Cargo and Border Security Screening

Scope and method

This note covers high-energy X-ray computed tomography (CT) applied to cargo and border-security screening — i.e., scanning trucks, shipping containers, air-cargo pallets, and vehicles to produce volumetric density maps that allow operators and algorithms to find contraband, weapons, explosives, drugs, and undeclared goods. It is not a baggage-aviation CT note; aviation checkpoints use lower-energy CT with different geometry and regulatory regimes, although the underlying reconstruction and detection-method literature overlaps.

The source tier rules for this note are:

• Tier 1: granted patents, peer-reviewed papers (DOI-verified), standards-body sources, regulator technical assessments.
• Tier 2: company technical disclosures, engineering blogs, named conference proceedings, and self-reported product pages.
• Prohibited: SEO content farms, anonymous benchmarks, AI-generated summaries, and sources quoting documents not retrievable.

I label any load-bearing claim that lacks two independent institutions as [SINGLE-SOURCE] and any number not verified as needs primary source validation.

Technical architecture and physics

Cargo CT systems differ from basic cargo radiography in a fundamental way: radiography projects a 3D object onto a 2D image, so dense objects hidden behind other dense objects can be invisible. CT rotates a source/detector array around or passes the cargo through a gantry while collecting many projection angles, then reconstructs a volumetric density map that separates overlapping materials along the beam path. The granted patent US 8,929,509 B2 describes an X-ray tomography inspection system with a rotating source/detector geometry, a conveyor for moving an object through the inspection tunnel, and processors that reconstruct tomographic images from collected data (Google Patents, US8929509B2, grant date 2015-01-06). The patent claims are a recorded architecture, not proof of achieved image quality or detection performance.

Cargo CT systems use high-energy X-ray photons — commonly reported in vendor disclosures in the 3 MeV to 9 MeV range — because a fully loaded 40 ft container can present more than 100 g/cm² of areal density in the densest ray paths. This vendor energy range is needs primary source validation unless a specific spec sheet is retrieved. The physics of attenuation is standard and can be calculated from NIST XCOM cross-sections:

\[

I/I_0 = \exp[-(\mu/\rho) \rho t]

\]

where \(I/I_0\) is transmitted fraction, \(\mu/\rho\) is mass attenuation coefficient, \(\rho\) is material density, and \(t\) is thickness. The NIST XCOM database provides photon cross-section data for iron, steel, and other materials (NIST XCOM, physics.nist.gov, accessed 2026-04-26).

The typical cargo CT chain is:

1. Source: high-energy X-ray tube or linear accelerator (linac); vendor-specific energy/current parameters are usually not in Tier 1 sources and should be treated as self-reported unless measured independently.
2. Pre-collimation and detector array: scintillator or direct-conversion detectors; some systems use multiple energy bins for material discrimination.
3. Gantry/translation motion: either the gantry rotates while the container is stationary and then translates, or the container moves through a rotating ring.
4. Reconstruction: filtered backprojection, iterative reconstruction, or model-based approaches produce a 3D voxel map.
5. Automated detection: machine-learning models flag anomalies for human review.

The multi-stage architecture is described in the granted patent family above but specific resolution, penetration, and false-alarm rates are not independently verified in this note (US8929509B2). [SINGLE-SOURCE for architecture/performance linkage]

Patents and prior art

Granted patent used:

• US 8,929,509 B2 — “X-ray tomography inspection systems”
• Assignee: Rapiscan Systems, Inc.
• Grant date: 2015-01-06
• URL: patents.google.com

This is a granted patent, not an application. The distinction matters: a granted patent is evidence that the patent office accepted claims at examination, but it is not evidence that the claimed system was built, sold, or performed at any particular throughput or false-alarm rate. The claims should be read as an intellectual-property boundary, not a benchmark.

Rapiscan is part of OSI Systems, Inc. (NASDAQ: OSIS). OSI Systems’ SEC filings show the company sells security inspection products through its Security division, including cargo, vehicle, and border inspection systems, and that the Security division is a major revenue driver (OSI Systems Form 10-K, SEC EDGAR, accessed 2026-04-26). However, the 10-K risk factors do not contain a CT-specific detection-rate or penetration benchmark; those figures appear in marketing collateral and are not uniformly present in the filed risk disclosure. This is a discrepancy worth noting: the technical moat is patent-visible, while the commercial claims are filed mainly as revenue and booking lines, not as measured spec sheets.

