ISO 13855 is an international standard that specifies how to calculate the minimum safety distance between protective devices, such as light curtains and safety sensors, and hazardous areas of machinery, based on the approach speed of the human body. It is one of the fundamental standards used when designing safety systems for presses, cutting machines, industrial robots, and many other types of production equipment. This article explains the core requirements of ISO 13855 and the role that Computer Vision (AI Cameras) can play in modern machinery safety applications. WHAT IS ISO 13855? ISO 13855 (full title: Safety of machinery — Positioning of safeguards with respect to the approach speeds of parts of the human body) is an international machinery safety standard that specifies how protective devices should be positioned so that a machine can stop before any part of a person’s body can reach the hazardous area. In Vietnam, the equivalent adopted standard is TCVN 7386:2011 (ISO 13855:2010) – Safety of machinery — Positioning of safeguards with respect to the approach speeds of parts of the human body, which belongs to the group of machinery safety standards issued by the Ministry of Science and Technology. This standard primarily applies to electro-sensitive protective equipment, such as light curtains, safety laser scanners, and similar presence-detection devices. SAFETY DISTANCE CALCULATION ACCORDING TO ISO 13855 The general formula specified in ISO 13855 is S = K × T + C, where S is the minimum safety distance, K is the approach speed, T is the total stopping time of the system, and C is an additional distance based on the detection capability of the protective device. * S - The minimum distance (mm) from the detection zone of the protective device to the hazardous area. * K - The assumed approach speed of the human body. The standard specifies two commonly used values: 2000 mm/s for hand/arm movement and 1600 mm/s for whole-body walking speed (applied when the distance calculated using K = 2000 mm/s exceeds 500 mm). * T - The total system response time, including the time required for the protective device to detect a signal, the processing time of the safety controller, and the time required for the machine to come to a complete stop. * C - An additional distance that depends on the detection resolution of the device, such as the spacing between beams in a light curtain. For whole-body detection using multiple beams, C is typically set to a minimum of 850 mm to compensate for arm reach. In other words, the slower the protective device responds or the longer the machine takes to stop, the farther the protective device must be installed from the hazardous area. HOW IS ISO 13855 RELATED TO OTHER FACTORY SAFETY STANDARDS? ISO 13855 is commonly used together with ISO 13857 (static geometric safety distances to prevent access to hazardous areas), ISO 13849-1 (design of safety-related control systems), and IEC 61496 (technical requirements for electro-sensitive protective equipment). A common mistake is using ISO 13857 to calculate safety distances based on approach speed. In practice, ISO 13857 only specifies static geometric distances, such as openings in protective fences, whereas ISO 13855 is specifically intended for active protective devices with response times, such as light curtains and laser scanners. The safety level (PL/SIL) of the controller under ISO 13849-1 also directly affects the response time T used in the formula above. WHAT CHALLENGES ARISE WHEN APPLYING ISO 13855 IN PRACTICE? Compliance with the safety distances specified by ISO 13855 can sometimes reduce operational convenience or require factories to use additional muting mechanisms to maintain productivity. For many production lines with limited floor space, the distance S calculated using the formula may be greater than the ideal working space available. When materials need to pass through the protected area of a light curtain during operation, factories often have to use a muting mechanism, which temporarily disables the detection zone to allow materials to pass through. If the muting window is configured incorrectly or extended to increase production speed, practical gaps may appear in the protected area even though the system still complies with the originally calculated safety distance. Another limitation is that devices covered by ISO 13855 typically detect intrusion across a fixed plane and cannot distinguish between a person’s hand and materials unless a separate processing mechanism is implemented. WHAT ROLE DOES COMPUTER VISION PLAY IN MACHINERY SAFETY? Computer Vision (AI Cameras) can add an intelligent detection layer alongside protective devices compliant with ISO 13855, helping distinguish people from materials, expand the monitored area, and provide data for safety analysis. However, it generally does not directly replace the certified machine-stopping function of these devices unless the Computer Vision system itself is certified according to the relevant safety standards. DISTINGUISHING PEOPLE FROM MATERIALS Light curtains and other sensors used under ISO 13855 treat all objects crossing the detection zone in the same way, which means muting mechanisms are required when materials need to pass through. AI Cameras analyze object shapes and can distinguish human hands or bodies from workpieces and materials, helping reduce the need for muting in many situations and allowing the protected area to remain continuously active for longer periods. EXPANDING THE MONITORING AREA Devices