Efficient on-demand image transmission in visual sensor networks

In a tracking system, an object of interest is monitored continuously in a sensor network. Information about the object is kept in the sensors and sensors transmit the information upon request. In this paper, we consider the scenario where all sensors around a targeted object capture images of it and these pictures will be sent to a mobile agent upon request. Due to the size and energy limitations in sensors, images kept in sensors are often small and highly compressed. We describe a framework to facilitate a mobile agent in the sensor network to request images of the object of interest. As sensors are limited in energy, it is desirable to reduce the energy used in transmitting the images. We observe that, in a sensor network that is sufficiently dense, images from neighbor cameras would likely overlap, and therefore intermediate sensors can process and combine overlapping portions so as to reduce the energy spent on image transmission. We develop a protocol for involved sensors to determine how to transmit the images they have kept to the mobile agent in an energy efficient manner. Our protocol is truly distributed and does not require any global information. We evaluate our protocol through extensive simulations.