Can OEM axle weight data and satellite imagery help identify cargo theft?

Class8 has released a study detailing how it analyzes commercial truck activity to identify unload events that may warrant further investigation. According to the report, the study evaluated 3,260,690 detected unload events collected from 68,340 commercial vehicles operating across 11,066 U. S. Department of Transportation entities.

The report states that each detected unload event was assigned a suspicion score ranging from 0 to 1 through a three-stage detection pipeline that begins with vehicle weight telemetry and incorporates satellite imagery analysis and behavioral enrichment.

Three stage detection pipeline According to the report, the first stage analyzes vehicle weight telemetry to detect significant weight reductions while a truck is stopped.

The report states that the system processes the telemetry to identify sustained load changes while filtering out normal sensor variation before determining whether an unload event should advance to the next stage of analysis.

Chris Atkinson, CEO of Class8, told FreightWaves the vehicle weight telemetry originates from the truck’s suspension load-sensing system. He said axle weight data provided by original equipment manufacturers is the primary metric that initiates the company’s data-driven investigation.

Atkinson said satellite imagery is used later in the analytical process, while axle weight data remains the primary metric used to initiate an investigation. He also said Class8 does not rely on axle weight as a standalone source of truth and instead validates the data against additional proprietary metrics.

The report states that the second stage evaluates each detected unload location using satellite imagery. According to the report, the imagery is analyzed to identify locations that differ from patterns commonly associated with warehouse depots and established freight facilities.

The report states that unload events occurring in remote areas, empty lots and roadside pull-offs generally receive higher anomaly scores than unload events associated with warehouses, industrial parks and distribution centers. According to the report, the third stage applies behavioral enrichment before assigning a final suspicion score.

The report states that additional weighting is applied for off-hours activity, geographic proximity clustering and unload locations that differ from a vehicle’s historical lane data. Each detected unload event is then assigned a continuous suspicion score ranging from 0 to 1, with the score determining its operational risk category.

Risk scoring and findings The report groups suspicion scores into five operational categories. According to the report, scores between 0. 8 and 1. 0 are classified as Critical risk, scores between 0. 6 and 0. 8 are classified as High risk, scores between 0. 4 and 0. 6 are classified as Moderate risk, scores between 0. 3 and 0.

4 are classified as Low risk and scores below 0. 3 are classified as Minimal risk. According to the report, 42,570 detected unload events met the system’s highest risk thresholds. The report states that 3,988 events were classified as Critical risk and 38,582 were classified as High risk.

Another 255,920 events were classified as Moderate risk, while 438,796 were categorized as Low risk. The remaining 1,755,118 events were classified as Minimal risk. Geographic and time-of-day analysis The report states that the distribution of High- and Critical-risk unload events was not uniform across the United States.

Based on average suspicion scores, Arizona, North Dakota and New Mexico ranked among the highest-risk states identified in the analysis. At the DAT Key Market Area level, Bismarck, North Dakota; Flagstaff, Arizona; Ontario, California; Green River, Wyoming; and Reno, Nevada recorded the highest average suspicion scores.

The report also analyzed unload events by time of day. According to the findings, unload events occurring between 8 p. m. and 6 a. m. exhibited both higher frequencies of High- and Critical-risk activity and higher average suspicion scores than unload events occurring during standard business hours.

The report includes satellite image examples comparing locations assigned high suspicion scores with locations assigned low suspicion scores.

According to the report, high-scoring locations were characterized by remote terrain, empty lots and roadside pull-offs, while low-scoring locations were associated with recognized freight facilities, truck stops, distribution centers and industrial parks.

Validation and next steps Atkinson told FreightWaves that Class8 has validated the model against a statistically significant number of confirmed cargo theft investigations. He declined to disclose the number of cases or provide specific examples.

He also said the company is moving toward working with law enforcement agencies but has not yet begun those efforts. The report states that the analytical framework is designed to identify unload events that may warrant further investigation by analyzing truck