used under ISO 13855 typically protect a single plane or approach direction. A camera can simultaneously monitor multiple approach angles, helping detect people approaching from the side or rear - directions that a standard light curtain may not cover. REDUCING THE RISK OF BYPASSING In production environments where there is strong pressure to maintain productivity, light curtain sensors may be covered or otherwise bypassed to avoid interrupting operations. AI Cameras do not rely on physical light beams and are therefore more difficult to disable in the same way. They can also retain images when interference or bypassing behavior is detected. PROVIDING DATA FOR SAFETY MANAGEMENT Devices used under ISO 13855 primarily generate immediate machine-stop signals without recording the context surrounding an event. AI Cameras can capture images whenever a risk is detected, allowing HSE teams to analyze near-miss incidents and continuously improve safety procedures over time. CAN COMPUTER VISION REPLACE ISO 13855-COMPLIANT PROTECTIVE DEVICES? No, unless the Computer Vision system is designed and certified to meet the corresponding safety level, such as Type 4, PLe, or SIL3, under the relevant standards. Type 4 light curtains compliant with IEC 61496, combined with safety controllers meeting PLe/SIL3 requirements under ISO 13849-1, have undergone rigorous testing for reliability and response time. These characteristics provide the basis for calculating safety distances according to ISO 13855. A conventional AI Camera system that has not been certified under the corresponding functional safety standards should therefore not be considered an independent, certified machine-stopping device. For this reason, the recommended approach is to build a multi-layered protection system: certified safety devices perform the emergency machine-stopping function in accordance with applicable standards, while AI Cameras serve as an additional monitoring layer—providing early detection, reducing false alarms, expanding the monitored area, and supplying data for safety management. WHERE SHOULD BUSINESSES START WHEN INTEGRATING COMPUTER VISION INTO EXISTING SAFETY SYSTEMS? Before adding AI Cameras, businesses should review whether their existing protective systems comply with the safety distances required by ISO 13855, particularly at locations where muting mechanisms are being used or where false alarms occur frequently. These are often the areas where an additional Computer Vision monitoring layer can provide the greatest value—not by replacing certified protective devices, but by reducing the gaps created by muting mechanisms or geometric blind spots. The next step is to clearly define the role of AI Cameras within the safety architecture: whether the system will only issue alerts and record data, or whether detection signals will be integrated into the machine-stopping chain through the existing safety controller. This choice determines which additional certification or verification requirements the system may need to meet and should be considered together with the factory’s engineering team and safety consultants from the design stage. EYEFIRE can support assessments of existing protective systems at individual production locations and recommend suitable AI Camera installation positions and configurations based on the characteristics of each production line—as an additional layer operating alongside existing certified protective devices, without changing the core safety architecture already established by the factory.

AI Camera or Passive Infrared (PIR) Sensor: Which is the more effective solution for hazardous-area intrusion warnings? In factories and logistics facilities, detecting intrusion into hazardous areas - around operating machinery, high-voltage zones, conveyor systems, or lifting equipment - is a fundamental requirement of occupational safety systems. Passive Infrared (PIR) sensors have long been a popular choice due to their low cost, while AI Cameras powered by computer vision are increasingly being considered by businesses. This article compares the two technologies to help businesses choose the most suitable solution. HOW DOES A PASSIVE INFRARED (PIR) SENSOR WORK? PIR (Passive Infrared) is a passive sensor that detects movement based on differences in infrared radiation between the human body and the surrounding environment. When a person moves through the sensor's detection area, this thermal difference causes the sensor to generate an electrical signal, triggering an alarm or a machine-stop command. The advantages of PIR sensors include their simple structure, low power consumption, low investment cost, and ability to operate in low-light conditions without requiring additional illumination. WHAT ARE THE LIMITATIONS OF PIR SENSORS IN FACTORIES? PIR sensors have difficulty distinguishing people from other heat sources, are susceptible to environmental temperature changes, cannot determine precise locations, and do not capture images for verification. CANNOT DISTINGUISH PEOPLE FROM OTHER HEAT SOURCES PIR detects differences in heat rather than analyzing the shape of an object. In factories with furnaces, forklifts, or heat-emitting equipment, these sources may cause false alarms or result in actual events being missed. SENSITIVE TO AMBIENT TEMPERATURE When the ambient temperature approaches human body temperature, or when an area is exposed to direct sunlight, the thermal difference between a person and the background becomes less distinct, reducing detection accuracy. CANNOT