No abstract

The development of a Passive Infra-Red (PIR) sens ing based intrusion detection system is presented here having the ability to reject vegetative clutter and distinguish between human and animal intrusions. This has potential application to reducing human-animal conflicts in the vicinity of a wildlife park. The system takes on the form of a sensor-tower platform (STP) and was developed in-house. It employs a sensor array that endows the platform with a spatial-resolution capability.Given the difficulty of collecting data involving animal motion, a simulation tool was created with the aid of Blender and OpenGL software that is capable of quickly generating streams of human and animal-intrusion data. The generated data was then examined to identify a suitable collection of features that are useful in classification. The features selected corresponded to parameters that model the received signal as the super imposition of a fixed number of chirplets, an energy signature and a cross-correlation parameter. The resultant feature vector was then passed on to a Support Vector Machine (SVM) for classification. This approach to classification was validated by making use of real-world data collected by the STP which showed both STP design as well as classification technique employed to be quite effective. The average classification accuracy with both real and simulated data was in excess of 94%.

No abstract

Robots that must operate in novel environments and collaborate with humans must be capable of acquiring new knowledge from human experts during operation. We propose teaching a robot novel objects it has not encountered before by pointing a hand at the new object of interest. An end-toend neural network is used to attend to the novel object of interest indicated by the pointing hand and then to localize the object in new scenes. In order to attend to the novel object indicated by the pointing hand, we propose a spatial attention modulation mechanism that learns to focus on the highlighted object while ignoring the other objects in the scene. We show that a robot arm can manipulate novel objects that are highlighted by pointing a hand at them. We also evaluate the performance of the proposed architecture on a synthetic dataset constructed using emojis and on a real-world dataset of common objects.