Description:FORM 2
THE PATENTS ACT 1970
(39 of 1970)
&
The Patents Rules, 2003
COMPLETE SPECIFICATION
(See section 10 and rule 13)
1. TITLE OF THE INVENTION: “REAL-TIME SEATBELT COMPLIANCE DETECTION SYSTEM AND METHOD THEREOF”
2. APPLICANTS:

(A) NAME : NERVANIK AI LABS PVT. LTD.
(B) NATIONALITY : INDIAN
(C) ADDRESS : A – 1111, WORLD TRADE TOWER
OFF. S G ROAD,
B/H SKODA SHOWROOM, MAKARBA
AHMEDABAD 380 051
GUJARAT, INDIA.

PROVISIONAL
The following specification describes the invention. þ COMPLETE
The following specification particularly describes the invention and the manner in which it is to be performed.

Field of invention
The present invention relates to vision based real-time seatbelt compliance detection system and method, particularly to a hybrid two-stage detection and classification architecture for adaptive and continuous monitoring of a seatbelt usage under challenging environmental and visual conditions.
Background of invention
Driver negligence in wearing seatbelts continues to be a significant contributor to injuries and fatalities in road traffic accidents. Although seatbelt use has been shown to reduce the risk of fatal injury by approximately 45–50% for front-seat occupants and by up to 25% for rear-seat passengers, compliance remains inconsistent globally. Despite governmental regulations and public awareness campaigns, the actual usage rates vary widely, with some regions reporting rates as low as 70%. One of the core issues underlying this non-compliance is the inadequacy of existing seatbelt monitoring systems, which fail to provide accurate, real-time detection under diverse and complex real-world conditions.
Traditional systems, which often rely on pressure sensors embedded in the seatbelt buckle or on basic computer vision techniques, are plagued by multiple limitations. These include environmental variability such as fluctuating lighting conditions, glare, and shadows that interfere with camera-based detection accuracy. Additionally, human factors such as variations in body type, seating posture, and clothing can mimic the appearance of a seatbelt, leading to frequent false positives or false negatives. For example, a bulky jacket or bag strap can be mistakenly identified as a seatbelt, while actual improper usage, such as wearing the belt under the arm or placing it behind the back, can go undetected. Furthermore, the technical architecture of many of these systems is limited, often depending on a single Module or sensor, which reduces their robustness in handling occlusions, overlapping objects, or incomplete visibility of seatbelt components—especially in crowded or complex interior vehicle configurations.
Existing patents have attempted to address these challenges to varying degrees. For instance, US20190274117A1 outlines a camera-based seatbelt detection system using machine learning, but it lacks adaptability to dynamic posture changes and lighting variations. US10377363B1 introduces capacitive sensors in combination with vision inputs; however, it is susceptible to interference from certain fabric types and physical degradation over time.
Given these challenges, there is a pressing need for a more intelligent and adaptable seatbelt detection system. Such a system should be capable of performing reliably in diverse environmental conditions, accurately distinguishing between genuine and false seatbelt appearances, and detecting improper usage. Moreover, it should support multi-occupant detection and operate in real-time to ensure consistent monitoring and enhanced passenger safety.
Hence, it is needed to invent seatbelt compliance detection system for dynamic and adaptive vehicular environment.
Object of Invention
The object of a present invention is to provide a real-time seatbelt compliance detection system and method that accurately identifies whether a driver is wearing a seatbelt, even under challenging environmental and visual conditions.
Further object of the seatbelt detection method and system is to provide a real-time seatbelt compliance detection system and method that is adaptive to diverse situations including, but not limited to, driver postures, body types, clothing styles, and vehicle interior designs.
Another object of the real-time seatbelt compliance detection system and method is to enhance road safety by enabling intelligent monitoring through improved detection fidelity and timely alerts or interventions.
Yet another object of the real-time seatbelt compliance detection system and method is to implement a hybrid two-stage detection and classification architecture to improve accuracy and reduce false positives.
Yet another object of the real-time seatbelt compliance detection system and method is to create a scalable and adaptable seatbelt detection framework that can be deployed across various vehicle types and integrated into broader automotive safety systems, supporting real-time decision-making, driver behavior analysis, and regulatory compliance.
These and other objects will be apparent based on the disclosure herein.

