Before configuring the smart video detection tools (see Creating and configuring the Tracker object), consider the following recommendations that could help to reduce the number of false positives:
- Configure the Tracker object iteratively and check its operation quality at each iteration.
- The parameters that affect the Tracker object operation quality the most are: Minimum size, Maximum size, and Sensitivity.
- It is recommended to set the Sensitivity parameter to zero (left end position of the slider). This will enable automatic sensitivity ranging to ensure reasonable operation quality of the tracker in changing lighting conditions.
- If lighting is stable, it could be reasonable to adjust the Sensitivity parameter manually. To detect the low contrast objects, set the Sensitivity value at about 35, for the high contrast ones—at about 15.
- Minimum size of an object must be selected so that it is slightly smaller than the size of a typical object on the image.
- Maximum size of an object must be selected so that it is slightly larger than the size of a typical object on the image, considering that the object can be joined with its shadow.
- Lowering the Waiting for loss parameter will allow you to exclude triggering on false static objects. Increasing the parameter will allow you, in some cases, not to break off overlapping or temporarily disappearing objects.
- To increase the VMDA detection tools accuracy, use an advanced object classifier, which is configured using the scripts and registry keys—see CAM_VMDA_DETECTOR VMDA detection.
- To increase the VMDA detection tools accuracy, the neural filter can be used. It is trained individually for each use case and provides the high accuracy of the objects classification—up to 99% (see Configuring the neural filter).
- To increase the operation stability of Axxon PSIM, you can separate the VMDA trackers operation into a separate detector_ext.run process using the VMDAEXT key (see Axxon PSIM base version). However, keep in mind that allocating trackers to a separate process increases the CPU and memory resources used.