{"id":922,"date":"2024-10-17T07:16:24","date_gmt":"2024-10-17T07:16:24","guid":{"rendered":"https:\/\/fleximodo.com\/blog\/?p=922"},"modified":"2026-10-09T07:21:31","modified_gmt":"2026-10-09T07:21:31","slug":"parking-sensor-accuracy-how-we-verified-99-96-detection","status":"publish","type":"post","link":"https:\/\/fleximodo.com\/blog\/2024\/10\/17\/parking-sensor-accuracy-how-we-verified-99-96-detection\/","title":{"rendered":"Parking Sensor Accuracy: How We Verified 99.96% Detection"},"content":{"rendered":"\n

Parking sensor accuracy is easy to claim and hard to prove. Almost every vendor prints a number on a datasheet, but few show how they measured it. So instead of just stating a figure, this article shows the method. Fleximodo verified 99.96% detection accuracy<\/strong> by checking sensor readings against recorded video, space by space and event by event. That is the difference between a marketing number and a measured one, and it is what separates the most accurate parking sensor from an average one.<\/p>\n\n\n\n

Because accuracy is the foundation of everything else, it deserves proof. A parking sensor<\/a> that misreads spaces breaks guidance, enforcement, and revenue all at once. So getting it right is not a detail. It is the product.<\/p>\n\n\n\n\n\n\n\n

Why parking sensor accuracy is the whole game<\/h2>\n\n\n\n

A parking system is only as trustworthy as its worst readings. For example, if a sensor reports a free space that is actually taken, a driver is sent to a dead end. And if it reports an occupied space that is actually free, the city loses a paying customer. Multiply either error across thousands of spaces, and confidence collapses. Therefore the goal is not “good enough” detection. It is detection you can rely on for every event, in every condition.<\/p>\n\n\n\n

How we measure parking sensor accuracy: the ViCheP method<\/h2>\n\n\n\n

To measure real accuracy, you have to compare what the sensor said with what actually happened. So Fleximodo built ViCheP (Video Check Portal)<\/strong>, a custom web app that lines up sensor data against recorded video of the same spaces. The team then reviews the footage and logs every real arrival and departure by hand.<\/p>\n\n\n\n

\"ViCheP<\/figure>\n\n\n\n
\n

“ViCheP wasn’t just about collecting data,” says Jozef Vindis<\/a>, the mechanical engineer who developed the software. “It let us overlay each sensor’s location onto the video feed, so the team could see exactly where every sensor sat relative to the vehicles. That made it far easier to spot when an error happened, and why.”<\/p>\n<\/blockquote>\n\n\n\n

The review itself is deliberately manual. The team goes through each 12-hour video segment<\/strong> and records every parking event they see, logging the exact time a vehicle enters or leaves each space. This hand-logged record becomes the baseline, the ground truth of what actually happened. Then we download the data straight from the sensors and compare it against those logged events. By finding the discrepancies, we can isolate the exact scenarios where a sensor diverged from reality and work out why.<\/p>\n\n\n\n

Because a human confirms each event, the comparison is exact. Every sensor reading is scored against ground truth: correct detection, missed event, or false trigger. In other words, the accuracy figure is not modelled or estimated. It is counted, event by event, from real parking behaviour.<\/p>\n\n\n\n

The challenges we had to solve<\/h2>\n\n\n\n

Real car parks are messy, and the visual overlay in ViCheP surfaced the exact situations that tripped up early detection. Two came up again and again:<\/p>\n\n\n\n