Object counting
Recording every object as it crosses a control line or passes through an analysis zone without stopping the conveyor.
Why it matters
Manual counting and simple sensors fail when the flow changes
On a conveyor, objects can move in dense flows, partially overlap, and differ by shape, color, packaging, orientation and speed. A simple photoelectric sensor counts a line crossing, but it does not understand the object class, duplicate counting, empty positions, rejects, packaging defects or mixed flows.
- quantity errors by shift, batch and line;
- no separation by class, type, color, geometry and quality attributes;
- label shifts, contamination, print defects, packaging damage and marking errors are not captured;
- the operator sees the issue too late, after the batch has moved further down the line.
STATANLY solution
Detector + counter + quality control + operator UI
The system analyzes video from an IP or industrial camera, detects objects, tracks their movement, removes duplicate counts, classifies products and captures defect indicators. Objects, analysis zones and inspection logic are configured in the interface: by line, area, control zone or production event.
- counting items, parts, packages, raw materials and semi-finished products;
- classification by type, color, shape, size and quality attributes;
- detection of packaging, marking, geometry, completeness and surface-condition defects;
- integration with conveyors, accounting systems, reject mechanisms and production equipment.
Capabilities
From a simple counter to in-flow product quality control
The solution can be deployed step by step: start with quantity control on one line, then add product classes, presence checks, quality attributes, defect detection, integrations and reports.
Product classification
Separating objects by type, color, size, shape, packaging, orientation, SKU and quality attributes.
Geometry and condition
Assessing dimensions, shape, position, skew, non-standard objects and clearly visible deviations.
Presence / absence
Checking empty positions, omissions, completeness and the presence of an item in a cell, box, blister or on the belt.
Rejects and visual defects
Detecting packaging damage, contamination, label shift, print defects, wrong shape and suspicious items.
Events and control
Sending counters, alerts, statuses and Warning / Reject / Alarm events to SCADA, MES, ERP, PLC, HMI, BI or reject systems.
Interface and analytics
The operator sees the flow, counters, product classes, defects and proof frames
The interface combines video streams, counting zones, event logs, class distribution, shift and line statistics, camera settings, defect frames and integration parameters.
Example scenarios
Counting bread, packages, parts and defect control in mixed flows
The system applies to food production, agriculture, warehouses, packaging and industrial assembly. In one contour it can count objects, classify them and detect rejects: damage, contamination, wrong orientation, missing components, print defects or non-standard shape.
White bread counting
The system records items in the flow, excludes duplicate counts and sends quantities by line, shift or batch. It can also flag non-standard shape, damage and visible rejects.
Counting dark bread and different product types
The AI detector separates product classes in one video stream, creates counters for each class and can highlight suspicious items for quality inspection.
Counting end pieces and non-standard objects
The solution works with objects of different shapes and orientations where simple sensors cannot classify reliably, and helps detect deformation or unusual geometry.
Counting chips, packages and small items
The scenario fits flow control before filling, packaging, sorting or transfer to the next area, including contamination, damage and incomplete package detection.
Counting chicks and moving objects
The system counts moving objects correctly even when they change position or briefly overlap each other.
Package counting and reject detection
Package counting can be combined with integrity checks, contamination detection, label shift, print defects and signs of seal failure.
Counting products and operations
The system counts packages on the production line, monitors basic sorting operations and helps capture distribution errors or class deviations.
Efficiency control, statistics and KPIs
Counting operations, shifts and batches by lines, operators and production modes, with the ability to highlight losses and deviation causes.
Quality control
Which defects can be detected together with counting
In packaging-control projects, the model can count an object and assess its condition at the same time. This creates a single loop: object detected, classified, checked for quality, event saved, and, if needed, a reject signal sent.
Integrity, contamination and sealing
Torn packaging elements, contamination, product residue on the outer surface, signs of seal failure and systematic packaging rejects.
Print, date and readable information
Defects or missing human-readable print, date and expiration marking errors, poor flexographic printing and label displacement.
Match, position and completeness
Label or insert mismatch with SKU, back-label displacement, missing component, wrong orientation, non-standard shape and suspicious object.
Hardware and software complex
Camera, lighting, compute node, database and integration module
Reliable counting and defect control depend not only on the model, but also on a stable inspection point: the right camera, lighting, protection, working distance, line synchronization and exception rules.
1
Cameras and lighting
IP or industrial cameras, lenses, lighting, protective housing and mounting for a specific inspection point.
2
AI flow analysis models
Detection, classification and quality-control models for selected objects, classes and production conditions.
3
Counter, rejects and reports
Line or zone counting, defect events, frame archive, statistics by shifts, batches, lines and product classes.
4
Integration
Data delivery to SCADA, PLC, MES, ERP, VMS, HMI, BI and internal accounting systems via standard protocols.
Deployment
From one counting point to a network of conveyor lines
We recommend starting with a pilot on one line: confirm accuracy, product classes, defect types, capture conditions and integration, then scale to other areas.
1
Survey
Define the counting object, product classes, defect types, line speed, capture conditions and reporting requirements.
2
Pilot
Collect examples, configure the model, analysis zones, counters, filters, defect indicators and operator interface.
3
Acceptance
Fix accuracy metrics, rules for ambiguous cases, exceptions, defect types and operation requirements.
4
Scaling
Add lines, classes, new scenarios, integrations and a unified production analytics center.
Software registry
System for detecting and recognizing characteristics of processes and objects
An open computer-vision algorithm library that can determine characteristics of objects and processes: dimensions, type, speed, color, geometric parameters, fonts, numbers, QR codes and other features.
Main classImage processing, analysis and recognition tools.
Other classesReal-time analytical processing tools (OLAP).
CapabilitiesVideo streams from web/IP cameras and files, classification, detection, segmentation, boxes, masks, labels, tracking, CPU/GPU/TPU.
LicenseApache License 2.0.
Rights holderSTATANLY LLC, Russia. OGRN 1237800072982, INN 7801724456.
CEOSergey Fedorov, General Director.
Documents and materials
Links to software registry information, documentation and installation materials.
Software registry extractOpen the entry in the Russian software registry
Software documentationDescription of processes and functional characteristics
Installation and operationDeployment and operation documentation
Software repositoryOpen source code and project materials
AccreditationState accreditation certificate
Rights holder charterSTATANLY LLC charter
Want to test counting and defect detection on your line?
Start with one inspection point: counting object, product classes, defect types, camera, lighting, pilot acceptance method, line integration and effect calculation before scaling.