casestudy : Rail
: OV at Industry Events
casestudy : Waterside Security
Learn More
What portion of the digital signal processing (DSP) chip is utilized by ObjectVideo OnBoard?
What DSP chip is required for ObjectVideo OnBoard?
Where is the rule inference performed in ObjectVideo OnBoard?
With all the video analysis and rule inference performed on the DSP, is there a limit on the number of rules?
What is the output of the ObjectVideo OnBoard library? What type of bandwidth is required to support this output?
How much memory is required to support the ObjectVideo OnBoard application?
Q: What portion of the digital signal processing (DSP) chip is utilized by ObjectVideo OnBoard™?
A: ObjectVideo OnBoard does not require a dedicated DSP.
- Original equipment manufacturers (OEMs) are able to run additional applications on the same DSP running the OnBoard software without impacting the performance of ObjectVideo’s library.
- When operating the full set of OV capabilities, ObjectVideo OnBoard requires less than 25 percent of a 600MHz DM642 (assumes processing a CIF or 320x240 video feed at 10 frames per second).
- If all OV features and functions are not needed, the DSP utilization drops to less than 20 percent.
Q: What DSP chip is required for ObjectVideo OnBoard?
A: OnBoard is designed and optimized for performance on the Texas Instruments DM64x+ DSP family.
- OnBoard is also compatible with any chip in the TI TMS320DM64X+ series, including the DaVinci™ technology platform.
- Development and performance requirements distributed by ObjectVideo are based on operation on the 600MHz DM642.
- Operation on a 720MHz or 1GHz DSP will result in even better performance.
Q: Where is the rule inference performed in ObjectVideo OnBoard?
A: OnBoard performs all video analytics and rule inference on the DSP. ObjectVideo does
not require additional backend servers for analysis or application of rules, filters, schedules, etc. This architecture enables truly intelligent devices that take in video and output alerts or other intelligent data.
Q: With all the video analysis and rule inference performed on the DSP, is there a limit on the number of rules?
A: There is
no limit on the number of rules or rule types.
- While each rule adds some additional processing and system memory requirements, the amounts are negligible and do not present any impact to users looking to activate a wide set of rules for concurrent real-time detections.
- ObjectVideo OnBoard includes a core set of DSP-optimized algorithms that initially analyze video independent of the rules created. As a result, activating new ObjectVideo rule types (e.g., tripwire, loitering) does not demand incremental add-on software or processing power.
Q: What is the output of the ObjectVideo OnBoard library? What type of bandwidth is required to support this output?
A: Because OnBoard performs all video analytics and rule inference on the DSP at the edge device, OnBoard outputs the “alerts” or behaviors detected by the system.
- An alert typically consists of alert data (e.g., camera name, rule violated, response message), and a pre-event and post-event alert image.
- Alert images include the rule overlay and object mark-up.
- The alert packet size, assuming two alert images, is approximately 80 KB.
- Alerts can also be configured to send alert data only, which is approximately 15 KB.
Q: How much memory is required to support the ObjectVideo OnBoard application?
A: The OnBoard library uses 18 to 24 MB (configurable) per CIF video channel. Therefore, ObjectVideo recommends at least 32 MB SDRAM per CIF video channel processed per DSP.