Counting rebar ends in a bundle
Counting a rebar bundle starts with the visible ends, not the outer bundle shape. Surface appearance can create false features.
Line solution
Test this on your samplesRebar counting system for bundles
// Short answer
A rebar counting system is a camera that faces the cut ends of a bundle and counts every bar before the bundle is tied, so each bundle leaves with a known count and a stored image. addaScan builds it on a HIKROBOT 20 MP camera running HIKROBOT VisionMaster deep-learning counting, with a screen where the operator confirms doubtful counts. An indicative installed budget for a typical application is R 430 000 – R 1 190 000 excluding VAT.
// At a glance
The station in three lines
- Faults caught
- Short and over-counted bundles; counts on the tag that do not match the bars; hand counts with no record
- Configuration
- Camera in a housing facing the bundle end, long bar lights, VisionMaster deep-learning count, operator screen
- Line link
- Count and hold signal to the bundling PLC; count and image to dispatch or MES
// Fit & limits
What this station will not do, and what it needs
What it will not do
- Hidden bars: a bar whose end is pushed back into the bundle, or hidden behind a bent end, may not be seen from the face.
- Mass, length and grade: the count does not weigh the bundle, measure bar length, or read rolled-in marks and grade.
- Coils and loose bars: coiled bar and loose bars on a cooling bed are different jobs; this station counts a formed bundle face.
- Nested pipes: a pipe inside a larger pipe is not separated by a face count.
What makes it work
- Visible ends: bar ends cut and roughly flush on the side that faces the camera.
- A fixed counting position: the bundle stops in the same place, square to the camera.
- Shade and shielding: no direct sunlight on the face, and heat and scale kept off the housing.
- Image samples: photos of real bundles of every bar size, including the untidy ones.
// On South African lines
Reinforcing bar in South Africa, and why bundles are counted
Reinforcing bar is made in South Africa: the South African Iron and Steel Institute lists it among the primary carbon steel products manufactured here, alongside billets, sections and bars (SAISI). Between the mill and the building site, bars are bundled and tied at the mill, then often split and re-bundled by merchants and cut-and-bend shops. Each of those points can be a place where bars are counted. Bar counting by hand means marking bar ends with chalk and counting again when two people do not agree.
There is no South African rule that requires a bar count, and this page does not suggest one. The reason to count is commercial. Where bars, tubes or pipes are dispatched or invoiced by count, a short bundle becomes a customer claim and an over-count is steel given away. A rebar counting machine with a camera counts every bundle the same way, before the ties go on, and keeps a marked image as the evidence behind the number on the tag or delivery note.
HIKROBOT lists rebar counting with deep learning as a HIKROBOT VisionMaster application case (HIKROBOT catalogue, p.65). The case gives no figures, and it is vendor material, not a South African reference; what a station counts reliably on your bar sizes is shown on your bundles first.
What the dispatch and bundling teams report, and what the station checks
| What the line reports | What a station checks | Where it stops |
|---|---|---|
| “The customer says the bundle was two bars short” | Every bar end in the bundle face is counted before tying, and the count is stored with a marked image against the bundle or order number. | A bar whose end sits deep inside the bundle, or behind a bent end, may not be seen; the image shows the operator where to look. |
| “We count with chalk on the ends, and recount when the numbers differ” | An automatic count at the bundling position, compared with the target count for that product; a mismatch holds the bundle. | It needs the cut ends roughly flush on the side facing the camera; a staggered face is counted with operator confirmation. |
| “The thin bars are impossible to count by eye” | A deep-learning model trained on each bar size finds touching round ends that look alike. | Small diameters across a large bundle face need enough resolution; each size is proven on sample images before the order. |
| “Tube and pipe bundles are miscounted as often as rebar” | The same station counts round tube and pipe ends, with a recipe per size. | Pipes nested inside larger pipes, or ends pushed back into the bundle, are not separated from the face alone. |
| “The tag says one count and the bundle has another” | The confirmed count can be sent to the tag printer or dispatch system, so the tag carries the counted number. | The station counts bars; it does not weigh the bundle, measure length or check grade. |
// The complete solution
Rebar counting system: camera, software and installation
The bundle stops at the counting position with its cut ends towards the camera. VisionMaster finds and counts every bar end, the PLC compares the count with the target, and the bundle is released for tying only when the count agrees or the operator has confirmed it on the screen.
