UC1 - Intelligent Traffic Surveillance

Wihin the use case, CAMEA is planning to investigate smart and mostly wireless sensors (cameras, radars, etc.) in terms of testing and verification of quality requirements such as reliability and security for applications built on the Unicam platform. The VALU3S framework shall be used as an input generator covering various potentially interesting situations as data collection in real traffic environments is costly and does not guarantee full coverage of the input space. This could be problematic when training ML-based algorithms for safety critical systems, i.e., novel approaches to support robustness must be explored. Based on this, within the framework, the behaviour of such systems will be analysed, tested and validated against defined targets (e.g. dependability or trustworthiness). This should result in decreased development effort and subsequent deployment, and result in better maintainability of systems in the field.
UC1
Automotive

Unicam is CAMEA’s most advanced and most complex product It is a state-of-the-art and field-proven platform for creation of multifunctional and scalable intelligent vision-based and signal processing solutions. The platform has been used in two key areas – intelligent transportation systems and industrial inspection systems. All key technologies used for creating the innovative products are continuously developed by CAMEA. While OEM components are available for integration into current systems, fully featured systems are also being provided.

The traffic monitoring systems based on the Unicam can be adjusted for one of many possible applications (and also adapted based on customer’s needs). The systems mostly rely on video cameras and advanced video detection using modern algorithms. The most frequently used is an algorithm for license plate (LP) detection and subsequent OCR. Alternatively, other sensors such as radars, inductive sensors, or piezoelectric sensors can be employed.

Picture1.png

The most typical examples of applications based on the Unicam platform are Spot Speed Enforcement, Section Speed Enforcement, Travel Time, Red Light Enforcement, or Weigh-in-Motion solutions. The Unicam systems (e.g. Unicam VELOCITY – section speed measurement) are composed as a combination of a local processing on-site (LP detection and OCR) with all the infrastructure around (video cameras, IR flashes, PC, networking, etc.), and background processing running on server side. As it can be seen, CAMEA uses multiple levels of processing. Part of the computation is done locally, and the rest is done on the server. Currently, on the sites Unicam systems are updated with smart cameras (manufactured by CAMEA) with ability of running LP video detection algorithms, i.e., edge AI. The detector also supports additional logic, for example, for controlling IR flashes (see Figure ‎1.2) on demand/trigger. The detection results are then sent to the server and processed in the meaning of the matching corresponding detection and calculating average speed. When speed limit is violated, detailed evidence images can be requested from the site (which saves network bandwidth). Furthermore, next generation applications for smart and sustainable urban mobility (e.g., traffic light control for flow optimization and assisted vehicle platooning) could be enabled using Unicam as a platform for system-of-system solutions.

Picture2.jpg

It is obvious that the data transferred from the node to the server must be credible as it serves as evidence for the various kinds of violations. At any time, we must prove the source of the data and time of the capture. We also must ensure that the data cannot be counterfeited at any time i.e., data provenance is of fundamental importance. Thus, we are aiming to implementing some data (LP string, image, etc.) signing mechanisms (with possible encryption) directly in the smart camera. This is planned to be done in cooperation with FIT and BUT. Trusted wireless communication is essential to enable data sharing between different stakeholders in the future software ecosystem, paving the way for open innovation and novel mobility solutions.

In CAMEA, due to our wide portfolio of products and orientation on customers, developers often have to cope with short deadlines and at this moment, the testing and verification is not the highest priority. Note that there is not so much time from the moment when a contract is signed to the moment when the system should be installed and running. As CAMEA does even installation and service of the systems, most of tests are done after installation during a testing period or even after that during regular operation. This is of course not ideal, and it puts a big load on technicians and even developers when resolving bugs. It results in increased costs and sometimes even penalties resulting from the contract signed. Using the VALU3S testing framework, integrated in state-of-the-art test automation workflows, the reliability and security of systems can be ensured before field deployment and thus catch most of the bugs introduced to the system during the design or redesign phases. This will surely result in reduced costs and effort spent on system maintenance and support after installation.

Use case Evaluation Scenarios
Workflows
VALU3S Framework
Contents
Radar/camera advanced detection and tracking

Developing radar/camera-based advanced detection and tracking algorithms in the context of applications for smart urban mobility - generating inputs for simulators; covering potentially problematic situations; testing accuracy and reliability of detectors; testing perception robustness under uncertainty; integrating simulation-based testing in continuous integration contexts

Radar + camera cooperation

Testing/validating radar + camera cooperation - testing communication between sensors; triggering camera based on the radar detection input; time criticality + detection reliability and accuracy

Node connection to cloud

Testing/verifying node connection to cloud - verifying reporting mechanism including connection to the server, detection result buffered for server query, testing proper functionality under various conditions