Citation: Shi, K.; Yan, J.; Yang, J. A
Semantic Partition Algorithm Based
on Improved K-Means Clustering for
Large-Scale Indoor Areas. ISPRS Int. J.
Geo-Inf. 2024, 13, 41. https://doi.org/
10.3390/ijgi13020041
Academic Editors: Wolfgang Kainz
and Hartwig H. Hochmair
Received: 11 November 2023
Revised: 19 January 2024
Accepted: 22 January 2024
Published: 27 January 2024
Copyright: © 2024 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
International Journal of
Geo-Information
Article
A Semantic Partition Algorithm Based on Improved K-Means
Clustering for Large-Scale Indoor Areas
Kegong Shi, Jinjin Yan * and Jinquan Yang
Qingdao Innovation and Development Center, Harbin Engineering University, Qingdao 266500, China;
kegong.shi@hrbeu.edu.cn (K.S.); yjq980314@hrbeu.edu.cn (J.Y.)
* Correspondence: jinjin.yan@hrbeu.edu.cn
Abstract: Reasonable semantic partition of indoor areas can improve space utilization, optimize
property management, and enhance safety and convenience. Existing algorithms for such partitions
have drawbacks, such as the inability to consider semantics, slow convergence, and sensitivity to
outliers. These limitations make it difficult to have partition schemes that can match the real-world
observations. To obtain proper partitions, this paper proposes an improved K-means clustering
algorithm (IK-means), which differs from traditional K-means in three respects, including the distance
measurement method, iterations, and stop conditions of iteration. The first aspect considers the
semantics of the spaces, thereby enhancing the rationality of the space partition. The last two increase
the convergence speed. The proposed algorithm is validated in a large-scale indoor scene, and the
results show that it has outperformance in both accuracy and efficiency. The proposed IK-means
algorithm offers a promising solution to overcome existing limitations and advance the effectiveness
of indoor space partitioning algorithms. This research has significant implications for the semantic
area partition of large-scale and complex indoor areas, such as shopping malls and hospitals.
Keywords: area semantic partition; improved K-means; large-scale indoor areas
1. Introduction
With the rapid urbanization and expansion of indoor environments in large-scale
settings [1,2], the partition of indoor areas has become a significant concern [3]. Reasonable
partitioning plays a crucial role in improving space utilization, optimizing management,
and enhancing safety and convenience. It has wide-ranging applications in various fields
such as indoor navigation [4,5], security monitoring [6], and resource management [7,8].
A semantic-based area partition is able to bring in numerous benefits to large-scale indoor
scenes (such as shopping malls and hospitals), including improved business operations,
enhanced customer service, and increased safety, leading to better experiences. For instance,
in shopping malls, a semantic-based area partition can assist new businesses in selecting
suitable shop locations based on attributes of their products. This allows the mall to effec-
tively manage the procurement, sales, and inventory of different product categories, thereby
enhancing overall operational efficiency. In particular, such a partition can assist shopping
malls in determining suitable locations for signage installation, which can effectively help
customers quickly locate the desired products and save shopping time. Analyzing rest
areas enables the placement of flammable and explosive items in safe zones away from
those areas, mitigating potential hazards and safety risks. In hospitals, dividing areas based
on different departments (semantic categories) improves the efficiency of patient diagnosis
and treatment. Patients can more accurately determine the areas they need to visit, like
inpatient wards and diagnostic sections, which helps in reducing wait times and confusion,
thereby improving the overall patient experience.
Indoor area partition can be seen as an application of regional clustering [9]. K-means
is a widely utilized machine learning algorithm among various partitioning methods [10].
ISPRS Int. J. Geo-Inf. 2024, 13, 41. https://doi.org/10.3390/ijgi13020041 https://www.mdpi.com/journal/ijgi