Port loading and unloading risk identification method and system based on image processing

Abstract

The invention relates to the technical field of image processing, and discloses a port loading and unloading risk identification method and system based on image processing. The method comprises the following steps: acquiring standardized image data through multispectral image acquisition and sea wind disturbance compensation processing, extracting multidimensional risk characteristics of suspension arm swinging, cargo deviation and personnel violation, performing dual evaluation of collision trajectory prediction and violation behavior detection, identifying comprehensive potential safety hazards by adopting a spatio-temporal correlation adaptive region growth algorithm, and determining whether the potential safety hazards exist or not. And generating port loading and unloading composite risk early warning information. The technical problems that multiple risk factors are difficult to accurately identify and the composite risk state cannot be effectively predicted in a complex marine environment in port loading and unloading operation are solved, and the environmental adaptability of port loading and unloading risk identification and the accuracy of composite risk assessment are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a port loading and unloading risk identification method and system based on image processing. Background Technology

[0002] Port loading and unloading operations are a crucial link in modern logistics and transportation. Traditional risk identification methods mainly rely on manual inspections and fixed sensor monitoring. This involves staff regularly patrolling the work site and monitoring the operational status of loading and unloading equipment and the safety of the working environment using pressure sensors, displacement sensors, and other equipment placed in key locations. Existing image processing technology also has some applications in port loading and unloading risk identification. It primarily uses single-spectrum visible light cameras for image acquisition, employs basic target detection algorithms to identify target objects such as personnel, loading and unloading equipment, and cargo, and uses simple area intrusion detection to determine whether there are any violations.

[0003] However, existing technologies have significant shortcomings: First, they have poor environmental adaptability. The port marine environment is subject to strong sea wind disturbances and reflective interference from the surface of metal equipment. Traditional image processing methods cannot effectively eliminate the impact of these environmental factors on image quality, leading to a decrease in target recognition accuracy. Second, they suffer from a lack of diversity in risk identification dimensions. Existing methods can typically only identify static violations or single types of equipment anomalies, lacking the ability to comprehensively analyze dynamic risk factors. Third, they suffer from insufficient accuracy in risk assessment. Existing technologies mainly adopt qualitative risk judgment methods, lacking quantitative risk assessment and prediction capabilities, and are unable to accurately predict the development trend and scope of impact of risks. Summary of the Invention

[0004] This application provides a port loading and unloading risk identification method and system based on image processing, which solves the technical problems of difficulty in accurately identifying multiple risk factors and effectively predicting complex risk states in complex marine environments during port loading and unloading operations. It significantly improves the environmental adaptability of port loading and unloading risk identification and the accuracy of complex risk assessment.

[0005] Firstly, this application provides a port loading and unloading risk identification method based on image processing, the port loading and unloading risk identification method based on image processing includes:

[0006] The original image sequence of the port loading and unloading operation site was acquired by multispectral image acquisition. The original image sequence was then subjected to composite environmental correction using sea breeze disturbance compensation and metal surface reflection suppression processing to obtain standardized image data.

[0007] Based on the target recognition results of crane boom, suspended cargo and operators in the standardized image data, multi-dimensional risk features are extracted from boom swing angle, cargo center of gravity shift and personnel violations to obtain composite risk feature data of loading and unloading operations.

[0008] The composite risk characteristic data of loading and unloading operations are used to predict collision trajectories and detect violations by the boundary of the port safety operation area. The prediction results are then used to determine the hazard level by dual assessment of swinging over the boundary and personnel intrusion, resulting in a graded composite risk assessment result.

[0009] An adaptive region growing algorithm based on spatiotemporal correlation is used to dynamically track the risk propagation range of the hierarchical composite risk assessment results. Combined with the superposition effect of swing collision risk and personnel injury risk, comprehensive safety hazards in loading and unloading operations are identified, and port loading and unloading composite risk early warning information is obtained.

[0010] Secondly, this application provides a port loading and unloading risk identification system based on image processing, the port loading and unloading risk identification system based on image processing includes:

[0011] The correction module is used to acquire the original image sequence of the port loading and unloading operation site through multispectral image acquisition, and to perform composite environmental correction on the original image sequence using sea breeze disturbance compensation and metal surface reflection suppression processing to obtain standardized image data.

[0012] The extraction module is used to extract multi-dimensional risk features of crane boom swing angle, cargo center of gravity shift and personnel violations based on the target recognition results of crane boom, suspended cargo and operators in the standardized image data, so as to obtain composite risk feature data of loading and unloading operations.

