The 3D calibration determines the sizes and is essential for classification and a number of filters. After calibration the video analytics has all references for determination of sizes depending on distances.
Determination of object sizes, depending on their area and speed is very important for many applications and contributes to more effective use of the system. After fixing the individual classification the system differentiates by i.e. animals, motorcycles, trucks and pedestrians or pedestrian groups.
3. Choosing the right Filter
There are various filters or filter combinations so individual filter can be used for any desired application. Think about effect and use of filters beforehand, in order to get the most efficient video analytics. Through a process of elimination you can satisfy even the most difficult conditions; planning is important!
4. Test Running
After calibration, classification and selection of filters test the result for a sufficient period of time! There will always be movements, processes, environmental influences, unforeseen events and not considered alarm trigger criteria, that needs to be remedied. Video analysis lives on it – and the more effective it becomes. Only after a thorough examination and test run one can assume that video analytics and alarming works correctly and provides what people understand under a good detection.
If your installation is outdoors make sure the system can handle environmental phenomena such as changing in lighting conditions, wind, rain and fog without generating false alarms or going ‘blind’. Ensure that sudden scene changes does not generate false alarms. Check that the system can learn the new scene following a sudden scene change quickly. Any longer than 20 seconds or so and there is a real possibility of missed detections.
Rules to improve precision and reliability
Too low installation or places where objects pass detection range to close should be avoided, as well as extreme backlighting. A reliable people counting will be achieved by directing the camera at a relatively steep downward angle.
Light chances and Light Amount
Video analytics is also working at very low light condition, however the result will not be as satisfactory as in bright light. It depends also on the type of detection. Also test the conditions as they will likely be in practice before the system is used.
RIVA Analytics is a learning system that adapts to environmental influences, rippling water, swaying trees, shadows and extreme lighting conditions. Nevertheless these disturbing influences should be within reasonable limits. If a false alarm by excessive movements cannot be avoided, the most sensitive areas should be marked in ignored.
Lens Selection and Observation angle
Choose a lens that offers a larger viewing angle: The system must be given sufficient time to detect objects properly. Detection of a passing car from a distance of 10 m with a focal length of 25 mm makes no sense; the result will be disappointing in any case.
RIVA Analytics is working with very precise algorithm; the system still has its limits, which become more closely with larger distances. Distant objects might still be detected, but do provide sufficient size to be successfully detected or classified. Please pay attention of minimum sizes, possibly change focal length.