Smart-Tunnel: Improving automatic incident detection in tunnels

Challenge

In many road tunnels worldwide, human operators monitor the flow of traffic through CCTV cameras from a central control room to ensure the users’ safety. However, scrutinizing large numbers of monitors over extended periods of time is a cognitive challenge. Automatic Incident Detection (AID) systems help by analysing multiple video streams in real time and automatically alerting operators to events which they may need to act upon. Current AID systems utilize traditional computer vision image processing algorithms, which are subject to performance limitations inherent to their functional principle.

Solution

Artificial neural networks can outperform traditional approaches in computer vision. Therefore, ISSD has augmented their AID solution with deep learning to increase detection accuracy and performance. Specialized hardware for the neural network inference process keeps the server cost competitive. The solution can be deployed to both new and – as a retrofit – to existing tunnel monitoring infrastructure.

Achievements

ISSD has created a deep-learning based detection engine for their AID product that generates real-time alerts when pedestrians or a stationary vehicles appear in a video stream. After outperforming the previous solution against a representative set of video segments recorded in several road tunnels, the new detection engine was integrated into a live AID installation for evaluation at one of ISSD’s customers in Turkey.

In this test, the false alarm rate was reduced about half from the legacy solution, with the new engine still detecting all the events that the legacy engine detected, and doing so faster. The deployment environment for the detection engine employs Intel Movidius Myriad X visual processing units (VPU) in a highly scaled high-density deep learning (HDDL) configuration, processing camera video at a lower hardware cost per stream than the previous platform despite the increased computational requirements.

FED4SAE Support

Intel benchmarked the solution across multiple hardware platforms (CPU, FPGA, VPU) to confirm ISSD’s selection, provided performance tuning support of the network and assistance on the server configuration using the HDDL boards. Leading the technical coaching on neural networks, fortiss supported ISSD in applying the Neural Network Dependability Kit (NNDK) and in interpreting its analysis results to evaluate the object detection network model and improve its robustness against various kinds of image noise.

Blumorpho provided business innovation coaching and pitch training. Several joint dissemination and networking opportunities were enabled by all the partners.

Impact

• After its maturity for production use is ensured, the improved detection engine will be an integral part of ISSD’s tunnel safety product.
• The new technology will also be applied to other traffic monitoring products in ISSD’s portfolio.
• Intel will partner with ISSD to assist them in scaling their portfolio of intelligent traffic management solutions to the global market, improving traffic safety worldwide.

Video