Developed Applications
Bright Vision Analyzer - AI-Powered CCTV, Computer Vision & Digital Forensic Analysis
Bright Vision Analyzer is an AI-powered computer vision and digital footage analysis application I developed to support intelligent CCTV examination, automated object detection, movement tracking, traffic monitoring, and digital forensic investigations. Built as a Windows desktop application using Python, OpenCV, YOLO11, and Claude AI integration, it combines computer vision, machine learning, investigative analytics, and evidence documentation within a unified platform.
The application provides automated detection and tracking of people, vehicles, motorcycles, bicycles, buses, and trucks in recorded CCTV and traffic footage. Its analytical capabilities include anonymous multi-object tracking, directional counting, movement trails, restricted-zone monitoring, dwell-time analysis, approximate speed estimation, activity heatmaps, and event timelines.
Additional features include AI-assisted image interpretation, natural-language analysis of investigation results, low-light video enhancement, motion-based frame filtering, privacy-preserving visual redaction, and structured case reporting. The system also supports evidence integrity documentation through SHA-256 file hashing, model an
Bright Vision Analyzer - AI-Powered Computer Vision, CCTV Analytics & Digital Forensic Investigation Platform
Project Overview:
Bright Vision Analyzer is an advanced computer vision and AI-assisted investigative software application that I conceived and developed to improve the examination, interpretation, and documentation of recorded CCTV footage, traffic surveillance videos, and digital images.
Developed using Python, OpenCV, YOLO11 object detection, and Claude AI integration, the application combines machine learning, automated video analytics, object tracking, image enhancement, and digital evidence documentation into a unified Windows desktop environment.
The project bridges the gap between conventional CCTV review and intelligent investigative analysis by providing tools for detecting objects, examining movement patterns, identifying activity within designated zones, and generating structured analytical outputs.
Bright Vision Analyzer is designed primarily for cybersecurity professionals, digital forensic investigators, law enforcement personnel, security analysts, researchers, and traffic monitoring practitioners.
Project Motivation and Problem Identification:
The development of Bright Vision Analyzer was motivated by practical challenges associated with the manual examination of large volumes of CCTV recordings and other digital visual evidence.
During criminal investigations, security personnel and digital forensic examiners may need to review hours of surveillance footage to locate vehicles, examine movements, identify relevant time intervals, and document events.
Traditional manual examination can be time-consuming, particularly when recordings have poor lighting, complex movement patterns, multiple objects, or extensive periods without relevant activity.
The project addresses these challenges by applying computer vision algorithms to assist with automated footage analysis, movement tracking, event detection, and structured investigative documentation.
Its design draws on my professional experience as a Cybercrime and Digital Forensics Investigator in the Ghana Police Service and my interest in applying artificial intelligence to real-world investigative problems.
Core Features and Technical Capabilities:
1. AI-Powered Object Detection
Bright Vision Analyzer incorporates YOLO11 object detection models for identifying supported object categories in images and recorded video footage.
The system includes selectable detection models and confidence thresholds, allowing analysts to balance processing performance and detection sensitivity.
For traffic and CCTV analysis, supported object categories include people, bicycles, cars, motorcycles, buses, and trucks.
2. Multi-Object Tracking and Movement Analysis
The application provides multi-object tracking through ByteTrack or a built-in intersection-over-union tracking mechanism, depending on the selected processing engine.
Tracked objects receive anonymous identification numbers to assist in monitoring their movements across video frames.
The analytical workflow supports movement trails, approximate distinct-object counts, and dwell-time measurements.
3. Traffic Monitoring and Directional Counting
The application includes virtual counting lines for monitoring the movement of detected objects across defined boundaries.
Analysts can configure directional counts to distinguish movement from one side of a line to another, supporting traffic-flow analysis, vehicle movement studies, and location-based surveillance review.
4. Zone Monitoring and Event Alerts
Bright Vision Analyzer provides configurable analytical zones for examining entry, exit, and duration of activity within selected areas of recorded footage.
Its zone analysis features are designed to flag restricted-zone entries and instances where detected objects exceed defined dwell-time thresholds.
These capabilities support investigative review of access points, monitored premises, restricted locations, and other security-sensitive environments.
5. Approximate Speed Estimation
The system includes movement-speed estimation expressed in pixels per second or, when suitable calibration information is provided, approximate kilometres per hour.
This feature is intended for analytical support and research rather than certified traffic-speed enforcement.
6. Activity Heatmaps and Event Timelines
The application provides heatmap and timeline functionality to help analysts examine where and when detected activity occurred.
These visual analytics can assist in identifying frequently occupied areas, movement concentrations, and potentially relevant periods within lengthy recordings.
7. Open-Vocabulary Object Search
Bright Vision Analyzer incorporates a YOLO-World open-vocabulary capability that allows analysts to specify object descriptions beyond a fixed set of predefined detection classes.
