Container counting Drone Guide

# Container Counting Drone Guide

Introduction

Container counting is a practical drone application for ports, inland terminals, depots and large logistics yards where operators need a rapid visual understanding of how many containers are physically present across a site.

Large terminals can contain thousands or even tens of thousands of containers distributed across multiple blocks, stack heights and operating zones. Terminal Operating Systems normally remain the authoritative source for inventory, but physical conditions can change quickly. Containers are constantly arriving, being moved, stacked, loaded, discharged and repositioned.

Drones provide a fast aerial method for creating an independent visual snapshot of the yard. High-resolution RGB imagery can be analysed manually or with AI to estimate the number of visible containers, while OCR can assist with reading container numbers where image quality and viewing angle allow.

The strongest use case is not replacing the terminal inventory system. It is comparing the digital inventory with the physical yard. If the terminal system indicates that 6,400 containers should be present, a drone survey can help determine whether the visible arrangement broadly supports that record and highlight areas requiring investigation.

Container counting can also support yard-capacity planning, congestion monitoring, inventory reconciliation, insurance surveys, depot audits and operational reporting.

Accuracy depends heavily on stack geometry, visibility, image resolution, flight planning and the quality of the AI model. Containers hidden beneath other units cannot be individually seen from directly overhead, so effective counting frequently requires a combination of stack detection, side imagery, known container dimensions and terminal data.

Why Use Drones for Container Counting?

Traditional container counts may involve terminal-system records, gate transactions, crane data, ground inspections and manual yard checks. These methods remain fundamental, but visually verifying a very large terminal can be time-consuming.

A drone can survey a large yard from above and provide a complete visual overview in a relatively short mission.

This is particularly useful when management needs a current estimate of yard occupancy or wants to verify a section of the terminal following an inventory discrepancy.

The aerial perspective also makes it easier to understand how containers are distributed across blocks and where congestion is developing.

Repeated surveys allow operators to compare physical yard conditions over time rather than relying on isolated inspections.

Container Yard Overview

Before individual containers are counted, the drone provides a useful high-level picture of the terminal.

Operators can see which blocks are heavily occupied, where empty slots exist and whether specific areas are being used differently from the planned yard layout.

This information can support operational planning even if the exact individual container count is obtained from the terminal management system.

Aerial mapping can also help identify temporary stacks, overflow areas and containers stored outside normal zones.

Top-Down Container Detection

The simplest approach uses near-vertical RGB imagery to identify visible container roofs.

Because standard ISO containers have broadly consistent dimensions, computer vision can detect rectangular patterns and estimate the number of visible top units.

This works well for single-layer storage areas or where stacks are relatively uniform.

The major limitation is obvious: a stack that is five containers high may appear as only one container from directly overhead.

Top-down detection alone therefore counts visible stack positions rather than total containers unless stack height information is also available.

Stack Height Estimation

To estimate the total number of containers, the system may need to determine how many units are stacked vertically.

This can be approached using oblique imagery, 3D reconstruction, LiDAR or known terminal stack information.

A photogrammetric model can provide an estimate of stack height. Because standard container heights are known, software may infer whether a stack contains one, two, three or more levels.

The calculation should account for terrain elevation and container type.

Errors can occur if containers have different heights, are stacked unevenly or if the surface model contains noise.

Combining Top-Down Count and Stack Height

A more advanced container-counting workflow combines the number of visible stack positions with estimated stack height.

For example, if software identifies 500 occupied ground positions and determines the approximate number of vertical tiers across each position, it can estimate the total number of containers.

This provides a more realistic yard count than simple roof detection.

However, the result should be presented with an accuracy or confidence range rather than assumed to be perfectly exact.

Oblique Imaging

Oblique imagery captures the sides of container stacks rather than only the roofs.

This can make individual tiers easier to recognise.

Side views may also reveal container doors, identification numbers, shipping-line logos and structural gaps between units.

A flight programme may therefore combine vertical mapping passes with selected oblique passes around stack blocks.

This provides stronger information for both counting and identification.

AI-Based Container Detection

AI can significantly reduce the amount of manual work involved in analysing terminal imagery.

Computer vision models can be trained to identify container roofs, side panels, stack boundaries and individual units.

The model can then process hundreds or thousands of images and create a provisional count.

AI is particularly useful for large terminals because manual counting from imagery becomes increasingly time-consuming as yard size grows.

The model should still be validated against real terminal data.

Container Segmentation

Rather than simply detecting rectangular objects, advanced AI systems can segment individual containers within an image.

This means the software identifies the boundary of each visible unit.

