A research project at the Indian Institute of Science (IISc) has come up with a sensor-based platform that can pick up movements and make out whether they were caused by a human or animal intruder, or simply by vegetative movement in the background.
In forested areas, it is often important to detect intrusions by humans to prevent illegal activities such as poaching, for which sensors need to be deployed. Sensors typically induce a current flow whenever there is a change in the radiation incident on the sensor. Such a change can happen due to any movement within the sensor's field of view. But, it is often difficult to ascertain whether the movement is by human beings or animals, or is simply a movement of the vegetation due to wind. This research project attempts to make this classification.
The platform used in the study is made of Passive Infra-Red (PIR) sensors, and the operating principle is based on techniques of signal processing. The system is mounted in the form of a tower, and consists of 8 sensors. The corresponding electronics and the packaging was designed and developed at the Department of Electronic System Design (DESE), IISc. The sensors are arranged in such a way so as to differentiate between translation movement by intruders, and oscillations of the vegetation. The system focusses on animals such as dogs, wolves, leopards and tigers, which are of lesser height in comparison with a human being. By estimating the height of the “intruder”, the system can distinguish between human and animal. So effectively, for every detected signal the system needs to do two rounds of classification: first to differentiate between vegetation and non-vegetation, and then between human and animal.
Data can be classified into different categories by various algorithms. But for that, the data must be represented mathematically. Such a representation is called “feature vector”. In this case, the data i.e. the radiation signal, is first converted into a feature vector by using a mathematical operation called “Chirplet Decomposition”. In addition, classification algorithms perform well only after they have been provided with enough examples of data, where the classification is known. In other words, they have to use a large number of sensor signals along with the information of what caused the signal—a human, an animal or simply, vegetative movement. Ideally, such examples should come directly from the sensor observations themselves. But since such data is hard to come by in large quantities, the researchers at IISc used animation-based techniques to simulate the received signal. They created realistic animations of humans and various animals, and simulated their movements to estimate the signals generated by the sensors. Features are identified using simulated data provided to the classification algorithm. "The use of techniques borrowed from animation to generate simulated signals allowed us to design and test classification algorithms”, said Tarun Choubisa, a researcher working on the project.
Some actual data was collected from sensors deployed in local forested areas within IISc campus, caused by intrusions or movements in the trees nearby. Additionally, data corresponding to animal intrusions was collected from a local dog training facility. The performance was validated on this real data. These sensor signals were correctly classified by the algorithms with very high accuracy. This validates the efficacy of these sensor systems, and indicates that it has potential for deployment in forests. In response to an unwanted intrusion within its field of vision, the PIR platform can be used to generate a warning signal to the forest staff, or record the event by an attached camera.
According to researcher Prof Vijay Kumar, “A principal contribution of the project is the design of an intelligent PIR-sensor platform that can distinguish between human, animal and vegetative motion. It was accomplished by careful design of a sensor-lens array with the necessary spatial resolution”.
According to Dr. T.V. Prabhakar from the research group, one advantage of their system is that PIR sensors can adjust their ranges “on-the-fly”, so that detection and classification can be improved. “Another associated novelty of this PIR system is the inclusion of energy harvesting circuits. With full charge and zero harvested energy, the system supports up to 48 hours of continuous operation under aggressive intrusion sensing and communication”, he added.
The reseachers feel that their research based on PIR sensors is an important step towards forest surveillance, but much more needs to be done for its ultimate success. According to Prof Vijay Kumar of IISc, “Every single sensing modality has its inherent strengths and limitations. We envisage that a collection of modalities, including the PIR sensor platform that we have designed, will be needed to significantly reduce human-animal conflicts”.
This work is part of a project jointly funded by the Department of Electronics and Information Technology under a Indo-US Pervasive Communications and Computing Collaboration (PC3) initiative. Apart from IISc, this project also involves researchers from the Indian Institute of Information Technology, Allahabad & the Wildlife Institute of India, Dehradun, from the Indian side and Ohio State & Cornell Universities from USA.
About the team:
From Electrical Communication Engineering Department:
Prof. Vijay Kumar (Principal Investigator)
Prof. Anurag Kumar (Co-Principal Investigator and Director, IISc)
Ajit Prabhu, V. S. Aswath, Christo K. Thomas, A. Praneeth, Raviteja Upadrashta, Siddhant Raman, Tarun Choubisa, Tony Gracious and S. Vikas (students under Prof. P. Vijay Kumar).
Abhishek Sinha, Anu Krishna, Arpan Chattopadhyay and Avishek Ghosh (students under Prof. Anurag Kumar).
Akhila S. Rao, S.V.R. Anand, Bharat Dwivedi, Easwar Vivek M., and Suman Kumar Panchal (work under Prof. Anurag Kumar).
From the Department of Electronic System Design:
Dr. T. V. Prabhakar (Senior Scientific Officer)
Mr. Ashwath Narayan Singh (Junior Scientific Assistant) and Mr. Chalapathi Rao (Principal Research Scientist)
Hari Prasad Gokul R., Madhuri S. Iyer and Sripad Kowshik (work under Dr. T.V. Prabhakar).
Tarun Choubisa email@example.com
Prof Vijay Kumar http://www.ece.iisc.ernet.in/~vijay/
ECE Office: 080 2293 2276; 080 2293 2278