This is paired with an AI-assisted search workflow intended to convert natural-language requests into object detection prompts.
8. Claude AI-Assisted Investigative Analysis
The application includes optional Claude AI integration for generating analytical summaries, examining detection results, producing draft case reports, and supporting question-and-answer interaction with available analysis information.
The AI assistant is intended to help analysts organise and interpret technical outputs while maintaining human responsibility for verifying findings.
AI-generated explanations are treated as investigative aids rather than conclusive forensic determinations.
9. Low-Light CCTV Enhancement
The application incorporates image enhancement techniques, including Contrast Limited Adaptive Histogram Equalization (CLAHE) and gamma correction.
These methods are intended to improve the visibility of recorded footage captured under difficult lighting conditions.
10. Motion-Based Frame Filtering
A motion-gating feature allows the application to skip selected static frames during processing.
This is intended to reduce unnecessary computational work when examining recordings containing lengthy periods of inactivity.
11. Privacy-Preserving Video Processing
Bright Vision Analyzer includes privacy-oriented visual redaction options for pixelating selected object classes in exported media.
This supports controlled presentation of investigative material and helps limit unnecessary disclosure of visually identifiable information.
12. Evidence Integrity and Reproducibility
A significant feature of the project is its focus on documented digital evidence processing.
The application provides for SHA-256 hashes of source and output files, processing settings, model information, and software version details in an analysis manifest.
It also supports HTML case reporting and is designed to keep original source media unmodified.
These functions assist with documenting how analytical outputs were generated and support review of the processing methodology.
13. Local Processing and Optional AI Integration
The application supports local computer vision processing using an OpenCV-based ONNX inference engine, with additional model configurations involving PyTorch.
Optional Claude AI features require an API connection. The interface provides controls for authorising the transmission of images or keyframes to the external AI service.
The project therefore distinguishes local media examination from optional cloud-based AI assistance.
14. Batch Processing and Command-Line Operation
Bright Vision Analyzer includes a command-line processing workflow for analysing recorded footage and collections of files.
This expands the project's potential use beyond interactive desktop review to repeatable analytical and research workflows.
My Role and Technical Contributions:
I served as the Software Developer, Product Designer, and Project Lead for Bright Vision Analyzer.
My responsibilities included conceptualising the software, defining its investigative requirements, designing the application interface, implementing computer vision workflows, integrating object detection and tracking technologies, and developing the evidence-oriented analytical features.
I also incorporated adjustable processing configurations, privacy controls, AI assistant integration, and documentation features intended to support reproducibility.
The project provided an opportunity to apply Python programming, computer vision libraries, machine learning models, and software engineering principles to problems encountered in CCTV examination and digital investigations.
Technical Architecture and Technologies:
The application's identified technical components include:
Python for the desktop application and analytical processing.
OpenCV for image processing, video handling, and computer vision operations.
YOLO11 models for object detection.
ONNX inference for lightweight model execution.
PyTorch-based processing options.
ByteTrack and intersection-over-union tracking methods.
YOLO-World for open-vocabulary object detection.
Claude AI integration for optional analytical assistance and report generation.
CLAHE and gamma correction for image enhancement.
SHA-256 hashing and JSON manifests for evidence integrity documentation.
HTML reporting and command-line batch processing.
Innovation and Project Significance
The distinctive contribution of Bright Vision Analyzer is the integration of AI-driven object detection, surveillance video analytics, privacy-conscious processing, and evidence documentation into a single investigative application.
Rather than treating object detection as an isolated AI demonstration, the project connects model inference with practical analytical activities such as movement tracking, directional counting, zone monitoring, evidence preservation, and reporting.
Its design also recognises the limitations of AI-based video analysis. Detection results may be affected by lighting, obstruction, camera quality, model accuracy, and scene complexity.
The application does not perform facial recognition or determine an individual's identity, intentions, or responsibility for an offence.
Current Development Status:
Bright Vision Analyzer has reached desktop version 0.3.1, with a Windows installation package and an integrated user interface for video analysis, image processing, AI assistance, and technical guidance.
The application is currently presented as a research prototype, with continued development and evaluation focused on detection reliability, processing efficiency, usability, reporting quality, and responsible AI implementation.
Professional and Research Relevance:
This project directly supports my academic and professional interests in artificial intelligence for cybersecurity, computer vision, digital forensics, cybercrime investigation, and secure investigative technologies.
It demonstrates my initiative in moving beyond the conventional use of forensic tools to designing and developing software that addresses practical investigative challenges.
Bright Vision Analyzer also provides a foundation for future research into automated CCTV examination, intelligent video analytics, reproducible forensic processing, human oversight of AI systems, and the reliability of machine learning in investigative environments.
Ultimately, the project represents my commitment to developing responsible, practical, and accessible technology that can strengthen digital investigative capabilities, support security research, and contribute to technological innovation in Ghana and beyond.