SKU: AD-0131

Grok visual detector

Drone identification by means of video analytics system Grok video detector.

Prime Consulting & Technologies is glad to offer our proprietary system of real-time detection of drones - Grok Video Detector (GVD).

System Application

Detection of UAVs in the video stream allows to solve two challenges. Firstly, it can be used as a stand-alone autonomous system for detecting drones and warning users about the appearance of a drone in the field of view of the video camera. This
solution allows to create the most inexpensive drone detection system as of available today on the market.

Secondly, this drone detection software can be used together with any other detection solutions, for example a radar, to reduce the number of false positives on birds. Grok Video Detector is able to distinguish birds from drones and minimize possible false alarms, which makes any detection system more reliable and eliminates the need for constant presence of an Operator to check any incoming alarm signals.

How it works

Video detection uses image recognition algorithms with the application of the in-house developed neural networks. Also, the pattern recognition algorithm helps to distinguish a UAV in the video stream from a bird or any other moving object. In the next software release, we are planning to add the algorithm of recognizing the model of a drone based on the collected database of images of various drones.

Detection of a UAV by the video stream is the cheapest and the most available technical solution for the moment. The cost of security cameras falls every year. The standard camera lens has an average viewing angle of 90 degrees. In other words, only four cameras are required to provide 360 degrees coverage and protection of one site. Of course, the range of a drone detection with this system is equal to a few hundreds of meters, which cannot be compared to more specialized solutions, such as radars or RF sensors. However, the installation of a bigger number of cameras solves this problem. At the same time, the radar, for example, has a very limited vertical field of view (VFOV) and has a limited line of sight. As a result, in order to cover the security zone in an urban area or with rough terrain, several radars are required, which significantly increases the budget of the drone detection system.

Grok Video Detector performs detection both above and below the skyline, which is especially important when using it in an urban area.

Fields of application

Grok Video Detector can be used as a stand-alone UAV detection system to protect any sites in an urban area or with rough terrain.

Grok Video Detector can also be used together with professional drone detection systems to eliminate false positives and automate the work of such systems. Also, the system can completely solve the problem of detecting UAVs at such facilities as an airport or a prison. For more complex projects, using Grok Video Detector with other technical solutions will help to enhance the detection system by covering numerous blind zones.

Main features

• Definitive drone\bird identification and classification before sending an alarm signal to the security system;
• Minimization of false alarms;
• Target tracking and automation of control over systems and equipment.

Explanation of video analytics performance

A radar detects a drone at the maximum possible distance. After the target was detected by the radar, there are two main risks:

1. a false target in the form of a bird, a ballon, etc.
2. the target can leave the radar detection zone pretty fast and enter the radar “blind zone”.

When a radar detects a target, it automatically transmits the coordinates of the target to the video surveillance system. The longer is the distance to the detected object, the bigger is the margin of error in accuracy of the target coordinates detection. The camera is pointed at the area where the target was detected, and then the target is detected and identified by the video analytics system Grok Video Detector. If the target was identified as a drone, the system issues an alarm signal and the camera starts tracking the drone, which allows not to lose the sight of drone in case it enters the radar “blind zone”.

Accurate identification of targets by means of the video analytics system Grok Video Detector also serves for automation of other systems operation, for example, operation of a jamming system, which will be automatically pointed at the target after its identification. This will reduce the reaction time of the security personnel and almost immediately activate the target neutralization upon command receipt.

Technical explanation

A. The video stream can be captured from a surveillance camera (or multiple cameras simultaneously) in real time, or it may be taken from a video file (avi, mp4, mkv etc.).
B. Next, each frame undergoes a number of transformations (binarization, filtering, etc.) for successful execution of drone search algorithms. [demonstrated in the video].
C. Drone detection is performed based on the processed frames by using Haar-like features and proprietary search algorithms that can accurately identify an object and essentially eliminate any false alarms.
D. In the real world when the video stream is captured from the surveillance camera, the camera is either static or rotating along a predetermined trajectory. Therefore, the background is static, so it can be subtracted during video stream processing, which makes it possible to detect objects with the level of reliability close to 100%.