Crane full stroke operation safety monitoring system and method

Abstract

The application discloses a crane full-stroke operation safety monitoring system and method, which is applied to the technical field of crane safety monitoring, and a start-stop control module determines the current load state of a crane based on current load contact signals and current electrical signals collected by a front sensing module, current operation state parameters and current operation environment parameters, and enables a safety monitoring module when the current load state is a loaded state; the safety monitoring module adopts a risk judgment model to perform target detection and risk judgment on crane operation area images collected by an image sensing module to obtain a personnel operation risk judgment result, and sends an operation risk alarm instruction to an alarm linkage module when the personnel operation risk judgment result represents that there is a personnel operation risk; and the alarm linkage module controls different types of alarm modes corresponding to the personnel operation risk to perform linkage alarm, so that the intelligent degree of safety monitoring and the operation safety of the crane are improved through collaborative operation among the modules.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of crane safety monitoring, and particularly relates to a crane full-stroke operation safety monitoring system and method. BACKGROUND

[0002] At present, the crane safety monitoring management system mainly relies on the monitoring of crane operation parameters, including weight, lifting height, stroke, wind speed, etc. The device networking and emergency braking are realized through 4G / 5G / WiFi. Some cranes are additionally equipped with a video system, which expands the operator's field of view through the camera and high-resolution monitor, realizes real-time monitoring of the crane working process, and can store a certain amount of video information in real time, which can be called and viewed at any time. However, the monitoring and the crane are not directly linked, but the operator judges the crane through the video information and then performs corresponding operations. SUMMARY

[0003] The present application provides a crane full-stroke operation safety monitoring system and method to solve the problem of insufficient intelligence of crane safety monitoring in the prior art. The technical scheme provided by the present application is as follows:

[0004] On the one hand, the present application provides a crane full-stroke operation safety monitoring system, which comprises a front sensing module, an image sensing module, a start-stop control module, a safety monitoring module and an alarm linkage module.

[0005] The front sensing module is used to collect the current load contact signal and the current electrical signal of the crane electrical cabinet, and the current operating state parameter and the current operating environment parameter of the crane, and send them to the start-stop control module.

[0006] The start-stop control module is used to determine the current load state of the crane based on the current load contact signal and the current electrical signal sent by the front sensing module, and the current operating state parameter and the current operating environment parameter. When the current load state is a loaded state, the safety monitoring module is enabled.

[0007] The image sensing module is used to collect the crane operation area image and send it to the safety monitoring module.

[0008] The safety monitoring module is used to adopt a risk judgment model to perform target detection and risk judgment on the crane operation area image to obtain a personnel operation risk judgment result. When the personnel operation risk judgment result represents that there is a personnel operation risk, the safety monitoring module sends an operation risk alarm instruction to the alarm linkage module.

[0009] The alarm linkage module is configured to control different types of alarm modes corresponding to the personnel operation risk to perform linkage alarm when receiving the operation risk alarm instruction sent by the safety monitoring module.

[0010] Optionally, the start-stop control module is configured to calculate a current load value of the crane by using a load evaluation model through a weighted fusion strategy based on the current load contact signal and the current electrical signal, and the current operation state parameter and the current operation environment parameter; determine a load state corresponding to a load value interval in which the current load value is located as the current load state of the crane based on a corresponding relationship between different load states and load value intervals; enable the safety monitoring module when the current load state is a loaded state, and disable the safety monitoring module when the current load state is an unloaded state.

[0011] Optionally, the start-stop control module is configured to preprocess the current load contact signal and the current electrical signal, and the current operation state parameter and the current operation environment parameter; input the preprocessed current load contact signal and the current electrical signal, and the current operation state parameter and the current operation environment parameter into the load evaluation model to obtain the current load value of the crane; wherein the load evaluation model uses a multi-parameter joint verification rule to perform credibility verification on the preprocessed current load contact signal to obtain a credibility verification result; if the credibility verification result represents that the current load contact signal is in an untrusted state, a current weight combination is determined using a weight mapping rule; if the credibility verification result represents that the current load contact signal is in a trusted state, the current weight combination is generated using a weight generation strategy; based on the current weight combination, the preprocessed current load contact signal and the current electrical signal, and the current operation state parameter and the current operation environment parameter are weighted and fused to obtain the current load value of the crane.

[0012] Optionally, the front sensing module is further configured to collect the weighing sensor signal and send the weighing sensor signal to the start-stop control module.

[0013] The image sensing module is further configured to send the crane operation area image to the start-stop control module.

[0014] The start-stop control module is further configured to, if the credibility verification result of the current load contact signal output by the load evaluation model represents that the current load contact signal is in an untrusted state, determine that there is a signal mismatch risk when it is detected that the weighing sensor signal represents that the weighing sensor output is zero, and the crane operation area image detected by the hoisted object detection model represents that there is a hoisted object below the hook, and send a signal mismatch alarm instruction to the alarm linkage module.

[0015] The alarm linkage module is configured to control different types of alarm modes corresponding to the signal mismatch risk to perform linkage alarm when receiving the signal mismatch alarm instruction sent by the start-stop control module.

[0016] Optionally, the start-stop control module is further configured to dynamically correct the load value interval of different load states based on historical load data of the crane, current operating environment parameters and current health state parameters of the crane through an adaptive adjustment strategy.

[0017] Optionally, the start-stop control module is configured to determine, based on a correspondence between the crane type and the load value interval of different load states, a load state corresponding to the load value interval in which the current load value of the crane is located as the current load state from the load value interval of different load states corresponding to the crane type of the crane.

[0018] Optionally, the image perception module is further configured to send the crane operating area image to the start-stop control module.

[0019] The start-stop control module is further configured to, when the current load state is the loaded state, detect, through the posture detection model, whether the operating personnel posture in the crane operating area image is a specific authorized posture, trigger the authorized shutdown safety monitoring module when detecting that the operating personnel posture in the crane operating area image is the specific authorized posture, and end the authorization when determining that the authorized end determination condition is met.

[0020] Optionally, the start-stop control module is configured to end the authorization after triggering the authorization and counting down for a first time length, or end the authorization after triggering the authorization and detecting that there is no operating personnel in the dangerous operating area for a second time length.

[0021] Optionally, the safety monitoring module is configured to input the crane operating area image into a risk determination model to obtain a personnel operating risk determination result, wherein the risk determination model extracts primary semantic features from the crane operating area image, performs multi-layer key channel enhancement and receptive field expansion processing on the primary semantic features to obtain multi-scale deep layer semantic enhancement features, performs multi-target detection based on the multi-scale deep layer semantic enhancement features to obtain human body bounding box detection results, human body key point detection results, safety helmet wearing preliminary detection results and sling bounding box detection results, and performs multi-type operating risk determination based on the human body bounding box detection results, the human body key point detection results, the safety helmet wearing preliminary detection results and the sling bounding box detection results to obtain the personnel operating risk determination result.

[0022] In another aspect, the present application provides a crane full-stroke operating safety monitoring method, comprising:

[0023] Collecting current load contact signals and current electrical signals of an electrical cabinet of the crane, and current operating state parameters, current operating environment parameters and a crane operating area image of the crane;

[0024] Determining a current load state of the crane based on the current load contact signals and the current electrical signals, and the current operating state parameters and the current operating environment parameters.

[0025] When the current load state is the loaded state, a risk determination model is adopted to perform target detection and risk determination on the crane operation area image to obtain a personnel operation risk determination result.

[0026] When the personnel operation risk determination result represents that there is a personnel operation risk, different types of alarm modes corresponding to the personnel operation risk are controlled to perform linkage alarm.

[0027] The beneficial effects of the present application are as follows:

[0028] The present application determines the current load state of the crane by combining the load contact signal with the electrical signal, the operation state parameter and the operation environment parameter, which can effectively avoid the load state misjudgment caused by factors such as contact oxidation, line loosening, motion interference and artificial short circuit, and improve the accuracy of load state determination, so as to improve the timeliness and accuracy of the enabling control of the safety monitoring module, and then effectively improve the intelligent degree of crane safety monitoring and the safety of crane operation through the timely and accurate personnel operation risk determination of the safety monitoring module based on the risk determination model for the crane operation area image.

[0029] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims, and the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0030] The drawings described herein are intended to provide further understanding of the present application, and form a part of the present application. The illustrations and descriptions used to explain the present application are intended to explain the present application, and do not constitute improper limitations on the present application. In the drawings:

[0031] Figure 1 The present application is a schematic diagram of the composition structure of the crane full-stroke operation safety monitoring system in the embodiment of the present application;

[0032] Figure 2 The present application is a schematic diagram of the composition structure of the crane full-stroke operation safety monitoring system in the embodiment of the present application;

[0033] Figure 3 The present application is a schematic diagram of the composition structure of the crane full-stroke operation safety monitoring system in the embodiment of the present application;

[0034] Figure 4 The present application is a schematic diagram of the composition structure of the crane full-stroke operation safety monitoring system in the embodiment of the present application; DETAILED DESCRIPTION

[0035] In order to make the purposes, technical solutions and beneficial effects of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with embodiments and drawings. Obviously, the described embodiments are only some of the embodiments of the present application, and not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0036] In order to make the purposes, technical solutions and beneficial effects of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with embodiments and drawings. Obviously, the described embodiments are only some of the embodiments of the present application, and not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0037] A crane is a mechanical device for realizing heavy object carrying through vertical lifting or vertical lifting and horizontal movement. According to structure and use, the crane can be divided into many types such as a bridge crane, a gantry crane, a tower crane, a mobile crane and a cantilever crane. In the present application, the crane refers to an indoor / semi-indoor fixed crane with high safety protection requirements for operating personnel, which is started and stopped by a safety monitoring module controlled by a load contact signal output by an electrical cabinet; for example, a bridge crane and a gantry crane.

