Visual Alarm Detection
Detection of intruders is the key to any perimeter protection. Be it guarding against trespassing, vandalism, or industrial espionage, first of all the intruder has to be detected. This software library offers distinct recognition and tracking of persons, animals or other objects. Mobile video recognition and tracking technology on unmanned vehicles is a key element for surveillance in open terrain. Our recognition and tracking software library serves for detection and verification of signals, and permits genuine alarms to be distinguished from deception alarms, or technical alarms. Determination of the real cause of the alarm is the prerequisite for assessing the situation and taking countermeasures, in particular in remote terrain where intervention by security personnel or police would involve high costs.
The software uses state-of-the-art image processing methods to detect movement in infrared camera image sequences, classify it appropriately according to its shape (for instance, human beings vs. other moving objects), and then determine that an alarm situation has occurred.
Fig. 1: Detecting human beings in infrared camera images
The objects are not only detected, but also tracked over time. Additionally, the software is capable of running on platforms that are moving themselves, such as mobile robots. This makes for a very potent solution for mobile surveillance.
Fig. 2: Detecting human beings with a mobile surveillance robot