The 4 Best Practices for Intelligent Video Applications
There are many variables to either implementing intelligent video in to your surveillance system or using it in practical applications.
To help sum up the recent intelligent video posts here now are the 4 Best Practices for IV Applications:
- Video Image Quality
- Efficient Algorithms
- Computer Processing Power
- System Configuration & Fine Tuning
Video Image Quality
In any kind of surveillance application getting the right video from your cameras is key and no more so than in intelligent video applications. There are 3 main factors to consider when working with image quality:
- Frame Rates & Resolution: Believe it or not, the highest frame rates and resolution aren’t always necessary. Rule of thumb is that 5 to 10 frames per second (FPS) in CIF resolution is best for most general applications. License plate or facial recognition may need even lower frame rates but work best with higher resolutions.
- Camera Position: Simply put – install your cameras in the best positions for the best results. People counting will need overhead cameras, license and facial recognition need frontal views, etc. Dedicate your cameras for your application’s specifications.
- Video Type: Consider the type of video that’s best for your IV application. Most IV software works best with uncompressed video to avoid loss of data. However, some applications might work better with black and white pictures versus color, or even a thermal image might be well suited distinguishing people and objects.
Complex mathematical alogorithms are the driving force behind processing video and images for IV applications. How well your application works depends on how well these algorithms perform. Only field testing your application can determine its accuracy.
There are bound to be “false positives” in intelligent video applications. That is, mistaking a harmless object or event for whatever it is that you’re on the lookout for. Therefore, 90% accuracy is most realistic number you can bank on.
Computer Processing Power
Because these mathematical algorithms are so complex they need a lot of power to operate. Their performace then is also dependent on the processors you’re using and the amount of available memory.
Generally speaking, the more computer processing power to run your IV application the better it will perform.
System Configuration & Fine-Tuning
Since every application is different you will need to configure and fine-tune the system based on your specific needs. No system is perfect but the goal should be to reduce the amount of false alarms to a minimum. Keep these things in mind when setting up your system:
- The system will never be 100% accurate.
- Optimizing the system takes longer when your application calls for many criteria to be accounted for.
- Changes in lighting will impact your results – monitor and ajust the system over a full 24 hour period.
- Exercise patience! Fully optimizing your system could take days or weeks depending on your application.