[0013] The detection module is used to predict the collision trajectory and detect violations by comparing the composite risk characteristic data of the loading and unloading operation with the boundary of the port safety operation area. The prediction results are used to determine the danger level through dual assessment of swinging over the boundary and personnel intrusion, so as to obtain the graded composite risk assessment results.

[0014] The tracking module is used to dynamically track the risk propagation range of the hierarchical composite risk assessment results using an adaptive region growth algorithm based on spatiotemporal correlation. It combines the superposition effect of swing collision risk and personnel injury risk to identify comprehensive safety hazards in loading and unloading operations and obtain port loading and unloading composite risk early warning information.

[0015] Thirdly, a port loading and unloading risk identification device based on image processing is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the port loading and unloading risk identification device based on image processing to execute the aforementioned port loading and unloading risk identification method based on image processing.

[0016] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the aforementioned image processing-based port loading and unloading risk identification method.

[0017] The technical solution provided in this application utilizes multispectral image acquisition technology to simultaneously acquire visible light and infrared thermal imaging information. Compared to traditional single-spectral image acquisition methods, it offers richer information dimensions. The introduction of sea breeze disturbance compensation and metal surface reflection suppression technologies effectively solves the technical challenges of unstable image quality and metal equipment reflection interference in port marine environments, significantly enhancing the adaptability and robustness of image processing algorithms in complex environments. The multidimensional risk feature extraction technology, through comprehensive analysis of boom swing angle, cargo center of gravity shift, and personnel violations, overcomes the limitations of traditional methods that can only identify single risk types. It achieves simultaneous identification and quantitative assessment of multiple risk factors in port loading and unloading operations. The dual assessment mechanism of collision trajectory prediction and violation detection can accurately predict the development trend of dynamic risks, demonstrating a significant forward-looking advantage compared to existing passive monitoring methods.

[0018] The application of the spatiotemporal correlation-based adaptive region growing algorithm in the field of port loading and unloading risk identification fully considers the spatial continuity and temporal evolution of risk propagation in the port operation environment. The algorithm's adaptive characteristics enable it to dynamically adjust growth parameters according to the actual risk distribution. Compared with traditional fixed-parameter region growing algorithms, it has higher identification accuracy and stronger environmental adaptability. The dynamic tracking function of risk propagation range can track the spatial path and time node of risk diffusion in real time, providing a scientific basis for port safety management personnel to formulate accurate emergency response strategies. The analysis of the superposition effect of swing collision risk and personnel injury risk accurately quantifies the comprehensive safety threat level under the composite risk state, making up for the shortcomings of existing technologies in the analysis of multiple risk synergy. The generation of port loading and unloading composite risk early warning information realizes the technological leap from single risk early warning to comprehensive risk early warning, and significantly improves the safety assurance level of port loading and unloading operations. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of an embodiment of the port loading and unloading risk identification method based on image processing in this application.

[0021] Figure 2 This is a schematic diagram illustrating the temporal changes in the extraction of multidimensional risk features in the embodiments of this application;

[0022] Figure 3 This is a schematic diagram of an embodiment of the port loading and unloading risk identification system based on image processing in this application.

[0023] Figure 4 This is a schematic block diagram of the port loading and unloading risk identification device based on image processing in an embodiment of the present invention. Detailed Implementation

[0024] This application provides a port loading and unloading risk identification method and system based on image processing. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data used can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0025] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the port loading and unloading risk identification method based on image processing in this application includes:

[0026] Step S101: Obtain the original image sequence of the port loading and unloading operation site through multispectral image acquisition, and perform composite environmental correction on the original image sequence using sea breeze disturbance compensation and metal surface reflection suppression processing to obtain standardized image data.

[0027] Step S102: Based on the target recognition results of the crane boom, suspended goods and operators in the standardized image data, extract multi-dimensional risk features of the boom swing angle, the center of gravity shift of the goods and the violation of personnel to obtain composite risk feature data of loading and unloading operations.

[0028] Step S103: The composite risk characteristic data of loading and unloading operations are compared with the boundary of the port safety operation area to predict the collision trajectory and detect violations. The prediction results are judged by the dual assessment of swinging over the boundary and personnel intrusion to obtain the graded composite risk assessment results.

[0029] Step S104: Using an adaptive region growth algorithm based on spatiotemporal correlation, the risk propagation range of the graded composite risk assessment results is dynamically tracked. Combined with the superposition effect of swing collision risk and personnel injury risk, comprehensive safety hazards in loading and unloading operations are identified, and port loading and unloading composite risk early warning information is obtained.

