The University of Pretoria has showcased advancements in radar and target tracking algorithms that can be used to improve anti-poaching and drone detection efforts.
University of Pretoria electronic engineering student Neil-John Lord, speaking at the recent SA Radar Interest Group conference at the Council for Scientific and Industrial Research, discussed the problem of wildlife poaching and described previous efforts at identifying poachers. These included the use of camera traps in the Pilanesberg game reserve and the use of patrols and other methods.
These methods were not satisfactory due to false alarms with the camera traps: wind moving trees, leaves falling in front of the sensors or the movement of animals. Of the useable 10 percent of data, only a small subset were poachers and not animals.
He said most motion-detection models were developed in urban environments, which did not translate well to the bush and the difficulty of operating at night exacerbated the problem.
A University of Pretoria team then augmented the camera traps with additional sensors including lidar and radar sensors to detect motion (such as that of poachers) at night. The added sensors could give the observers distance to target, velocity and other extended information (size, orientation and shape) of the target.
Lord explained that with earlier radars, a target was picked up as a single return. With the advent of millimetre band radar, numerous returns were received from the same target. The target points on the radar screen were further clarified by use of an algorithm to create a ‘point cloud’. This can be averaged to get a very accurate measurement of where the target is at a given time. But a lot more can be done.
Target tracking and target classification are critical problems in battlefield surveillance systems, Lord explained. Traditionally, military radars detected and tracked targets, but classifying them is a new development in the digital age.
In World War II, for example, radars could detect enemy aircraft, but could not tell the number or type, only height and vector (size and direction). Modern systems, using improved radar and digitised software, including artificial intelligence (AI), can separate out the height, vector and often, the general type of aircraft, such as a fighter, passenger aircraft, cargo plane, or a missile. The software’s ability to ‘classify’ radar targets can make all the difference to success in the battlespace.
Getting back to the poaching problem, Lord explained that the difficulty was that, unlike ships or aircraft, sizes of animals and poachers were unknown. He said that sizes varied drastically, from elephants to impalas, but both are animals. He pointed out that separate classes could be created for separate species, but ‘that could get complicated quite quickly’.
Lord and his team discovered that the main difference between humans and animals was that humans exhibited bipedal motion, giving a different radar return to animals with quadrupedal motion. Once this difference was fed into the radar computer, it could create two ‘classes’, one for humans, one for animals. Two possible classifications were created, one based on orientation, and one on size.
Lord said he hoped the findings could be incorporated into existing poacher detection and tracking systems.
Drone detection
William Bourn, representing the University of Cape Town, told conference delegates of the promising use of millimetre band radar, specifically Frequency Modulated Continuous Wave (FMCW) radar, to detect drones and small targets.
With the increased use of drones used by terrorists, armed forces and civilian troublemakers, such as those who disrupted air traffic at Gatwick Airport, London in 2018, drone detection has become a priority, he explained.
He described existing challenges as ‘teaching’ software with machine learning to distinguish between small objects, such as small drones and birds. While much work needs to be done, the study of millimetre band radar promises solutions for a host of tasks such as assisting civil aviation at airports, catching poachers or preventing terrorists using drones to reconnoitre or attack military or civilian targets.
