Object detection

Man walking outside a big building with object detection frame around him.

Detect. Recognize. Act.

AI-powered object detection and recognition that turn video into searchable events and actionable insight.

Object detection explained

Two security guards watching surveillance footage with the help of AXIS Analytics used in cameras.

Object detection combines several analytical capabilities. It determines whether an object is present, where it appears, and describes it using object class, object type, and attributes. It can also provide contextual information such as time, location, and movement.

The result: cameras, radar, and other sensors don’t just capture a scene. They generate structured, lightweight data about objects in the scene that systems can use to automate actions or that operators can search and investigate.

The evolution of object detection

Road intersection with multiple cars. AXIS Q1728 with AXIS Object Detection identifying both humans and vehicles.

Two decades ago, object detection meant watching for pixel changes. Shadows, headlights, and weather regularly triggered false alarms, and systems had no way of knowing what they were seeing.

Deep learning changed that. AI models trained on annotated data can now detect, classify, and track objects reliably across a wide range of real-world conditions, not just register that something moved.

At the same time, advances in hardware have moved processing closer to the source. Tasks that once required powerful centralized servers can now run directly on cameras and edge devices, enabling faster responses and more scalable deployments.

How object detection works

When cameras, radars, or sensors are equipped with AI analytics, they can process the data stream in real time. They can classify objects, activities and movement as they occur.

When something relevant happens, the system generates an event based on pre-defined action rules. This event can trigger actions such as starting a recording, activating lights, or a siren, or notifying an operator. Your team is alerted when it matters most.

When analytics run at the edge, processing occurs where the data is captured. This enables faster responses, reduced bandwidth usage, and a system that scales efficiently without imposing unnecessary strain on infrastructure.

Object detection determines whether an object is present in a scene and where it appears.
Object class describes what broad category the detected object belongs to, such as a human, a vehicle, or an inanimate object.
Object type provides a more detailed description within a class. For example, within the vehicle class, types may include cars, trucks, or bicycles.

Object attributes and characteristics provide additional contextual detail about detected objects. For people, this could include features such as clothing color, hats, or bags. For vehicles, it could include attributes such as color, make, and model. These attributes support more precise search, filtering, and analysis and help match or correlate objects across cameras or over time.

Identification differs from detection and classification. In video analytics, it means assigning a unique identity to a specific person or object using a distinct identifier, such as a face or license plate.

How AI improves object detection

AXIS Scene Metadata used to identify metadata of two people walking on a crosswalk as well as moving vehicles.

Traditional motion detection had several limitations: it could, for example, mistake a passing shadow, headlights or a gust of wind for something worth flagging.

AI changes that. Instead of simply reacting to motion or sound, it understands what it’s seeing, hearing or otherwise sensing. It can distinguish a person from a swaying tree branch, track the same vehicle across cameras and perform more accurate classifications in low light, crowded scenes or bad weather. The result is fewer false alarms, more reliable detection, and less noise for your team to deal with.

Practical detection scenarios

Object detection supports a wide range of configurable scenarios, from basic intrusion detection to complex operational workflows. Here are some of the most common.

An intruder trespassing a gated area caught on a security camera with AXIS Object Analytics.

Security and safety

Track security-related events such as:

Intrusion detection
Area protection
Loitering detection
Unauthorized parking detection
Occupancy monitoring
Personal Protective Equipment (PPE) detection

A queue of customers in front of the checkout desk.

Operational efficiency

Measure activity and automate workflows based on:

People counting
Flow monitoring
Occupancy levels
Dwell time analysis
Queue monitoring
Wrong-way detection
Tailgating detection

Town square in Copenhagen with people walking around. Shopping buildings are lined up both to the left and right.

Business intelligence

Aggregated metadata from multiple sensors can be visualized in dashboards to reveal trends, patterns, and anomalies, such as:

Peak and off-peak activity
Occupancy trends
Crowd density
Object types present

AXIS P3268-SLVE mounted on a metal pole inside a food processing plant.

Customizable scenarios

Run multiple detection scenarios simultaneously, including:

Object in area detection
Line crossing detection
Dwell time
Crossline counting
Occupancy monitoring

Beyond camera-based analytics

Cameras are powerful, but not always sufficient. In challenging environments, such as low light, harsh weather, or complex scenes, technologies such as radar and Light Detection and Ranging (LiDAR) complement camera-based analytics. They provide reliable distance and motion data and enhance detection and situational awareness, either independently or alongside cameras.

