Emerging technologies are creating a real-time, high-bandwidth, global sensor network. The most visible component, the Internet, has become fundamental to 21st century business. The evolution of low-cost, networked sensors, often directly Internet-enabled, is bringing sensors out of their traditional closed-loop realms into the rest of our reality. Don’t believe me? Consider cell phones: There are 1.4 billion active cell phones in use today with more than half a billion units sold last year. As cameras become a standard cell phone feature, we’re becoming the most connected and instrumented people in history.
As sensor and communications technology continues to develop, we can envision a very different Internet than the one we use today. Rather than sending messages and browsing Web pages, we may experience new interactions such as experience sharing and browsing reality.
Data mining, defined broadly as extracting useful information and insights from data, may be the untold half of the sensor networks story. Given the potentially huge amount of data streamed by live sensors, algorithms to fuse, interpret, augment, and present information will become an increasingly important part of networked sensor applications. In this article, we’ll show examples of data integration, analysis, and visualization of sensor information.
We call the data mining of sensor streams “reality mining” to emphasize the direct mining of insight from operations-relevant sensor data streams. Reality mining provides an insight infrastructure between detection and action, allowing businesses and other organizations to use sensor data in valuable ways. For example, adding sensors to stands of trees would allow experts in a wood products company to monitor tree growth for operational efficiency and yield. Combining these sensor data with models of tree growth and projections of product markets as the trees mature could let the company make resource allocation decisions today to maximize profits later.