[0030] It is understood that the executing entity of this application can be a port loading and unloading risk identification system based on image processing, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiment uses a server as an example for illustration.

[0031] Specifically, multispectral image sequences of the port loading and unloading site are simultaneously acquired using a binocular stereo vision camera and an infrared thermal imaging camera. The binocular stereo vision camera captures visible light images to obtain spatial geometric information, while the infrared thermal imaging camera captures thermal infrared images to identify equipment temperature distribution and personnel thermal characteristics. The acquired image sequences are first time-stamp aligned to ensure the synchronization of different spectral images along the timeline. Subsequently, composite environmental corrections are performed to address the unique characteristics of the port's marine environment. Sea wind disturbance compensation quantifies image displacement deviation by calculating the gradient of pixel grayscale value changes between consecutive frames. This deviation reflects the vibration impact of sea wind on the camera equipment, and this vibration interference is eliminated through image stabilization processing. Metal surface reflection suppression addresses the strong reflection problem caused by the large number of metal equipment in the port. By identifying high-brightness pixels and performing intensity attenuation processing, the pixel values ​​of the reflective areas are smoothly fused with the surrounding normal areas to obtain standardized image data free from environmental interference.

[0032] Multi-target recognition is performed based on standardized image data. Contour feature extraction algorithms are used to identify the steel structure contour of the crane boom, the geometric shape of suspended goods, and the human contour of workers. Each target object is segmented and labeled to form a classification target recognition result. For the crane boom, its spatial coordinates are extracted to calculate the angle between the boom and the vertical baseline. The angle change values ​​between consecutive frames are used to statistically analyze the swing amplitude and frequency, forming the boom swing angle parameter. For suspended goods, the offset distance of the cargo's center of gravity relative to the lifting point is calculated by tracking the trajectory of the center of gravity coordinate change. The offset direction and velocity are processed through vector analysis to obtain the cargo center of gravity offset parameter. For workers, their position coordinates are spatially compared with the boundary of a preset safety zone to determine the type of violation of personnel entering the danger zone and to statistically analyze the duration. Finally, the boom swing angle parameter, cargo center of gravity offset parameter, and violation statistics are fused to form composite risk feature data for loading and unloading operations.

[0033] The motion trajectories of the crane boom and cargo are modeled and calculated using physical dynamics. The periodic swing motion of the boom is modeled using a sine function, and the coordinate sequence of the boom's motion trajectory is calculated by combining the boom length and swing angle. Simultaneously, the inertial swing of the suspended cargo is modeled using damped vibration, and the coordinate sequence of the cargo's swing trajectory is calculated considering the cargo's mass and wind load coefficient. After transforming the two trajectory coordinate sequences into a coordinate system, the intersection of the two trajectory coordinate sequences with the geometric boundary line of the port's safe operation area is determined, resulting in a set of trajectory boundary intersection coordinates, forming a collision trajectory prediction result. Simultaneously, time-series analysis is performed on the behavior patterns of personnel entering the hazardous area, comparing and detecting personnel position coordinates with the pre-set restricted area boundary in real time, forming a violation detection result. By weighted and fused calculation of the swing boundary crossing risk level from the collision trajectory prediction result and the personnel intrusion risk level from the violation detection result, the fused risk value is divided into threshold intervals and calibrated to obtain a graded composite risk assessment result.

[0034] First, pixels with risk levels exceeding the safety threshold are selected as initial growth seeds. A risk propagation seed point set is formed by calculating the spatiotemporal neighborhood correlation of these seed points. The algorithm analyzes the temporal changes based on the risk correlation between adjacent seed points. When the correlation exceeds the propagation threshold, the corresponding area is adaptively expanded to form a risk propagation and diffusion area. Within this area, the swing collision risk and personnel injury risk are superimposed to calculate the effect. The swing collision risk is quantified by collision speed, collision area, and equipment mass, while the personnel injury risk is quantified by the number of personnel, exposure time, and protection level. The superimposed composite risk value is weighted according to spatial distance, forming a spatially weighted comprehensive risk intensity value. Finally, the risk intensity data is spatially mapped according to geographic coordinates using color coding and contour lines to create a comprehensive safety hazard distribution map and generate port loading and unloading composite risk early warning information.

