Background and problem
Video surveillance is a good way to keep city safe. For example, when police notice an abnormal event on the video, they can immediately go to the place to solve the problem. However, video surveillance is a time-consuming task, because the police will always need to look through all of the videos carefully. To use the video data effectively, we need to understand the events to which occurred in the video automatically and categorize them to see easily.
Image recognition APIs will be useful to resolve the situation. Those APIs return object names, which the image recognition algorithm detects from the image data. My application shows categories of object names and the details of the results from those APIs.
Police always don't have to look at all of surveillance video using my application, because my application automatically categorizes the events and they look only the events. Therefore police can focus on other tasks to keep their city safe and their city will be more safe than before.
How does it work?
GE Predix has the service to get captured video data from street lights installed in San Diego through public safety API of Intelligent Environments service. Using the image recognition API, my application uses 2 APIs, Computer Vision API from Microsoft Cognitive Services and Watson Visual Recognition API on IBM Bluemix. My application retrieves the information from the video and categorizes the information on web user interface. In the user interface, user can see the details of the recognition results in each video.
Predix services and external APIs
I utilized the following Predix services and external APIs.