[0038] A load assessment model is a model for comprehensively assessing a current load value of a crane by integrating a current load contact signal, a current electrical signal, a current operating state parameter and a current operating environment parameter.

[0039] A multi-parameter joint verification rule is a rule for verifying the credibility of a current load contact signal based on parameter correlation.

[0040] A weight mapping rule is a weight mapping table constructed based on weight combinations of different parameter combinations, which is used to quickly locate a more accurate weight combination when the load contact signal is determined to be untrustworthy; wherein each parameter combination includes a load contact signal, an electrical signal, an operating state parameter and an operating environment parameter.

[0041] A weight generation strategy is a weight generation algorithm constructed based on index contribution and expert experience, which is used to quickly generate a more accurate weight combination when the load contact signal is determined to be trustworthy.

[0042] An adaptive adjustment strategy is a strategy for dynamically correcting a load value interval of a crane under different load states based on historical load data, a current operating environment parameter and a current health state parameter.

[0043] A dynamic validity generation strategy is a strategy for dynamically generating a posture authorization effective duration based on a current operating state parameter, a current operating environment parameter and a current brake state parameter of the crane.

[0044] After introducing the technical terms involved in the present application, next, the application scenario and design idea of the present application are briefly introduced.

[0045] To solve the problem of insufficient intelligence of crane safety monitoring, in the present application, the image perception module collects the crane operation area image and sends it to the safety monitoring module; the front perception module collects the current load contact point signal and the current electrical signal in the crane electrical cabinet, and the current operating state parameters and the current operating environment parameters of the crane, and sends them to the safety monitoring module; the start-stop control module determines the current load state of the crane based on the current load contact point signal and the current electrical signal, and the current operating state parameters and the current operating environment parameters, and enables the safety monitoring module when the current load state is the loaded state; the safety monitoring module uses a risk judgment model to detect and judge the target of the crane operation area image to obtain the personnel operation risk judgment result, and sends the operation risk alarm instruction to the alarm linkage module when the personnel operation risk judgment result represents that there is personnel operation risk; when the alarm linkage module receives the operation risk alarm instruction sent by the safety monitoring module, it controls the different types of alarm modes corresponding to the personnel operation risk to link alarm.

[0046] In this way, the current load state of the crane is determined by combining the load contact point signal with the electrical signal, the operating state parameters and the operating environment parameters, which can effectively avoid the misjudgment of the load state caused by factors such as contact oxidation, line loosening, motion interference and artificial short circuit, improve the accuracy of load state judgment, and thus can improve the timeliness and accuracy of the enabling control of the safety monitoring module, and further can effectively improve the intelligence of crane safety monitoring and the safety of crane operation through the timely and accurate personnel operation risk judgment of the safety monitoring module based on the risk judgment model for the crane operation area image.

[0047] After introducing the application scenario and design idea of the present application, the technical solutions provided by the present application are described in detail below.

[0048] The present application embodiment provides a crane full-stroke operation safety monitoring system, as shown in Figure 1 The crane full-stroke operation safety monitoring system 100 provided by the present application embodiment includes a front perception module 101, an image perception module 102, a start-stop control module 103, a safety monitoring module 104 and an alarm linkage module 105;

[0049] The front perception module 101 is used to collect the current load contact point signal and the current electrical signal of the crane electrical cabinet, and the current operating state parameters and the current operating environment parameters of the crane, and send them to the start-stop control module 103;

[0050] The start-stop control module 103 is configured to determine a current load state of the crane based on the current load contact signal and the current electrical signal sent by the front-end perception module, and the current operating state parameter and the current work environment parameter, and enable the safety monitoring module 104 when the current load state is a loaded state.

[0051] The image perception module 102 is configured to collect a crane work area image and send the crane work area image to the safety monitoring module 104.

[0052] The safety monitoring module 104 is configured to use a risk judgment model to perform target detection and risk judgment on the crane work area image to obtain a personnel work risk judgment result, and send a work risk alarm instruction to the alarm linkage module 105 when the personnel work risk judgment result indicates that there is a personnel work risk.

[0053] The alarm linkage module 105 is configured to control different types of alarm modes corresponding to the personnel work risk to perform linkage alarm when the work risk alarm instruction sent by the safety monitoring module 104 is received.

[0054] In the embodiment, the start-stop control module 103 can determine that the current load state is a loaded state when the current load contact signal is a high-level signal, and determine that the current load state is an unloaded state when the current load contact signal is a low-level signal. This load state judgment method based on the current load contact signal may have the problem of load state misjudgment due to oxidation of the load contact, looseness of the line, motion interference, and artificial short circuit, and the like. Therefore, the front-end perception module 101 can also be used to synchronously collect the current load contact signal and the current electrical signal, and the current operating state parameter and the current work environment parameter, and the start-stop control module 103 can determine the current load state of the crane based on the current load contact signal and the current electrical signal, and the current operating state parameter and the current work environment parameter, so as to determine the current load state of the crane by combining the load contact signal with the electrical signal, the operating state parameter, and the work environment parameter, thereby effectively avoiding the load state misjudgment caused by factors such as oxidation of the contact, looseness of the line, motion interference, and artificial short circuit, and improving the accuracy of load state judgment.

[0055] In a possible implementation, the start-stop control module 103 and the safety monitoring module 104 can be two independent modules, or two units integrated in the main control module.

[0056] In a possible implementation, the crane full-stroke operation safety monitoring system 100 provided by the embodiments of the present application further comprises a clock synchronization module 106; the clock synchronization module 106 is configured to output a clock signal to the front sensing module 101 and the image sensing module 102 under the control of the start-stop control module 103, so that the front sensing module 101 synchronously collects the crane operation area image, the current load contact signal and the current electrical signal in the electrical cabinet of the crane, and the current running state parameter and the current operation environment parameter of the crane.

[0057] In a possible implementation, the start-stop control module 103 is configured to calculate a current load value of the crane by using a load evaluation model through a weighted fusion strategy based on the current load contact signal and the current electrical signal, and the current running state parameter and the current operation environment parameter; determine a load state corresponding to a load value interval in which the current load value is located as the current load state of the crane based on a corresponding relationship between different load states and load value intervals; enable the safety monitoring module 104 when the current load state is a loaded state, and disable the safety monitoring module 104 when the current load state is an unloaded state.

[0058] In the embodiments of the present application, when the current load state is a loaded state, the safety monitoring module 104 is enabled, which can realize timely monitoring of the operation risk of the operation personnel, and improve the operation safety of the crane; when the current load state is an unloaded state, according to the operation safety specification, there is no lifting object below the hook in the unloaded state (i.e., the empty hook state), and the hook needs to be lifted according to the regulation, and the lifting height is much higher than the body height of a human being, at this time, the operation personnel will not cause a safety hazard, and the operation risk is low, so the safety monitoring module 104 can be disabled to reduce unnecessary running time and resource consumption of the safety monitoring module 104, and reduce the downtime loss caused by false alarms, and ensure efficient operation of the crane in a risk-free state.

[0059] In a possible implementation, the start-stop control module 103 is configured to preprocess the current load contact signal and the current electrical signal, and the current operating state parameter and the current working environment parameter; input the preprocessed current load contact signal and the current electrical signal, and the current operating state parameter and the current working environment parameter into the load assessment model to obtain the current load value of the crane; wherein the load assessment model adopts a multi-parameter joint verification rule to perform credibility verification on the preprocessed current load contact signal to obtain a credibility verification result; if the credibility verification result indicates that the current load contact signal is in an untrustworthy state, a weight mapping rule is adopted to determine a current weight combination; if the credibility verification result indicates that the current load contact signal is in a trustworthy state, a weight generation strategy is adopted to generate the current weight combination; and based on the current weight combination, the preprocessed current load contact signal and the current electrical signal, and the current operating state parameter and the current working environment parameter are weighted and fused to obtain the current load value of the crane.

[0060] In a possible implementation, the front-end perception module 101 is further configured to collect the weighing sensor signal and send the weighing sensor signal to the start-stop control module 103.

[0061] The image perception module 102 is further configured to send the crane working area image to the start-stop control module 103.

[0062] The start-stop control module 103 is further configured to, if the credibility verification result of the current load contact signal output by the load assessment model indicates that the current load contact signal is in an untrustworthy state, determine that there is a signal mismatch risk when it is detected that the weighing sensor signal indicates that the weighing sensor output is zero and the crane working area image is detected by the load detection model to indicate that there is a load under the hook, and send a signal mismatch alarm instruction to the alarm linkage module 105.

[0063] The alarm linkage module 105 is configured to, when receiving the signal mismatch alarm instruction sent by the start-stop control module 103, control different types of alarm modes corresponding to the signal mismatch risk to perform linkage alarm.

