking environments with low light or illumination alternations, in this paper, we propose a direct monocular vSLAM method that can handle lowlight or illumination changes by using an improved local interpretable model-agnostic explanation (LIME) algorithm. The core of our proposed method is to improve the quality of mapping and camera pose localization accuracy under low-light parking scenes and to estimate the parameters of the improved LIME algorithm. We compare the performance of the improved LIME algorithm and other low-light image processing algorithms on the photos from the high dynamic range (HDR) dataset and our self-collected dataset. In addition, we also compare the performance of our proposed monocular vSLAM method with that of the vSLAM pre-processed by the simultaneous reflection and illumination estimation (SRIE) image processing algorithm. The experimental results show that the proposed method performs better in environment mapping and camera pose localization under parking environments with low light and significant illumination changes than other methods. |
| Key Words:
Parking environment, Low light, Monocular visual simultaneous localization and mapping (vSLAM), Environment mapping |
|
|
|
Corresponding Author.
Rui He , Email. 
Preview
Full text via DOI
Download Citation 
Print 
