Case Studies
Computer vision system for windshield assembly
Key Results
- ~94% defect detection accuracy
- ~40% less manual inspection
- ~18% fewer false positives
About the project
A commercial vehicle manufacturer engaged PerformaCode to develop a computer vision system for automated windshield assembly inspection. The platform was designed to verify the quality of critical installation steps directly within the production process and reduce dependence on manual visual checks.
The system evaluated multiple aspects of windshield assembly, including glass defects, seal installation, primer application, and adhesive quality. Inspection results were recorded to provide traceability and support manufacturing quality-control processes.
PerformaCode was responsible for the complete software solution, including system architecture, computer vision components, web-based operator interfaces, testing infrastructure, deployment environments, and production integration.
The project also included hardware evaluation, image-acquisition design, validation activities, and production handover to support long-term operation within the customer’s manufacturing environment.
6
engineers
10
months
FT
delivery model
Client challenges
The manufacturer was dealing with a significant volume of windshield leakage complaints and warranty claims. Existing inspection procedures were not providing sufficient control over glass condition, seal installation, primer application, and adhesive quality, so automated inspection had to be added directly to the assembly process.
The main technical constraint was image reliability. Windshield glass created reflections and glare that could obscure scratches, chips, and cracks. Primer and adhesive inspection added another challenge: dark materials had to be detected and measured on dark surfaces under factory lighting conditions.
The system also had to fit an existing production process. Camera placement, lighting, inspection timing, and data capture had to work without disrupting assembly operations. At the same time, the platform had to support 24/7 operation, maintain inspection history, and integrate later with a manufacturing execution system (MES).
Tasks performed
- Collected and analyzed production process requirements for windshield installation quality control, including glass defects, seal orientation, primer application, and adhesive placement.
- Defined functional requirements for the inspection system, web application, backend services, operating modes, availability behavior, deployment model, and future manufacturing execution system integration.
- Specified hardware requirements for cameras, lighting, mounting, workstation components, and supporting infrastructure.
- Tested demonstration equipment on a stand to validate whether the customer’s inspection requirements could be implemented under production-like constraints.
- Selected and evaluated computer vision equipment based on field-of-view coverage, lighting control, image stability, mounting constraints, and compatibility with the existing assembly process.
- Designed the system architecture covering image acquisition, recognition components, backend services, web-based workstations, data storage, monitoring, deployment, and recovery scenarios.
- Created development and test environments for server components, web interfaces, recognition algorithms, logging, monitoring, and automated deployment.
- Developed image acquisition methods for reflective automotive glass, including synchronized lighting and capture sequences to reduce glare and obtain analyzable images.
- Developed primer and adhesive inspection methods for detecting defects in dark materials on dark backgrounds.
- Developed glass defect detection methods to separate scratches, chips, cracks, and scuffs from reflections and lighting artifacts.
- Developed seal-orientation recognition methods to verify correct installation of the windshield seal.
- Implemented recognition components for primer defects, adhesive defects, glass defects, and seal installation issues.
- Built backend services for image processing, inspection data handling, event logging, component availability monitoring, and inspection history storage.
- Developed web-based operator workstations for inspection review, system operation, quality-control workflows, and administrative functions.
- Implemented access control and user scenarios for production, quality-control, and administrative users.
- Automated CI/CD processes for server-side and web application components.
- Integrated the system into the customer’s production network infrastructure and configured operational logging, monitoring, backup, and recovery procedures.
- Prepared manual and automated test procedures for the application server, web application, recognition components, and production workflows.
- Performed stabilization and defect correction before system handover.
- Prepared acceptance documentation, test reports, installation requirements, training materials, and handover documentation for production operation.
Project results
94% detection accuracy
The system detected scratches, chips, cracks, seal installation errors, primer gaps, and adhesive defects by combining controlled image capture with synchronized lighting and OpenCV-based processing.
40% less manual inspection
Automated inspection reduced operator workload by validating windshield assembly quality indicators that previously required repeated visual checks.
18% fewer false positives
The inspection pipeline reduced false positives by separating real defects from glare, reflections, and low-contrast artifacts during image post-processing.
24/7 operation
The platform supported continuous production use through component monitoring, event logging, and recovery scenarios for server, web, and recognition modules.
Inspection traceability
The system stored vehicle-level inspection images and defect results, creating a quality history for production review and warranty investigation.
100% open-source software
The platform avoided proprietary machine-vision licensing by using OpenCV and open-source components for recognition, processing, and deployment workflows.
MES-ready architecture
The system supported standalone operation and future manufacturing execution system (MES) integration through separate operating modes and integration-ready architecture.
Production-line integration
The inspection setup was deployed within the existing assembly process by adapting camera placement, lighting, and capture timing to line layout and vehicle positioning tolerances.
Value we bring
Hardware selection, validation, and prototype development
Selecting hardware from a catalog is rarely enough for systems that depend on optics, sensing, measurement, or environmental conditions. PerformaCode evaluates candidate components, builds test stands, validates technical approaches, and develops working prototypes before final architecture decisions are made. This allows hardware, software, and deployment constraints to be understood using measured results rather than assumptions.
Open-source platforms for production systems
Open-source software can reduce licensing costs, but production systems still require reliability, maintainability, monitoring, deployment, and long-term support. PerformaCode develops complete software platforms using open-source technologies while meeting requirements for scalability, integration, operational stability, and future system evolution.
Engineering for real-world operating conditions
Many systems operate in environments that are difficult to model in specifications: reflective surfaces, low-contrast materials, variable lighting, sensor noise, temperature variation, vibration, manufacturing tolerances, or changing operating conditions. PerformaCode combines software engineering, signal processing, image processing, hardware integration, and system validation to develop solutions that remain reliable under real operating conditions.
Technologies
- Windows Server
- Python
- C#
- JavaScript
- OpenCV
- .NET Core
- HTML
- CI/CD
Other Case Studies
Industrial asset monitoring and risk forecasting platform
Legacy industrial asset monitoring system modernization for teleme...