[0064] In the embodiments of the present application, when the start-stop control module 103 detects whether there is a load under the hook in the crane working area image by the load detection model, the crane working area image can be input into the load detection model to obtain a load detection result for indicating whether there is a load under the hook; wherein when the load detection model detects whether there is a load under the hook in the crane working area image, the following methods can be used, but are not limited to:

[0065] The visual feature data is extracted from the crane working area image by a feature extraction network (such as a convolutional neural network);

[0066] The visual feature data of the crane operation area image is matched (such as similarity calculation) with the anchor frame features in the hook detection frame parameter set through the hook area positioning module, candidate regions that meet the first feature matching threshold (such as a similarity threshold) are screened out, and the position of the candidate region is corrected through a bounding box regression algorithm to obtain the hook area and locate the spatial coordinate information (including but not limited to the vertex coordinates and center point coordinates of the bounding box) of the hook area; wherein the hook detection frame parameter set is a structured detection parameter learned by the hook object detection model in the training stage through a large number of hook sample images (covering different types of crane hooks, hook images under different light / angle / occlusion working conditions), including but not limited to the anchor frame size range, feature matching threshold, and bounding box regression parameter of different types of crane hooks, and the anchor frame size range is adapted to the actual physical size ratio of the crane hook and the image resolution;

[0067] Based on the spatial coordinate information of the hook area, the center point of the hook area is taken as the reference to extend downward by a preset spatial range (such as a rectangular area with a preset length and width extending downward along the vertical direction from the hook center point), to locate the hook object detection associated area below the hook;

[0068] The visual feature data in the hook object detection associated area is matched (such as similarity calculation) with the visual feature data of different hook object types in the hook object feature template library to detect whether there is a hook object type that meets the second feature matching threshold (such as a similarity threshold) in the hook object detection associated area; wherein the hook object feature template library is constructed in the following way: in the training stage of the hook object detection model, input hook object image samples containing different hook object types (such as steel members, containers, and material bundled bodies) and different working conditions (such as different light, occlusion, and angles) into the hook object detection model, learn and extract common visual features (such as contour shape, texture distribution, size ratio, and local structure features) and specific visual features of different hook object types through a feature extraction network, to form visual feature data of different hook object types and store them in the feature matching module of the hook object detection model;

[0069] If there is a hook object type that meets the second feature matching threshold (such as a similarity threshold) in the hook object detection associated area, output the hook object detection result representing the existence of the hook object type below the hook, otherwise output the hook object detection result representing the absence of hook objects below the hook.

[0070] In one possible implementation, the start-stop control module 103 is also configured to dynamically correct the load value interval of different load states through an adaptive adjustment strategy based on the historical load data of the crane, the current operation environment parameters, and the current health state parameters.

[0071] In a possible implementation, the start-stop control module 103 is configured to determine, based on the correspondence between the crane type and the load value intervals of different load states, a load state corresponding to a load value interval in which a current load value of the crane is located as a current load state, from the load value interval of the different load states corresponding to the crane type of the crane.

[0072] In a possible implementation, the image perception module 102 is further configured to send the crane operation area image to the start-stop control module 103.

[0073] The start-stop control module 103 is further configured to, when the current load state is the loaded state, detect, by using the posture detection model, whether a posture of the operation personnel in the crane operation area image is a specific authorized posture, trigger the authorized stop safety monitoring module 104 when it is detected that the posture of the operation personnel in the crane operation area image is the specific authorized posture, and determine that an authorized end condition is met when the authorized end condition is met.

[0074] In the embodiments of the present application, the specific authorized posture can be that the arms are opened upward and the included angle θ ∈ [75°, 105°]. Based on this, the start-stop control module 103 can input the crane operation area image into the posture detection model to obtain an authorized posture detection result. The posture detection model identifies the operation personnel in the crane operation area image and extracts the spatial coordinate information of the skeletal key points (such as the left shoulder joint, the right shoulder joint, the left elbow joint, the right elbow joint, the left wrist joint, the right wrist joint, etc.) of the operation personnel. Based on the spatial coordinate information of the skeletal key points of the operation personnel, the spatial posture of the arms of the operation personnel is determined, and the included angle θ between the arms is calculated with the midpoint of the line connecting the left and right shoulder joints as the vertex, and the line connecting the left shoulder joint and the left wrist joint and the line connecting the right shoulder joint and the right wrist joint as the two sides. It is determined whether the spatial posture of the arms of the operation personnel is the specific spatial posture of being opened upward, and whether the included angle θ between the arms is in the preset angle interval [75°, 105°]. If the above conditions are all met, an authorized posture detection result for representing that the operation personnel has appeared in the specific authorized posture is output. If the above conditions are not all met, an authorized posture detection result for representing that the operation personnel has not appeared in the specific authorized posture is output.

[0075] Further, when it is detected by the posture detection model that the worker is in a specific authorized posture (e.g., the arms are opened upwards and the included angle θ ∈ [75°, 105°]), and the worker in the specific authorized posture has authorized permission, it is determined that the posture authorization deactivation risk monitoring is effective, at which time the authorized deactivation safety monitoring module 104 can be triggered, leaving a time window for the worker to perform necessary operations such as lifting object binding and hook positioning, thereby effectively preventing accidents caused by unauthorized personnel or illegal operations, and indirectly improving economic benefits by precise authorization to reduce monitoring interference and speed up the operation process. In addition, in an optional embodiment, on the basis of posture authorization, voiceprint recognition (specific password such as "apply authorization") and face recognition (permission white list) can be combined to form three-factor authentication of password, posture and identity. Specifically, a lightweight audio encoder (such as LCNN) and a face ReID feature extractor are embedded in the posture detection model, and the cross-modal attention mechanism is used to align the time consistency of voiceprint, face and posture features to prevent false triggering (such as non-worker raising arms) and malicious bypassing. In another optional embodiment, on the basis of posture authorization, hierarchical authorization is performed based on the operation task and the authorized posture. Specifically, the specific authorized posture is subdivided into first-level authorization (allowing entry into a dangerous area) and second-level authorization (allowing the overhead traveling crane to run at high speed), wherein the first-level authorization and the second-level authorization can be distinguished by the arm spread angle (e.g., 75°-105° and 150°-180°).

[0076] In a possible embodiment, the start-stop control module 103 is configured to determine, based on the camera installation height (usually the camera is installed on the trolley, or the trolley height can also be used), the posture detection model corresponding to the current height scene from the posture detection models corresponding to different height scenes, and use the posture detection model corresponding to the current height scene to detect whether the worker posture in the crane operation area image is a specific authorized posture.

[0077] In the embodiments of the present application, for different height scenes, the training samples (including historical crane operation area images collected on the operation site and their worker posture annotation results) adapted to the height scene are used to train the posture detection model to obtain the posture detection model adapted to the height scene. In this way, when the posture detection model adapted to the current height scene is used to detect whether the worker posture in the crane operation area image is a specific authorized posture, the accuracy of authorized posture detection can be improved. Moreover, during the crane full-stroke operation safety monitoring process, the worker postures of each crane operation area image collected during the crane operation process can be annotated, and the posture detection model can be continuously optimized based on each crane operation area image and its worker posture annotation result, thereby further improving the accuracy of the posture detection model in authorized posture detection.

[0078] In a possible implementation, the safety monitoring module 104 is further configured to control the crane to run at a low speed after determining that the posture authorization takes effect, i.e., after triggering the safety monitoring module 104 to be deactivated, and record an authorized alarm cancellation log for later safety operation tracing, so as to effectively ensure the safety of the necessary operations such as the binding of the hoisted object and the positioning of the hook during the posture authorization process.

[0079] In a possible implementation, the start-stop control module 103 is configured to end the authorization after a first time length after triggering the authorization, or end the authorization after detecting that there is no operation personnel in the dangerous operation area for a second time length.

[0080] In a possible implementation, the safety monitoring module 104 is further configured to send prompt information indicating that the authorized alarm cancellation has taken effect through the alarm linkage module 105 when the authorized alarm cancellation takes effect, and send prompt information indicating that the authorized alarm cancellation has failed through the alarm linkage module 105 when the authorized alarm cancellation fails.

[0081] In a possible implementation, the first time length and the second time length can be predefined time lengths. Preferably, the first time length can also be a posture authorization effective time length dynamically generated by a dynamic time limit generation strategy based on current operating state parameters, current operation environment parameters and current brake state parameters of the crane; wherein the current operating state parameters include at least one of current hoisting height, current operating speed, current operating position, current brake state and current track flatness data; the current operation environment parameters include at least one of current wind speed and direction, current environment temperature, current environment humidity, current environment dust concentration and current environment light intensity data; and the current brake state parameters include at least one of brake coil on / off state, brake coil current real-time value, brake pad wear, brake response time, brake torque measured value, and brake cumulative action frequency data.

[0082] In a possible implementation, the start-stop control module 103 is configured to determine a current operation risk level based on current operating state parameters, current operation environment parameters and current brake state parameters of the crane; determine a posture authorization reference time length corresponding to the current operation risk level as a current posture authorization reference time length based on a corresponding relationship between the operation risk level and the posture authorization reference time length; determine a current risk level coefficient based on the current operation environment parameters; determine a current motion influence coefficient based on the current operating state parameters; determine a current brake effectiveness coefficient based on the current brake state parameters; and calculate a posture authorization effective time length based on the current posture authorization reference time length, the current risk level coefficient, the current motion influence coefficient and the current brake effectiveness coefficient.