DETERMINE PRECISE LOCATIONS PIR can only determine whether there is movement within its detection area. It cannot identify a person's exact location, direction of movement, or distance from a hazardous zone. In areas that require multiple warning levels, or where hazardous zones change according to equipment movement, such as around overhead cranes, PIR sensors are difficult to apply effectively. PRONE TO BLIND SPOTS Each sensor has a limited detection angle and range. To cover a large or complex-shaped area, businesses need to install multiple sensors, increasing both equipment costs and calibration efforts. CANNOT STORE IMAGES FOR INVESTIGATION When a PIR sensor triggers an alarm, the system only records that "movement was detected" without providing actual images. This makes it difficult for HSE teams to verify alerts or analyze incidents. HOW DOES AN AI CAMERA DETECT INTRUSION INTO HAZARDOUS AREAS? AI Cameras use Computer Vision to analyze images in real time, accurately detect people, determine their locations, and compare those locations with predefined warning zones, rather than simply detecting generic movement signals. DISTINGUISHES PEOPLE FROM OBJECTS AND OTHER HEAT SOURCES By analyzing object shapes and visual characteristics, AI Cameras can distinguish people from equipment or heat-emitting materials, significantly reducing false alarms compared with PIR sensors. DETERMINES LOCATION AND SUPPORTS MULTIPLE WARNING ZONES AI Cameras can define multiple warning zones—such as an early-warning zone, hazardous zone, and emergency-stop zone—within the same field of view. In areas where hazardous zones change dynamically, such as beneath a moving overhead crane, the system can also dynamically adjust warning zones according to the equipment's position. WIDER MONITORING COVERAGE A wide-angle camera can cover a significantly larger area than a single PIR sensor, helping reduce the number of devices required and minimizing the risk of blind spots. STORES IMAGES FOR VERIFICATION AND INVESTIGATION Each alert can be accompanied by an image or video recording of the detection event, providing valuable data for HSE teams to verify incidents and analyze near-misses over time. ONE PLATFORM FOR MULTIPLE USE CASES The same camera infrastructure can be expanded to support PPE monitoring, detection of people in hazardous areas around overhead cranes and forklifts, or abnormal behavior detection without requiring a separate system for each use case. DOES AI CAMERA TECHNOLOGY HAVE ANY LIMITATIONS? AI Camera performance depends on lighting conditions, installation angles, and typically requires a higher initial investment than PIR sensors. In extremely dark or dusty environments, infrared (IR) cameras or additional lighting may be required to maintain sufficient image quality. If a person is completely obstructed by a large object and is outside the camera's field of view, the system cannot detect them until they become visible. In addition, the initial investment—including cameras and AI processing infrastructure—is generally higher than that of a single PIR sensor, although the actual cost per monitoring point should be evaluated based on coverage area and system scalability. QUICK COMPARISON OF PASSIVE INFRARED (PIR) SENSORS AND AI CAMERAS WHERE SHOULD EACH TECHNOLOGY BE APPLIED IN A FACTORY? PIR is suitable for: secondary walkways, warehouse doors, low-traffic areas, and locations where the only requirement is to determine whether "someone is present" and the investment budget is limited. AI Camera is suitable for: stamping machines, cutting machines, industrial robots, overhead cranes, and forklifts - areas where hazardous zones involve multiple risk levels, frequently change position, or where businesses require visual data for incident investigation and HSE reporting. For factories with different types of equipment and varying levels of risk, combining both technologies according to the requirements of each area often provides better investment efficiency than relying on a single solution throughout the entire facility. SO, SHOULD BUSINESSES CHOOSE PIR OR AI CAMERAS? PIR is suitable for areas with limited budgets, stable environmental temperatures, few sources of interference, and where the only requirement is to determine whether "someone is present" at a fixed point. AI Cameras are more suitable when high accuracy, multiple warning zones, visual data for investigations, or a scalable safety monitoring platform for future use cases are required. In many cases, the two technologies can complement each other: PIR provides fast, low-cost warnings at secondary monitoring points, while AI Cameras focus on critical areas that require higher accuracy and more advanced analytical capabilities. HOW DOES EYEFIRE APPROACH THIS USE CASE? EYEFIRE applies Computer Vision AI to detect people entering hazardous areas in real time, define warning zones according to multiple risk levels, and trigger appropriate alerts for different situations. Cameras can be installed at fixed locations in high-risk areas or mounted on moving equipment such as overhead cranes and forklifts, allowing the monitoring zone to move together with the equipment. Detected events are stored together with images, enabling HSE teams to verify alerts, analyze near-misses, and continuously improve safety procedures. If a business is considering PIR sensors, AI Cameras, or a combination of both, EYEFIRE can assess the actual operating environment and recommend an appropriate configuration for each area of the factory.

