Graph-Based Non-Linear Least Squares Optimization for Visual Place Recognition in Changing Environments

Schubert, Stefan; Neubert, Peer; Protzel, Peter

doi:10.1109/lra.2021.3052446

Cited by 8 publications

(31 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A subsequent optimization of the graph leads to significantly improved place recognition results. The experimental evaluation in [32] demonstrated a very good place recognition performance of the graph-based approach, which significantly outperformed multiple sequence-based methods from the literature.…”

Section: Introductionmentioning

confidence: 85%

“…However, most methods focus on only one type of additional knowledge, even if more information would be available or use them successively rather than in combination. To address this problem, we proposed in our recent work [32] a graph-based framework for place recognition that fuses additional structural knowledge from different sources in a single graph. A subsequent optimization of the graph leads to significantly improved place recognition results.…”

Section: Introductionmentioning

confidence: 99%

“…A drawback of our graph-based method from [32] is the relatively long runtime and memory inefficiency: In our experiments, we measured a maximum runtime of 5.3 sec per query image and had to use partial optimizations of the graph as in [32] due to memory limitations. The reason for the long runtime and memory inefficiency is the used non-linear least squares optimization (NLSQ) of the graph.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, we adopt our graph-based approach from [32] with its superior place recognition performance. We contribute…”

Section: Introductionmentioning

confidence: 99%

“…• the sequence-based method SeqConv which is used during optimization with ICM (Sec. VII) • an extensive experimental evaluation of our approach with comparisons to [32] and six sequence-based methods from the literature on four different descriptors and 21 datasets (Sec. VIII) Further, we provide a short introduction to the graph-based approach from [32] (Sec.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Fast and Memory Efficient Graph Optimization via ICM for Visual Place Recognition

Schubert

Neubert

Protzel

2021

Robotics: Science and Systems XVII

Self Cite

View full text Add to dashboard Cite

Visual place recognition is the task of finding same places in a set of database images for a given set of query images. This task becomes particularly challenging if the environmental condition changes between database and query, for example from day to night. In this paper, we build upon our recent work on graph optimization for place recognition, where a graph was used to model additional structural knowledge like sequences. A subsequent non-linear least squares optimization (NLSQ) improved the place recognition performance. While this approach achieves very high performance, it is quite slow and memory inefficient. This paper addresses the long runtime and the high memory usage in order to obtain the same or better place recognition performance faster on larger problems. We propose a novel graph optimization procedure that is based on Iterated Conditional Modes (ICM). In addition, we investigate a new cost function for an edge in the graph. Our novel ICMbased approach achieves 9.1msec maximum runtime per query, which is 260× faster than the minimum runtime with NLSQ. Moreover, with ICM we can optimize problems that are not feasible with NLSQ on a full graph due to memory limitations. To demonstrate the superior performance of our ICM-based method, we provide extensive experimental evaluations with the essence of 987 precision-recall curves: Our proposed ICM-based method is compared to the NLSQ-based method as well as to six sequence-based approaches from the literature on 21 sequence combinations from five datasets with four different image descriptors. Our experiments show that our ICM-based method with sequence-exploitation not only improves the NLSQbased performance by 10% on average while being 385× faster and using more than 60× less memory. It also significantly outperforms all six sequence-based methods from the literature by at least 32% on average with the NetVLAD descriptor while using comparable runtime and memory. Code is available online 1 .

show abstract

Section: Introductionmentioning

confidence: 85%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…In this paper, we adopt our graph-based approach from [32] with its superior place recognition performance. We contribute…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Fast and Memory Efficient Graph Optimization via ICM for Visual Place Recognition

Schubert

Neubert

Protzel

2021

Robotics: Science and Systems XVII

Self Cite

View full text Add to dashboard Cite

show abstract

Optimal Densely Connected Networks with Pyramid Spatial Matching Scheme for Visual Place Recognition

Sasikumar¹,

Sathiamoorthy²

2022

Pervasive Computing and Social Networking

View full text Add to dashboard Cite

Open-source optimization method for android smartphone single point positioning

et al. 2022

View full text Add to dashboard Cite

Nowadays, a chip-scale Global Navigation Satellite System (GNSS) receiver is ubiquitous in smartphones. In a smartphone GNSS receiver, the least square (LS) or Kalman Filter (KF) is implemented to estimate the position. With the aim to improve the smartphone GNSS position accuracy, we propose a position-smoothing method considering more historical information than the traditional methods, i.e., LS and KF. More past states are regarded as unknowns, and a cost function is constructed to optimize these states. An open-source smartphone dataset from Google was used for testing the proposed method. The experimental results indicate that the proposed method outperforms the other conventional methods in position errors. In addition, we open the source code. We expect that the optimization method implemented in the smartphone GNSS position smoothing application can be an illustrative example to clearly introduce such an optimization method and a reference for its implementation, which might inspire some other meaningful and exciting applications in GNSS.

show abstract

Graph-Based Non-Linear Least Squares Optimization for Visual Place Recognition in Changing Environments

Cited by 8 publications

References 18 publications

Fast and Memory Efficient Graph Optimization via ICM for Visual Place Recognition

Fast and Memory Efficient Graph Optimization via ICM for Visual Place Recognition

Optimal Densely Connected Networks with Pyramid Spatial Matching Scheme for Visual Place Recognition

Open-source optimization method for android smartphone single point positioning

Contact Info

Product

Resources

About