Data Association and Localization of Classified Objects in Visual SLAM
- PDF / 2,767,487 Bytes
- 18 Pages / 595.224 x 790.955 pts Page_size
- 8 Downloads / 233 Views
Data Association and Localization of Classified Objects in Visual SLAM Asif Iqbal1
· Nicholas R. Gans1
Received: 9 September 2019 / Accepted: 20 March 2020 © Springer Nature B.V. 2020
Abstract Maps generated by many visual Simultaneous Localization and Mapping algorithms consist of geometric primitives such as points, lines or planes. These maps offer a topographic representation of the environment, but they contain no semantic information about the environments. Object classifiers leveraging advances in machine learning are highly accurate and reliable, capable of detecting and classifying thousands of objects. Classifiers can be incorporated into a SLAM pipeline to add semantic information to a scene. Frequently, this semantic information is conducted for each frame of the image, but semantic labeling is not persistent over time. In this work, we present a nonparametric statistical approach to perform matching/association of objects detected over consecutive image frames. These associated classified objects are then localized in the accrued map using an unsupervised clustering method. We test our approach on multiple data sets, and it shows strong performance in terms of objects correctly associated from frame to frame. We also have tested our algorithm on three data sets in our lab environment using tag markers to demonstrate the accuracy of classified object localization process. Keywords Localization and mapping · Data association · Semantic mapping · Unsupervised learning
1 Introduction The capability of a robot to localize itself as it moves through an unknown environment while building a distinct map of the surroundings is essential for full autonomy. Known as Simultaneous Localization and Mapping (SLAM), this problem has been an important research topic with numerous notable contributions. Excellent overviews with extensive references were provided by Thrun [44], Durrant-Whyte and Bailey [15], and others. Vision-based SLAM (VSLAM) uses cameras as sensors (e.g., [17, 20]). An important area of current research is adding semantic information in the environment map, which can be used for navigation, localization, retrieval, etc.
This work was supported by the Advanced Driver Assistance System (ADAS) group at Texas Instruments (TI) in Dallas, TX Nicholas R. Gans
[email protected] Asif Iqbal [email protected] 1
University of Texas at Arlington Research Institute, University of Texas at Arlington, Fort Worth, TX, 76118, USA
Most SLAM algorithms generate a map consisting of estimated geometric features such as points, lines or planes. These maps do not include semantic meaning or information [7, 13]. We seek to add semantic information to maps in the form of labeled, persistent objects present in the environment, such as furniture, office equipment, kitchen items, etc. It is not sufficient for objects to be detected and classified in an image, but objects must be matched/associated with the correct objects in previous images over time. This remains a difficult problem in SLAM and object classification.
Data Loading...
