View source for Logistics Safety AI

== Event Gates Logistics Safety AI ==
'''Event Gates Logistics Safety AI''' is an advanced industrial vision system designed to enhance safety in busy logistics environments. By combining deep-learning image analytics with real-time monitoring, it detects potential hazards—such as collisions between forklifts, trucks, pedestrians, or falling loads—and intelligently triggers vehicle deceleration before an accident occurs. The system also ensures compliance by monitoring the use of personal protective equipment (e.g., helmets, hi-vis vests) and vehicle safety measures like curtains and wheel chocks, automatically logging safety violations and issuing notifications as needed.

The solution is tailored to high-risk zones like loading docks, sideloading bays, and high-rack corridors, where precise coordination between people, vehicles, and equipment is crucial. It integrates seamlessly with existing infrastructure—such as vehicle restraint systems and automated alerts—without requiring wearable tags or manual intervention. This enables organizations to proactively manage workplace safety, reduce accident risk, and ensure operational continuity—all while collecting valuable video and event data for audit and continuous improvement.

=== Features ===
Logistics Safety AI is a [[:Category:VIS|VIS]] solution therefore it has all [[:Category:VIS Features|VIS features and functions]] where applicable:

* Disconnected solution, deployed to the edge devices on the crane trolley or bridge, does not need internet or corporate network access.
* Proprietary high performance person detection model Event Gates PF, specially trained for top down view in industrial environments.
* [[Vislet Model Fusion]] allows detections of multiple models interplay together.
* Built-in trustworthiness features: [[Heartbeat]], [[Short Term Memory (STM)|Short Term Memory]], [[IPIO#Emergency Preset|Emergency Preset]] of [[IPIO]].
* [[Logging]] features including [[System Logs|system logs]], [[Event Logs|event logs]] and optional emailing features.
* All flexibility of [[Regions of Interest (ROI)|regions]](ROI), including polygons and multiple regions.
* Built-in [[IPIO#Mute Input and Preset|mute]] with [[IPIO]].
* Tracking
* Person counting helps secure unmonitored blind spots.
* UWB based wireless device to slow down vehicles such as electric forklifts or alert individual operators inside the vehicle.<ref>Subject to vehicle capability.</ref>

=== Risk assessment ===

==== Forklift–pedestrian collisions ====
Forklifts and other industrial vehicles often operate in shared spaces with on-foot personnel, creating constant collision risks. These accidents typically occur in intersections, corners, or blind spots where the operator’s visibility is limited. Even when lanes are marked, distractions, poor lighting, or noncompliance with walking routes can lead to serious or fatal accidents. Logistics Safety AI continuously monitors these zones and detects the presence of both pedestrians and vehicles in proximity, enabling early intervention such as deceleration triggers or alerts before a collision occurs.

==== Truck drivers reversing into docks without visibility ====
When truck drivers approach loading docks in reverse, they often lose visual contact with the area behind the vehicle. This becomes extremely dangerous if pedestrians or forklifts are moving or working between the dock and the trailer. Human presence in this blind spot is nearly invisible to the driver, especially if mirrors or cameras are not used properly. Logistics Safety AI detects humans in this zone during reversing maneuvers and can trigger alerts, vehicle restraints, or automatic deceleration to prevent deadly accidents.

==== Obstructed vision during side loading/unloading ====
Side-loading operations, particularly in narrow bays or tightly packed environments, reduce the forklift operator’s visibility of the space behind the truck. This is especially hazardous when workers are performing tasks in those areas or passing through unexpectedly. Even with mirrors and training, obstructions caused by trailer walls, loads, or warehouse structure can hide human presence. Logistics Safety AI identifies these occluded zones and monitors for pedestrians near side-loading operations, generating warnings or logging events when danger conditions are met.

==== Falling objects caused by order pickers ====
In logistics centers with high-rack storage, order pickers—manual or robotic—operate at heights to retrieve or place items. Improper handling or unstable stacking may cause packages, boxes, or equipment to fall into adjacent aisles or onto unsuspecting personnel below. These accidents are difficult to anticipate and often go undocumented. Logistics Safety AI uses spatial awareness to detect nearby people, identify hazardous object motion, and log or alert on events where falling objects may pose a safety risk.

==== Personnel working behind or between goods ====
In some loading scenarios, workers position themselves behind or between stacked pallets or cargo to inspect, adjust, or prepare items for transport. These individuals are often fully or partially obscured from the forklift operator’s view, particularly when large items block lines of sight. This creates a high-risk situation where someone may be struck or pinned during routine loading or unloading. Logistics Safety AI monitors these high-obstruction zones and detects human presence even when visibility is minimal, helping prevent crushing or impact injuries.

==== Non-use of personal protective equipment (PPE) ====
Warehouse and logistics operations frequently require the use of safety gear such as high-visibility vests, helmets, or protective footwear. When this gear is not worn, workers become less visible and more vulnerable—both to vehicle operators and to automated safety systems. In dimly lit or fast-paced environments, missing PPE can be the difference between detection and accident. Logistics Safety AI includes PPE compliance monitoring and can detect missing or incorrect gear in real time, issuing alerts or logging violations as needed.

==== Unsecured vehicles at loading bays ====
Trucks parked at docks must be properly restrained to prevent unintended movement during loading or unloading. If wheel chocks or dock-lock systems are not used—or are removed too early—the vehicle may shift unexpectedly, causing serious injury or damage. This risk is especially high when heavy forklifts are entering or exiting the trailer. Logistics Safety AI verifies the presence of safety measures like wheel chocks or locks and checks whether vehicle movement occurs in an unsafe state, enabling automated alerts or integration with restraint systems.

