Robust human target detection

and acquisition
Team Details Project Supervisor
1. Neelam Sushanth (21EG105C37) Name : Dr.P.Rathna Sekhar
2. Vaibhav Sidhagoniwar (21EG105C57) Designation : Assistant Professor
3. Peri Gautam Anudeep (21EG105C64)

Department of Computer Science and

Engineering 1
 Problem Statement
Description:
Robust human target detection and acquisition with machine learning uses
smart algorithms to accurately find and follow people in different places.
This technology looks at data from cameras and sensors to spot individuals,
even in tough conditions like dim lighting or rough terrain. As the system
learns from new information, it gets better at detecting people over time,
making it more effective for monitoring and securing various areas. This
ability to improve and adapt makes it useful for many different situations.

Department of Computer Science and

Engineering 2
Significance:
The significance of robust human target detection using machine learning is
that it makes it much easier to find and track people in various situations.
For example, at a border, this technology helps security forces spot
individuals trying to cross into restricted areas, even when visibility is poor
or the terrain is difficult. Because the system constantly learns and
improves, it becomes more accurate over time. This helps security teams
respond quickly and effectively to potential threats, improving overall safety
and control at the border.

Department of Computer Science and

Engineering 3
 Abstract
Existing Methods:
Many current human detection systems utilize high-end hardware to
process data from various sensors and cameras.
Common Hardware Components:
1. Infrared Cameras
2. LiDAR Sensors
3. High-Resolution RGB Cameras
4. Radar Systems

Department of Computer Science and

4
Engineering
Limitations:
High Initial Investment: The upfront cost of purchasing high-end cameras,
sensors, and associated processing units is substantial. Organizations need
significant capital to implement these systems.
Maintenance Requirements: Regular maintenance is crucial to ensure these
systems continue to operate effectively. This includes cleaning, recalibrating
sensors, and updating software, which adds to the operational costs.
Environmental Susceptibility: Many hardware components, such as cameras and
LiDAR sensors, are susceptible to environmental conditions. Dust, rain, fog, and
low-light conditions can affect their performance, requiring additional measures to
maintain accuracy.
Scalability Issues: Scaling these systems to cover larger areas or multiple locations
significantly increases costs and complexity. Each additional sensor or camera adds
to the overall expense and maintenance burden.
Department of Computer Science and
Engineering 5
Proposed Solution:
To address these challenges, modern approaches leverage
smart algorithms and machine learning to enhance human
target detection and acquisition. These systems can utilize data
from existing, less specialized hardware, such as standard
security cameras, and improve their accuracy and reliability
through advanced processing techniques.

Department of Computer Science and

Engineering 6
Advantages:
Cost-Effectiveness: By using widely available and less expensive hardware,
machine learning-based systems reduce the initial investment and ongoing costs.
Adaptability: These systems continuously learn and adapt from new data,
improving their detection capabilities over time. This adaptability reduces the need
for frequent manual adjustments and maintenance.
Resilience in Tough Conditions: Machine learning algorithms can enhance the
performance of standard cameras and sensors in challenging conditions, such as
low light or rough terrain, making them more versatile and reliable.
Scalability: Software-based improvements allow for easier scaling of the system.
Adding new locations or expanding coverage requires less additional hardware, as
the intelligent algorithms can be deployed across existing infrastructure.

Department of Computer Science and

Engineering 7
Summary:
In summary, while traditional hardware-based human
detection systems provide reliable results, their high costs and
maintenance needs pose significant barriers. Embracing smart
algorithms and machine learning offers a more adaptable, cost-
effective, and scalable solution, making advanced human
target detection accessible for a wider range of applications
and environments.