The patent reference should not be confused with pending applications in the same family. Applications may be cited for direction but are weaker evidence of established claims. For this note, only the granted patent is used as claim evidence.

Peer-reviewed evidence and algorithm risk

For automated threat detection in X-ray imagery, the strongest peer-reviewed source retrieved is:

• Akcay, S., Kundegorski, M. E., Willcocks, C. G., Breckon, T. P., “Using Deep Convolutional Neural Network Architectures for Object Classification and Detection Within X-Ray Baggage Security Imagery,” IEEE Transactions on Information Forensics and Security, vol. 13, no. 9, pp. 2203–2215, 2018.
• DOI: 10.1109/TIFS.2018.2812196
• URL: doi.org

This is a peer-reviewed journal paper with a DOI. It is not a preprint. Its experiments are on X-ray baggage-security imagery, not cargo-container CT volumes. The paper is relevant because it shows convolutional neural networks can classify and detect objects in security X-ray data with published performance on benchmark imagery, but the results do not transfer automatically to cargo CT. Cargo CT has different energy spectra, scatter, metal artifacts, container clutter, and much larger object volumes.

The correct claim is therefore narrow: published peer-reviewed evidence shows automated detection methods exist for security X-ray imagery; their extension to cargo CT is a separate technical problem. Vendors may claim trained “AI” detection, but this note does not retrieve a cargo-CT-specific peer-reviewed detector evaluation. The Akcay et al. paper is one lineage from Durham University; any company claim relying on it would need independent cargo validation. [SINGLE-SOURCE for method transfer]

Regulatory, safety, and standards

Cargo CT scanners are radiation-producing machines. In the U.S., regulatory jurisdiction can be shared by state radiation-control programs and federal agencies, but this note does not retrieve a specific FDA/FCC cargo-CT technical assessment. Safety standards for non-medical security X-ray systems are generally covered by documents such as ANSI/HPS N43.17, but I did not verify a retrievable cargo-CT-specific standard for this note. The absence here should be marked: no standards-body cargo-CT performance standard was retrieved.

The load-bearing physics parameter — photon mass attenuation coefficient — is sourced from NIST XCOM, which is a standards-body/physics reference database and qualifies as Tier 1 for cross-section data (NIST XCOM). It is not, however, a cargo-CT operational standard.

Worked quantitative calculations

6.1 Attenuation of 10 cm of iron at approximately 6 MeV

Using the NIST XCOM exponent:

\[

I/I_0 = \exp[-(\mu/\rho)\rho t]

\]

Assumptions/inputs:

• \(\mu/\rho\) for iron at ~6 MeV ≈ 0.030 cm²/g — approximate NIST XCOM value; exact 6 MeV value should be spot-checked against the database and is marked needs primary source validation for the precise energy (NIST XCOM).
• Iron/steel density \(\rho = 7.87\ \text{g/cm}^3\) — standard material property.
• Thickness \(t = 10\ \text{cm}\).

Calculation:

\[

(\mu/\rho)\rho t = 0.030 \times 7.87 \times 10 = 2.361

\]

\[

I/I_0 = e^{-2.361} = 0.0943

\]

Thus approximately 9.4% transmission for a 10 cm iron/steel path at a representative 6 MeV photon energy.

Sensitivity:

• If thickness is halved to 5 cm: \(e^{-1.1805}=0.307\), so ~30.7% transmission.
• If thickness doubles to 20 cm: \(e^{-4.722}=0.0089\), so ~0.9% transmission.

This shows a 10 cm incremental thickness difference changes transmission by roughly 21 percentage points — a strong contrast signal for CT if sufficient photons are collected. It also shows why dense cargo paths require high-energy sources and substantial detector dynamic range.