[0083] In a possible implementation, the safety monitoring module 104 is configured to input the crane operation area image into a risk judgment model to obtain a personnel operation risk judgment result; the risk judgment model extracts initial layer semantic features from the crane operation area image; the initial layer semantic features are subjected to multi-layer key channel enhancement and receptive field expansion processing to obtain multi-scale deep layer semantic enhancement features; multi-target detection is performed based on the multi-scale deep layer semantic enhancement features to obtain a human body bounding box detection result, a human body key point detection result, a safety helmet wearing preliminary detection result, and a sling bounding box detection result; and multi-type operation risk judgment is performed based on the human body bounding box detection result, the human body key point detection result, the safety helmet wearing preliminary detection result, and the sling bounding box detection result to obtain the personnel operation risk judgment result.

[0084] Based on the above embodiments, the crane full-stroke operation safety monitoring method provided in the embodiments of the present application has the following general process: Figure 2

[0085] Step 201: Collecting the current load contact signal and the current electrical signal in the electrical cabinet of the crane, and collecting the current operating state parameters and the current operation environment parameters of the crane.

[0086] In the embodiments of the present application, after the crane is started, the crane can enter a self-control mode. In the self-control mode, the front perception module 101, the image perception module 102, the clock synchronization module 106, and the start-stop control module 103 are automatically enabled. The front perception module 101 and the image perception module 102 perform data synchronization collection under the triggering of the clock signal output by the clock synchronization module 106. At the same time, the start-stop control module 103 enables the laser lamp to detect the human body. If the human body is detected, load judgment is performed. If the human body is not detected, the laser lamp is turned off.

[0087] In specific implementation, when the front perception module 101 synchronously collects the current load contact signal and the current electrical signal in the electrical cabinet of the crane, and synchronously collects the current operating state parameters and the current operation environment parameters of the crane, the clock signal output by the clock synchronization module 106 is taken as a triggering reference, and the collection specifically includes:

[0088] The current load contact signal output by the electrical cabinet of the crane is collected through the load contact signal collection terminal. For example, when the load contact is closed (indicating that the load is on), the signal collection terminal outputs a high-level signal as the current load contact signal. When the load contact is disconnected (indicating that the load is off), the signal collection terminal outputs a low-level signal as the current load contact signal.

[0089] ​The current current signal output by the electric cabinet of the crane is collected by a Hall current sensor, the current voltage signal output by the electric cabinet of the crane is collected by a voltage transformer, the current contactor on-off signal is collected by the auxiliary contact of the contactor (such as the auxiliary contact of the lifting / lowering contactor), and the current frequency and torque output by the frequency converter are collected by the communication interface of the frequency converter. At least one of the current current signal, the current voltage signal, the current contactor on-off signal, the current frequency and torque output by the frequency converter is used as the current electrical signal.

[0090] The current lifting height of the hook is measured by a height sensor, the current running speed of each mechanism (such as the lifting mechanism, the trolley mechanism and the car mechanism) is measured by an encoder (such as a rotary encoder), the current running position of the trolley mechanism and the car mechanism is detected by a position sensor or a laser ranging sensor, the current brake state for indicating whether the brake is completely released or tightly gripped is detected by a brake travel switch or a pressure sensor, and the current track flatness is calculated based on the track vibration signal collected by a track vibration sensor and the track surface image collected by a visual sensor. At least one of the current lifting height, the current running speed, the current running position, the current brake state and the current track flatness is used as the current running state parameter.

[0091] The current wind speed and direction are collected by an array of ultrasonic anemometers, the current environmental temperature is collected by a temperature sensor, the current environmental humidity is collected by a humidity sensor, the current environmental dust concentration is collected by a laser dust sensor, and the current environmental light intensity is collected by a photoelectric light sensor. At least one of the current wind speed and direction, the current environmental temperature, the current environmental humidity, the current environmental dust concentration and the current environmental light intensity is used as the current working environment parameter. In an indoor / semi-indoor scene, there are often local airflow / sudden airflow disturbances (such as the start of a ventilation system, the opening of a door or window, and sudden changes in port wind). By introducing the wind speed and direction as the working environment parameter, the influence of local airflow / sudden airflow on load determination and work safety can be effectively reduced.

[0092] Step 202: Based on the current load contact signal and the current electrical signal, and the current running state parameter and the current working environment parameter, the current load state of the crane is determined.

[0093] In the embodiment of the application, the pre-sensing module 101 synchronously collects the current load contact signal and the current electrical signal in the electric cabinet of the crane, and the current running state parameter and the current working environment parameter of the crane, and sends them to the start-stop control module 103. When the start-stop control module 103 determines the current load state of the crane based on the current load contact signal and the current electrical signal, and the current running state parameter and the current working environment parameter, the following methods can be used, but are not limited to:

[0094] Firstly, the current load contact signal and the current electrical signal, and the current running state parameter and the current working environment parameter are preprocessed; wherein the preprocessing includes but is not limited to filtering, denoising, normalization processing, etc., to eliminate interference and unify the dimension.

[0095] Then, based on the preprocessed current load contact signal and the current electrical signal, and the current running state parameter and the current working environment parameter, the current load value of the crane is calculated by using a load evaluation model through a weighted fusion strategy. Specifically, the preprocessed current load contact signal and the current electrical signal, and the current running state parameter and the current working environment parameter are input into the load evaluation model to obtain the current load value of the crane; wherein the load evaluation model uses a multi-parameter joint verification rule to verify the credibility of the processed current load contact signal to obtain a credibility verification result; if the credibility verification result represents that the current load contact signal is in an untrusted state, a weight mapping rule is used to determine the current weight combination; if the credibility verification result represents that the current load contact signal is in a trusted state, a weight generation strategy is used to generate the current weight combination; based on the current weight combination, the preprocessed current load contact signal and the current electrical signal, and the current running state parameter and the current working environment parameter are weighted and fused to obtain the current load value of the crane.

[0096] Finally, based on the correspondence between different load states and load value intervals, the load state corresponding to the load value interval in which the current load value is located is determined as the current load state of the crane. Specifically, in order to adapt to different crane types, the load value intervals of each load state corresponding to different crane types are established in advance. Based on this, when the load state corresponding to the load value interval in which the current load value is located is determined as the current load state of the crane based on the correspondence between different load states and load value intervals, the load state corresponding to the load value interval in which the current load value is located can be determined as the current load state based on the correspondence between the crane type and the load value interval of different load states, from the load value interval of different load states corresponding to the crane type of the crane, the load state corresponding to the load value interval in which the current load value is located is determined as the current load state.

[0097] It is worth mentioning that in the embodiments of the present application, the load evaluation model also outputs the credibility verification result of the current load contact signal; the pre-sensing module 101 also synchronously collects the weighing sensor signal and sends it to the start-stop control module 103; the image sensing module 102 also sends the crane operating area image to the start-stop control module 103; the start-stop control module 103 can also perform signal mismatch risk detection based on the weighing sensor signal collected by the pre-sensing module 101 and the crane operating area image collected by the image sensing module 102 when the credibility verification result of the current load contact signal output by the load evaluation model indicates that the current load contact signal is in an untrustworthy state, that is, when it is detected that the weighing sensor signal indicates that the weighing sensor output is zero and the crane operating area image is detected by the load detection model. When the hook below exists, it is determined that there is a signal mismatch risk, and the alarm linkage module 105 controls the different types of alarm modes corresponding to the signal mismatch risk to perform linkage alarm. In this way, when the load contact signal is determined to be in an untrustworthy state, by enabling the redundant verification process based on the weighing sensor signal and the hook image, when the weighing sensor signal is zero (empty load) and the load is visually detected (loaded), it is determined that there is a signal mismatch risk and an alarm is triggered. It can further reduce the misjudgment rate while effectively avoiding heavy object falling or equipment collision accidents caused by signal mismatch, significantly improving the safety of crane operation.

[0098] In addition, in the embodiments of the present application, in order to avoid the problem that the load contact signal is not trustworthy due to factors such as human short-circuit contact, contact oxidation, and loose circuit, a dual-contact redundancy design can also be used to realize accurate acquisition of the load contact signal. Specifically, two independent load contacts are arranged in the electrical cabinet of the crane, that is, a main load contact and a standby load contact, which are respectively connected to different signal acquisition terminals; when the load contact signal output by the main load contact is not trustworthy, the standby load contact can be switched to reacquire the load contact signal for load value prediction.

[0099] Step 203: When the current load state is a loaded state, a risk judgment model is used to perform target detection and risk judgment on the crane operating area image to obtain a personnel operation risk judgment result.

[0100] In the embodiments of the present application, the start-stop control module 103 disables the safety monitoring module 104 when the current load state is an empty load state, and enables the safety monitoring module 104 when the current load state is a loaded state, thereby realizing accurate control of the safety monitoring module 104, reducing the probability of false operation of the safety monitoring module 104, and achieving the dual consideration of safety and low energy consumption of crane operation.

[0101] When the current load state is the loaded state, the start-stop control module 103 enables the safety monitoring module 104, and the safety monitoring module 104 can perform personnel operation risk judgment by using a risk judgment model; specifically, the safety monitoring module 104 can input the crane operation area image into the risk judgment model to obtain a personnel operation risk judgment result; wherein the risk judgment model extracts primary semantic features from the crane operation area image; the primary semantic features are subjected to multi-layer key channel enhancement and receptive field expansion processing to obtain multi-scale deep layer semantic enhancement features; multi-target detection is performed based on the multi-scale deep layer semantic enhancement features to obtain human body bounding box detection results, human body key point detection results, safety helmet wearing preliminary detection results and sling bounding box detection results; multi-type operation risk judgment is performed based on the human body bounding box detection results, the human body key point detection results, the safety helmet wearing preliminary detection results and the sling bounding box detection results to obtain the personnel operation risk judgment result.

