---
title: "Optical Potato Sorter Acceptance: False Rejects, Missed Defects and Representative Samples"
url: https://www.spudmate.com/ur/optical-potato-sorting-acceptance-false-rejects/
description: "A convincing sorting demonstration can hide unnecessary crop losses or missed defects. This guide explains how to label representative potato samples,…"
published: 2026-10-10
updated: 2026-10-08
author: "xtlaigc"
site: "Tianjin Goldilocks Co., Ltd."
---

# Optical Potato Sorter Acceptance: False Rejects, Missed Defects and Representative Samples

## Key takeaways

- Use acceptable input potatoes for the false-reject denominator and defective input potatoes for the missed-defect denominator.
- Keep representative production samples separate from deliberately defect-rich challenge sets.
- Reconcile every output and distinguish inspection problems from failed diversion where observation allows.
- Agree limits before testing and accept only the crop conditions and operating window demonstrated.

## Why a clean accepted stream is not enough

An optical sorter can produce a visually convincing accepted stream while discarding too many saleable potatoes. Conversely, a small reject pile can conceal defects that remain in the accepted crop. The buyer’s problem is therefore not simply whether the machine rejects something: it is whether the agreed quality separation is achieved without unacceptable losses, under the intended operating conditions. We recommend defining both errors before arranging a demonstration.

A useful acceptance test follows labeled input potatoes through the sorting process and independently checks their destinations. It also records the crop condition, feed rate and sorting settings that produced the result. Without that context, a demonstration on clean, evenly spaced potatoes says little about a different incoming lot. Keep three questions separate: what was presented for inspection, what quality decision was appropriate, and where the potato actually went.

## Define false rejects and missed defects with different denominators

**False-reject rate = acceptable potatoes sent to rejection ÷ acceptable potatoes entering the test × 100%.** **Missed-defect rate = defective potatoes sent to acceptance ÷ defective potatoes entering the test × 100%.** These denominators must remain separate. The percentage of the reject stream that is actually acceptable answers a different question: reject-stream composition. Likewise, the defective proportion of the accepted stream describes outgoing quality, not the fraction of incoming defects missed.

In a **hypothetical** trial containing 800 acceptable potatoes and 200 defective potatoes, rejecting 40 acceptable potatoes gives a 5% false-reject rate. Accepting 20 defective potatoes gives a 10% missed-defect rate. These numbers illustrate arithmetic, not a performance target. Use individual-potato counts for classification errors and record mass separately for crop loss and material reconciliation. State the counting convention for potatoes with multiple defects so they are not accidentally counted twice in the overall defective denominator.

1. Grade representative samples
2. Run recorded conditions
3. Collect every output
4. Score separate error rates

Conceptual acceptance workflow, not a machine layout or a drawing to scale.

## Build a reference sample that represents the buying decision

Write the reference grading rules before labeling the sample. Define acceptable variation, each reject category and how borderline cases will be handled. Have assessors grade independently of the sorter result, retaining images or written descriptions where useful. Resolve disagreements before scoring, or keep unresolved cases in a separately reported group. Labels and tracking methods must not obscure the surface or introduce visual cues that influence sorting; verify the chosen identification method during preparation.

Include ordinary acceptable potatoes, difficult but acceptable appearances, clear defects and borderline examples. Cover the size range and washed or unwashed presentation intended for the project. A deliberately defect-rich challenge set helps investigate weak categories, but its accepted-stream defect proportion does not represent routine production unless its composition matches incoming crop. Use a representative lot for operational acceptance and a separately reported challenge set for diagnosis. Choose sample quantities to support category-level conclusions, rather than relying on a large total dominated by easy examples.

## Choose the inspection and rejection mechanism to test

The [AI intelligent sorting machine CASL-2X-1800](https://www.spudmate.com/products/ai-intelligent-sorting-machine-casl-2x-1800/) combines AI vision, near-infrared detection and high-pressure air blowing, with a listed capacity of 40 t/h. Its stated detection scope includes soil, rock blocks, green heads, decay, deformities, mechanical injuries, length, diameter and size characteristics in washed or unwashed root and stem products. Acceptance should challenge the selected quality categories and verify actual output destinations, not assume that detection alone proves successful removal.

The [Roller-type optical sorting machine CASV-2X-1200](https://www.spudmate.com/products/roller-type-optical-sorting-machine-casv-2x-1200/) uses AI vision and short-wave near-infrared dual-source detection. Stainless rollers rotate material for 360° inspection, and a plastic finger device removes non-conforming products; its listed capacity is 20 t/h. Rotation changes surface presentation, while the rejection mechanism changes how separation occurs. Neither establishes a universal defect result. Near-infrared terminology also does not establish measurement of internal chemistry. Compare mechanisms against your actual samples and operating requirement.

| Comparison | CASL-2X-1800 | CASV-2X-1200 | Meaning for acceptance planning |
| --- | --- | --- | --- |
| Listed capacity | 40 t/h | 20 t/h | Use the listed capacity to frame the operating-rate requirement; score quality separation during the trial. |
| Detection | AI vision and near-infrared detection | AI vision and short-wave near-infrared dual-source detection | Test the required defect categories rather than treating detection terminology as a quality guarantee. |
| Rejection mechanism | High-pressure air blowing | Plastic finger device | Observe actual diversion and the destination of acceptable neighbors. |
| Material presentation statement | Detection in washed or unwashed root and stem products | Stainless rollers rotate the material for 360° inspection | Keep surface condition and inspection presentation explicit in the test record. |
| Stated material scope | Potatoes, carrots, ginger, sweet potatoes, cassava and other root and stem products | Seed potatoes, commercial potatoes and onions | Select trial material within the stated scope and evaluate the specific crop requirement. |

[![AI intelligent sorting machine CASL-2X-1800 showing the equipment's external construction.](https://www.spudmate.com/wp-content/uploads/2026/10/spudmate-ai-intelligent-sorting-machine-casl-2x-1800-1024x512.jpg)](https://www.spudmate.com/products/ai-intelligent-sorting-machine-casl-2x-1800/)

CASL-2X-1800 — optical sorting with high-pressure air rejection.

