Nano Rover: A Multi-Sensory Full-Functional Surveillance Robot with Modified Inception-Net

Paper ID- 292

​

Presenter:

Sheekar Banerjee

AI-ML Software Engineer, Cisscom LLC, California, USA

​

​

Co-authors:

Aminun Nahar Jhumur

Senior Lecturer, Department of Computer Science and Engineering

International University of Business Agriculture and Technology

​

Md. Ezharul Islam

Associate Professor, Department of Computer Science and Engineering, Jahangirnagar University

​

​

​

​

Contents

Surveillance Robot is one kind machinery which is utilized for the monitoring of behavior, many activities, or information for the purpose of information gathering, influencing, managing or directing.

​

Image Source: Hanson Robotics

Introduction�

Different Types of Surveillance Robots

Image Source: Boston Dynamics

• Tank-Based Military Robot (TBM)

​

• Indonesian research effort
• New features of object detection and target locking systems
• Normally initiated with the help of OpenCV

Literature Review�

Literature Review..�

• Semi-Autonomous Gun and Movement Detection Robot (SAG)

​

• Included a portable command station.
• Invested prime concentration in the main structure of the robot very precisely

Literature Review…�

• Weapon Detection Robot (WDR)

​

• Concerns about the implementation of several computer vision algorithm approach for the various types of weapons recognition and detection at the same time.
• Almost correct model training accuracy for the weapons image datasets

Motivation

• A speedy and quick response

​

• Number of customers can deal with immediately

​

• Resolve many queries at the same time.

​

• The new revolution for customer service, reducing the impact on humans.

​

• There are security vulnerabilities in chatbot that can allow hackers to assault websites and hack content such as email addresses, content, and so on

Methodology

​

• Architecture for navigation and sensors

​

• Real time assembly for navigation

​

• Architecture for modified Inception-Net

​

• Model training and hyper-parameter tuning

​

• Keras (Backend), Tensorflow JS and TF-Lite

Schematic diagram for navigation of Nano Rover

Circuit diagrams for navigation and sensors of Nano Rover

Real time assembly of Nano Rover

LIDAR Algorithm:

The architecture of modified Inception-Net with 20 layers for Image Processing task of Nano Rover

Modified Training:

- 60000 Images for our custom build dataset

​

- Labeling and Annotating Images

​

- Primary Training (20 Layers)

​

- Data Augmentation and Image Generation

​

- Re-scaling (0.1/255)

​

- Hyper-parameter Tuning

LIDAR Algorithm:

Result Analysis

Real time visual data transmission and facial detection preciseness of Nano Rover

Result Analysis (Con’t)

Real time 3D observation of LIDAR Sensor and live GPS location streaming of Nano Rover

Result Analysis (Con’t)

Real time detection of Ak-47 and Benelli M4

Result Analysis (Con’t)

Real time detection of Glock-19 and HK-MP5

Result Analysis (Con’t)

Real time detection of Sig Sauer 226 semi automatic pistol and UZI 9mm sub machine gun

Result Analysis (Con’t)

Result Analysis (Con’t)

Result Analysis (Con’t)

Limitations

• Slower Run Time
• Less amount of GPU
• Combination of more modified Neural Architecture is necessary
• Limited Memory Space of Hardware Machinery.

Conclusion

• Expand the boundaries of robotics approach into the field of engineering for surveillance, reconnaissance and security of life.
• Threat mitigation with Technology.
• Focusing to take our research at further level of implementation of new cutting-edge Neural Network Architectures (Res-Net, Mobile-Net, Efficient-Net etc.) which will sharpen our visionary perception in the field of Artificial Intelligence, Sensor Fusion and Robotics

Any

Question?

Thank

You!