[Schematic]Plan view, not to scale
Swipe sideways to see the whole drawing →
[1]: The bundling PLC or a sensor signals that the bundle has stopped at the counting position.
[2]: Long bar lights, angled at the bundle face, light every cut end.
[3]: A 20 MP camera in a housing faces the bundle ends from the side of the transfer.
[4]: The PLC compares the count with the target; a mismatch holds the bundle and asks the operator to confirm on screen.
[5]: Only a confirmed bundle is released; its count goes to the tag or to dispatch.
Hardware
| Component | HIKROBOT example | Role in this station |
|---|---|---|
| Camera | HIKROBOT MV-CS200-10GM, 20 MP mono GigE area scan camera: one in the low case, two for large bundles or two counting positions | Images every bar end across the bundle face |
| Lens | HIKROBOT MVL-HF2524M-10MP; MVL-KF1224M-25MP lenses in the high case | Fits the whole bundle face into the image from a safe distance |
| Processing | HIKROBOT VC3000 vision controller (MV-VC3302P-128G60; MV-VC3704H-128G60 in the high case) running HIKROBOT VisionMaster with deep learning | Runs the counting model, keeps the image archive and drives the operator screen |
| Lighting | HIKROBOT MV-LLDS-H-1400-40-W bar lights, two or four | Even light on cut steel ends across a large face |
| Housing and shield | Not HIKROBOT: IP67 camera housing, plus a local heat and dust shield | HIKROBOT lists the CS-series camera at IP40, so in a mill bay it goes in a housing |
| Smaller bundles (alternative) | HIKROBOT SC3000X vision sensor with learning-based count, or an SC6000 smart camera | One device in place of camera and controller where the bundle face is small and fixed |
| Trigger, mount and screen | Existing bundling PLC signal or a HIKROBOT MV-PE photoelectric sensor; local supply: pole or gantry mount, touch screen | Starts the count; holds the camera square to the face; lets the operator confirm |
Software and vision tools
| Tool or software | What it does here |
|---|---|
| HIKROBOT VisionMaster deep-learning object detection | Finds each bar end as an object and counts them, including touching ends of the same size; trained on images of your bar sizes and bundle shapes. |
| VisionMaster blob tools | For clean, separated tube and pipe ends a rule-based count can be enough, and it is simpler to maintain. |
| Operator confirmation screen | Shows the bundle image with every counted bar marked and low-confidence areas highlighted; the operator adds or removes a mark and confirms. |
| Recipe per bar size and bundle | Target count, bar size and image area per product, selected by the PLC or taken from the order. |
| Count and image record | Count, time, bundle or order number and the marked image saved for every bundle, ready for a customer query. |
Installation and integration
01
Mounting and enclosure
Camera and lights on a pole or gantry square to the counting position, out of the path of the crane and the bundles; the camera in an IP67 housing behind a heat and dust shield.
02
Lighting and shrouding
Long bar lights at an angle to the bundle face so every cut end returns light; a hood or roof sheet keeps direct sunlight off the face in open bays.
03
Trigger and tracking
The bundling PLC signals that the bundle has stopped; the bundle or order number comes from the PLC or MES so each count is stored against it.
04
PLC link
Count, target and hold signal by digital I/O from the VC3000 controller to the bundling PLC; count and image reference to dispatch or MES over Ethernet, in a format agreed per site.
05
Hold and confirmation
There is no rejector: a count that does not match holds the bundle at the tying position until the operator confirms or corrects it. If the vision system is not ready, the bundle is held as uncounted, never released as counted.
