The solution we proposing is to create an additional service on CitySense platform. This means CitySense platform would have virtual extension for street markings detection and evaluation. CitySense addresses various urban challenges by implementing a dynamic and flexible mobile sensing network. Every mobile node has processing core and several sensors including camera, GNSS and IMU. Mobile nodes placed on city-owned or partner vehicles continuously collect data as they move, ensuring comprehensive coverage without the need for fixed sensor installations. This approach reduces the need for manual inspections and reporting, minimizing human error and increasing reliability. This collected data is then processed locally on the CitySense Unit before being securely transmitted through an IOT communication network to a secure cloud database system. This cloud infrastructure serves as the central hub for data storage, where the information is organized, analyzed, and made accessible.
The goal of the extension service is to process video stream from camera – part of CitySense Unit mounted to the vehicle and using AI and computer vision algorithms identify the marking pattern on the road. We assume that there is ground trues images of road markings with exact location of these road markings provided by city municipalities. This information allows us to compare processed by AI algorithms images with ground truth images and make comparison. The challenges for this solution is to synchronize GNSS data with video processed stream and match these processed images with ground truth data. After synchronization autoencoder based AI algorithm could be applied in order to identify
the goodness of the processed road markings. We can create several levels of marking condition: “excellent”, “very good”, “good”, “bad”, “no existing”. Thus, we classify by autoencoder output to which level current identified marking belongs to. Other challenges for this solution are different weather conditions (puddles, leaves, dust, snow etc.), which will lead to difficulties for recognizing the marking pattern by video processing. CitySense platform will collect a lot of images from the same place (fleet of vehicles regularly move along the city), thus statistically improve the accuracy of detection. Collecting and storing data from diverse sensors monitoring the urban environment is pivotal to fostering data-driven decision-making and enabling city planners to proactively address challenges. This comprehensive data platform serves as the foundation for informed city management, facilitating the delivery of optimized services and creating a more sustainable and resilient urban environment.
