Useful information
Drivers needed to know where a vacant space was, not simply whether a car park had spare capacity. Facility teams needed a current spatial view they could use operationally.
Parking intelligence without bay-by-bay infrastructure
Zambeel developed a low-infrastructure parking technology stack that progressed from standalone wireless bay sensors to computer vision, geolocation, operator tools and outdoor driver guidance.
OriginZambeel technology initiative
DevelopmentSensor and vision platforms developed
ValidationField deployment and vision study
Current stageOutdoor pilot deployments
Engagement routeDevelop
Zambeel originated and developed the technology from field sensing through vision, spatial software and a licensable operating platform.
How engagements workThe challenge
Real-time parking guidance usually becomes an infrastructure project.
Many parking-occupancy systems place a powered sensor at every bay. At useful scale, wiring, civil work and the cost of each installed point can become more significant than the sensing technology itself.
Lower installation burden
The system needed to avoid extensive wiring and remain viable across existing outdoor parking areas that had not been designed around smart infrastructure.
The development questionHow much new infrastructure and installation effort could be avoided while still locating individual vacant spaces reliably enough to guide a driver?
Hardware path
A two-stage bay sensor reduced the energy cost of reliable detection.
The first development path used a standalone wireless unit at each space. A very-low-energy sensor watched for change. A more capable radar sensor was activated only when that change needed to be verified.
Detection in two stages
Keeping the higher-power sensor dormant for most of the operating cycle preserved the accuracy benefit of radar without the continuous battery consumption it would otherwise require.
Wireless and self-contained
The sensing and communication architecture avoided power and data wiring at each bay, with the field units reporting occupancy to the parking service.
Vacancy monitoring without civil works
The field deployment demonstrated real-time, bay-level vacancy monitoring using self-contained wireless sensors, without power wiring or civil modifications at each space.
Energy evidenceEight months of field operation on one charge demonstrated the low-energy architecture in use. Power-consumption measurements supported a projected battery life beyond the three-year design target.
Vision evolution
As vision models improved, one camera could replace many physical sensing points.
The project shifted toward scene-level computer vision when segmentation and contextual methods became capable enough to classify spaces from a wider view. Development then moved from a sensor at each bay to a connected platform built around cameras, cloud processing and spatial software.
Fixed or existing cameras
Initial trials used cameras selected and positioned for the task. Current development can also work from suitable existing CCTV views.
Cloud inference
Current frames are transmitted for server-side processing, allowing the model and management tools to develop without replacing field compute.
Map-ready space positions
Camera location and learned scene understanding estimate the position of each space, with human-supervised training refining placement during onboarding.
Internal GIS tools
Zambeel built graphical tools for onboarding lots, reviewing detections and managing the platform across its spatial and technical layers.
Validation
The early vision study established a measurable baseline and a clear next task.
A 30-day evaluation covered 55 parking spaces. System classifications for each bay were compared with a human-annotated reference dataset drawn from images captured every 30 seconds.
Early vision baseline
Across all bay-level decisions in the study, the system assigned the correct vacant-or-occupied status 84% of the time. This established a measured starting point for subsequent dataset and model improvements.
Error analysis
The study confirmed that the approach could work and identified where additional data and model refinement were needed.
Driver benefit evaluation
Current pilots are designed to compare parking-search time and reported convenience between drivers using the guidance service and those finding spaces without it.
Deployment model
The technology became a licensable system rather than a single parking installation.
The platform combines field capture, vacancy classification, verified map positions, outdoor guidance and administrative control. Parking solution providers can use the components under licence within their own customer offering.
Current pilots
The platform is being piloted with parking solution companies in North America, the UK and Asia.
Image handling and privacy
The system uses low-resolution views that do not allow faces, vehicle number plates or other personal information to be identified. Images are processed without retaining an image history.
Current development status
Development currently focuses on outdoor parking areas. Performance with existing CCTV and the benefit to drivers are being evaluated before the platform is offered at more locations and to more drivers.
What changed
A narrow sensing problem became a broader spatial-intelligence platform.
The sensor development proved that a self-contained parking detector could operate for months without wired infrastructure and identified the practical factors that mattered to field reliability.
The vision platform changed the economics and reach of the proposition by allowing one view to cover many spaces and by making use of cameras already present at a site.
Mapping, onboarding and driver guidance turned occupancy detection into an operational service. The current pilots are now testing whether this service materially reduces search time and frustration.
Capabilities involved
Connected hardware, visual computing and spatial software developed as one system.
Developing the platform required Zambeel to move between low-power electronics, field deployment, model development, mapping interfaces, cloud operations and pilot deployment without separating them into disconnected technical exercises.
AI, Vision and Connected Systems
Dataset collection, occupancy classification, scene understanding, camera integration and cloud inference.
ExploreHardware Engineering and Systems Integration
Low-power sensing, radar verification, communications, battery architecture and field-ready assemblies.
ExploreSoftware Platforms and Spatial Interfaces
GIS-style onboarding, parking maps, driver guidance, administrative tools and deployment infrastructure.
ExplorePrototyping, Test and Validation
Instrumented field pilots, uptime analysis, labelled vision studies and benefit-measurement design.
ExploreRelated work
Other systems built around sensing, intelligence and real operating environments.
Discuss a requirement
Tell us what you need
Bring us the physical environment and the decision your system needs to make.
We can help determine the right combination of sensing, vision, connected hardware, software and field evidence.