Segmentation may improve counting in dense yards where stacks overlap visually.

Shadows, cranes, vehicles and adjacent container blocks can still create classification errors.

OCR and Container Numbers

Container identification numbers can add another layer of value.

ISO containers normally display unique identification markings on their doors and sides.

High-resolution imagery and OCR software may allow some of these numbers to be captured automatically.

This can help connect visual observations with the terminal database.

OCR is most reliable when the container face is clearly visible, the camera angle is favourable and the marking is clean.

Limitations of OCR

Container OCR should not be assumed to work perfectly.

Numbers may be obscured by other containers, dirt, damage, shadows or poor lighting.

Aerial viewing angles may also make characters difficult to read.

AI can occasionally confuse similar characters such as zero and the letter O, or five and the letter S.

Any automatically recognised container number should therefore be validated before it is used for an important operational decision.

Matching Drone Data with the Terminal Operating System

The greatest value comes from linking drone observations to the Terminal Operating System.

The TOS already contains information about expected container location, status and movement.

Drone imagery provides the physical visual layer.

Software can compare what the system says should be present with what the aerial survey appears to show.

This creates an inventory reconciliation process rather than a completely independent inventory system.

Inventory Discrepancy Detection

If the expected and observed conditions differ, the system can flag the relevant yard block.

For example, a block may physically contain more stack positions than expected, or a container may appear to be stored in a location not reflected in the database.

The discrepancy should then be investigated using terminal records and ground verification.

The drone should not automatically alter the official inventory.

Empty Container Depots

Empty-container depots are particularly suitable for drone counting because large numbers of similar units may be stored in dense stacks.

Operators often need to understand depot occupancy and available capacity.

Aerial surveys can provide a rapid estimate of how many containers are present and how much storage area remains.

Oblique imagery can help estimate stack height.

Integration with depot management systems improves accuracy.

Inland Container Terminals

Container counting is not limited to seaports.

Rail terminals, inland ports and logistics hubs may also store large numbers of containers.

Drones can survey these facilities using similar methods.

Rail operations, overhead power lines and moving equipment should be considered carefully during flight planning.

Intermodal Yards

Intermodal sites may contain containers, swap bodies, trailers and other cargo units.

Computer vision models should be trained to distinguish between these categories.

Otherwise, trailer roofs or other rectangular objects may be incorrectly counted as containers.

Human review can help validate ambiguous detections.

Yard Occupancy Measurement

In many cases, the most valuable metric is not simply total container count but yard occupancy.

The drone can help determine how many storage positions are occupied compared with the available capacity.

This provides management with a clear visual picture of congestion.

Occupancy information can support decisions about stack strategy, overflow areas and equipment allocation.

Capacity Planning

Historical drone surveys can show how yard occupancy changes over weeks or months.

This can help identify recurring peaks and provide evidence for infrastructure expansion.

The imagery may also show whether some areas are consistently underused.

This makes container counting useful for strategic planning as well as daily operations.

Congestion Monitoring

When terminal occupancy becomes high, operational efficiency can decline.

More container reshuffling may be required, and vehicle movements can become increasingly complex.

A drone survey can show where dense stacks are developing and whether certain yard blocks are approaching practical capacity.

This should be used alongside operational data from cranes and vehicles.

Container Block Counting

Large terminals are normally divided into defined blocks.

Instead of generating only a site-wide total, drone analytics can calculate counts by block.

This makes results much more useful.

Operations teams can compare occupancy between different sections and identify local discrepancies.

Container Type Classification

Computer vision may also assist with broad classification.

The system may distinguish standard dry containers, refrigerated containers, tank containers or open-top units where the visual characteristics are sufficiently clear.

This is more difficult than simple counting and should be treated cautiously.

Official container type should come from logistics records.

20-Foot and 40-Foot Containers

Different container lengths affect counting.

A 40-foot container occupies approximately twice the longitudinal space of a 20-foot unit.

Software can use visible dimensions and yard-slot geometry to help distinguish them.

Perspective distortion and partial obstruction can create errors, so the result should be compared with terminal records.

High-Cube Containers

High-cube containers are taller than standard units.

This matters when stack height is estimated using 3D geometry.

A system assuming every container has exactly the same height may overestimate or underestimate vertical tiers.

Known container-type information from the TOS can improve the calculation.

Reefer Container Counting

Reefer yards may need dedicated counting because refrigerated containers often occupy specific powered positions.

Aerial imagery can show the number of visible units and general occupancy of the area.

The drone cannot confirm whether each reefer is connected correctly or operating at the required temperature.