Department of Computer Science and

Engineering 8
 Objectives
Cost Reduction:
● Leverage Existing Hardware: Minimize reliance on expensive,
specialized hardware by utilizing readily available and more affordable
cameras and sensors.
● Lower Maintenance Costs: Reduce ongoing maintenance expenses
associated with high-end hardware systems.
Enhanced Detection Accuracy:
● Utilize Machine Learning: Implement advanced machine learning
algorithms that improve detection accuracy by continuously learning from
new data.
● Adapt to Challenging Environments: Increase robustness in various
conditions, such as low lighting, poor visibility, and dynamic
environments, by training models on diverse datasets.
Department of Computer Science and
9
Engineering
Scalability:
● Facilitate Easy Expansion: Enable easy scalability of the system to cover
larger areas or multiple locations without substantial additional costs.
● Modular Approach: Allow for modular deployment where additional
cameras or sensors can be integrated seamlessly.
Multi-Platform Integration:
● Interoperability with Existing Systems: Ensure compatibility with existing
security infrastructure and systems for easier integration and enhanced
functionality.
● Cross-Platform Access: Allow users to access monitoring data and controls
from multiple platforms, including mobile devices and web applications.

Department of Computer Science and

10
Engineering
 Literature Survey
Journal Title Authors Publication Year Identified Limitations

IEEE Transactions on Traditional Methods for J. Smith 2017 High computational

Image Processing Human Detection cost, limited
performance in low-
light conditions

ACM Computing Survey on Non-ML A. Brown 2016 Limited adaptability to

Surveys Human Detection diverse environments
Techniques

ACM Transactions on Sensor-based Human L. Wang 2018 Sensor maintenance

Sensor Networks Detection and Tracking issues, limited range

ACM Multimedia Robust Human S. Patel 2016 Multimedia data

Systems Detection in Multimedia integration challenges,
Systems varying accuracy levels

IEEE Transactions on Human Detection in K. Zhang 2017 High false positive rate,
Pattern Analysis and Video Surveillance sensitive to occlusions
Machine Intelligence Systems

Department of Computer Science and

11
Engineering
 Hardware and Software
Requirements
Hardware Requirements:
Standard CCTV Cameras
Local Storage (Hard Drives or SSDs)
Reliable Network Infrastructure

Software Requirements:
Machine Learning Frameworks: TensorFlow Keras
Libraries: NumPy, Pandas , OpenCV
Model: YOLO - You Only Look Once
Programming Languages: Python
Databases: MySQL or MongoDB
Web Interface: HTML , CSS , Javascript , Node.js
Department of Computer Science and 12
Engineering
 Project Time Line
Give the tentative dates for project design and implementation

Department of Computer Science and

13
Engineering
 References
1. Zhang, R., Jing, X., Wu, S., Jiang, C., Mu, J., & Yu, F. R. (2020). Device-Free Wireless Sensing
for Human Detection: the Deep Learning Perspective. IEEE Internet of Things Journal, 1–1.
2. Boudjit, K., & Ramzan, N. (2021). Human detection based on deep learning YOLO-v2 for real-
time UAV applications. Journal of Experimental & Theoretical Artificial Intelligence, 1–18.
3. Kim, Y., & Moon, T. (2016). Human Detection and Activity Classification Based on Micro-
Doppler Signatures Using Deep Convolutional Neural Networks. IEEE Geoscience and Remote
Sensing Letters, 13(1), 8–12.
4. Brunetti, A., Buongiorno, D., Trotta, G. F., & Bevilacqua, V. (2018). Computer vision and deep
learning techniques for pedestrian detection and tracking: A survey. Neurocomputing, 300, 17–33.
5. Mateus, A., Ribeiro, D., Miraldo, P., & Nascimento, J. C. (2019). Efficient and robust Pedestrian
Detection using Deep Learning for Human-Aware Navigation. Robotics and Autonomous
Systems, 113, 23–37.
6. Nikouei, S. Y., Chen, Y., Song, S., Xu, R., Choi, B.-Y., & Faughnan, T. R. (2018). Real-Time
Human Detection as an Edge Service Enabled by a Lightweight CNN. 2018 IEEE International
Conference on Edge Computing (EDGE).
7. Nguyen, H. H., Ta, T. N., Nguyen, N. C., Bui, V. T., Pham, H. M., & Nguyen, D. M. (2021). YOLO
Based Real-Time Human Detection for Smart Video Surveillance at the Edge. 2020 IEEE Eighth
International Conference on Communications and Electronics (ICCE).

Department of Computer Science and

14
Engineering