Summary of invention
The present invention discloses a real-time seatbelt compliance detection system and method. The real-time seatbelt compliance detection system comprises dual-camera setup, including a road-facing camera to capture traffic conditions and a driver-facing camera to focus on the driver’s upper body. Said system comprises various modules including, but not limited to, a data acquisition module, face detection module, object detection module, classification Module, real-time monitoring and alert module, the adaptability module, and performance validation module to execute necessary tasks. The seatbelt is detected using a YOLOv9-based object detection module, followed by a classification module that determines whether the seatbelt is correctly fastened, improperly fastened, or unfastened. The system operates in real time, providing continuous monitoring and issuing alerts when improper seatbelt usage is detected. The system provides high accuracy, robust detection under diverse conditions, adaptability to varying driver profiles and environments, and low false positive rates.
The seatbelt compliance detection system for dynamic vehicular environment provides multiple benefits to the end users which are described in the following pages of specification.
Brief description of drawings
Other objects, advantages and novel features of the invention will become apparent from the following detailed description of the present embodiment when taken in conjunction with the accompanying drawings. Fig. 1 illustrates a block diagram depicting various modules of the real-time seatbelt compliance detection system, in accordance with some embodiments of the present disclosure.
Fig. 2 illustrates a flowchart, depicting a method of operation of the real-time seatbelt compliance detection system, in accordance with some embodiments of the present disclosure.
Fig. 3 illustrates a brief flow diagram, depicting an operation of object detection and classifier module, in accordance with some embodiments of the present disclosure.
Detailed Description of Invention
Before explaining the present invention in detail, it is to be understood that the invention is not limited in its application to the details of the construction and arrangement of parts illustrated in the accompany drawings. The invention is capable of other embodiment, as depicted in different figures as described above and of being practiced or carried out in a variety of ways. It is to be understood that the phraseology and terminology employed herein is for the purpose of description and not of limitation.
It is to be also understood that the term "comprises" and grammatical equivalents thereof are used herein to mean that other components, ingredients, steps, etc. are optionally present. For example, an article "comprising" (or "which comprises") components A, B, and C can consist of (i.e., contain only) components A, B, and C, or can contain not only components A, B, and C but also contain one or more other components.
The present invention relates to a real-time seatbelt compliance detection system (100) that utilizes a dual-camera configuration and advanced modules to monitor and assess seatbelt usage with high precision and robustness. The system (100) is particularly suited for deployment in vehicles and is designed to enhance driver safety by ensuring timely detection and reporting of seatbelt non-compliance under diverse and challenging conditions.
As shown in fig. 1, the real-time seatbelt compliance detection system comprises a plurality of interconnected hardware and software modules that function in unison to detect, classify, and monitor seatbelt usage in vehicular environment (10) having a driver (1). The hardware setup includes, but not limited to, a road-facing camera (101a) and a driver-facing camera (101b). The road-facing camera (101a) captures external driving conditions, such as road events and traffic flow, to provide contextual awareness of the driving environment. Simultaneously, the driver-facing camera (101b) focuses on the driver’s (1) upper body, including the face, chest, and shoulders, enabling focused analysis of seatbelt usage and posture. In some embodiments, real-time seatbelt compliance detection system (100) includes various modules including, but not limited to, a data acquisition module (101), a face detection module (103), an object detection module (105), a classification module (107), a real-time monitoring and alert module (109), an adaptation module (111), and performance validation module (113) for performing various task like acquiring, processing and providing various data.
As shown in fig. 2, system is initiated by capturing synchronized video streams using a road-facing camera (101a) and a driver-facing camera (101b) by a data acquisition module (101) that is configured to collect synchronized real-time captured data (201) in form of sequence of images or video streams from both cameras. Said module ensures a continuous feed of image data necessary for live monitoring. Further, a face detection module (103) processes the driver-facing camera (101b) feed to accurately detect and track the driver's (1) face, even in adverse lighting or partially occluded conditions in vehicular environment (10). The face detection module (103) is also configured for dynamic isolation of a Region of Interest (ROI). The ROI comprises, but not limited to, the chest and shoulder region of the driver (1), while temporarily filtering out irrelevant parts of the frame, such as the passenger area or vehicle background.
