Mike A.
Known around here
- May 6, 2017
- 4,163
- 6,969
Dahua's guidance for IVS:
I think at least part of it is the complexity of the scene and little residence time for recognition given that as they pass through quickly. Also geometry as far as how the box/lines are drawn vs the scene.
Try making a zig-zag tripwire with a little more horizontal coverage angling from the front of the mat up to the top of the door.
Try setting target filter with human turned on.
- The target should be less than 10% of the image.
- The target size in the image should not be less than 50×50 pixels.
- Brightness difference between the target and the background should not be less than 10 greyscale.
- The target should be continuously present in the image for no less than two seconds, and Event 134 the moving distance of the target should be larger than its width and no less than 15 pixels (CIF image).
- Reduce the complexity of surveillance scene as possible as you can. Intelligent analysis function is not recommended to be used in scene with dense targets and frequent illumination change.
- Avoid areas such as glass, reflective ground, water surface, and areas interfered by branch, shadow and mosquito. Avoid backlight scene and direct light.
I think at least part of it is the complexity of the scene and little residence time for recognition given that as they pass through quickly. Also geometry as far as how the box/lines are drawn vs the scene.
Try making a zig-zag tripwire with a little more horizontal coverage angling from the front of the mat up to the top of the door.
Try setting target filter with human turned on.