Radar detection

Radar uses radio waves to detect and track objects, providing information such as distance, speed, and direction of motion.

Because it doesn’t rely on visible light, it performs reliably in darkness, fog, rain, or snow. This makes it well-suited for perimeter protection and traffic monitoring across large areas.

Radar doesn't capture visual detail. Its accurate motion detection and speed measurement make it a strong complement to camera-based analytics.

LiDAR remote sensing technology

AXIS Q1686-DLE mounted on top of a highway to monitor traffic and hazardous vehicles during nighttime.

LiDAR sensors use laser pulses to measure distance and create a three-dimensional representation of the surrounding environment.

By calculating how long it takes for pulses to return, LiDAR generates precise spatial data about objects, including their shape and position.

This makes it particularly useful where accurate depth and spatial awareness are critical. Examples are traffic systems, industrial automation and advanced monitoring. Combined with camera-based analytics, LiDAR can enhance detection accuracy and provide a more complete picture of a scene.

What object detection and analytics enable

An analytics solution can do more than monitor. It can also trigger actions based on predefined rules. Here’s what that means in practice.

Faster response and improved situational awareness

The system can notify operators without requiring continuous screen monitoring. For instance, when a person enters a restricted area, a queue exceeds a defined threshold, or a vehicle crosses a defined line. Events are generated when something relevant happens, helping teams focus on what matters.

Fewer false alarms

By distinguishing relevant events from background activity, object detection helps reduce unnecessary notifications. This leads to a more focused workflow. It allow operators to spend less time filtering out noise and more time responding to what matters.

Faster search and investigation

Investigators can search by object class, object type, attributes, and other parameters such as time, location, or direction of movement, instead of manually reviewing hours of footage. This enables them to find what they need in seconds rather than hours.

Improved security, safety and operational efficiency

Object analytics supports a wide range of scenarios, helping your organization automate processes, reduce risk and make better decisions. From perimeter protection and access control to occupancy monitoring, counting, and workflow optimization.

Exploring object detection solutions

AXIS Object Analytics

An AI-powered multi-purpose analytics that detects, classifies, tracks, and counts objects directly on compatible Axis devices, at the edge. Configure detection scenarios, set up real-time alerts and access structured insights.

A black car exiting a dark parking garage.

AXIS License Plate Verifier

AXIS License Plate Verifier identifies license plates in moderate and high-speed traffic. It supports traffic management, access control, parking, and vehicle search. It also recognizes vehicle type, color, make, and model for more precise identification.

A woman captured on camera with AXIS Analytics walking in a airport with her luggage in hand.

AXIS Scene Intelligence

Axis Scene Intelligence combines AI-driven analysis with advanced imaging to turn cameras into intelligent tools, automating actions, enabling fast search and delivering insight that scales with operational needs.

Object detection across segments

Analytics and object detection deliver value across a wide range of industries. Here are some examples.

AXIS P1518-LE mounted high in a traffic area in the city during dusk.

Smart cities and traffic management

Traffic doesn’t manage itself, but analytics can help optimize how it flows. By detecting and classifying vehicles, monitoring pedestrians, and identifying congestion patterns, cities gain critical insights. This data enables dynamic signal control, faster incident response, and smarter long-term infrastructure planning.

A female customer scanning her goods in the self-checkout. A staff member is in the background of the grocery store.

Store performance and optimization

Understanding how customers move through a store is the first step to improving their experience. Analytics count visitors, analyze flow and monitor queue lengths, giving retailers the data to optimize layouts, staffing and operations, and spot trends before they become problems.

AXIS Q3546-LVE with a weathershield mounted on a pole inside a fenced area.

Perimeter security

At airports, industrial sites, and data centers, you can't afford to miss what matters. Analytics detect and classify objects, track movement across defined zones, and surface events that need attention, so operators stay focused on genuine threats, not noise.

Close up of a worker holding a tablet in an assembly line.

Manufacturing and production

Downtime is expensive. Analytics help monitor production, detect anomalies, and support safety by identifying line stops, misplaced objects, or unsafe entry into restricted areas. Edge-based processing flags issues the moment they occur, before they escalate.

