Technology Overview
Kajima Corporation, in collaboration with Pluszero, announced on January 20, 2026, the development of an AI model that automatically classifies and quantifies the operational activities of backhoes using footage from standard dashboard cameras. This innovative system categorizes machinery movements into eight specific actions—including excavation, loading, leveling, and standby time—by analyzing video data directly from the equipment. By integrating expert knowledge into the AI logic to correct common classification errors and filter unrealistic transitions, the system achieves a high accuracy rate of 97.1% for detecting standby periods. This allows site managers to pinpoint inefficiencies and optimize heavy equipment allocation without the need for manual labor-intensive monitoring. The technology is currently being deployed at large-scale earthwork projects to enhance overall productivity through data-driven management.
Source: https://www.kajima.co.jp/news/press/202601/20c1-j.htm
Background
Traditionally, monitoring the efficiency of heavy machinery required site engineers to perform manual “cycle time” measurements using stopwatches. In large-scale land development projects involving dozens of backhoes, tracking every unit’s activity is practically impossible due to the immense labor and time required. Furthermore, manual data collection is often subjective, leading to inconsistent results that make precise productivity analysis difficult. As the construction industry faces critical labor shortages and a pressing need for digital transformation (DX), there is an urgent demand for automated, objective data collection. Kajima’s AI model addresses these challenges by repurposing existing dashboard camera footage into actionable data. This eliminates the physical burden on young engineers, who previously spent hours on-site for manual tracking, and provides a transparent, data-driven foundation for optimizing earthwork operations and reducing “hidden” idle time.
Advantages
This AI-driven approach significantly outperforms manual methods by providing 24/7 comprehensive data across all active machinery with high objectivity and minimal labor cost.
| Feature | Conventional Method (Manual) | AI Dashboard Camera Analysis |
|---|---|---|
| Measurement Method | Stopwatches and visual observation | AI analysis of camera footage |
| Required Labor | 1-2 dedicated staff members | Zero (Automated) |
| Scope | Limited to 1-2 units at a time | All active units simultaneously |
| Data Accuracy | Subjective and prone to human error | Objective and high (97% for standby) |
| Data Integration | Requires manual data entry | Automatic quantification and linking |
Potential Applications
The potential for this technology extends beyond simple efficiency tracking; it paves the way for fully autonomous construction management. By accumulating machine learning data across various sites, the AI’s precision will continue to improve, eventually allowing for real-time adjustments to heavy equipment fleets. In the future, this data could be integrated with autonomous machinery or digital twins to simulate and execute the most efficient construction sequences automatically. Furthermore, automating daily reporting based on AI-generated logs will significantly reduce administrative overhead, allowing engineers to focus on high-value decision-making and safety management rather than repetitive data entry.
日本語ページはこちら： https://hatara-kuma.com/20260108-2/