Object Detection For Indoor Localization System
Object Detection For Indoor Localization System
Volume 9, Issue 9, September – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1488
Object Detection for Indoor Localization System
Ravi Bagade Kavita Killiketar
Dept. of Electronics and Communcation Engg., Dept. of Electronics and Communcation Engg.,
JCET, Hubbali KLEIT, Hubballi
Abstract:- Indoor localization systems have gained obstacles that are very close to the person. This brings up the
significant attention in recent years due to their issue of safety of the blind and visually impaired.
applications in various fields such as smart homes, retail Independent mobility also is matter of concern. An object
environments, and healthcare facilities. This paper detection system is the need of the hour to combat the safety
presents an innovative approach to indoor localization issues and also enable them to have mobility that is not
through the integration of object detection techniques, dependent on anyone. The object detection system has to be
aiming to enhance accuracy and efficiency in identifying a simple and most importantly user friendly. It also should be
and locating objects within indoor spaces. We explore the budget friendly so that it can be afforded by them. We are
use of advanced deep learning algorithms, particularly fortunate to have technologies that can make a huge
convolutional neural networks (CNNs), to detect and difference in the life of the blind and visually impaired. Some
classify objects in real-time. Our methodology involves advanced technologies like computer vision, object detection
collecting a comprehensive dataset of indoor and object identification are available in today’s world. The
environments, training a robust object detection model, camera of a smart phone can be used as the vision apparatus
and implementing it in a localization framework that which helps in the object detection system[2].
utilizes both visual and spatial data. The experimental
results demonstrate that our proposed system achieves Indoor localization or indoor positioning system is a
high detection accuracy and reduced localization errors, known as a process of detecting position of any object or
outperforming traditional methods. Furthermore, we people inside a building or room by different sensory data
discuss the potential of leveraging object recognition to collected from different devices using different techniques
improve user experience and navigation in complex such as radio waves, magnetic fields, acoustic signals or other
indoor settings. This research contributes to the evolving procedures. However, lacking of a standard localization
field of indoor localization and offers a foundation for system is still a very big concern. Solution of this issue can
future developments in intelligent indoor navigation be very beneficial for people in many cases but it can be
systems. especially very beneficial for the visually impaired people
[3].
Keywords:- Indoor Localization, Object Detection,
Convolutional Neural Networks (CNNs). II. RELATED WORKS
I. INTRODUCTION Joseph Redmon in [4], presents a groundbreaking
method that significantly speeds up object detection by
The advent of Artificial Intelligence (AI) and Machine unifying it into a single neural network model. While it
Learning (ML) technologies has ushered in a new era of introduces some trade-offs in accuracy, especially with
possibilities for enhancing the lives of individuals with visual smaller objects, its real-time capabilities and simplified
impairments. The development and refinement of assistive architecture have made it a foundational work in the field of
devices powered by sophisticated AI algorithms, particularly computer vision and object detection.
in the domain of object detection and identification, have
shown immense potential in bridging the gap between the Mark Sandler in [5], introduces the core design
capabilities of the visually impaired and the demands of innovations inverted residuals and linear bottlenecks. It
navigating through a visually oriented world. These focuses on the architecture, explaining how it achieves
technologies leverage the power of Convolutional Neural efficiency in terms of computation and memory usage by
Networks (CNNs) and innovative models like You Only using specific building blocks, whereas this paper uses a
Look Once (YOLO) to provide real time information about pre-trained MobileNetV2 model (pre-trained on
the surrounding environment, thereby empowering users with ImageNet) via TensorFlow's keras.applications module.
crucial data to make informed decisions and interact more This model is already trained and optimized.
freely and safely with their surroundings[1].
The challenges that the completely blind and visually
impaired people face directly impact the quality of their lives
and also how do they perform their daily activities. The
traditional method is that of a supportive guide. A person is
needed who can efficiently guide them in travelling from one
place to another. Image processing can only help identify
IJISRT24SEP1488 www.ijisrt.com 2824
Volume 9, Issue 9, September – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1488
III. METHODOLOGY D. Define Function: Predict_Image:
To implement the object detection system, we have Input: Path to the Image File (img_path).
used the algorithm as below.
Call load_and_preprocess_image to get the preprocessed
A. Import Libraries: image array.
Use the pre-trained model to predict the class of the
Import necessary libraries: tensorflow, numpy, image.
matplotlib, and Keras modules for image processing and Decode the predictions to retrieve the top 3 predicted
MobileNetV2. classes along with their probabilities.
Print the predicted objects and their corresponding
B. Load Pre-Trained Model: probabilities.
Load the original image and display it using matplotlib.
Load the MobileNetV2 model pre-trained on ImageNet
data, including the weights. E. Main Execution:
C. Define Function: Load_And_Preprocess_Image: Specified the path to the image that has to be classified.
Call the predict_image function with the specified image
Input: Path to the Image File (img_path). path to make predictions and display the results.
Load the image from the specified path with a target size IV. RESULTS
of (224, 224) pixels.
Convert the loaded image to a NumPy array. The proposed system is capable of detecting the objects
Expand the dimensions of the image array to match the present in the surrounding environment with good speed and
input shape expected by the model (adding a batch accuracy. It detects objects effectively the indoor
dimension). environment with good accuracy.
Preprocess the image using MobileNetV2’s
preprocessing function to normalize the pixel values.
Return the preprocessed image array.
Table 1: Accuracy of Python-based Object Detection in Indoor Environment
Sl no Object Accuracy
1 wardrobe: 78.99%
2 Chair: 61.84%
3 Table: 57.44%
IJISRT24SEP1488 www.ijisrt.com 2825
Volume 9, Issue 9, September – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP1488
V. CONCLUSION
In this study, MobileNetV2 was applied for detecting
indoor objects; however, the model's performance yielded
lower than expected accuracy. The reduced accuracy could
be attributed to challenges such as limited dataset diversity,
variations in indoor lighting, and the complexity of the
environment. Future work may focus on improving accuracy
by fine-tuning the model with a more diverse dataset or
exploring advanced architectures such as EfficientNet or
ResNet.
REFERENCES
[1]. Beeta Narayan, B. Noble, and Aiswariya Binu,
“Survey paper on indoor object detection and voice
feedback system”, International Research Journal of
Engineering and Technology (IRJET), Volume: 11
Issue: 04 | Apr 2024
[2]. Hiren Kumar Thakkar and Suresh Merugu, “Object
Detection System for Visually Impaired Persons
Using Smartphone,” in ResearchGate , January 2022.
[3]. Tasnia Ashrafi Heya,Sayed Erfan Arefin,"Image
Processing Based Indoor Localization System for
Assisting Visually Impaired People",IEEE,2018.
[4]. Joseph Redmon, Santosh Divvala, “You Only Look
Once: Unified, Real-Time Object Detection”, IEEE
Conference on Computer Vision and Pattern
Recognition, 2016.
[5]. Mark Sandler, Andrew Howard, “MobileNetV2:
Inverted Residuals and Linear Bottlenecks”,
Conference on Computer Vision and Pattern
Recognition, 2018.