[0102] Step 204: When the personnel operation risk judgment result represents that there is a personnel operation risk, control different types of alarm modes corresponding to the personnel operation risk to perform linkage alarm.

[0103] In the embodiments of the present application, the alarm linkage module 105 mainly consists of a laser lamp, a voice alarm, a display and a linkage control interface; wherein the laser lamp can project two light circles vertically downward to the ground, the outer circle is a warning circle (for example, green, diameter 6m), and the inner circle is a danger circle (for example, red, diameter 4m), the laser lamp is in a constant light state after the crane is started, and in the trigger alarm state; the voice alarm plays different voices according to different alarm types when the alarm is triggered; the display continuously outputs the alarm message when the alarm is triggered. The linkage control interface receives the alarm instructions of the start-stop control module 103 or the safety monitoring module 104, and sends a trigger signal to the laser lamp, the voice alarm or the display to realize the alarm. Specifically, different types of alarm modes can be used for linkage alarm according to different alarm instructions, wherein the different types of alarm modes include but are not limited to: only sound and light alarm, sound and light alarm + reverse travel low-speed operation, sound and light alarm + reverse travel braking, etc. For example, when the signal mismatch alarm occurs, the laser lamp is used for light alarm and the voice alarm is used for voice alarm (i.e. sound and light alarm), and at the same time the display continuously presents the signal mismatch risk alarm message; for example, when the safety helmet alarm occurs, the laser lamp is used for light alarm and the voice alarm is used for voice alarm (i.e. sound and light alarm), and at the same time the display continuously presents the signal mismatch risk alarm message; for example, when the personnel intrusion or personnel overrun alarm occurs, the laser lamp is used for light alarm and the voice alarm is used for voice alarm (i.e. sound and light alarm), and at the same time, the reverse travel low-speed operation and the display continuously present the personnel operation risk alarm message and the related crane operation area image or the related dynamic video. For example, when the sling is severely inclined, the travel speed is limited to 30% of the rated value, and the laser lamp is used for light alarm and the voice alarm is used for voice alarm (i.e. sound and light alarm); when the sling is slightly shaken, the travel acceleration is limited, and the vibration log is recorded for subsequent maintenance analysis.

[0104] Next, the risk determination model adopted by the safety monitoring module 104 in the embodiments of the present application is described in detail.

[0105] The risk determination model is an improved YOLOv5 network, as shown in Figure 3 The risk determination model includes an input layer, a feature extraction layer, a multi-branch detection layer and a multi-task output layer connected in turn; wherein:

[0106] The input layer is used to receive the crane operation area image;

[0107] The feature extraction layer comprises a backbone network and a neck network; the backbone network is configured to extract initial layer semantic features from the crane operation area image; and the neck network is configured to perform multi-layer key channel enhancement and receptive field expansion processing on the initial layer semantic features to obtain multi-scale deep layer semantic enhancement features. The backbone network comprises a CBS layer and a CSP layer connected in sequence; the CBS layer is configured to extract basic semantic features from the crane operation area image; and the CSP layer is configured to perform gradient shunting and residual mapping processing on the basic semantic features output by the CBS layer to obtain the initial layer semantic features. The neck network comprises a plurality of deep layer feature extraction layers connected in sequence, and each deep layer feature extraction layer comprises a channel attention unit and a spatial attention unit connected in sequence; the channel attention unit is configured to perform key channel enhancement processing on input features, wherein the input features of the channel attention unit in the first deep layer feature extraction layer are the initial layer semantic features output by the backbone network, and the input features of the channel attention unit in the non-first deep layer feature extraction layer are the output features of the spatial attention unit in the previous deep layer feature extraction layer; and the spatial attention unit is configured to perform receptive field expansion processing on the key channel enhancement features output by the channel attention unit, wherein the output features of the spatial attention units in the plurality of deep layer feature extraction layers constitute the multi-scale deep layer semantic enhancement features. In the embodiment of the application, the spatial attention unit mainly introduces a hollow convolution to replace the original convolution. In each deep layer feature extraction layer, the original convolution is replaced by the hollow convolution, which can expand the receptive field of the original convolution layer, strongly cover the surrounding situation of the occlusion area, and thus effectively reduce the precision loss caused by the occlusion. In addition, the spatial attention unit can dynamically adjust the hollow convolution rate according to the occlusion, specifically, the spatial attention unit comprises an occlusion discrimination layer (a binary classification network) and a hollow convolution layer connected in sequence; the occlusion discrimination layer (the binary classification network) performs occlusion determination on the key channel enhancement features output by the channel attention unit to output an occlusion density map; the hollow convolution layer dynamically adjusts the hollow convolution rate according to the occlusion density map; wherein the occlusion density and the hollow convolution rate are in a positive correlation relationship (for example, in a high-density area, the hollow convolution rate = 5, and in a low-density area, the hollow convolution rate = 2), so as to realize intelligent adaptation of the receptive field and balance the calculation efficiency and the context coverage.

[0108] The multi-branch detection layer comprises a key point detection head, a safety helmet detection head, a sling detection head, and a human body detection head; the human body detection head is configured to perform human body detection based on multi-scale deep layer semantic enhancement features to obtain a human body bounding box corresponding to each human body target as a human body bounding box detection result; the key point detection head is configured to perform human body key point detection based on multi-scale deep layer semantic enhancement features to obtain a human body key point set corresponding to each human body target as a human body key point detection result; the human body key point set comprises at least a head key point and a chest key point; the safety helmet detection head is configured to perform safety helmet wearing detection based on multi-scale deep layer semantic enhancement features to obtain a binary classification preliminary detection result for indicating whether a safety helmet is worn corresponding to each human body target as a safety helmet wearing preliminary detection result; and the sling detection head is configured to perform sling detection based on multi-scale deep layer semantic enhancement features to obtain a sling direction angle and a sling bounding box as a sling bounding box detection result.

[0109] The multi-task output layer comprises a post-processing unit and a fusion output unit connected in sequence; the post-processing unit is configured to associate the human body bounding box corresponding to each human body target output by the human body detection head, the human body key point set corresponding to each human body target output by the key point detection head, and the safety helmet wearing preliminary detection result corresponding to each human body target output by the safety helmet detection head to obtain the human body bounding box, the human body key point set, and the safety helmet wearing preliminary detection result corresponding to each human body target; perform personnel intrusion determination, safety helmet wearing determination, and number of people over-limit determination based on the human body bounding box, the human body key point set, and the safety helmet wearing preliminary detection result corresponding to each human body target to obtain a personnel intrusion detection result, a safety helmet wearing detection result, and a number of people over-limit detection result; and detect whether the sling has a posture anomaly based on the sling direction angle and the sling bounding box of the sling and a preset standard posture bounding box to obtain a sling posture anomaly detection result; and the fusion output unit is configured to integrate the personnel intrusion detection result, the safety helmet wearing detection result, the number of people over-limit detection result, and the sling posture anomaly detection result to form a work risk determination result and output the work risk determination result.

[0110] In a specific implementation, when the post-processing unit performs personnel intrusion determination, safety helmet wearing determination, and number of people over-limit determination based on the human body bounding box, the human body key point set, and the safety helmet wearing preliminary detection result corresponding to each human body target, the following methods can be used, but are not limited thereto:

[0111] For each human body target, based on the human body bounding box of the human body target and the preset danger circle, whether the human body target is located in the preset danger circle is detected to obtain a personnel intrusion detection result of the human body target; when a head key point exists in the human body key point set of the human body target, a safety helmet is detected in a region of interest centered on the head key point to obtain a safety helmet wearing re-inspection result, and when the head key point does not exist in the human body key point set of the human body target and a chest key point exists, the safety helmet is detected in a region of interest centered on the chest key point to obtain a safety helmet wearing re-inspection result, and a safety helmet wearing detection result of the human body target is determined based on the safety helmet wearing preliminary inspection result and the safety helmet wearing re-inspection result; and based on the human body bounding box of each human body target and the preset danger circle, whether the number of human body targets in the preset danger circle exceeds a preset number threshold is detected to obtain a number overrun detection result.

[0112] In this way, by adopting the improved YOLOv5 as the risk judgment model to perform target detection and risk judgment on the crane operation area image, four pieces of information of the human body, the human body key point, the safety helmet and the lifting tool can be output simultaneously, and the end-to-end delay is reduced. Moreover, after the backbone network extracts the initial layer semantic features from the crane operation area image, the initial layer semantic features are subjected to multi-layer key channel enhancement and receptive field expansion processing by the neck network, which can effectively highlight the key channel features, improve the target capture capability, and reduce the missed detection rate of the occluded workers and safety helmets and the like, thereby improving the accuracy of the operation risk judgment, and effectively eliminating the false positives caused by occlusion or normal operation, and significantly reducing the supervision fatigue. In addition, in the safety helmet detection, a double detection mode based on the safety helmet detection head preliminary inspection and the post-processing unit re-inspection is adopted, and in the safety helmet re-inspection, based on whether the head key point is visible, the region of interest centered on the head key point or the chest key point is dynamically selected for the safety helmet re-inspection, which can effectively cope with the problem of invisible head caused by dense occlusion in the industrial scene, control the safety helmet detection error in the head occlusion condition within a lower pixel, and improve the safety helmet detection recall rate in the occlusion scene.

