Controlled capture
Camera position, lighting geometry, reflections and physical presentation of the target had to produce inputs the vision teams could tune and trust.
Hardware platforms for industrial vision AI
Across three automotive quality-inspection and worker-support projects, Zambeel developed the physical systems that allowed vision-software teams to create controlled images, deploy models at workstations and collect evidence in real factory conditions.
Client contextVision-AI software companies
EngagementHardware development
DeliveryThree hardware platforms developed
OutcomeDemonstrated, deployed and factory-evaluated
Engagement routeDevelop and Extend
Zambeel developed purpose-built hardware around client-owned software and provided a defined hardware workstream as the applications progressed.
How engagements workShared need
The software teams knew what the cameras needed to capture. The hardware had to produce consistent images.
The underlying algorithms needed clear, consistently illuminated views of automotive surfaces, components or worker activity. Moving from that software requirement to useful equipment meant resolving lighting, imaging, compute, operator interaction and the constraints of each working environment together.
Operational fit
Each system also had to work around operators, factory routines, remote software updates and the practical limits of installation, maintenance and daily use.
The design questionHow should imaging, lighting, compute and operator interaction be arranged so the software could progress from controlled experiments into useful physical systems?
Design approach
The hardware had to remain adjustable as the algorithms developed.
Rather than freeze one arrangement too early, Zambeel exposed the physical and electronic parameters the vision teams needed to test, tune and revise.
Software-controlled imaging
Camera and lighting parameters were made available to the software layer where remote tuning or changing models required it.
Application-specific platforms
Camera stations, inspection equipment and wearable hardware were developed around the needs and constraints of each application.
Compute and interfaces
Onboard and workstation compute, Ethernet, CAN and hardware interfaces were selected around the operational setting rather than treated as isolated electronics.
The three products
Three products developed around automotive vision AI.
The shared requirement connected the work, but each product addressed a different inspection or worker-support need.
Product 01
Automotive inspection tunnel
A sequence of camera and inspection-lighting platforms culminated in a walk-through tunnel for automotive final-assembly inspection.
Development path
Across approximately one year, Zambeel developed standalone camera posts, camera stations paired with zebra-pattern inspection panels, and a walk-through tunnel that integrated vision sensors and controlled lighting around the vehicle.
Hardware around the software
Camera and lighting arrangements were tailored as the inspection models developed, while physical trade-offs remained clear to the software engineers.
Result
The completed tunnel gave the software company a working automotive-inspection demonstrator for model development and customer demonstrations.
Product 02
Fuse-box inspection terminal
A benchtop terminal presented completed automotive fuse boxes to the vision software, returned an immediate operator result and connected each inspection to the wider production environment.
Physical inspection system
A wear-resistant receptacle, controlled lighting and imaging tube, presence sensing and an eye-level display formed a robust workstation for repeated handling and clear acceptance or rework decisions.
Software and factory interface
Software-controlled imaging parameters, onboard compute and remote configuration allowed vision engineers to tune the system without returning to the factory. Zambeel trained the software company’s team to integrate the hardware and supported deployment remotely.
Result
The terminal entered regular factory use. Vision engineers retained remote access for model updates.
Product 03
Worker-support wearable
A head-worn camera and haptic-feedback system provided a physical platform for evaluating real-time assembly-error prevention while workers performed normal factory tasks.
Ergonomic design
Zambeel moved vision processing to the workstation and streamed camera data wirelessly, reducing headset weight, heat and battery demand.
Evaluation hardware
Lighting, batteries, wireless communication and haptic feedback were integrated into a full-day wearable designed for factory evaluation and data collection.
Result
Ten wearable units enabled a vision-AI factory pilot. The hardware was used for data collection as development work for model training continued.
What changed
Three platforms moved vision software into three different kinds of physical proof.
The inspection tunnel gave a software company a substantial demonstration environment it could use to develop its proposition and open conversations with further automotive customers.
The fuse-box terminal progressed into regular production use. In later discussions, the software team reported that supply rejections had become virtually non-existent because internally detected errors were no longer reaching the receiving inspection stage.
The wearable platform enabled pilot evaluation and data collection while making the remaining software-reliability problem explicit. This distinction protected the credibility of the hardware work without overstating the maturity of the complete solution.
Capabilities involved
Physical product development around visual intelligence.
The work connected imaging and algorithms to the physical interfaces, environments and operating patterns that determined whether each proposition could be demonstrated or used.
Hardware Engineering and Systems Integration
Lighting, imaging, sensing, compute, power, communication and mechanical architecture developed as complete systems.
ExploreProduct Definition and Industrial Design
Operator ergonomics, physical presentation of target components and wearable form developed around actual use.
ExploreAI, Vision and Connected Systems
Capture architecture, software-controlled parameters, datasets, network interfaces and deployment-aware compute decisions.
ExplorePrototyping, Test and Validation
Progressive rigs, demonstrators, production equipment and pilot units used to resolve uncertainty at each stage.
ExploreRelated work
Other connected systems developed around real operating environments.
Discuss a requirement
Tell us what you need
Bring us the software proposition and the physical situation around it.
We can help determine what the algorithms need from imaging, hardware, operators and the operating environment.