## Control presentation before judging the sorting decision

Start the trial with clean optical surfaces and record their condition according to the equipment’s operating instructions. Record surface soil, wetness, size distribution and material spacing at the inspection zone. These variables can change which features are visible and whether neighboring potatoes interfere with observation or separation. If washed and unwashed material are both project requirements, treat them as distinct test conditions rather than pooling their results into one average.

Observe feed stability and potato presentation independently during each run. For roller inspection, examine whether the trial material rotates as intended; a stated 360° inspection arrangement is not a reason to ignore poor presentation in a particular lot. For air rejection, observe whether the selected item separates without taking acceptable neighbors with it. For finger rejection, observe the selected item’s diversion and destination. These are commissioning checks, not claims of additional installed monitoring features. Relate them to the surrounding flow through our [potato processing solution planning](https://www.spudmate.com/#solutions).

[![Roller-type optical sorting machine CASV-2X-1200 showing the equipment's external construction.](https://www.spudmate.com/wp-content/uploads/2026/10/spudmate-roller-type-optical-sorting-machine-casv-2x-1200-1024x512.jpg)](https://www.spudmate.com/products/roller-type-optical-sorting-machine-casv-2x-1200/)

CASV-2X-1200 — roller-type optical inspection with plastic finger rejection.

## Account for every stream and distinguish detection from diversion

Collect accepted and rejected outputs separately, including any additional discharge paths present in the trial arrangement. Reconcile the input count with all recovered potatoes, and record retained, spilled, damaged or untraceable material separately. Do not silently exclude missing items to improve either error rate. If unresolved destinations prevent reliable scoring, repeat the affected run after correcting collection or tracking. Also report rejected foreign material separately from potato classification errors.

Use independent observation of selected test items to distinguish an apparent inspection failure from a diversion failure where practicable. A correctly identified reject that reaches the accepted outlet suggests a different investigation from a defect that was not identified. If the available observations cannot separate those events, report an end-to-end missed defect without assigning a cause. Video or direct observation can support diagnosis, but neither should be described as a standard machine feature unless explicitly included in the project configuration.

## Separate tuning runs from the scored acceptance trial

Use an initial sample for setup, then score a separate labeled sample without further adjustment during that run. Record the selected settings, feed condition, measured test throughput, elapsed time and output collection method. Begin with a controlled feed to establish traceability, then test the proposed operating condition. Listed capacity is not evidence that the agreed error limits will hold for every crop mixture; demonstrate the required separation at the intended rate and presentation.

Repeat scored runs with fresh representative material where possible, changing only one diagnostic variable at a time. Record the numerator and denominator for each defect category and each false-reject subgroup, not percentages alone. A category represented by only a few potatoes provides limited evidence even when every item is correctly sorted. Use the sampling matrix below to plan observations, and use our [AI vision potato sorting procurement checklist](https://www.spudmate.com/ai-vision-potato-processing-procurement-checklist/) to connect the acceptance method with the purchasing specification.

| Sample category | Reference preparation | Measure or observation | Diagnostic use |
| --- | --- | --- | --- |
| Ordinary acceptable potatoes | Include the intended size and surface-condition range. | Acceptable potatoes rejected / acceptable potatoes entered. | Establish baseline false rejects under representative feed conditions. |
| Difficult but acceptable appearances | Document why each potato remains acceptable under the agreed grade. | False rejects within this subgroup. | Identify acceptable variation being confused with a reject category. |
| Defects with agreed reference criteria | Label by category and document multi-defect potatoes. | Missed defects / input defective potatoes, reported by category. | Locate weak categories that an overall result could conceal. |
| Borderline potatoes | Resolve assessor disagreement or retain a separate unresolved group. | Destination counts and agreement with the reference decision. | Separate unclear grading rules from sorting errors. |
| Washed and unwashed presentations | Prepare separate groups when both are required for the project. | Error rates, optical cleanliness and visible surface condition by group. | Check whether a presentation change alters the result. |
| Foreign material and collection exceptions | Record foreign material separately; identify every collection path. | Foreign-material destinations and unreconciled potato counts. | Avoid mixing contamination removal with potato error denominators. |

## Accept a defined operating window, not an unrestricted promise

Before the scored trial, agree the acceptable false-reject and missed-defect limits, category requirements, sample quantities, repeat-run requirements and treatment of unresolved cases. Do not create limits after seeing the results. If a setting reduces missed defects but increases acceptable crop loss, document that trade-off against the buyer’s grading objective. A single combined accuracy percentage hides this decision and can favor the most common category while concealing poor performance on a less frequent defect.

The acceptance record should identify the tested crop, surface condition, size range, defect definitions, operating rate and settings, together with all output counts and exceptions. State which conditions were not represented. Recheck performance when incoming presentation or the grading objective changes materially. We treat acceptance as evidence for a defined operating window: it supports a configuration decision, but does not guarantee identical results on every harvest lot or establish capabilities for untested internal defects.

## Plan a representative potato sorting acceptance trial

[Discuss your potato sorting acceptance trial with our engineering team](https://www.spudmate.com/#contact). Send crop samples covering acceptable potatoes, relevant defects and borderline grades, together with washed or unwashed condition, size range, expected defect mix, target throughput, grading rules, proposed false-reject and missed-defect limits, site layout and available utilities. We can use these inputs to define the configuration review and trial conditions.