Emza Video Analytics for Remote Site Monitoring
Emza offers a remote site intrusion detection solution by applying video analytics in a very specific and optimized manner for the application. Remote sites like substations and water plants contain valuable resources yet are routinely unguarded and have limited communication infrastructure. A traditional video surveillance solution may be significantly expensive ($10,000 USD or more) while requiring significant bandwidth to access video streams.
Video Analytics Sensor - Not Traditional Surveillance Camera
What's interesting about Emza is that it is a video analytics sensor [link no longer available] that sends image but is not a regular surveillance camera. The sensor can send stream JPEG images on demand and based on events detected by the on-board video analytics. However, it is not a typical surveillance camera that uses H.264 or MPEG-4, etc.
Benefits for Remote Site Monitoring
This design decision makes Emza less expensive, simpler and easier for deployment in remote sites for intrusion detection. Compared to smart cameras such as ioimage, Emza is up to 70% less expensive. Unlike many other lower cost video analytics (e.g, via:sys), Emza's analytics are built-in and optimized for its own hardware. Finally, for low speed networks at remote sites, sending only images is optimal.
Limitations on Emza's Application
Emza is not a general replacement for surveillance cameras with video analytics; Emza is instead optimized around identifying and recording intrusions, not people counting nor abandoned objects. If you need to record ongoing daily events for later investigations and are willing to spend additional money for recorders, networking infrastructure, etc., a traditional surveillance camera is more appropriate. Similarly, if you have existing cameras deployed, you may find it more economical to add video analytics to those cameras.
Product Test Results
I conducted an in-house product test of the Emza sensors. Here are our key findings:
- The sensors require basic optimization. Optimization is accomplished by adjusting a few basic parameters including depth, areas to alert and areas to exclude. Accomplishing this does require knowledge and some hands-on experience. However, it does not require complex configuration. (As a side note, I am deeply skeptical of any system that requires no configuration and consider moderate GUI driven configurations to be the current state of the art in video analytics).
- The sensors generated only a handful of false alerts over a one week period. The sensor was able to detect events over 100 yards away consistently. This included two periods of moderate rainfall as well as night time with street lights.
- I did not have any periods where the sensors consistently triggered false alerts (a common problem within video analytics).
- I was not able to test the unit in snow, heavy rain or zero light conditions. Also, the period of the test does not allow me to tell whether seasonal changes would eventually require tuning. [Emza says this is not necessary but I cannot verify]