Our team has developed IoT parking sensors that boast a best-in-class accuracy rate of over 99%. This exceptional performance is the result of extensive testing and the development of a proprietary detection algorithm.

We conducted rigorous testing on over 50,000 real-world parking events, utilizing custom software to validate accuracy. This iterative process enabled us to refine our sensors’ capabilities and ensure their reliability in diverse operating conditions.

The following sections will detail our testing methodology, the unique features of our detection algorithm, and the benefits of our high-accuracy IoT parking sensors.

Methodology

Rigorous Testing with ViCheP: Validating Sensor Accuracy

To accurately analyze and validate our IoT parking sensor data, we developed a custom web application called ViCheP (Video Check Portal). This tool was instrumental in allowing our team to perform precise, visual comparisons between real-world parking events and sensor data, a process essential for achieving high accuracy.

ViCheP allowed us to upload video recordings directly from parking lots equipped with our test sensors. The application would overlay the parking lot layout onto the video, marking the locations of each sensor and the parking space boundaries. This visualization provided a clear, real-time mapping of sensor placements relative to parked vehicles.

“ViCheP wasn’t just about collecting data,” explains Jozef, the mechanical engineer responsible for developing the software.

It allowed us to overlay sensor locations on the video feed, so our team could observe exactly where every sensor was positioned relative to vehicles. This made it much easier to spot when and why errors were happening.

Manual Event Recording for High-Precision Data

The Fleximodo team then manually reviewed each 12-hour video segment. During these sessions, team members recorded every parking event they observed on video, logging specific details such as the exact time when a vehicle entered or exited each space. This manual review was crucial for establishing an accurate baseline against which sensor performance could be compared.

After logging the video observations, we downloaded data directly from the sensors and ran detailed comparisons with the manually recorded events. By identifying discrepancies, we could analyze specific scenarios where sensor performance diverged from actual events and determine the underlying causes.

Identifying and Addressing Challenges in Detection

Through ViCheP’s precise visual and data overlay capabilities, we discovered several common situations that initially challenged our sensors, such as:

  • Quick Transient Events: These included vehicles that briefly occupied a space (e.g., turning around) without actually parking, or quick successions where one vehicle departed and another arrived within seconds.
  • Extended Adjustment Periods: Situations where a driver spent time adjusting their parked position also created initial difficulties for the sensor in finalizing event status.

In each case, our team analyzed the root cause of errors, often adjusting sensor sensitivity or detection parameters to improve performance under these complex, real-world conditions.

Combining Magnetometer and Nano-Radar

Originally, our sensors used only magnetometers (Mag-only), which alone could reach around 95% accuracy. However, during development, a compact, low-power nano-radar component became available, and we began experimenting with its integration. By combining the nano-radar with the magnetometer, we significantly improved accuracy to over 99%.

“Reaching this level of accuracy was a game-changer,” shares Bohdan, responsible for Business Development at Fleximodo.

When we started talking to city officials and commercial clients, the feedback was overwhelmingly positive. They need this level of reliability to confidently implement smart parking systems, and now we can assure them of that.

The radar component excelled at tracking the presence and movement of vehicles, particularly under dynamic conditions where the magnetometer alone was insufficient. The combination of radar and magnetometer created a system where each sensor type balanced the other’s weaknesses, such as compensating for rapid movements or short-duration events.

Results: Reaching 99%+ Accuracy

After extensive testing and adjustments, our IoT parking sensors demonstrated a verified event accuracy rate of 99.29% under real-world conditions. This high degree of accuracy was achieved by systematically refining our detection algorithm and optimizing sensor parameters based on data insights from ViCheP. This accuracy rate was not only a milestone but also an indicator of the system’s reliability in real-world deployments.