[0113] Next, the load evaluation model adopted by the start-stop control module 103 in the embodiment of the present application will be described in detail.

[0114] The load evaluation model is a hybrid neural network and fuzzy logic model, as shown in Figure 4 The load evaluation model includes an input layer, a multi-parameter joint verification layer, a hybrid processing layer, a fusion prediction layer and an output layer; wherein:

[0115] The input layer includes a plurality of nodes for receiving the pre-processed current load contact signal, the current electrical signal, the current operating state parameter and the current operating environment parameter respectively;

[0116] The multi-parameter joint verification layer is configured to perform signal consistency analysis and state logic verification on the preprocessed current load contact signal, the preprocessed current electrical signal, the current operating state parameter and the current working environment parameter by using a multi-parameter joint verification rule, so as to evaluate the credibility of the current load contact signal and output a credibility verification result. Specifically, the time sequence characteristics of the current load contact signal are compared with the time sequence characteristics of the current electrical signal to obtain a change trend matching degree between the current load contact signal and the current electrical signal, a first credibility prediction value is determined based on the change trend matching degree, for example, the change trend matching degree M is directly used to represent the first credibility prediction value, which ranges from 0 to 1, and the higher the value, the higher the credibility; a physical logic consistency verification rule is used to verify the physical logic consistency between the current load contact signal and the current operating state parameter, and a second credibility prediction value is obtained by comprehensively scoring the verification results of all physical logic consistency verification rules, for example, each physical logic consistency verification rule is verified, and if it is consistent, it is recorded as 1, otherwise it is recorded as 0, if there are n physical logic consistency verification rules, and k physical logic consistency verification rules are consistent, the second credibility prediction value ranges from 0 to 1, and the higher the value, the higher the credibility; the exceeding degree of each working environment parameter is calculated based on each working environment parameter contained in the current working environment parameter and its corresponding safety range, and the third credibility prediction value is obtained by using a linear decay method to decay the preset basic credibility (for example, “1”) based on the exceeding degree of each working environment parameter, for example, the calculation method of the exceeding degree is , and the calculation method of the third credibility prediction value is , the weights are allocated according to the importance of each working environment parameter to the safety of the work, and the sum of the weights is 1; the first credibility prediction value, the second credibility prediction value and the third credibility prediction value are weighted and averaged to obtain a current credibility prediction value, the weights are allocated according to the importance of each evaluation index, and the sum of the weights is 1; the current credibility verification result is determined based on the matching relationship between the current credibility prediction value and the credibility interval of different credibility levels, for example, if the current credibility prediction value is greater than or equal to the credibility threshold, it is determined that the current load contact signal is in a credible state, otherwise, it is determined that the current load contact signal is in an untrustworthy state.

[0117] The hybrid processing layer comprises a weight determination unit and a fuzzy inference unit; the weight determination unit is configured to, if the credibility check result indicates that the current load contact point signal is in an untrustworthy state, call a weight mapping table comprising different weight combinations of parameters, which is constructed based on at least one of historical data statistics, physical relationship model analysis and expert knowledge calculation, and determine the current weight combination of the current parameter combination through fuzzy matching; if the credibility check result indicates that the current load contact point signal is in a trustworthy state, analyze the discreteness and conflict of the current load contact point signal, the current electrical signal, the current operating state parameter and the current operating environment parameter by using the CRITIC algorithm and quantify the index contribution degree as an objective weight, analyze the causal relationship between the current load contact point signal, the current electrical signal, the current operating state parameter and the current operating environment parameter by using the DEMATEL algorithm and quantify the index influence degree as an experience weight, and generate the final weight combination by weighted average of the objective weight and the experience weight, the weights being allocated according to the importance of the index contribution degree and the index influence degree, and the sum of the weights being 1. The fuzzy inference unit is configured to use a fuzzy rule base constructed based on expert experience to describe the influence relationship of the load contact point signal, the electrical signal, the operating state parameter and the operating environment parameter on the crane load value, and perform fuzzy processing on the current load contact point signal, the current electrical signal, the current operating state parameter and the current operating environment parameter by using a membership function to obtain the fuzzy rule and the membership degree to which the current load contact point signal, the current electrical signal, the current operating state parameter and the current operating environment parameter belong. The membership function defines the degree to which each input value belongs to a certain fuzzy rule in the fuzzy rule base.

[0118] The fusion prediction layer comprises a weighted fusion unit and a deep regression unit; the weighted fusion unit is configured to perform weighted fusion on the preprocessed current load contact point signal, the current electrical signal, the current operating state parameter and the current operating environment parameter of the crane according to the current weight combination to obtain an initial load value of the crane. The deep regression unit comprises a plurality of hidden layers, each hidden layer comprising a plurality of neurons, and the neurons are connected through a nonlinear activation function; the deep regression unit is configured to perform deep-level nonlinear feature extraction, nonlinear transformation and regression prediction on the initial load value and the fuzzy set and membership degree of the influence of the current load contact point signal, the current electrical signal, the current operating state parameter and the current operating environment parameter on the load value to obtain the current load value of the crane.

[0119] The output layer comprises one node configured to output the final predicted current load value of the crane.

[0120] In this way, the hybrid neural network and fuzzy logic fusion model is adopted as the load evaluation model, which has both the powerful feature learning ability of neural network and the processing advantage of fuzzy logic for uncertainty, so that the start-stop control module 103 can more accurately and intelligently evaluate the current load value of the crane in a complex crane operation environment, and thus can improve the evaluation accuracy of the current load state when determining the current load state based on the current load value and the load value interval of different load states, thereby providing strong support for the start-stop control of the safety monitoring module 104.

[0121] In addition, in order to further improve the evaluation accuracy of the load state and adapt to different working conditions, the start-stop control module 103 can also dynamically correct the load value interval of different load states based on the historical load data, current operation environment parameters and current health state parameters of the crane by using an adaptive adjustment strategy through timing triggering or event triggering. The event triggering includes but is not limited to: the mutation amplitude of the operation environment parameters of the crane exceeding a preset threshold (such as real-time average wind speed mutation ≥ 5 m / s, environmental visibility mutation ≥ 200 m, etc.), or the crane state parameters appearing abnormal (such as brake clearance mutation ≥ 0.5 mm, motor current distortion rate exceeding 5%, etc.); the historical load data includes but is not limited to: the cumulative working hours of the crane in the last N operation cycles, the measured load value at each historical time, and the original load contact signal, original electrical signal, original operating state parameter, original operation environment parameter recorded at the same time as the measured load value, etc.; the current operation environment parameters include but are not limited to: wind speed and direction, humidity, temperature, dust concentration, light intensity, etc.; the current health state parameters include but are not limited to: motor temperature rise, brake pad wear, steel wire rope diameter reduction, etc. Specifically, when dynamically correcting the load value interval of different load states based on the historical load data, current operation environment parameters and current health state parameters of the crane using an adaptive adjustment strategy, the following methods can be used but are not limited to:

[0122] After removing the outliers in the historical load data, current operation environment parameters and current health state parameters using the 3σ criterion, the missing data is completed using the linear interpolation method, and the historical load data, current operation environment parameters and current health state parameters after the outlier removal and missing data completion are linearly normalized to obtain standardized historical load data, standardized operation environment parameters and standardized health state parameters;

[0123] Based on the standardized historical load data, the clustering method is used to cluster by load state to obtain the initial load value interval [L min , L max]; wherein the clustering method can be but is not limited to a density peak-adaptive bandwidth kernel density estimation (DPC-AKDE) algorithm, and a bandwidth coefficient exponentially decays with the cumulative working hours of the crane to compensate for the load zero point drift caused by mechanical wear;

[0124] The standardized operating environment parameters are constructed as a multi-dimensional environment vector, and the multi-dimensional environment vector is input into a pre-trained lightweight neural network model (such as a lightweight environment shift perception model, i.e., an ESN model) to obtain an environment shift coefficient λ E , wherein the lightweight neural network model is trained in an offline manner on big data of the same type of machine, and a loss function is a transfer maximum mean difference (MMD) loss, which is used to quantify the distribution shift of the multi-dimensional environment vector on the measured load value; the environment shift coefficient λ E is multiplied by the upper and lower bounds of the initial load value interval of each load state to obtain a first corrected load value interval [λ E ·L min of each load state; E ·L max ];

[0125] Each health state parameter in the current health state parameter is constructed as a device health vector, and a mechanical degradation coefficient λ H is obtained in a table lookup manner in a degradation coefficient library; wherein the degradation coefficient library records the degradation rules of different health state parameters; for example: the mechanical degradation coefficient λ H decreases by 0.02 for each 10℃ increase in motor temperature rise; the mechanical degradation coefficient λ H decreases by 0.03 for each 1mm increase in brake pad wear; the mechanical degradation coefficient λ H decreases by 0.05 for each 2% decrease in steel wire rope diameter; and the lower limit of the mechanical degradation coefficient λ H is 0.7; the mechanical degradation coefficient λ H is multiplied by the upper and lower bounds of the first corrected load value interval of each load state to obtain a second corrected load value interval [λ H ·λ E ·L min of each load state; H ·λ E ·L max ];

[0126] An age decay coefficient λ T =1-α·exp(-t / β) is calculated, wherein t is the number of days from the last calibration, α and β are obtained by exponential regression fitting of a large sample and reliability-cost double constraint optimization, α is a decay amplitude coefficient, for example α=0.15, and β is a decay time constant, for example β=30; and the age decay coefficient λ TThe upper and lower limits of the secondary modified load value interval of each load state are multiplied to obtain the final load value interval [λ T ·λ H ·λ E ·L min , λ T ·λ H ·λ E ·L max ] of each load state.