[0035] In one specific embodiment, a sequence of original images of the port loading and unloading operation site is acquired through multispectral image acquisition. The original image sequence is then subjected to composite environmental correction using sea breeze disturbance compensation and metal surface reflectivity suppression processing to obtain standardized image data, including:

[0036] Simultaneous image acquisition using a binocular stereo vision camera and an infrared thermal imaging camera; time-stamping and aligning visible light and thermal infrared images of the port loading and unloading site to obtain a multispectral raw image sequence.

[0037] Based on the gradient of pixel grayscale value changes in the multispectral original image sequence, the wind disturbance amplitude of the image displacement deviation between consecutive frames is calculated, and the calculation results are subjected to image stabilization processing to obtain the wind compensation and correction image.

[0038] Based on the pixel brightness distribution characteristics of the metal surface region in the sea breeze compensation and correction image, reflective areas are identified and intensity attenuation is performed on high-brightness pixels. The attenuated pixel values ​​are then smoothly fused with the surrounding normal areas to obtain a reflection suppression processed image.

[0039] The image with reflection suppression is subjected to global contrast adjustment, and the adjusted image is then subjected to noise filtering and edge sharpening to obtain standardized image data.

[0040] Specifically, data fusion is achieved using a binocular stereo vision camera and an infrared thermal imaging camera. The binocular stereo vision camera uses two cameras spaced a certain baseline distance to simulate human vision and acquire depth information of objects. The infrared thermal imaging camera acquires temperature distribution information by detecting the infrared radiation emitted by objects. During synchronous image acquisition, both cameras simultaneously capture images of the port loading and unloading site. The binocular stereo vision camera captures visible light images containing RGB three-channel information, while the infrared thermal imaging camera captures thermal infrared images reflecting temperature differences. Timestamp alignment is performed by adding precise timestamps to each frame of the image, matching images acquired by different sensors on the timeline to ensure that subsequent analysis uses multispectral information from the same moment, thus forming a multispectral original image sequence.

[0041] The calculation of sea breeze disturbance amplitude is based on the quantization of image displacement deviation using pixel grayscale value change gradients. Pixel grayscale value change gradients refer to the difference in grayscale values ​​between adjacent pixels. The overall image displacement is detected by calculating the difference in grayscale values ​​of pixels at the same position between consecutive frames. Sea breeze disturbances cause slight vibrations in the camera equipment, which manifest as overall image displacement jitter. By comparing the positional changes of the same feature points in two consecutive frames, the horizontal and vertical displacement amounts are calculated. Image stabilization processing employs a motion compensation algorithm. Based on the calculated displacement amounts, the current frame image undergoes reverse offset correction. That is, if a certain number of pixels are detected as shifted to the right, the entire image is shifted to the left by the corresponding distance, thereby eliminating the image jitter caused by sea breeze disturbances and obtaining a sea breeze-compensated corrected image.

[0042] The identification of reflective areas on metal surfaces is based on the detection of high-brightness pixels using pixel brightness distribution characteristics. Pixel brightness distribution characteristics refer to the statistical distribution of brightness values ​​of each pixel in an image. In port environments, a large number of metal devices will produce strong reflections under sunlight. These reflective areas appear as areas with abnormally high local pixel brightness values ​​in the image. Reflective area identification filters out pixels with brightness values ​​exceeding the normal range by setting a brightness threshold. These pixels are usually in a near-white high-brightness state. Intensity attenuation processing reduces the brightness value of these high-brightness pixels by a certain proportion, the degree of reduction being determined based on the brightness difference between the pixel and its surrounding normal pixels. Gray-scale smoothing fusion processing uses a weighted average algorithm to fuse the attenuated pixel value with the average value of its eight neighboring pixels. The fusion weight is allocated according to distance, with closer pixels having a higher weight, thereby achieving a natural transition between the reflective area and the surrounding normal area, resulting in a reflection-suppressed image.

[0043] Global contrast adjustment enhances the visual effect of an image by stretching its dynamic range of brightness. Contrast refers to the difference in brightness between the brightest and darkest areas of an image. Global contrast adjustment remaps the brightness distribution of the entire image, making dark areas darker and bright areas brighter, thus enhancing the image's sense of depth. Noise filtering uses a median filtering algorithm to eliminate random noise points in the image. Median filtering replaces the value of a pixel with the median of all pixels in its neighborhood; this method effectively suppresses salt-and-pepper noise. Edge sharpening uses the Laplacian operator to enhance image edges. The Laplacian operator detects edges by calculating the second derivative of each pixel, then superimposes the detected edge information onto the original image, making object outlines clearer and obtaining standardized image data.