Implementation considerations

Great analytics start with great deployment. Get these factors right from the beginning, and your system will deliver reliable results today and scale with your needs tomorrow.

Edge, server, cloud, or hybrid – where should data be processed?

AXIS Q6325-LE PTZ Camera mounted in a city square. Big building to the left and open area to the right.

Where your analytics run affects everything – how fast your system responds, how much bandwidth it uses and how well it scales. Most deployments combine multiple approaches.

Edge-based analytics run directly on the camera or device, processing data at the point of capture. This enables real-time detection and response, reduces bandwidth and storage requirements, enhances data privacy, and supports scalable, resilient system design.
On-prem server-based analytics process data centrally from multiple cameras and sensors, enabling system-wide coordination and analysis. This makes them well-suited for large deployments or scenarios that require cross-device insights and more compute-intensive processing.

Cloud-based analytics offer flexibility and scalability, making it easy to scale and access data across sites. They typically require stable connectivity and sufficient bandwidth, particularly for real-time or data-intensive use cases. Processing large volumes of data entirely in the cloud can also increase bandwidth and storage costs.

In practice, hybrid architectures are often the preferred approach. Edge analytics enable real-time detection and response directly on the device, while server or cloud solutions support more advanced analytics across sites. Together they provide a scalable, flexible architecture that balances performance, cost, and operational needs.

Environmental conditions and accuracy

Even the most advanced AI analytics depend on the right foundation. Performance is shaped by the entire solution – sensor quality, image technology, system-on-chip (SoC), and the device’s placement and configuration.

Get these fundamentals right from the start, and the system can handle challenging conditions such as busy scenes, vibration, and variations in angle, scale, and partial visibility more reliably.

In demanding environments, additional sensors such as radar or LiDAR can complement camera-based analytics, adding an extra layer of robustness where needed.

Explore AXIS Scene Intelligence

Metadata and system integration

Object detection delivers full value when it connects to the systems that act on the data. Open standards and structured scene metadata make it straightforward to integrate with access control, alarm systems, and business intelligence tools. This enable detections to automatically trigger the right response.

Privacy and responsible use

Analytics that process video and audio data – and therefore potentially personal data – come with responsibility. Edge processing helps limit unnecessary transfer of personal data, while privacy masking supports local regulations. The goal is always to assist human decision-making, not replace it.

Explore AXIS Live Privacy Shield

Scalability and long-term flexibility

As needs change, your analytics solution should be ready to adapt. Scalable architectures, flexible deployment options, and support for future software updates protect your investment and make expansion straightforward as your operations grow.

Future trends in object detection and intelligent video analytics

AI, connectivity and edge computing are pushing analytics beyond detection, toward systems that understand, anticipate, integrate and adapt. Here's where things are heading.

Predictive analytics
Aggregated metadata can be used to identify patterns over time, detect anomalies earlier, and flag risks before they escalate. The shift is from reacting to events to getting ahead of them.

More efficient processing
Advances in computational power, specialized hardware, and model optimization mean that sophisticated analytics can now run on a wider range of devices. This makes intelligent monitoring accessible across more environments, not just in large, well-resourced deployments.

Integration with IoT systems
Analytics are increasingly integrated across multiple data sources and sensors, including cameras, environmental sensors, and acoustic sensors. Correlating these sources provides broader situational awareness and enables faster, more informed responses.

Privacy and responsible use
As capabilities expand, responsible deployment becomes increasingly important. Future systems will rely on privacy-by-design principles, including anonymization, metadata-driven workflows, and limited data sharing. Processing data at the edge can also help reduce exposure to sensitive personal information and support regulatory compliance.

AV1 and efficient video streaming
Modern video encoding standards such as AV1 significantly reduce bandwidth and storage requirements while maintaining high image quality. This enables more efficient video transmission and storage, making it easier to scale analytics across sites and systems without increasing infrastructure costs.

The object detection orchestra

We put our AI-based analytics technology to the test by forming the world’s first orchestra of video surveillance cameras. Experience their epic performance and see the tech in action.

A city over the canal viewed from a magnifying glass during sunset. Everything outside the magnifying glass is blurred.