Further, within this isolated ROI, a YOLOv9-based object detection module (105) is employed. Said module is configured by training on a dataset of annotated seatbelt instances, identifies and localizes the seatbelt within the ROI by drawing bounding boxes around potential seatbelt regions providing Bounding Box Regression and assigning object confidence score. The object detection module’s (105) custom backbone architecture enhances feature extraction by emphasizing spatial and positional relationships critical for accurate detection. In initial detection phase, the object detection module (105) broadly categorizes seatbelt presence as either "present" or "absent," based on its position relative to the driver’s upper body.
Subsequently, the system employs the YOLOv9-based classifications Module (107) that receives the detection results and performs a more granular analysis. The classification Module (107) is configured and trained to distinguish between three categories of seatbelt usage: "correctly fastened," "improperly fastened" (e.g., worn under the arm or behind the back), and "unfastened". Said module focuses on identifying subtle inconsistencies, including slack, misalignment, or partial occlusion of the seatbelt. In some embodiment, the object detection module’s (105) and the classification module (107) may work simultaneously to provide end results, as shown in fig. 3.
The system further includes a real-time monitoring and alert module (109), which provides an alerts when improper or absent seatbelt usage is detected for a predefined duration (e.g., 60 seconds), said module may trigger audio-visual alerts or can also include other means of alerts to prompt user compliance.
Subsequently, the adaptability module (111) enhances the real-time seatbelt compliance detection system's (100) usability across a variety of conditions, accommodating different driver (1) physiques, apparel, seating positions, and variable environmental lighting conditions such as shadows or glare in vehicular environment (10). Such environmental condition adaptation keeps enhancing the real-time seatbelt compliance detection system's (100) accuracy as per driver’s (1) usage.
The performance validation module (113) ensures reliability by evaluating performance across real-world scenarios using performance validation metrics such as precision, recall, and F1 score. The standard data of precision metric stays around 92.3%, Recall metric stays around 90.7% and F1 Score metric stays around 91.5%. These figures may vary slightly depending on deployment conditions and data quality, but they represent typical performance under standard test scenarios. The real-time seatbelt compliance detection system (100) is configured and trained on a curated dataset of over 100,000 annotated images containing diverse seatbelt types, driver postures, and lighting environments. Data augmentation techniques, including rotation, scaling, and brightness variation, were applied to simulate challenging real-world scenarios.
For object detection module (105) training, the object detection Module (105) uses the YOLOv9 architecture, incorporating bounding box regression loss resulting in bounding box regression, object confidence score, and class probability functions to achieve precise localization and broad classification. The classification module (107) is fine-tuned from a pre-trained YOLOv9 network and optimized for accuracy in nuanced seatbelt state distinctions. Training datasets included, but not limited to, labeled examples of each seatbelt condition, even under partial occlusion or poor lighting.
The hybrid approach, leveraging both detection and classification phases significantly improves accuracy and robustness. The real-time seatbelt compliance detection system (100) achieves a detection accuracy of 95%, classification accuracy of 93%, and an overall system accuracy of 94%, with a false positive rate below 3%. Additionally, the system supports real-time processing at 30 frames per second, making it suitable for in-vehicle deployment on edge processing devices. Such edge processing devices are a computing unit capable of performing tasks on-device without dependence on cloud-based computation. Edge processing device is configured to run the trained face detection model being run by The Face Detection Module (103) locally and interface with image capture modules such as a camera. The face detection module (103) is optimized to ensure real-time performance and low-latency response suitable for embedded or resource-constrained environments. The architecture of the edge processing device may vary and is not limited to a specific processor type, allowing flexibility in hardware integration. In some embodiments, system may include cloud connectivity (111) as optional feature that allows for long-term data logging, compliance analytics, and integration with vehicle telematics systems.
The seat-belt compliance detection method and system (100) has beneficial advantages that said system provides a novel solution to address the persistent issue of driver negligence in seatbelt usage, enabling improved compliance monitoring, real-time safety interventions, and potential integration into broader vehicular safety and monitoring ecosystems.
Field testing results validate the real-time seatbelt compliance detection system's effectiveness under various real-world conditions, with 94% accuracy during daytime, 91% in low-light environments, 89% with partial occlusion, and 95% under standard seating conditions.
The invention has been explained in relation to specific embodiment. It is inferred that the foregoing description is only illustrative of the present invention and it is not intended that the invention be limited or restrictive thereto. Many other specific embodiments of the present invention will be apparent to one skilled in the art from the foregoing disclosure.
