Contour/edge detection aims to find the boundaries between areas where sensor readings change from true to false. There are two main tasks:
1. Individual sensors determining if they are interior, exterior, or on the edge without communicating. This helps sensors choose sleep schedules.
2. Obtaining an explicit geometric description of the contour/edge to communicate to the user. This requires determining the shape and communicating overhead depends on complexity - a circle needs 3 parameters but a polygon needs 2n for n points.
Localized edge detection has each node determine if it is on the boundary using information from neighbors. This is important for boundary tracking over time as the phenomenon evolves and edge nodes alert neighbors.
Contour/edge detection aims to find the boundaries between areas where sensor readings change from true to false. There are two main tasks:
1. Individual sensors determining if they are interior, exterior, or on the edge without communicating. This helps sensors choose sleep schedules.
2. Obtaining an explicit geometric description of the contour/edge to communicate to the user. This requires determining the shape and communicating overhead depends on complexity - a circle needs 3 parameters but a polygon needs 2n for n points.
Localized edge detection has each node determine if it is on the boundary using information from neighbors. This is important for boundary tracking over time as the phenomenon evolves and edge nodes alert neighbors.
Contour/edge detection aims to find the boundaries between areas where sensor readings change from true to false. There are two main tasks:
1. Individual sensors determining if they are interior, exterior, or on the edge without communicating. This helps sensors choose sleep schedules.
2. Obtaining an explicit geometric description of the contour/edge to communicate to the user. This requires determining the shape and communicating overhead depends on complexity - a circle needs 3 parameters but a polygon needs 2n for n points.
Localized edge detection has each node determine if it is on the boundary using information from neighbors. This is important for boundary tracking over time as the phenomenon evolves and edge nodes alert neighbors.
- Some sensor network applications require detection of contours or edges. Consider, as an example, a large field of chemical sensors. In case of an accident, it is important to get an idea of the position, extent, and shape of a toxic plum. Problem description - The goal of an edge detection algorithm is to find the boundaries between areas where the predicate, evaluated by perfect sensors without any measurement errors, evaluates to true or false. - In edge detection, there is an explicit notion of interior and exterior points; in contour detection, this is not the case.
distinguish different tasks in edge/contour detection:
A single sensor wants to determine whether it is an interior, exterior, or edge sensor but there is no immediate need to communicate this result further to any other node. For example, an exterior sensor might choose longer sleep periods than an interior sensor. A user wants obtain an explicit geometric description of the edge/contour. Accordingly, this shape must be determined and communicated to the user. The complexity of a shape description relates directly to the communication overhead, as the following examples illustrate: - If the network designer assumes beforehand that all contours have circular shape, three parameters suffice to describe a contour in the plane – the center point (x and y coordinates) and the radius of the circle determined by the protocol. Accordingly, the whole description can be encapsulated into a single small packet. - If the contour is described by a polygon with n points, a number of 2n values must be transported to the sink node. The number n depends on the number of sensor nodes in the vicinity of the contour and on the number of individual points each node contributes. The edge/contour detection problem has some similarities to edge or contour detection in the computer vision/image processing field. However, there are also important differences: Image processing algorithms work on pixels that are nicely arranged in a grid. They can therefore rely on techniques that require this regularity, for example, Fourier transform techniques. It is however, reasonable to assume that such a grid placement is not the dominant case in sensor networks. Instead, the edges/contours have to be estimated from irregularly placed points. • The sensor readings can be noisy.
Que: Write a short note on
i) Localized edge detection [S/16, W/16, S/17, S/18, W/18, S/19] Localized edge detection scheme: a technique by which each node locally determines (perhaps by gathering in- formation from other nodes within its neighborhood) whether it lies on or near the boundary specified by the query. Localized edge detection will be an essential component of boundary tracking as well; as the phenomenon evolves with time, nodes at the edge may alert neighboring nodes (in a manner similar to target tracking).