DVIDP: A dual video stream-based drone intruder detector with the ability to both scan the horizon and classify the intruder using the PTZ platform.
The DVIDP is an early detection sensor, with two operational states (i) sense objects at the horizon and (ii) detect and recognize the object for deploying countermeasures. The sensor contains four components, namely (i) situation awareness module, (ii) Pan-Tilt-Zoom (PTZ) platform, (iii) multi-class drone classifier using deep-learning, and (iv) alert command center. The situation awareness module is operates based on the static camera videos to monitor intruders. The PTZ platform consists of a pan-tilt platform with the camera that supports programmatic control of the focal length of the lens. The PTZ signals are processed by the Raspberry Pi for activating the appropriate binary bits interfaced with the servo motors for accurately positioning the platform. The continuous media stream captured by the PTZ camera is transmitted to the analytic component, where the media is processed using the deep-learning network framework for detecting drone and subsequently identify the intruders at the horizon.
DVIDP offers an advanced capability for critical infrastructure operators to monitor the horizon for intruder objects using low-cost physical sensing device. The unique coordination of the deep-learning component feedback to control the PTZ platform offers an extra layer of security against false positives often countered in competing technologies based on radio communication. The multi- stream analytics processor integrates the best-in class technology for identifying objects at the horizon which is used to trigger and control the operation of the PTZ camera for gaining additional video evidence against the intruder. The threat levels are reported to the command center in real-time.
Dr. Krishna Chandramouli
Defender deliverable: D4.1, D4.2, D5.4 and D5.5. All the deliverables are classified as EU RESTRICTED