Make sure the following requirements for abandoned object detection are met:
- Camera requirements:
- Resolution must be not less than 640х480 pixels.
- Color: analytics works both with color and monochromatic images.
- It is not allowed to the image shift be more than 1% from the frame size due to camera shake.
- Lighting requirements:
- Moderate lighting. The quality of analytics performance can be lower in case of low lighting (night) or over lighting (overexposure).
- There must be no dramatic changes in lighting.
- Scene and camera angle requirements:
- Background is mostly static and is not changed.
- Incorrect analytics performance can be caused by reflective surfaces.
- Abandoned objects are hidden by moving objects not longer than 10% of the time.
- Object images requirements:
- An abandoned object must be visible on the image.
- The digital noise and compression-related artifacts are minimal.
The width and height of the objects in the image must be not less than 1% of the frame size (if resolution is over 1920 pixels) or 15 pixels for lower resolution.
The width and height of the objects in the image must not exceed 25% of the frame size.
If these conditions are in place, the Abandoned Object detection tool is guaranteed to:
- Detect 92 items out of 100
- Keep false positives to 20 out of 100.