Page 348 - Big data - Concept and application for telecommunications

6 Big data - Concept and application for telecommunications

8.1.2 Non-real-time data

FUN003: VSBD is required to support analysis of events (intrusion, stranded, etc.) and use of the resulting
information.
FUN004: VSBD is required to support collection of alarming information.
FUN005: VSBD is recommended to support acquisition and analysis of historical case information databases.

FUN006: VSBD is recommended to support acquisition and use of characteristic image databases (crime
picture, etc.).
FUN007: VSBD can optionally support use of digital map information databases.

FUN008: VSBD can optionally support using traffic signal databases.

8.2 Data-processing requirement
Based on computer image-processing technology and image recognition algorithm, a visual surveillance
system based on big data application can have the capability to understand and analyse the content of videos
or images, generate description information from the content in a structured or semi-structured description
language, and provide the basis for video big data analytics and application.

8.2.1 Image processing and recognition requirement

FUN009: VSBD is required to extract key frames from videos and restructure the synthesis of new videos.
FUN010: VSBD is recommended to automatically monitor moving objects in videos, extract object
trajectories, and combine them into concentrated videos optimally.
FUN011: VSBD is recommended to unify the format and structures of object information in the video in a
timely fashion, and extract important information from objects in the video as the meta-data to save, such
as colour, people/car classification and perimeter prevention rules.

8.2.2 Parallel computing or distributed processing requirement

FUN012: VSBD is required to support parallel computing or distributed processing technology, so as to
provide real-time batch processing and extract-transform-load (ETL) operations.

8.3 Data storage requirement
FUN013: VSBD is required to support persistent storage and operations for massive data.
FUN014: VSBD is required to support storage of various types of structured data, including image index
information, vehicle characteristics, human characteristic information, facial feature information and all
kinds of characteristic information for the subject research-analytics.
FUN015: VSBD is required to support the storage of a feature-related object image or video in order to assist
display and analytics in characteristic information retrieval processes.
FUN016: VSBD is recommended to support localized storage with external or auxiliary data sources.

9 Performance requirements

PRF001: VSBD is required to support object detection index: minimum detectable object size, detection rate,
miss rate, error recognition rate and recognition speed.
PRF002: VSBD is required to support analytical prediction: prediction accuracy, mistaken correlated event
rate.
PRF003: VSBD is required to support intelligent analytics of high-throughput real-time video. The average
processing and response time is recommended to be less than 1 s.
PRF004: VSBD is required to support parallel analytics of single videos by loading multiple algorithms.

340 Application