This information should come from the terminal's reefer-monitoring system.

Damaged Container Identification During Counting

The same imagery used for counting may reveal visible damage.

AI could flag containers with obvious roof deformation, open doors or unusual positioning for human review.

This provides additional value from the same survey.

Visible external damage does not determine whether the cargo inside has been affected.

Misaligned or Unusual Stack Detection

AI can identify containers or stack sections that appear visually different from surrounding units.

This might include an unusual gap, shifted unit or irregular alignment.

Such observations may warrant closer ground inspection.

The aerial system should not independently determine whether the stack is unsafe.

Photogrammetry

Photogrammetry can support both counting and stack-height estimation.

Overlapping aerial images are processed into a 3D surface model.

The resulting geometry allows software to estimate the height of container stacks above the known yard surface.

Photogrammetry also produces an orthomosaic that can be used for visual review.

LiDAR

LiDAR provides another method of measuring stack geometry.

It can create dense 3D point clouds showing the shape and height of container blocks.

LiDAR may be particularly valuable where strong geometric repeatability is required.

It is normally more expensive than a standard RGB mapping payload, so the business case should reflect the required level of accuracy.

RTK and PPK

RTK and PPK positioning improve the geospatial consistency of repeated surveys.

This is useful when container blocks need to be matched accurately with terminal grid locations.

Precise positioning also improves the alignment of drone results with existing GIS and terminal maps.

Digital Yard Maps

A drone survey can be displayed over a digital map of the terminal.

Each container block can then be associated with an estimated count or occupancy percentage.

This provides managers with a visual dashboard rather than only a numerical report.

GIS Integration

GIS allows the terminal to combine container counts with roads, cranes, rail lines, gates and infrastructure.

The same environment can also contain inspection, security and construction data.

This makes the drone survey useful beyond the immediate counting task.

Automated Change Detection

Repeat surveys can show how container distribution has changed since the previous flight.

Software may highlight areas where stacks have grown, reduced or moved.

This can support inventory reconciliation and operational analysis.

Because normal terminal activity creates constant changes, the system should distinguish between expected movement and unexplained discrepancies.

Drone-in-a-Box

Automated drone stations may be useful for regular container counting at large terminals.

A drone can perform a predefined mapping mission at selected times and automatically upload the imagery for processing.

This allows the terminal to create recurring physical snapshots without requiring a manual launch every time.

Operations must still be coordinated safely with cranes, vehicles and other terminal activity.

Scheduled Counting

Some facilities may benefit from daily or weekly counts.

Others may only need periodic audits or surveys during high-congestion periods.

The appropriate frequency depends on the purpose.

If the TOS is functioning accurately, continuous aerial counting may provide limited additional benefit.

Targeted surveys often produce better value.

Event-Triggered Counting

A drone survey may be performed after a system outage, inventory discrepancy, storm, operational disruption or terminal handover.

This can quickly establish a new physical baseline.

Counting During Low-Activity Periods

Surveying during quieter operational periods often improves both safety and data quality.

There are fewer moving cranes and vehicles, and container positions are less likely to change during image capture.

Early morning, scheduled maintenance windows or other low-activity periods may therefore be suitable.

Moving Equipment

Straddle carriers, reach stackers, gantry cranes and terminal tractors can appear within images.

AI needs to distinguish this equipment from containers.

Moving machinery may also create photogrammetry artefacts.

Flight planning should account for terminal operations.

Crane Obstruction

Large gantry cranes may obscure sections of container stacks.

Different camera angles or repeated passes can reduce the problem.

A completely unobstructed aerial view may not always be possible.

Shadows

Tall stacks and cranes create strong shadows.

These can reduce object-detection performance.

Survey timing and camera exposure settings can improve consistency.

Container Colour and Appearance

Containers vary widely in colour, age and condition.

AI systems should therefore be trained on diverse real-world examples.

A model trained only on clean, brightly coloured containers may perform poorly in an operational terminal.

Weather

Strong wind, rain and fog can prevent reliable mapping.

Poor lighting may also reduce OCR performance.

Repeatable counting programmes should have suitable weather criteria.

Saltwater and Industrial Environment

Ports expose drones to salt air, dust and industrial contamination.

Regular maintenance is important where counting missions are repeated frequently.

GNSS and Steel Infrastructure

Large ships, cranes and steel container stacks can affect local navigation conditions.

RTK can improve positional accuracy but does not eliminate every operational challenge.

Operators should maintain appropriate margins around structures.

Thin Obstacles

Crane cables, wires and antennas may be difficult for obstacle-detection systems to identify.