Overhead cranes are widely used in steel, mechanical engineering, automotive, cement, and other heavy industrial facilities to transport heavy materials. Because both the crane and its load continuously move throughout the facility, the hazardous zone also changes accordingly and can appear in different locations. AI Cameras for overhead crane safety monitoring can help detect people beneath suspended loads or entering hazardous zones in real time. The system uses Computer Vision to identify people, determine their position relative to the load, and trigger alerts when a potential risk is detected. Unlike conventional surveillance cameras that simply record footage, AI Cameras can actively participate in risk detection. This approach is particularly suitable for areas where hazardous zones continuously change as the overhead crane moves. WHAT ARE THE COMMON SAFETY RISKS WHEN OPERATING OVERHEAD CRANES? During operation, overhead cranes may transport loads weighing several tons in different directions across a facility. Loads can change position, swing while moving, or obstruct visibility between the crane operator and the area below. Hazardous situations can occur when workers stand or walk beneath suspended loads, move too close to a moving load, or enter areas that the crane operator cannot see clearly. When the distance between a person and a load changes rapidly, the available time to recognize the danger and respond also becomes shorter. At a steel processing facility where EYEFIRE implemented its solution, heavy steel coils were continuously transported horizontally and vertically throughout the production area. Detecting workers beneath the crane or entering hazardous zones around the steel coils was therefore an important requirement for the safety system. WHAT IS A DYNAMIC HAZARDOUS ZONE AROUND AN OVERHEAD CRANE? A dynamic hazardous zone is an area of risk whose position changes according to the movement of the overhead crane and its load. As the load moves, the area that needs to be monitored moves with it rather than remaining fixed at one location on the factory floor. This is what makes overhead crane safety different from many types of stationary industrial equipment. With a press machine or machining equipment, businesses can define a hazardous zone in advance and use barriers, light curtains, or sensors to control access. With an overhead crane, a location that is safe at one moment can become hazardous only a few seconds later when the load moves into that area. The monitoring system therefore needs to determine not only whether a person is present in the facility, but also where that person is in relation to the current position of the load. WHY DOES TRADITIONAL OVERHEAD CRANE SAFETY MONITORING STILL HAVE LIMITATIONS? In many factories, overhead crane safety still depends on the experience of crane operators, personnel observing from the factory floor, radio communication, warning alarms, and other safety devices. These measures remain necessary, but their ability to detect hazards may depend heavily on human visibility and concentration. Large loads, factory structures, and production equipment can also create blind spots. As a result, crane operators may not always be able to see the entire area beneath and around the suspended load. Fixed sensors can provide an additional layer of protection, but they are generally designed to monitor a specific location or predefined area. When the crane and its load continuously change position, monitoring the entire operating area becomes more complex. HOW DOES AN AI CAMERA DETECT PEOPLE IN AN OVERHEAD CRANE HAZARDOUS ZONE? AI Cameras detect people in hazardous zones by analyzing images in real time, identifying people, and determining their position relative to the warning zone around the load. When a person enters an area defined as hazardous, the system can automatically trigger an alert. Cameras can be installed directly on the overhead crane so that their field of view moves together with the equipment. Camera footage is transmitted to an AI processing system, where Computer Vision models continuously analyze the presence and location of people. In the solution implemented by EYEFIRE at a steel processing facility, AI automatically creates hazardous warning zones around steel coils and divides them into multiple alert levels. The cameras are installed on the overhead crane, allowing them to move together with the equipment during operation. As a result, the system does not simply answer the question, “Is there a person in the factory?” More importantly, it determines “Is anyone currently inside the hazardous zone around the load?” CAN AI CAMERAS TRACK HAZARDOUS ZONES WHILE THE OVERHEAD CRANE IS MOVING? Yes. When cameras are appropriately positioned and move together with the overhead crane, the monitored zone can be defined according to the actual position of the equipment and its load rather than remaining fixed at a specific location on the factory floor. As the overhead crane changes position, the system continues analyzing images to detect people within the monitored area. This is how AI Cameras can support dynamic hazardous zone monitoring during overhead crane operations. This approach is particularly useful in facilities where overhead cranes operate across large areas. Instead of creating numerous fixed warning zones, the system can monitor risks associated with the actual position of the moving load. HOW CAN AI CAMERAS REDUCE BLIND SPOTS BENEATH OVERHEAD CRANES? A single camera may not always be sufficient to observe the entire area beneath a large overhead crane. The load itself, machinery, or structural elements within the facility may obstruct people from view and create blind spots. One approach is to install multiple cameras facing different directions to expand