Your Paragraph Text

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Vehicle Alteration Detection

Using ML and Image


processing
Team members :
Muhammed Adil S Project Guide:
Johan C Jacob
Ms. Poorna B.R
Rezil Robin
Vaisakh S
INTRODUCTION
Vehicle alteration detection using ML and image processing identifies modifications or
tampering in vehicles by analyzing images with advanced algorithms. The system detects
changes like body modifications, paint alterations, or accessory additions by training
models on diverse image data. This approach enhances accuracy and adaptability,
offering a cost-effective solution that improves vehicle security and compliance while
reducing manual inspection efforts.
ABSTRACT
This project aims to develop an AI-powered application that detects unauthorized vehicle
alterations by comparing real-time images with a manufacturer-provided dataset. Key phases
include dataset acquisition, training a Convolutional Neural Network (CNN) to identify alterations,
and deploying the model on a cloud platform for scalable image processing. The app will feature a
user-friendly interface for capturing images, with real-time processing via a cloud-based model.
The project seeks to enhance vehicle safety and inspection efficiency through automated
alteration detection.
USE CASE DIAGRAM
LITERATURE REVIEW
SL NO. TITLE METHODOLOGY RESULTS LIMITATIONS

The authors developed a deep convolutional neural The model achieved The method struggled with
Automatic Vehicle state-of-the-art vehicles that had undergone
network (CNN) trained on the CompCars dataset to significant alterations or
Make and Model accuracy in recognizing
identify vehicle make and model from images. The modifications, such as
1 Recognition with
model utilized layers designed for feature
vehicle make and model
across different
custom paint jobs or
Deep Convolutional aftermarket parts, which
extraction, enabling accurate recognition across viewpoints and lighting
Neural Networks were not well represented in
different vehicle viewpoints and lighting conditions. conditions. the training dataset.

The model’s performance


Vehicle This study introduced a dual CNN architecture, with The integrated approach
was heavily dependent on
Identification Based improved the robustness of
one network focusing on license plate recognition the quality of the license
vehicle identification,
on License Plate and (LPR) and the other on vehicle body recognition. The plate images. In cases of
2 Vehicle Body Using combined outputs from both networks were used to
particularly in scenarios
extreme alteration or
where the license plate
Convolutional occlusion of the plate, the
improve overall vehicle identification accuracy using a was altered but the vehicle
accuracy dropped
Neural Networks labeled dataset of vehicle images. body remained unchanged.
significantly.
The survey categorizes deep learning-based vehicle
The survey compares these Challenges include handling
Survey of Deep re-identification methods into local features-based, methods, highlighting their environmental variations,
Learning-Based representation learning-based, metric learning-based, strengths and performance scalability issues with
3 Vehicle Re- unsupervised learning-based, and attention on public datasets, and labeled data, and
Identification mechanism-based approaches, each with distinct identifies which methods difficulties in dense traffic
Methods techniques for distinguishing and re-identifying are most effective under or with partially occluded
various conditions. vehicles.
vehicles.
SL NO. TITLE METHODOLOGY RESULTS LIMITATIONS

The authors utilized YOLO (You Only Look Once) for The system While effective for detection,
Real-Time Vehicle the system did not explicitly
real-time vehicle detection and Kalman filtering for successfully detected
Detection and and tracked vehicles in
focus on detecting alterations
4 tracking in traffic surveillance videos. The system was in vehicles. The tracking
Tracking for tested in a real-world setting, focusing on the real-time, providing performance could also
Traffic continuous detection and tracking of vehicles across valuable data for degrade in dense traffic
Surveillance traffic management. conditions or when vehicles
multiple frames. were partially occluded.

The authors applied machine learning algorithms, The model was limited by
Anomaly Detection The models achieved
the quality and diversity
including support vector machines (SVM) and high accuracy in
in Vehicle of the training data. VINs
5 random forests, to detect anomalies in Vehicle detecting tampered
Identification from certain regions or
Identification Numbers (VINs). The model was VINs, with random
Numbers (VIN) manufacturers were
trained on a dataset of valid and tampered VINs, forests performing
underrepresented,
Using Machine slightly better than
with random forests performing slightly better than affecting the model’s
Learning SVM. SVM.
generalizability.
TIMELINE
REFERENCES
[1]Yongguo Ren and Shanzhen Lan, "Vehicle make and model recognition based on convolutional neural networks," 2016 7th IEEE
International Conference on Software Engineering and Service Science (ICSESS), Beijing, China, 2016, pp. 692-695, doi:
10.1109/ICSESS.2016.7883162.

[2]Rafique, M.A., Pedrycz, W. & Jeon, M. Vehicle license plate detection using region-based convolutional neural networks. Soft Comput 22,
6429–6440 (2018). https://doi.org/10.1007/s00500-017-2696-2

[3]H. Wang, J. Hou and N. Chen, "A Survey of Vehicle Re-Identification Based on Deep Learning," in IEEE Access, vol. 7, pp. 172443-
172469, 2019, doi: 10.1109/ACCESS.2019.2956172

[4]Kumar, S.; Singh, S.K.; Varshney, S.; Singh, S.; Kumar, P.; Kim, B.-G.; Ra, I.-H. Fusion of Deep Sort and Yolov5 for Effective Vehicle
Detection and Tracking Scheme in Real-Time Traffic Management Sustainable System. Sustainability 2023, 15, 16869.
https://doi.org/10.3390/su152416869

[5]Alfardus, A.; Rawat, D.B. Machine Learning-Based Anomaly Detection for Securing In-Vehicle Networks. Electronics 2024, 13, 1962.
https://doi.org/10.3390/electronics13101962
THANK YOU!

You might also like