Back to overview

How video analytics led to enhanced user notifications without having it to cost a fortune

Amsterdam, December 3, 2014

Cloud-based video surveillance has gotten more and more advanced over the last few years. Thanks to further developed video analytics, video surveillance is not depending anymore on people constantly monitoring video streams. Video analytics are already implemented in a mixture of embedded equipment such as: (IP) surveillance cameras, DVR’s and video servers.

This article shows how video analytics have improved the surveillance industry, and how we, at Panasonic Cameramanager, revolutionized the surveillance industry by applying human- and object detection to our cloud-based video surveillance solution.

Understanding the content of video
Video analytics, in general, involves analyzing video images automatically to define its content and extracting meaningful or relevant information. It uses computer vision algorithms which enables a system to perceive or see. Video analytics are not only interesting for the surveillance industry, but are also implemented in health care systems and the automotive-, transport- and retail industry. Think about situations where people counting, heat mapping, etc. are relevant.

There are multiple different functions in the category ‘video analytics’: motion detection, facial recognition, object detection, shape recognition, and many more. However, the concept of video analytics is not only referring to the detection or recognition of motion or objects. It refers to the understanding that motion is caused by an object moving around a static background and the capability of tracking the cause of motion around the scene or classify the type of motion as a human, object or animal.

Segmentation versus classification
Segmentation and classification are two important aspects within video analytics. Segmentation refers to the process of distinguishing differences, whereas classification is the process in which these differences are qualified in separate categories. Within the classification process there are two types of distinguishing differences: binary and multiclass classifiers.

While binary classifiers can distinguish a human blob (a collection of connected pixels) from a non-human one, multi class classifiers can separate non-human blobs from each other and placing them in other categories like objects, animals, etc. This gives the surveillance industry substantial reasons to implement video analytics for multiple purposes, but also make video analytics interesting for other uses for instance in the retail industry where analytics can serve purposes like people counting, age- and gender recognition, etc.

Developing CLVR to meet the needs of the market
At Panasonic Cameramanager, we have combined binary classification aspects to develop innovative cloud-based video analytics software that make it possible for cameras to distinguish humans from other objects. We do so under the name CLVR (pronounced as ‘clever’).

Implementing this smart analyzing software onto our cloud service, makes it possible to detect more relevant events than mainstream motion detection features. With CLVR we are able to filter out ‘motion’ that is detected by, for instance, lights switching on/off in an office environment or trees swooping in front of a window. Instead, only events that contain a human or object are registered and sent as an alarm to the user.

Applying features like CLVR make it possible to instantly receive relevant notifications via the cloud when a human or object is detected, without any recording devices or onsite servers needed to conduct the algorithms. We do so by combining the motion detection filter from our Cloud Camera with an on-cloud algorithm. With CLVR users will receive 66% more relevant notifications as opposed to earlier motion detection solutions. This way, bandwidth use is dramatically decreased. Which in the world of cloud-based surveillance solutions means less storage & less costs, because the camera only needs to record when triggered by motion making continuous recording to the cloud unnecessary.

Continue reading about video analytics and CLVR in our white paper.