All substitution, alterations and modification of the present invention which come within the scope of the following claims are to which the present invention is readily susceptible without departing from the invention. The scope of the invention should therefore be determined not with reference to the above description but should be determined with reference to appended claims along with full scope of equivalents to which such claims are entitled.

List of Reference Numerals
1 Driver
10 Vehicular Environment
100 Real-time seatbelt compliance detection system
101 Data Acquisition Module
101a Road-Facing Camera
101b Driver-Facing Camera
103 Face Detection Module
105 Object Detection Module
107 Classification Module
109 Real-time Monitoring and Alert Module
111 Adaptation Module
113 Performance Validation Module
201 Real-time Captured Data
, Claims:We Claim:
1. A real-time seatbelt compliance detection system, comprising;
a road-facing camera (101a) configured to capture external traffic conditions and driving context;
a driver-facing camera (101b) configured to capture images or video data of the driver’s (1) upper body including face, chest, and shoulder regions;
a data acquisition module (101) configured to receive and synchronize real-time Captured Data (201) from both the road-facing camera (101a) and driver-facing camera (101b);
a face detection module (103) configured to identify the face of a driver (1) in the driver-facing video feed and dynamically isolate a region of interest (ROI) corresponding to the chest and shoulder area of the driver (1);
a YOLOv9-based object detection module (105) configured to provide bounding box regression and assigning object confidence s to detect and localize the seatbelt within the ROI and broadly classify seatbelt presence as “present” or “absent”;
a YOLOv9-based classification Module (107) configured to further analyze the ROI and classify seatbelt usage based on class probability , into one of at least three states namely “correctly fastened”, “improperly fastened”, or “unfastened”;
a real-time monitoring and alert module (109) configured to issue an audio and/or visual alert when improper or absent seatbelt usage is detected for a predefined duration;
an adaptation module (111) configured to accommodate diverse driver (1) profiles and varying vehicular environmental conditions including variable lighting, occlusion, and driver posture;
a performance validation module (113) configured to compute system performance using performance validation metrics made and selected from the group comprising precision, recall, and F1 score.
a non-transitory computer-readable storage medium having instructions stored thereon that, when executed by a processor, causes real-time seatbelt compliance detection system to perform the steps of the method.
2. The system as claimed in claim 1, wherein the face detection module (103) is configured to handle partial occlusion and varying lighting conditions.
3. The system as claimed in claim 1, wherein the region of interest (ROI) isolation is dynamically adjusted based on the location of the driver's (1) face to focus processing on chest and shoulder region.
4. The system as claimed in claim 1, wherein the real-time monitoring and alert module (109) is configured to trigger alerts when non-compliance persists beyond customizable threshold duration.
5. The system as claimed in claim 1, wherein the adaptability module (111) supports variations including driver’s (1) body size, clothing, and seating posture along with vehicular environmental (10) factors like variable lighting, and occlusion.
6. The system as claimed in claim 1, wherein the system is configured to be deployed on an edge processing devices for real-time processing at a minimum frame rate of 30 frames per second.
7. The system as claimed in claim 1, further comprises cloud connectivity for uploading compliance data, generating analytics reports, and integrating with external vehicle telematics systems.
8. A method for detecting seatbelt compliance in real-time comprising,
capturing synchronized real-time data (201) using a data acquisition module (101);
detecting the driver’s (1) face from the driver-facing video feed and isolating a region of interest (ROI) based on the face location using a face detection module (103);
detecting and localizing the seatbelt within the ROI using a YOLOv9-based object detection module (105);
classifying the seatbelt usage into “correctly fastened”, “improperly fastened” or “unfastened” using a YOLOv9-based classification module (107);
issuing an alert incase the seatbelt is improperly fastened or unfastened for a predefined time duration using real-time monitoring and alert module (109);
adapting the detection and classification steps for variations in driver (1) profile and environmental conditions using adaptation module (111)
computing system performance through performance validation metrics made and selected from the group comprising precision, recall, and F1 score using performance validation module (113).
9. The method as claimed in claim 11, wherein detecting and localizing the seatbelt within the ROI by a YOLOv9-based object detection module (105) involves the bounding box regression and object confidence score for detection and localization.

Dated this on 04th August, 2025.