6533b825fe1ef96bd1281aa0
RESEARCH PRODUCT
Vehicle localization based on scene perception
Linrunjia Liusubject
Feature representationDynamic objects removal[SPI.OTHER] Engineering Sciences [physics]/OtherImage enhancementSuppression d'objets dynamiquesPlace recognitionImage blurringFlou d'imagesAmélioration d'imagesReconnaissance de lieuxReprésentation d'entitésdescription
The vision-based vehicle localization task is tackled as an image-based place recognition problem in this thesis. As image representation is an important process to place recognition, we proposed place recognition methods focusing on increasing the quality of image features. First, the dynamic object removal step is proposed to remove the dynamic objects, such as vehicles and pedestrians, of an image by semantic segmentation method and restore their background information by image inpainting method. Second, reducing instead of removing the noisy information of an image is proposed. Using the image blurring method to reduce the image noise, the performance of place recognition is improved with less time consuming compared to the dynamic objects removal step. Third, the feature similarity achievement method is proposed for the day and night place recognition problem. Blurring the daytime images and enhancing the nighttime images to achieve a balance in their feature quality is the key to improve the place recognition performance. The images generated by the proposed methods are evaluated in the same place recognition step and are proved their effectiveness by compared with other advanced place recognition methods.
| year | journal | country | edition | language |
|---|---|---|---|---|
| 2021-01-01 |