==== Congestion in high-traffic corridors ====
Logistics environments often include narrow aisles or intersections where pedestrians, forklifts, AGVs, and order pickers converge. These areas become safety-critical due to the density of movement and the complexity of interactions. Visual obstructions, noise, and fast-paced operations make it hard for workers to anticipate others’ movements. Logistics Safety AI detects objects and people in shared zones, identifies risky proximity or convergence, and supports alert systems to reduce the chance of collisions or unsafe crowding.

==== Unnoticed persons in restricted or automated zones ====
Certain warehouse zones are designated for robots, AGVs, or heavy equipment and are not intended for human access. However, unauthorized or accidental entry by workers into these areas can result in near misses or serious injuries. Traditional access control measures (like warning signs or physical barriers) are not always effective. Logistics Safety AI continuously monitors these zones for unauthorized human presence and can trigger alarms, stop automation, or log the intrusion for further action.

=== Solutions ===
[[File:Forklift – pedestrian proximity.png|thumb|'''Forklift – pedestrian proximity''']]

==== Forklift–pedestrian proximity detection ====
This solution dynamically links a virtual safety zone to each detected forklift. The region moves with the vehicle and is continuously evaluated for overlap with detected pedestrians. If a person’s bounding box intersects with the forklift’s dynamic zone, VIS can trigger alerts or initiate automated slow-down actions on the vehicle. This approach is particularly effective in large operational areas and high-traffic junctions where blind spots exist for both the forklift operator and nearby pedestrians. Overhead or wall-mounted cameras provide the external viewpoint needed to observe both parties clearly and mitigate blind-zone risk.
[[File:Co-operation area protection.png|thumb|Co-operation area protection]]

==== Co-operation area protection ====
In designated co-operation zones—where vehicles and people are expected to work in close proximity—fixed virtual regions are configured to monitor shared activity. When both a vehicle and a person are detected within the same region, the system triggers an alert or deceleration action to reduce the likelihood of a collision. This solution is ideal for loading/unloading platforms, sorting areas, or compact warehouse zones where humans and machines must operate side by side.

==== Multi-angle region grouping ====
VIS supports grouping regions from multiple camera views to create a logical shared zone, even when individual cameras cannot cover the entire area. This allows risk detection in spatially complex environments—such as both sides of a truck, or the front and rear of an order picker—where visibility from a single viewpoint is not sufficient. Region groups are evaluated collectively to identify cross-camera presence of people and vehicles, enabling a more complete and accurate understanding of the situation.

==== On-vehicle threat detection ====
For vast or open-air logistics environments where static infrastructure is limited, VIS can operate on vehicle-mounted cameras to detect people or hazards directly from the moving vehicle’s perspective. This mobile setup allows real-time threat detection in container yards, outdoor storage, or large indoor areas with dynamic layouts. It enables smart forklifts, tuggers, or other logistics vehicles to become self-aware of their surroundings and respond to nearby human presence.
[[File:Counter-region.gif|thumb|Counter region]]

==== Perimeter tracking and person counting ====
In restricted-access zones or areas without full camera coverage, VIS uses perimeter-based tracking and entry/exit counting to maintain situational awareness. By monitoring movement across defined entry lines and keeping a live count of people present within a closed area, the system can infer whether personnel remain inside—even if visual contact is lost. This is critical in sealed docks, trailer interiors, or back-of-rack zones where direct monitoring is difficult but worker presence must still be accounted for.

Information on video section displays the person (object) count inside a specific region.

== Use Cases ==

=== Side loading bays ===

=== Loading docs and ramps ===
[[File:Loading docks and ramps.jpg|thumb|Loading dock]]
Loading docks are critical transfer points where trucks reverse into position so forklifts can load or unload goods directly through the back of the trailer. The dock entrance typically includes a raised platform, and personnel work on the indoor side to prepare or handle materials. During loading operations, forklifts frequently move in and out of the truck from this platform.

One of the most serious safety risks arises when a truck approaches the dock in reverse while personnel are still present in or near the path of the vehicle—especially near the dock entrance or between the dock and the trailer. This can occur if workers are finishing up tasks, inspecting the ramp, or simply passing by. Since truck drivers cannot see behind their vehicle during this maneuver, a person in the danger zone may go unnoticed until it’s too late.

To address the risk of trucks reversing into occupied areas, Logistics Safety AI monitors the space in front of the dock using '''one camera mounted above the dock door''', facing outward. This perspective allows the system to detect both incoming trucks and any personnel present in the approach zone.

Two conditions must be met for an '''alarm''': a person must be detected in the predefined danger zone '''and''' a truck must be reversing or present. If only one of these elements is detected—either a person or a truck—the system does not raise an alarm, but can still activate a separate '''non-critical output'''. This output can be used by external automation systems, for example to display ramp status, log activity, or prepare preemptive safety measures without triggering a full alert. This layered approach helps balance safety with operational flexibility.
[[File:Pop up barrier.gif|thumb|Pop up barrier]]
When both conditions are met, the system triggers configurable countermeasures, such as flashing lights, sound alarms, or an automatic barrier that rises to block the truck’s path. These interventions aim to stop the vehicle before it reaches the dock and ensure the safety of anyone in the area.

This setup works autonomously, without requiring manual confirmation or sensor input from the truck. It’s especially suitable for environments where drivers or visitors are unfamiliar with the site and may not expect pedestrian activity near the loading zone.

=== High rack corridors ===

== Notes ==
<references />
[[Category:Safety AI]]