Synesis introduced a high precision AI solution which enables regular CCTV cameras to count the number of available and occupied parking spots in the visible area in real time. Unlike existing analogues, the Smart Parking module based on the bound of deeply trained neural networks and computer vision allows to solve three major problems simultaneously: a real-time monitoring and notification, compatibility with parking lots of any scale and configuration, and minimization of control resources.
Synesis Smart Parking has a great potential for further use and is implemented both by parking space operators (city administrations, shopping and entertainment centers, hotels, parking lots, etc.) and drivers.
Using computer vision, the module analyzes the location and number of cars in the frame with a specified time span
Accurately Count Parking Spaces
The solution works as follows: using a special software module, parking space operator allocate parking zones and their maximal capacity. Using computer vision, the module analyzes the location and number of cars in the frame with a specified time span - for example, 5 or 600 seconds.
The analysis performance by Synesis’ layers of deeply trained neural network and own dataset of multiple vehicle masks excludes scenario in which any other objects – people, tree crowns, etc. – can be falsely indicated as a car. Сomputer vision can accurately count parking spaces even when parking striping layout is covered with snow or when it’s absent at all.
The accuracy rate of technology is significantly higher than that of similar solutions, reaching out to 95%. After testing the technology in real conditions at the streets of Minsk, Belarus, the developers’ team plan to continue the neural network training for further accuracy rate improvement.
Integration Into Third-Party Services
The module can be used both to improve the level of urban environmental comfort and safetyThe solution was developed as a part of Synesis’ Kipod Smart City platform. However, the software company does not exclude that in future the module can be integrated into third-party services for transmitting information to online maps, mobile applications, chatbots, etc.
As a part of Kipod cloud platform, the module can be used both to improve the level of urban environmental comfort and safety. It allows parking operators and drivers to set special rules for recognizing and notifying events such as a car’s unplanned departure from a parking spot, trespassing, suspicious crowding and even emergencies (smoke/fire, shooting sound, broken glass sound, etc.). Such Kipod module is widely used in the United Kingdom for autonomous management of filling station networks.
Kipod’s ‘traffic’ module had also been trained to indicate parameters such as vehicle’s type, color, speed, parking legitimacy. It allows to determinate parking and traffic violations, as well as to investigate traffic accidents as quickly as possible. It also simplifies the search of vehicles with missing or not readable license plates.