Question
What to do when MIC IP camera is unable to perform the Object Classification in IVA detection?
Answer
Based on examples of target objects and non-target objects, the Camera Trainer program uses machine learning to allow the user to define objects of interest and generate detectors for them.
Using Configuration Manager, you can configure the Camera Trainer program using both live video as well as recordings available through your MIC IP cameras.
The camera also includes the latest release of Intelligent Video Analytics application for use on prepositions. Designed for mission-critical applications, the video analytics can reliably detect, track, and analyze moving objects while suppressing unwanted alarms from spurious sources in the image, even in harsh weather conditions. Advanced tasks like multiple line crossing, loitering, idle / removed object detection, crowd density estimation, occupancy and people counting are available for live alarming and forensic search. Object filters based on size, speed, direction, aspect ratio, and color can be defined.
In order to perform Object Classification correctly, please make sure that you have correctly:
set the correct Orientation
Save all the pre-positions again if the orientation stetting is changed
calibrated the camera
A simplified calibration mode reduces installation time significantly because you only need to enter the installation height once for each camera, independent of pre-positions.
After the camera is calibrated, the analytics engine can automatically classify objects as upright person, car, bike, or truck.)
enabled 3D tracking
set it up the tasks and followed the VCA manual.
Please see an overview of object classification in the following article:
How to set up object classification with video analytics
View full article