Accuware Sentinel is a cameras based Real-Time Locating System (RTLS) that recognizes people inside video streams and is able to follow their movement through indoor and outdoor venues over time. Applying object recognition techniques, Sentinel recognizes human shapes from images in video streams based on their appearance, including features like bulk, clothes color and height. This enables Sentinel to discern people from the background and to enable implementing capabilities such as:

  • people counting.
  • tracking people’s movements.
  • searching for a like match among individuals previously “seen” by the video cameras.

WARNING: note that Sentinel’s functionality is not face recognition but a complementary technique that can recognize and follow individuals even when they face away from the camera.

How it works

Sentinel processes digital video streams from one or more cameras deployed around the monitored area. It is often possible for the system to use cameras already installed in a venue. Recognizing people in video streams is done as follows:

  1. Sentinel analyzes images from the video streams, recognizing people, assigning each of them a unique Person ID, as well as a Feature Vector that encodes the person’s appearance.
  2. The actual image of each individual recognized by the system is saved as a tracklet. All tracklets are collected into a database, each of them identified by a Tracklet ID.
  3. For as long as a person remains in the camera’s field of view, their Person ID remains unchanged. That person’s location and movements are then tracked over time.
  4. Sentinel creates a record for each tracklet. The record includes: timestamp, Frame Number, Person ID, Feature Vector and coordinates which identify the specific tracklet within a frame. This data enables historical search of individuals whose appearance matches those in a given tracklet.
  5. Periodically, Sentinel outputs a CSV file containing data from records of people recognized by the system during the previous time window. The time window covered by a CSV file is configurable. By default, CSV files are output every 5 minutes.
  6. An external system may periodically upload and convert these CSV files issued by Sentinel to build other applications. For example, a dashboard with search functionality based on individuals’ appearances.

Overview of the setup steps

Here is an overview of all the steps that must be performed to setup Sentinel:

  1. Read carefully the introductory page about the features and limits in order to understand what to expect from this technology.
  2. Test the system in few minutes by uploading a sample video to the Accuware Sentinel Test Page.
  3. Plan the installation by computing the correct amount of cameras.
  4. Configure the server that will run the Accuware Sentinel software.
  5. Configure the cameras by making the video streams accessible by the server.
  6. Verify the data collected using the Accuware Sentinel dashboard.
  7. Integrate the data inside your application through the Sentinel CSV files.

No platform specific library or SDK is required. Any Web or Mobile application, capable of parsing CSV files, can benefit of the data produced by Accuware Sentinel.

Test Accuware Sentinel on your video samples: to test the capabilities of the Accuware Sentinel algorithm on some video samples without having to setup a local server please use the Accuware Sentinel – Test Page after having read the information about the features and limits.

Support requests

In order to get the most out of this technology, please read carefully the whole documentation available in the left sidebar. If you have problems or doubts, please submit a support request using this form. We will be pleased to help you achieve your goals and get the most out of the Accuware technologies!