the overall field of view. In an actual EYEFIRE deployment, the system uses four AI Cameras positioned in four directions beneath the overhead crane, helping continuously monitor the area around the load and reduce blind spots during operation. Because the cameras are installed directly on the overhead crane, the monitoring system moves together with the equipment. This differs from relying solely on fixed surveillance cameras to monitor the factory floor. HOW DOES AN AI CAMERA ISSUE ALERTS WHEN SOMEONE ENTERS A HAZARDOUS ZONE? Not every position around a suspended load represents the same level of risk. A person approaching a load may require a different warning from someone standing directly within a high-risk area. AI Camera systems can define multiple levels of warning zones around the load. As a person moves from a safe zone into a warning or hazardous zone, the system can generate an alert corresponding to the level of risk. In the system implemented by EYEFIRE, hazardous zones around steel coils are divided into multiple levels to support appropriate responses for different situations. This approach allows the system not only to detect people but also to determine the level of attention required based on their position. CAN AI CAMERAS BE INTEGRATED WITH AN OVERHEAD CRANE CONTROL SYSTEM? The camera serves as the image acquisition layer, while analysis is performed by the AI processing system. Images can be transmitted to an AI Hub, where the AI Engine detects people, identifies hazardous zones, and generates warning signals. Depending on the system design and the factory's safety requirements, warning signals can be sent to the control system to support appropriate actions. Integration needs to be designed and evaluated according to the existing control system, operational requirements, and specific safety standards of each facility. The processing flow can be illustrated as follows: Camera → AI Hub → Person Detection → Hazardous Zone Identification → Alert → Control System As a result, AI Cameras do more than simply record footage for review after an incident. The system can become an active detection layer that helps factories identify risks as soon as they emerge. DO AI CAMERAS REPLACE EXISTING OVERHEAD CRANE SAFETY DEVICES? AI Cameras should not be considered a complete replacement for existing safety devices and procedures. For overhead cranes and other heavy lifting equipment, a multi-layered safety approach remains important. Measures such as operating procedures, safety zoning, audible warnings, travel limits, and control systems continue to serve their own functions. AI Cameras add the ability to identify people and analyze their position relative to the actual hazardous zone. By combining these layers of protection, businesses can reduce their dependence on a person having to observe the entire area at all times. At the same time, the system gains an additional capability to detect situations that fixed monitoring methods may find difficult to identify. HOW CAN AI CAMERAS HELP MANAGE NEAR-MISSES BENEATH OVERHEAD CRANES? The value of AI Cameras goes beyond real-time alerts. Detected events can be stored so that HSE teams can review situations in which people frequently approach or enter hazardous zones. When this data is aggregated over time, factories can identify areas where warnings frequently occur, periods with higher numbers of hazardous interactions, and situations that repeatedly happen. This provides valuable data for analyzing near-misses and identifying their causes before a serious incident occurs. If alerts repeatedly occur in the same location, the problem may not simply be related to worker behavior. The layout of pedestrian routes, lifting procedures, or the movement patterns of people and equipment may also need to be reviewed. HOW HAS EYEFIRE IMPLEMENTED AI CAMERA MONITORING FOR OVERHEAD CRANE SAFETY? At a facility specializing in steel coil processing, EYEFIRE implemented an AI Camera system for overhead crane safety monitoring to detect people in areas beneath suspended loads. Four cameras were positioned in four directions and moved together with the overhead crane, while AI automatically created warning zones around the steel coils. Images were transmitted to the AI Hub for real-time analysis. When a person was detected within a hazardous zone, the system generated alerts according to the defined risk level and could send signals to the control system based on the deployment configuration. The solution enables continuous monitoring of the area beneath the overhead crane, reduces dependence on manual supervision, and supports faster responses when hazards are detected. This model can also be considered for similar lifting equipment in factories, ports, warehouses, logistics facilities, and other industrial environments. WHEN SHOULD BUSINESSES CONSIDER AI CAMERAS FOR OVERHEAD CRANE SAFETY MONITORING? AI Cameras are worth considering when overhead cranes operate across large areas, suspended loads frequently move through locations where people work, or the facility contains multiple blind spots. They are also suitable for businesses seeking to add real-time person detection capabilities to their existing safety systems. Before deployment, businesses should assess the crane's operating range, load types, worker locations, blind spots, lighting conditions, and existing control systems. These factors determine the appropriate number of cameras, installation positions, warning zones, and integration approach. EYEFIRE can assess actual operating environments and develop AI Camera solutions tailored to the specific characteristics of each facility. The objective is not only to detect people beneath overhead cranes but also to provide an additional layer of proactive monitoring throughout lifting operations.