6.2 Throughput calculation for a 12.2 m container

Throughput formula:

\[

\text{Containers/hour} = \frac{3600}{L/v + t_{\text{overhead}}}

\]

Inputs:

• \(L = 12.2\ \text{m}\) — typical 40 ft container length, assumed.
• \(v = 0.1\ \text{m/s}\) — representative translation speed, assumed; vendor product pages may report faster or slower, needs primary source validation.
• \(t_{\text{overhead}} = 20\ \text{s}\) — assumed loading/reconstruction/decision overhead.

Base calculation:

\[

\text{Scan time} = 12.2/0.1 + 20 = 122 + 20 = 142\ \text{s}

\]

\[

\text{Throughput} = 3600 / 142 = 25.4\ \text{containers/hour}

\]

Sensitivity:

• At \(v = 0.05\ \text{m/s}\): scan time \(= 12.2/0.05 + 20 = 264\ \text{s}\), throughput \(= 3600/264 = 13.6\) containers/hour.
• At \(v = 0.2\ \text{m/s}\): scan time \(= 12.2/0.2 + 20 = 81\ \text{s}\), throughput \(= 3600/81 = 44.4\) containers/hour.

The calculation illustrates the operational tension: higher linear speed raises throughput but can degrade X-ray photon statistics and reconstruction quality. Without a vendor-verified speed/spec curve, the exact trade-off remains needs primary source validation.

Parameter table

Unit economics / business model

The cargo/border X-ray CT value chain monetizes mainly through capital equipment sales plus services. From the SEC filings:

• OSI Systems / Rapiscan (OSIS): sells complete inspection systems to government customs and border agencies; revenue includes product sales and follow-on service, maintenance, training, and upgrades. The 10-K reports Security division revenue and operating income separately; exact current-year figures should be pulled from the filed document and are needs primary source validation for this note (OSI Systems 10-K, EDGAR).
• Leidos (LDOS): sells integrated security and detection systems, including VACIS non-intrusive inspection and related cargo screening systems, often as part of larger government programs. Its filings disclose government contract dependence and recurring service revenue (Leidos 10-K, SEC EDGAR, accessed 2026-04-26).
• Varex Imaging (VREX): supplies X-ray tubes and digital detector components used in cargo, industrial, and security imaging; it is a component supplier rather than a full-system maker. Its filings describe industrial/security segment revenue and component sales (Varex Imaging 10-K, SEC EDGAR, accessed 2026-04-26).

The business model is not a subscription or royalty model at the system level; it is closer to “razor + service”: high-priced capital systems followed by multi-year maintenance agreements, spare parts, and software updates. Margins on service are typically higher than on hardware, but exact segment margins are needs primary source validation unless read directly from the 10-K line items. Value capture is split: full-system integrators capture system and service revenue; component makers capture tube/detector margin; government customers capture operational security benefit.

Disconfirming Evidence

The technology thesis would be wrong or weakened if any of the following were demonstrated with verifiable evidence:

• Throughput falls below operational targets. If a cargo CT system cannot scan containers quickly enough — e.g., below the 25 containers/hour illustrative rate in this note — ports and border agencies may prefer faster dual-energy radiography or targeted secondary CT. The assumed 0.1 m/s translation speed is needs primary source validation; vendor performance could be slower in practice.
• False-alarm or missed-detection rates are unacceptable. CT’s volumetric advantage matters only if it separates clutter from threats. If independent tests show that cargo CT false-alarm rates are not materially better than radiography, the premium price is hard to justify. No independent cargo-CT false-alarm benchmark was retrieved.
• Cost and radiation-safety constraints cap adoption. High-energy linac-based systems are expensive and require shielding, site licensing, and trained operators. If procurement budgets or safety reviews restrict deployment, volume will remain low.
• The patent is invalidated or narrowly construed. Rapiscan’s granted patent protects an architecture, not the concept of cargo CT. If competitors design around it, the moat weakens; the patent also may not cover later multi-energy or deep-learning enhancements.
• Component supply shifts. Varex and other component makers can supply competing integrators; if tubes/detectors become commodity, system integrator margin may shift to software and service, changing the value capture shown in the ticker list.
• Government procurement changes. The filings repeatedly note dependence on government contracts. A shift to low-cost dual-energy radiography, budget sequestration, or reshaped U.S. Customs and Border Protection acquisition programs would directly reduce cargo CT adoption (sec.gov ).