[0127] In the embodiments of the present application, in order to further refine the start-stop management of the safety monitoring module 104 and reduce false positives and interference, the start-stop control module 103 can also control the start and stop of the safety monitoring module 104 through the posture authorization chain when the current load state is a load state, specifically including:

[0128] detecting whether the posture of the worker in the worker image is a specific authorized posture through the posture detection model; when it is detected that the posture of the worker in the worker image is a specific authorized posture, triggering the authorized deactivation of the safety monitoring module 104, and ending the authorization when it is determined that the authorized end determination condition is met. Specifically, ending the authorization after a first time length after triggering the authorization; or, ending the authorization when it is detected that there is no worker in the dangerous operation area for a second time length.

[0129] In the embodiments of the present application, the first time length and the second time length can be predefined time lengths. Preferably, the first time length can also be a posture authorization effective time length dynamically generated through a dynamic time limit generation strategy based on the current operating state parameters, the current operation environment parameters and the current brake state parameters of the crane.

[0130] In specific implementation, when the start-stop control module 103 dynamically generates the posture authorization effective time length based on the current operating state parameters, the current operation environment parameters and the current brake state parameters of the crane through the dynamic time limit generation strategy, the following methods can be used, but are not limited to:

[0131] determining the current operation risk level through an operation risk assessment model based on the current operating state parameters, the current operation environment parameters and the current brake state parameters of the crane;

[0132] based on the corresponding relationship between the operation risk level and the posture authorization reference time length, the posture authorization reference time length corresponding to the current operation risk level is the current posture authorization reference time length; wherein, the operation risk level and the posture authorization reference time length are in a negative correlation relationship; the higher the operation risk level, the shorter the posture authorization reference time length;

[0133] determine a current risk level coefficient based on the current working environment parameters, wherein the current risk level coefficient is a correction coefficient for quantifying the influence degree of the working environment parameters on the posture authorization effective duration; specifically, each working environment parameter in the current working environment parameters is preprocessed by outlier rejection, linear interpolation completion, linear normalization, etc. to obtain each standardized working environment parameter; the current environment risk weight of each standardized working environment parameter is determined based on an analytic hierarchy process (such as DEMATEL algorithm and / or CRITIC algorithm); based on the current environment risk weight of each standardized working environment parameter, the current environment risk comprehensive value R env is calculated by weighted summation env ; based on the segmented mapping relationship between the environment risk comprehensive value and the risk level coefficient, the risk level coefficient corresponding to the environment risk interval where the current environment risk comprehensive value R env is located is taken as the current risk level coefficient K env ; for example: when R env ∈[0,0.3], K env =1.0; when R env ∈(0.3,0.6], K env =0.8; when R env ∈(0.6,0.8], K env =0.6; when R env ∈(0.8,1.0], K

[0134] determine a current motion influence coefficient based on the current running state parameters, wherein the current motion influence coefficient is a correction coefficient for quantifying the influence degree of the crane running state parameters on the posture authorization effective duration; specifically, each running state parameter in the current running state parameters is preprocessed by moving average filtering, linear normalization, etc. to obtain the standardized running state parameter of each running state parameter; the current motion risk weight of each standardized running state parameter is determined based on an analytic hierarchy process (such as DEMATEL algorithm and / or CRITIC algorithm); based on the current motion risk weight of each standardized running state parameter, the current motion risk comprehensive value R mov is calculated by weighted summation mov ; based on the segmented mapping relationship between the motion risk comprehensive value and the motion influence coefficient, the motion influence coefficient corresponding to the motion risk interval where the current motion risk comprehensive value R mov is located is taken as the current motion influence coefficient K mov ; for example: when R mov ∈[0,0.2], K mov =1.0; when R mov ∈(0.2,0.4], K mov =0.9; when R mov=0.7; when R mov When K ∈ (0.6, 0.8], mov =0.5; when R mov When K ∈ (0.8, 1.0], mov =0.2;

[0135] The current braking efficiency coefficient is determined based on the current brake state parameters, which include at least one of the following: brake pad temperature, brake clearance, and brake response delay. The current braking efficiency coefficient is a correction coefficient used to quantify the influence of the crane brake state parameters on the effective duration of attitude authorization. Specifically, each brake state parameter in the current braking parameters undergoes preprocessing such as first-order low-pass filtering and linear normalization to obtain standardized brake state parameters. The current efficiency weight of each standardized brake state parameter is determined based on the analytic hierarchy process (AHP) (such as the DEMATEL algorithm and / or CRITIC algorithm). Based on the current efficiency weights of each standardized brake state parameter, the current braking efficiency comprehensive value R is calculated by weighted summation. brk Based on the piecewise mapping relationship between the comprehensive braking performance value and the braking performance coefficient, the current comprehensive braking performance value R is... brk The braking efficiency coefficient corresponding to the current braking efficiency range is used as the current braking efficiency coefficient K. brk For example: when R brk When K ∈ [0, 0.2] brk =1.0; when R brk When K ∈ (0.2, 0.4], brk =0.9; when R brk When K ∈ (0.4, 0.6], brk =0.7; when R brk When K ∈ (0.6, 0.8], brk =0.4; when R brk When K ∈ (0.8, 1.0], brk =0.1;

[0136] The effective duration of attitude authorization is calculated based on the current attitude authorization reference duration, the current risk level coefficient, the current motion impact coefficient, and the current braking effectiveness coefficient. For example, the effective duration of attitude authorization is the product of the current attitude authorization reference duration and the current risk level coefficient, the current motion impact coefficient, and the current braking effectiveness coefficient.

[0137] In this way, through the intelligent linkage of load state and posture detection, combined with the dynamic time limit generation strategy, the on-demand start and stop of the safety monitoring module 104 is realized, which optimizes resource utilization and system reliability under the premise of ensuring safety, specifically in the following aspects: by closing the safety monitoring module 104 when it is idle, unnecessary running time and resource consumption of the safety monitoring module 104 can be reduced, and downtime losses caused by false alarms can be reduced, ensuring efficient operation of the crane in a risk-free state; by detecting a specific authorized posture (such as opening the arms upward by 75-105°) when it is loaded, and closing the safety monitoring module 104 only when the specific authorized posture detection is confirmed to be passed, unauthorized personnel or illegal operation can be effectively prevented from causing accidents, and through precise authorization and dynamic adjustment, monitoring interference can be reduced, the work process can be accelerated, and economic benefits can be indirectly improved; by re-enabling the safety monitoring module 104 when posture detection fails or the posture authorization validity period expires, abnormal behavior (such as overload, collision, personnel intrusion, etc.) can be captured in time to reduce the risk of accidents; by intelligently adjusting the posture authorization validity period according to the current operating state parameters, current operating environment parameters, and current brake state parameters of the crane, the posture authorization validity period can be extended in a stable operating environment to reduce frequent detection, and the posture authorization validity period can be shortened in a complex environment to enhance the monitoring effect of the safety monitoring module 104, avoid false positives or false negatives caused by fixed time limit strategies, and improve the accuracy and reliability of the safety monitoring module 104.

[0138] In the implementation of the present application, in order to further improve the safety of crane operation, the front sensing module 101 can also collect wind speed data within a set time range at a set collection frequency through an ultrasonic anemometer array and send it to the start-stop control module 103, and the start-stop control module 103 can also control the start and stop of the safety monitoring module 104 in combination with the low wind resistance mode, specifically including:

[0139] Based on the wind speed data collected within a set time range (such as 10 minutes) at a set collection frequency by the ultrasonic anemometer array, the average wind speed within the set time range (such as 10 minutes) is calculated;

[0140] When the average wind speed within the set time range (such as 10 minutes) is less than the wind speed threshold and the current load state is an idle state, the safety monitoring module 104 is disabled to reduce invalid alarms, and the low wind resistance mode is enabled to optimize energy efficiency and reduce wind sway interference; under the low wind resistance mode, the following operations are performed: (1) a safety speed limit message is issued to the frequency converter through the 5G-TSN network to limit the operating speed of the crane to α% of the rated speed, where α% is an empirical value; (2) a carbon fiber drag reduction wing panel is popped up at the front end of the boom to reduce wind-induced sway amplitude.

[0141] When the average wind speed in a certain time range (e.g., within 10 minutes) is greater than or equal to the wind speed threshold and the current load state is the empty load state, the safety monitoring module 104 can be enabled, the low wind resistance mode is disabled, and prompt information (such as voice prompts, ringing prompts, etc.) indicating that the safety monitoring module 104 has been started is output, so that the worker can close the safety monitoring module 104 according to the work demand through the gesture authorization.

