Sumitomo Heavy Industries (SHI) and NEC Corporation are collaborating on a near-miss detection system for construction sites that analyzes excavator camera footage and sensor data to spot hazards in real time. The companies aim to decrease accidents, strengthen safety management and give project teams reliable evidence for investigations and training.
The system processes video from cameras mounted on hydraulic excavators and pairs it with operational telemetry from the machines. SHI supplies the data through its SHICuTe ICT and IoT platform, while NEC contributes methods that turn raw footage into structured incident records with time and location context. The approach has been designed to surface moments that matter, like a bucket swing that passes close to a worker, low‑light movement around tracks or machine behavior that looks abnormal.
Early testing supports the concept. In September 2025, a proof of concept demonstrated that the system could identify and report near‑miss events from excavator‑mounted cameras, align them with machine state and generate concise summaries for review. That is a practical step from lab to jobsite since it confirms video quality, data synchronization and the ability to produce human‑readable outputs that safety teams can act on without sorting through hours of footage.
The collaboration relies on SHI’s operational expertise to focus detection on realistic risks. The system checks potential incidents against SHI’s knowledge of equipment behavior, unsafe actions and workflow patterns as well as historical accident data. Companies will be able to tune parameters to match site rules and local practices, which helps reduce false alarms and spotlights the risks that matter for a specific project type, such as confined urban digs or high‑reach demolition.
The roadmap sets development to begin this month, April 2026, with validation throughout fiscal 2026 and a practical rollout targeted for fiscal 2027. Initial detection will focus on worker‑machine interactions along with machinery behavior, then expand to conditions that commonly go unnoticed on first pass, such as unstable terrain or signaling issues. As the models take into account environmental cues and workflow violations, site teams can move from post‑incident reviews to timely interventions.
For project managers and safety leads, the value is clear. Reliable near‑miss reporting lowers dependence on witness statements and paper logs. Consistent video evidence tightens investigations, improves toolbox talks and gives legal teams better documentation when they need it. Over time, the data can inform refresher training and method statements, support insurance discussions, and guide procurement of camera and sensor packages that deliver the right coverage.
Contractors considering near‑miss detection should define success metrics early, whether that is a percentage reduction in recordable incidents, shorter investigation cycles or a set number of high‑quality training cases per month. Camera placement and data quality matter, so confirm fields of view around swing radii, blind spots and spoil heaps, and verify that telemetry feeds align to timestamps. Insist on site‑specific tuning so that alerts reflect your work patterns. Establish data governance up front, including retention, access control, and privacy policies for video and machine data. Validate that your policy complies with labor agreements and local regulations. An incremental rollout, starting with a small fleet on one or two representative job sites, allows teams to fine-tune thresholds and processes without disrupting operations.
SHI and NEC frame this as a functional safety upgrade that turns existing excavator cameras and sensors into a consistent near‑miss reporting tool. If the rollout meets its targets, contractors could achieve a clearer view of risk in daily operations and a tighter loop between field events, corrective actions and training.
(Note: AI assisted in summarizing the key points for this story.)