Automated missions should be planned with this in mind.

Airspace and Port Permissions

Container terminals may be close to airports or heliports.

Routine automated operations may therefore require aviation permissions and coordination with the port.

Flight planning should also account for local operational restrictions.

Privacy

Container counting normally focuses on cargo rather than people.

However, aerial imagery may still capture workers and vehicles.

Data collection should remain proportionate to the operational purpose and comply with applicable privacy requirements.

Cybersecurity

Container data can be commercially sensitive.

High-resolution yard maps combined with inventory information should therefore be protected.

System access, cloud processing and data sharing should follow suitable cybersecurity controls.

Accuracy Measurement

A professional counting programme should measure its own accuracy rather than simply assume the AI is correct.

A sample of drone results can be compared with verified terminal records or manual counts.

Metrics might include total-count accuracy, container-detection accuracy, OCR accuracy and stack-height classification accuracy.

This allows operators to understand whether the system is suitable for operational, planning or audit purposes.

Confidence Scores

Instead of presenting every detected container as equally certain, software can assign confidence scores.

Low-confidence detections can then be reviewed by a human operator.

This is particularly useful around partially hidden containers or complex stack edges.

Human Verification

Human review remains an important part of container-counting workflows.

Operators can examine flagged sections and correct obvious AI errors before the report is finalised.

This hybrid approach often provides better reliability than fully automated counting.

Reporting

A container-counting report should state the survey date, terminal area, flight conditions, sensor, processing method and expected accuracy.

The report can provide totals by yard block, estimated occupancy and a map showing counted areas.

Any sections with poor visibility or uncertain results should be clearly identified.

If the result is an estimate rather than a verified inventory, the report should say so.

For example, it may state that the aerial survey identified an estimated 4,820 visible and inferred container units across the surveyed yard, subject to uncertainty from stack height, partial obstruction and AI classification.

This is more appropriate than presenting the number as an exact official inventory unless it has been reconciled and verified.

Benefits of Container Counting with Drones

The greatest advantage is speed and visibility.

A drone can create a complete aerial snapshot of a large container yard far more efficiently than a manual visual walk-through.

AI can reduce the time required to analyse the imagery, while photogrammetry or LiDAR can help estimate stack height.

The same dataset can also support yard occupancy analysis, congestion monitoring, damage screening, security, asset mapping and operational planning.

When integrated with a TOS, drone counting becomes especially valuable as an independent physical verification layer.

Challenges and Limitations

Container counting from the air is more complex than simply detecting rectangles.

Containers may be stacked several levels high, partially hidden by cranes or other units and mixed with equipment that looks similar from above.

OCR can fail because container markings are obscured or viewed at poor angles.

Photogrammetry and AI both introduce uncertainty.

The drone should therefore complement rather than replace the terminal's official inventory and logistics systems.

The strongest applications are reconciliation, auditing, occupancy analysis and exception detection.

The Future of Container Counting

Container counting is likely to become increasingly automated as computer vision, 3D modelling and terminal-system integration improve.

Drone-in-a-Box systems may conduct scheduled surveys across large yards, while AI automatically detects stack boundaries, estimates tier heights and compares the observed physical arrangement with the TOS.

Rather than producing only a total container number, future systems may create a continuously updated visual model showing occupancy by block, expected versus observed inventory, container type and recent changes.

OCR and asset recognition may increasingly connect individual visible containers with logistics records, while low-confidence matches are automatically sent for human review.

AI could also identify unusual stack arrangements, visible damage or areas of developing congestion during the same processing workflow.

The long-term direction is toward a digital yard-verification system in which the terminal operating system provides the official inventory, drones provide the physical visual snapshot, and AI continuously compares the two to highlight discrepancies and support better operational decisions.

Conclusion

Container counting is a valuable drone application for ports, inland terminals, empty-container depots and intermodal logistics facilities.

Drones equipped with high-resolution RGB cameras, RTK or PPK positioning and, where required, photogrammetry or LiDAR can provide rapid physical snapshots of large container yards.

AI can assist with container detection, stack-height estimation, OCR and change analysis, reducing the manual effort required to interpret thousands of containers.

The greatest value is not in replacing the Terminal Operating System. It is in providing an independent visual layer that can verify yard occupancy, support inventory reconciliation and identify areas where the physical site may not match the digital record.

Used with professional validation and appropriate system integration, container-counting drones can provide faster physical inventory checks, stronger yard-capacity information, improved discrepancy detection and a more accurate visual understanding of container-terminal operations.