[0142] When the current load state is the loaded state, the start and stop state of the safety monitoring module 104 is controlled based on the gesture authorization chain; wherein the gesture authorization chain includes:

[0143] Detect whether the worker gesture in the worker image is a specific authorized gesture through the gesture detection model; for example, through the gesture detection model to track the skeletal key points of the worker, detect whether the worker appears a specific authorized gesture of opening both arms upward and the included angle θ ∈ [75°, 105°];

[0144] If the worker gesture is detected as a specific authorized gesture in a plurality of consecutive frames of worker images, a gesture authorization token PAT is generated, the gesture authorization token PAT is signed using a physically unclonable function to prevent replay attacks, and a gesture authorization effective duration T is generated through a dynamic time-to-live generation strategy based on the current operating state parameters of the crane, the current work environment parameters, and the current brake state parameters;

[0145] The gesture authorization token PAT and the gesture authorization effective duration are written into the blockchain light node, and the smart contract is started to count down the gesture authorization effective duration T. Within the gesture authorization effective duration, the safety monitoring module 104 is turned off, and the low wind resistance mode is disabled at the same time because the pop-up of the carbon fiber drag reduction wing plate may interfere with the operation of the worker, and speed limiting will reduce work efficiency;

[0146] Within the gesture authorization effective duration, if the current load state changes to the empty load state or the number of people detected entering the dangerous work area exceeds the limit, the gesture authorization token PAT is immediately revoked, the safety monitoring module 104 is started, and the low wind resistance mode is disabled, and a three-level cascading emergency operation is triggered: the first level emergency operation is to power off the crane to prevent the crane from continuing to run in an abnormal state, the second level emergency operation is to realize zero gravity hovering of the lifted object based on the superconducting magnetic damper to avoid the swing of the boom caused by the sudden disappearance of the load, and the third level emergency operation is to form a physical isolation barrier through a laser grid within a certain time range (e.g., within 50 ms) to quickly isolate the dangerous area.

[0147] In this way, when the average wind speed is less than the wind speed threshold and the current load state is the empty load state, by closing the safety monitoring module 104 and enabling the low wind resistance mode, wind swing interference can be reduced, energy waste can be reduced, and work efficiency can be improved; when the average wind speed is greater than or equal to the wind speed threshold and the current load state is the empty load state, by automatically enabling the safety monitoring module 104 and disabling the low wind resistance mode, work risks caused by hook swinging due to strong wind can be effectively avoided; when the current load state is the loaded state, by triggering the posture authorization token through a specific authorized posture (such as opening the two arms upward by 75-105°) and combining the physical unclonable function anti-replay attack, the authorization uniqueness can be ensured, and by writing the posture authorization token and the posture authorization effective time into the blockchain light node and using the smart contract countdown management, the authorization process can be ensured to be transparent and tamper-proof, in addition, within the posture authorization validity period, by automatically closing the safety monitoring module 104, false alarm interference can be avoided, and the low wind resistance mode is disabled, which can prevent carbon fiber wing plate interference operation and speed limit to reduce work efficiency, in addition, within the posture authorization validity period, when the load is empty or the number of people entering the dangerous operation area exceeds the limit, by immediately revoking the posture authorization token, starting the safety monitoring module 104, and disabling the low wind resistance mode, and triggering the three-level cascading emergency operation, the influence range of the accident can be minimized through power-off, hovering, isolation and other step-by-step protection, and the safety of the operating personnel and equipment can be ensured.

[0148] Although the preferred embodiments of the present application have been described, those skilled in the art who are familiar with the basic inventive concept can make additional changes and modifications to the embodiments. Therefore, the appended claims are intended to include the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0149] Obviously, those skilled in the art can make various modifications and variations to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A crane full-stroke operation safety monitoring system, characterized in that, It includes a front-end sensing module, an image sensing module, a start / stop control module, a safety monitoring module, and an alarm linkage module; The image sensing module is used to acquire images of the crane's operating area and send them to the start / stop control module and the safety monitoring module; The front-end sensing module is used to collect the current load contact signal and current electrical signal of the crane electrical cabinet, as well as the current operating status parameters and current working environment parameters of the crane, and send them to the start-stop control module. The start / stop control module is used to input the preprocessed current load contact signal, current electrical signal, current operating status parameters, and current working environment parameters into the load assessment model to obtain the current load value of the crane; based on the correspondence between different load states and load value intervals, the load state corresponding to the load value interval where the current load value is located is determined as the current load state of the crane; when the current load state is an unloaded state, the safety monitoring module is disabled; when the current load state is a loaded state, the safety monitoring module is enabled, or, by using a posture detection model to detect whether the posture of the operator in the crane's working area image is a specific authorized posture, when the posture of the operator in the crane's working area image is detected to be a specific authorized posture, the start / stop control module is activated. The authorization process involves disabling the safety monitoring module, ending the authorization when the termination condition is met, and then reactivating the safety monitoring module. The load assessment model uses a multi-parameter joint verification rule to verify the reliability of the processed current load contact signal, obtaining a reliability verification result. If the reliability verification result indicates that the current load contact signal is in an unreliable state, a weight mapping rule is used to determine the current weight combination. If the reliability verification result indicates that the current load contact signal is in a reliable state, a weight generation strategy is used to generate the current weight combination. Based on the current weight combination, the preprocessed current load contact signal, current electrical signal, current operating status parameters, and current working environment parameters are weighted and fused to obtain the current load value of the crane. The safety monitoring module is used to perform target detection and risk assessment on the crane operation area image using a risk assessment model to obtain a personnel operation risk assessment result; when the personnel operation risk assessment result indicates that there is a personnel operation risk, it sends an operation risk alarm command to the alarm linkage module. The alarm linkage module is used to trigger a linkage alarm by controlling different types of alarms corresponding to the work risk when it receives the work risk alarm command.

2. The crane full-stroke operation safety monitoring system as described in claim 1, characterized in that, The front-end sensing module is also used to collect weighing sensor signals and send them to the start / stop control module; The start / stop control module is also used to determine that there is a signal mismatch risk and send a signal mismatch alarm command to the alarm linkage module if the reliability verification result of the current load contact signal output by the load assessment model indicates that the current load contact signal is in an unreliable state, and when the weighing sensor signal indicates that the weighing sensor output is zero, and the load detection model detects that there is a load under the hook in the crane operation area image, the module detects that there is a load under the hook. The alarm linkage module is used to control different types of alarms corresponding to the signal mismatch risk to trigger a linkage alarm when it receives the signal mismatch alarm command sent by the start / stop control module.

3. The crane full-stroke operation safety monitoring system as described in claim 1, characterized in that, The start / stop control module is also used to dynamically adjust the load value range for different load states based on the crane's historical load data, current operating environment parameters, and current health status parameters through an adaptive adjustment strategy.

4. The crane full-stroke operation safety monitoring system as described in claim 1, characterized in that, The start / stop control module is used to determine the load state corresponding to the load value interval where the current load value of the crane is located from the load value intervals of different load states corresponding to the crane type, based on the correspondence between the crane type and the load value intervals of different load states.

5. The crane full-stroke operation safety monitoring system as described in claim 1, characterized in that, The start / stop control module is used to end authorization after a countdown of one time after authorization is triggered; or, after authorization is triggered, authorization is ended when no personnel are detected in the hazardous work area for a continuous second time.

6. The crane full-stroke operation safety monitoring system as described in claim 1, characterized in that, The safety monitoring module is used to input the image of the crane operation area into the risk assessment model to obtain the personnel operation risk assessment result. The risk assessment model extracts initial semantic features from the crane operation area image; performs multi-layer key channel enhancement and receptive field expansion processing on the initial semantic features to obtain multi-scale deep semantic enhancement features; performs multi-target detection based on the multi-scale deep semantic enhancement features to obtain human bounding box detection results, human key point detection results, initial safety helmet wearing inspection results, and lifting equipment bounding box detection results; and performs multi-type operation risk assessment based on the human bounding box detection results, human key point detection results, initial safety helmet wearing inspection results, and lifting equipment bounding box detection results to obtain the personnel operation risk assessment result.

7. A method for monitoring the safety of a crane throughout its entire stroke operation, characterized in that, include: Collect the current load contact signals and current electrical signals of the crane's electrical cabinet, as well as the crane's current operating status parameters, current operating environment parameters, and images of the crane's operating area; The preprocessed current load contact signal, current electrical signal, current operating status parameters, and current working environment parameters are input into the load assessment model to obtain the current load value of the crane. The load assessment model uses a multi-parameter joint verification rule to perform a reliability verification on the processed current load contact signal to obtain a reliability verification result. If the reliability verification result indicates that the current load contact signal is in an unreliable state, a weight mapping rule is used to determine the current weight combination. If the reliability verification result indicates that the current load contact signal is in a reliable state, a weight generation strategy is used to generate the current weight combination. Based on the current weight combination, the preprocessed current load contact signal, current electrical signal, current operating status parameters, and current working environment parameters are weighted and fused to obtain the current load value of the crane. Based on the correspondence between different load states and load value ranges, the load state corresponding to the load value range where the current load value is located is determined as the current load state of the crane. When the current load state is no load, the safety monitoring module is disabled; When the current load state is a loaded state, the safety monitoring module is activated, or the posture detection model is used to detect whether the posture of the operator in the crane operation area image is a specific authorized posture. When the posture of the operator in the crane operation area image is detected to be a specific authorized posture, the authorization is triggered to disable the safety monitoring module. When the authorization end judgment condition is met, the authorization is terminated and the safety monitoring module is activated. A risk assessment model is used to perform target detection and risk assessment on the crane operation area image to obtain the personnel operation risk assessment result. When the personnel operation risk assessment result indicates the existence of personnel operation risk, the system will trigger a coordinated alarm for different types of alarms corresponding to the personnel operation risk.

Images