[0044] In one specific embodiment, based on the target recognition results of the crane boom, suspended cargo, and operators in standardized image data, multi-dimensional risk features are extracted from the boom swing angle, cargo center of gravity shift, and personnel violations to obtain composite risk feature data for loading and unloading operations, including:

[0045] Based on the contour features in standardized image data, target classification and recognition are performed. The steel structure contour of the crane boom, the geometric shape of the suspended goods, and the human body contour of the workers are segmented and labeled to obtain the target recognition results.

[0046] Based on the spatial coordinates of the crane boom in the classification target recognition results, the angle between the boom and the vertical baseline is calculated geometrically. The angle change values ​​between consecutive frames are statistically analyzed for swing amplitude and swing frequency to obtain the boom swing angle parameters.

[0047] Based on the trajectory of the change of the center of gravity of the suspended cargo in the classification target recognition results, the offset distance of the cargo's center of gravity relative to the lifting point is measured and calculated in real time. The offset direction and offset speed are processed by vector analysis to obtain the cargo's center of gravity offset parameters.

[0048] The spatial relationship between the location coordinates of the workers in the classification target identification results and the boundary of the preset safety area is determined. The violation type and duration of the personnel entering the dangerous area are identified and statistically analyzed. The crane swing angle parameters, cargo center of gravity offset parameters and violation statistics are fused to obtain composite risk characteristic data of loading and unloading operations.

[0049] Specifically, the algorithm distinguishes three target objects—crane booms, suspended goods, and workers—by analyzing the boundary shape features of different objects in standardized image data. Contour features refer to the geometric shape information of object boundaries, including attributes such as the curvature, angle, and length of the boundary lines. The steel structure contour of a crane boom exhibits a polygonal structure with distinct straight lines and regular angles. The geometric shape of suspended goods typically presents regular geometric features such as rectangles or cylinders. The human contour of workers exhibits typical human features such as a round head, rectangular torso, and striped limbs. Region segmentation employs a connected component analysis algorithm, grouping pixels belonging to the same target object into a connected region. A connected region refers to a set of spatially adjacent pixels with the same attributes in an image. Labeling is then performed, assigning a unique identifier to each identified target object and establishing a correspondence between the target object and its spatial location and geometric attributes, thus forming the classification target recognition result.

[0050] The calculation of the boom swing angle parameter is based on the geometric angle measurement of the spatial position coordinates of the crane boom in the classification target recognition results. The spatial position coordinates refer to the pixel position of the boom in the image coordinate system, which is converted into actual coordinates in three-dimensional space through the principle of binocular stereo vision. The vertical baseline is set as the reference line of the direction of gravity, that is, a vertical line from the root of the boom downwards. The geometric angle calculation adopts the vector angle formula, which mathematically calculates the angle value by performing a mathematical operation between the boom vector and the vertical baseline vector. The boom vector points from the root of the boom to the end of the boom, and the vertical baseline vector is a fixed downward unit vector. The angle change value between consecutive frames is obtained by subtracting the angle of the previous frame from the current frame angle. The swing amplitude refers to the difference between the maximum and minimum angle change values ​​within a continuous time period. The swing frequency refers to the number of swing cycles completed per unit time, which is calculated by counting the number of peaks and troughs in the angle change curve.

[0051] The calculation of cargo center of gravity offset parameters is based on spatial displacement analysis of the temporal variation trajectory of the cargo's center of gravity coordinates. The center of gravity coordinates refer to the pixel position of the cargo's geometric center of gravity in the image, obtained by calculating the weighted average of the coordinates of all pixels within the cargo's outline. The suspension point refers to the fixed connection point where the cargo is suspended, typically located at the crane hook directly above the cargo. The offset distance is calculated using the Euclidean distance formula, determining the straight-line distance between the center of gravity coordinates and the suspension point coordinates. This distance reflects the degree of deviation of the cargo's center of gravity from its ideal vertical suspension position. The offset direction is represented by the azimuth angle of the center of gravity coordinates relative to the suspension point coordinates, which is the angle rotated clockwise from true north to the offset direction. The offset velocity is obtained by dividing the change in offset distance between consecutive frames by the time interval. Vector analysis combines the offset direction and velocity into a vector containing both magnitude and direction information, forming the cargo center of gravity offset parameters.

