How Artifical Intelligence is improving video surveillance technology
Deep learning is a subset of Artificial Intelligence. Deep learning gives machines the ability to process data in the same way as the human brain. By using deep learning, machines can begin to predict patterns and behaviours using information it has learnt previously. This creates a machine which can essentially, think for itself and learn by example.
Arthur Samuel defined machine learning as a “field of study that gives computers the ability to learn without being explicitly programmed.”
How is deep learning being used with CCTV & Surveillance?
HikVision showcased their latest deep learning CCTV technologies at the China Public Security Expo 2017. Deep learning is hugely valuable to the field of security as it provides benefits to applications such as facial recognition, people counting and detecting vehicles. This will create a change in CCTV surveillance security where incidents can be detected and then dealt with in real time. Therefore, instead of going back through old footage to pin point a particular incident or individual, the CCTV will detect the individual whilst it is recording live.
But does it work? – The test
Can you evade the Chinese authorities? After passing a photo of BBC reporter John Sudworth to the Chinese authorities, it took them just seven minutes to locate and apprehend him.
“DeepInView” CCTV camera & “DeepInMind” Recorders
The DeepInView and DeepInMind CCTV surveillance products were announced by HikVision in one of their corporate press releases. The equipment is able to carry out functions such as: object detection, facial recognition, people counting and vehicle management.
DeepInView is the name given to the technologically advanced CCTV camera. This high tech CCTV camera features a panoramic lens on the top and a PTZ (Pan Tilt Zoom) lens at the bottom.
According to the press release, this is the world’s first ever smart outdoor dual sensor “Deep Learning” camera.
The DeepInView Camera can:
- Accurately detect, recognise and analyse human and vehicle ranging up to 40 meters with an accuracy level of around 99%
- Integrated with the Movidus Myriad 2 Vision Processing Unit (VPU) – This is a type of processor that uses parallel cores to run algorithms like object detection and facial recognition
- Capture sharp images of objects both in distant and close up. Resolution is 1920 x 1080 resolution
DeepInMind is the name given to the Network Video Recorder (NVR) that works alongside the DeepInView CCTV camera. The NVR is the first to benefit from deep learning. Whilst the camera acts as the eyes, the NVR represents the analytic and storage capabilities of the brain.
Previously, it would have taken 1,000 devices with 16,000 CPUs (Central Processing Unit) to simulate a neural network. Now this can be achieved with just a few GPUs (Graphics Processing Unit). HikVision have also worked to reduces transmission bandwidth and data storage capacity requirements. This means, no loss of quality even though the data being shared is exponentially higher.