06
Recipes and operator screen
One recipe per bar size and target count, chosen by the PLC or from the order; the screen shows the last bundles with their marked images.
// Indicative cost in South Africa
What this station costs, excluding VAT
Indicative budget for a typical application of this kind. Every project involves custom requirements (your products, line speed, layout, PLC and acceptance criteria), so the final price differs from case to case. A firm price follows the sample test.
- Low
- R 430 000
- one camera, one station
- High
- R 1 190 000
- two cameras with MES link
- Basis
- Typical application of this kind, indicative
- South African rand, excluding VAT
- Prices as at 2 October 2026
- Travel (Gauteng): outside Gauteng quoted separately
// Included and excluded
The low figure is one camera at one bundling position, with deep-learning counting and an operator confirmation screen, signalling the existing bundling PLC. The high figure is two cameras (large bundles or two counting positions), operator confirmation, a link to MES and heavier protection for a mill bay. Even the one-camera case needs a deep-learning licence and mill-environment mounting, so this work sits outside the standard single-view range on the cost guide and is therefore priced here separately. Changes to your bundling line, cradle or transfer, utilities and VAT are excluded.
Hardware is priced from public list prices converted at the ECB reference rate, plus freight, clearance and margin; engineering at R 900 – R 1 300 per hour. How these ranges are built →
Swipe the table sideways to see the high figure and the basis →
| Cost block | Low | High | Basis |
|---|---|---|---|
| Vision hardware (landed, with margin) | R 128 600 | R 318 400 | Low: one 20 MP mono camera, lens, two long bar lights, VC3000 controller, IP67 housing. High: two 20 MP cameras with 25 MP lenses, four long bar lights, larger VC3000 controller, two housings |
| Software licence | R 101 200 | R 150 000 | HIKROBOT VisionMaster with deep learning; high case on the PRO deep-learning tier |
| Local materials and reject | R 60 000 | R 150 000 | Low: heavy-duty pole mount, heat and dust shield, count confirmation screen. High: structural mounts and heat shields, panel with screen and UPS, long cable runs |
| Engineering (recipes, PLC link) | R 90 000 | R 312 000 | 100 h at R 900 to 240 h at R 1 300 |
| Installation, FAT, SAT, training, documents | R 36 000 | R 124 800 | 40 h to 96 h |
| Project management | R 14 400 | R 62 400 | 16 h to 48 h |
| Travel (Gauteng) | R 4 000 | R 10 000 | Outside Gauteng quoted separately |
| Total (rounded) | R 430 000 | R 1 190 000 | Low: one camera, one station. High: two cameras with MES link |
// Related technology
Which technologies this scenario uses, and why
The technologies behind this station, how much each matters here, and where each is explained in depth.
| Technology | Role | Why it matters in this scenario | Read more |
|---|---|---|---|
| Deep learning (classification, segmentation, DL OCR) | Core | Touching round ends of the same size and colour are where rule-based blob counting fails; a trained object detection model separates them. | Rule-based vs deep learning |
| Presence detection and counting | Core | The result is one number per bundle, compared with the target count for the product or order. | Presence and counting |
| Area scan cameras with a vision controller | Supporting | A 20 MP camera on a controller resolves small bar ends across a large face, and the controller keeps the image record. | Camera and lens selection |
| Lighting technique | Supporting | Cut steel ends reflect unevenly and mill bays let in daylight; angled bar lights and a hood keep every end visible. | Machine vision lighting |
| PLC link and reject with confirmation | Supporting | Nothing is rejected: a miscount holds the bundle at the tying position until the operator confirms it. | PLC and reject integration |
// Acceptance
How the station is proven before handover
Every station is accepted on your products, against numbers agreed in writing before the build. How we deliver →
01
Sample test
Images of your bundles for every bar size, taken with the proposed camera and lights at the counting position, including the worst faces; the result says which sizes count reliably and where operator confirmation is needed.
