Media Server
Images and video are examples of unstructured information that represent a vast quantity of data. Media Server enables you to maximize the utility of this data by extracting meaningful information about its content.
For example, Media Server can:
- run optical character recognition, to read text in a scanned document or subtitles in a video.
- identify a person who appears in an image or video by matching their face to a database of known faces.
- identify logos and other objects when they appear in images and video.
- identify the exact time of a scene change in video.
- determine whether a video contains speech, and convert the speech into text.
You can deploy Media Server for broadcast monitoring purposes, to identify when individuals or organizations appear in television broadcasts and extract information about these appearances. You might also use Media Server to catalog an existing archive of audio and video clips.
In security and surveillance deployments, Media Server can help security personnel. Human operators can become overwhelmed by the amount of data available from CCTV cameras. Media Server reduces the operator's workload with automatic processing, such as reading the number plates on vehicles and automatically identifying suspicious events.
Media Server can transform the metadata that it extracts into many output formats, and send the data to many systems.
Media Server can be used to analyze images, video, and audio for a custom application that you develop. You can also use Media Server as part of a larger IDOL deployment, to run analysis on files that your connectors extract from your organization's data repositories, from web sites, and from social media. If you index information from Media Server into an IDOL Content text index, you can use IDOL to search within your media. Searches will return images such as scanned documents that contain the search terms. You can search ingested video for specific events, such as the appearance of a particular speaker or discussion of a particular subject. If you are running Media Server for broadcast monitoring you can run sentiment analysis to determine whether the appearance of an individual or organization contained positive sentiment. If you are cataloging clips, you can use IDOL to categorize the clips into a custom taxonomy.
The following sections provide more information about Media Server features.
Ingest Media
Ingestion is the process of bringing media into Media Server so that it can be processed and analyzed. Media Server can ingest the following media:
- image files.
- office documents such as PDF files that contain embedded images.
- video files.
- video from IP streams. Many devices, for example IP cameras, network encoders, and IPTV devices can produce IP streams.
- video from cameras and third-party video management systems.
Analyze Media
Media Server can run many types of analysis, including:
- number plate recognition
- barcode recognition
- color analysis
- face detection, face recognition, demographic analysis and expression analysis
- scene analysis
- keyframe extraction
- object class recognition
- object recognition
- image classification
- optical character recognition
- speech analysis including language identification, speech-to-text, and speaker identification
For more information about the types of analysis that you can run, see Analyze Media.
Encode Media
Media Server can write a copy of ingested video to disk. You can encode video in several formats, segment the video, and change the size and bit rate of the video to create files that are optimized for your use case.
Media Server can also write video to rolling buffers, fixed-size storage areas on disk where the oldest content is discarded to make space for the latest. For example, if you deploy Media Server for video surveillance, you could configure a rolling buffer to store the last seven days of video from a camera.
Media Server can also write images to disk. You can encode copies of ingested images, or use the image encoder to encode still images from video. For example, if you run face recognition on a video file, you might want to encode images of the recognized faces.
You can also create a live UDP stream of the content ingested by Media Server.
Event Stream Processing
You can configure Media Server to filter, deduplicate, and find combinations of events in analysis results. For example, you could use Event Stream Processing (ESP) rules to identify events where the text "Breaking News" appears in a television broadcast and the newsreader speaks the words "election results" within 10 seconds. You can add custom logic by writing scripts using the Lua scripting language.
Output Information
Media Server can output the metadata that it extracts to many formats and systems, including:
- Connector Framework Server (CFS)
- IDOL Server
- Vertica databases
- XML
- Milestone XProtect
OEM Certification
Media Server works in OEM licensed environments.