AI and Video Analytics

Human, Vehicle and Object Detection on Cameras

We test person and vehicle classification together with the chosen zone and direction rules. Camera angle, the target’s size in the image, lighting and occlusion can affect the result. The test report shows true alarms, false alarms, missed events and notification delay separately. With zone, direction, duration and object size rules, we aim to reduce the number of events sent to the operator.

Person classificationVehicle classificationObject classesZone rulesDirection and durationLoitering / abandoned objectPixel densityNight profile

Service definition

What is person, vehicle and object detection, and who is it for?

Object detection is the analytics function that finds an object’s position in the image and classifies it as a person, vehicle, bicycle and so on. Rules such as zone entry, line crossing, loitering longer than a set time or wrong-way movement are built on top of this classification. The model alone doesn’t decide a rule’s success; the target’s pixel size, camera angle, lighting and scene complexity do.

01

Should object detection be enabled on all cameras?

No. Given processing resources and operator workload, it should be enabled only on cameras with a defined scenario.

02

Is abandoned object detection reliable?

It depends heavily on the scene and crowding; in busy areas it’s prone to false alarms. The decision is made from pilot measurements.

03

Do animals trigger alarms?

Models try to distinguish people and vehicles. Even so, large animals or night conditions can cause misclassification.

Project stages

How do we proceed?

01 / STAGE

Scenario

Classes and rules.

02 / STAGE

Suitability

Camera and angle check.

03 / STAGE

Configuration

Rules and profiles.

04 / STAGE

Measurement and tuning

Improvement with pilot data.

Video Surveillance

Know not that something moved, but what moved

Let’s talk about the current situation and goals; we’ll come back with a written, concrete proposal covering the scope and deliverables.