02
FAT with challenge bundles
Before delivery, a test bundle of short bar offcuts with known counts is counted in the workshop, with bars added and removed, and the model is run on a held-back set of your bundle images; every count is compared with a hand count.
03
SAT on the line
At your bundling position, an agreed number of bundles of each size is counted by hand and by the station; miscounts released, correct bundles held in error and operator corrections are recorded over the agreed run.
// FAQ
Questions about rebar counting
How does a rebar counting system count bars in a bundle?
A camera faces the cut ends of the bundle while it stands at the counting position. Deep-learning object detection in HIKROBOT VisionMaster finds each bar end, including ends that touch, and counts them. The PLC compares the count with the target for that product and holds the bundle if they differ. The operator sees the image with every counted bar marked, confirms or corrects it, and the count and image are stored against the bundle. The same principle serves as a bar counting machine for round bar and other solid sections.
How much does a rebar counting system cost in South Africa?
An indicative installed budget is R 430 000 – R 1 190 000 excluding VAT, at 2 October 2026 prices. The low figure is one camera at one bundling position with deep-learning counting and an operator screen; the high figure is two cameras with a link to MES and heavier mill protection. Both include hardware, software, engineering, installation, FAT and SAT in Gauteng. A firm price follows a sample test on images of your bundles. This is a typical application: every project involves custom requirements, so the final price depends on your products, line and acceptance criteria.
Can the same station work as a pipe counting machine for tubes?
Yes, where the tube or pipe ends face the camera and are not nested. The camera then works as a pipe counting camera or tube counting machine: round hollow ends are counted with their own recipe, and clean separated ends can often be counted with simpler rule-based tools. Plastic pipe ends reflect light differently from cut steel, so they are part of the sample test. Pipes inside larger pipes, and ends pushed back into the bundle, cannot be separated from the face.
Is a rebar counting machine different from a rebar counting system?
Suppliers use both names for the same job. A rebar counting machine usually means a standalone counting unit; a rebar counting system or station, as on this page, is fitted at the counting position of your existing bundling line.
Is a camera count accurate enough to invoice from?
That is shown on your bundles, not promised in advance. The station counts what it can see on the face, which is why the counting position, the lighting and tidy ends matter. Doubtful counts are held for the operator rather than passed, and every bundle keeps a marked image, so a customer query can be answered with the picture. The SAT records miscounts on an agreed run of your production.
Does it check bar grade, length or mass?
No. The station counts bars in the bundle face. It does not read rolled-in marks, check grade, measure length or weigh the bundle; mass stays with your weighbridge or crane scale, and grade with your mill records. If the count and the mass disagree, the stored image helps find out which one is wrong.
Can it count bundles outdoors in a merchant’s yard?
Only at a fixed counting position with some shelter. Direct sunlight moving across the bundle face changes the image through the day, so the station needs a roof or hood over the face, a housed camera and a place where bundles stop square to it. A counting point at the cut-and-bend bay or the loading bay usually suits better than open ground.
Sources
- SAISI: primary steel industry overview (products manufactured in South Africa include reinforcing bar) (industry body page, checked 2026-10-02)
- HIKROBOT Machine Vision Smart Product Catalog 2025Q4, p.65: rebar counting application case (vendor catalogue, checked 2026-10-02)
- HIKROBOT VisionMaster (algorithm platform, deep-learning modules including object detection) (vendor page, checked 2026-10-02)
- HIKROBOT SC3000X vision sensor (learning-based count) (vendor product page, checked 2026-10-02)
- HIKROBOT MV-CS050-10GC area scan camera (CS series, IP40) (vendor product page, checked 2026-10-02)
Related:
More line solutions in building materials and metals:
// Next step
Send us photos of your bundle ends, every bar size
Describe the line, the products and the fault. We test on your good and faulty samples and reply with a feasibility note and a budget.