[0052] Violation detection identifies worker behavior patterns entering hazardous areas by judging spatial relationships. Preset safety zone boundaries are established according to port operation safety regulations, including the crane operating radius, cargo swing range, and equipment running tracks. Spatial relationship judgment uses a point-within-a-polygon algorithm, calculating the relative positional relationship between the worker's coordinates and the safety zone boundary to determine if boundary crossing has occurred. Violation type identification categorizes violations based on the specific type of hazardous area entered, including entering the crane operating range, approaching operating equipment, and entering cargo stacking areas. Duration statistics calculate the duration of violations by recording the start and end times of entry into the hazardous area. Data fusion processing employs a weighted fusion algorithm, combining crane swing angle parameters, cargo center of gravity offset parameters, and violation statistics with different weighting coefficients. These weighting coefficients are set according to the importance of each parameter to the overall risk assessment, forming composite risk characteristic data for loading and unloading operations.

[0053] Figure 2 This is a schematic diagram illustrating the temporal changes of multidimensional risk feature extraction in this embodiment of the application. The diagram shows the dynamic trends of multidimensional risk feature parameters over a 150-second time period during port loading and unloading operations. Solid circles represent the temporal changes in the boom swing angle, reaching a maximum of 18 degrees at 75 seconds; solid triangles represent the cargo center of gravity offset distance, peaking at 15 centimeters at 75 seconds; dashed squares represent the cumulative number of personnel violations, showing a trend of first increasing and then decreasing; dashed diamonds represent the comprehensive risk index calculated based on the first three parameters, reaching a maximum value of 92 at 75 seconds. As can be observed from the diagram, the four risk parameters exhibit a clear correlation over time; when the boom swing and cargo offset reach their peaks, the comprehensive risk index also reaches its highest point.

[0054] In one specific embodiment, collision trajectory prediction and violation detection are performed by combining the composite risk characteristic data of loading and unloading operations with the boundary of the port's safe operation area. The prediction results are then used to determine the hazard level through a dual assessment of boundary crossing and personnel intrusion, resulting in a graded composite risk assessment result, including:

[0055] Based on the boom swing angle parameters and cargo center of gravity offset parameters in the composite risk characteristic data of loading and unloading operations, physical dynamics modeling and calculation of the motion trajectory of the boom and cargo are performed. The spatial intersection analysis of the calculated trajectory coordinates with the boundary of the port safe operation area is conducted to obtain the collision trajectory prediction results.

[0056] Based on the statistical results of violations in the composite risk characteristic data of loading and unloading operations, a time series analysis is conducted on the behavioral patterns of workers entering dangerous areas. The personnel location coordinates are compared and detected in real time with the boundaries of the preset prohibited areas to obtain the violation detection results.

[0057] The risk level of swinging and crossing the boundary in the collision trajectory prediction results and the risk level of personnel intrusion in the violation detection results are weighted and fused together. The fused risk value is then divided into threshold ranges and graded to obtain a dual-assessment hazard level.

[0058] Based on the numerical distribution characteristics of the dual-assessment hazard levels, different risk types are prioritized and classified, and the ranking results are hierarchically classified and stored according to the urgency of the risk, resulting in a graded composite risk assessment result.

[0059] Specifically, physical dynamics modeling refers to using Newtonian mechanics principles to describe the motion of an object under the action of forces, including the mathematical relationships between motion parameters such as displacement, velocity, and acceleration. Crane trajectory modeling employs the principle of simple harmonic motion, treating the crane's swing as a damped pendulum motion, and calculating the spatial coordinates of the crane's end at different times using parameters such as crane length, swing angle, and swing frequency. Cargo trajectory modeling considers the effects of inertial oscillation and wind loads, treating the cargo's center of gravity shift as a point mass motion under external forces, and calculating the cargo's position coordinates at various time points using parameters such as mass, gravity, and wind force. Trajectory coordinate calculation uses a time-step recursive method, starting from the current moment and extrapolating forward at fixed time intervals. Each time step calculates the next motion state based on the previous step's position and velocity. Spatial intersection analysis performs geometric intersection judgments between the predicted motion trajectory coordinates and the port's safe operation area boundaries. These boundaries include polygonal boundary lines such as building outlines, equipment restricted areas, and personnel passage paths. The positional relationship between points and polygons determines whether the trajectory crosses the boundary, forming a collision trajectory prediction result.

[0060] Time series analysis extracts temporal features from the statistical results of violations to identify the behavioral patterns of workers entering hazardous areas. Time series analysis involves statistically analyzing data arranged in chronological order to identify temporal characteristics such as trends, periodicity, and abrupt changes. Worker behavior pattern analysis calculates behavioral parameters such as movement speed, direction of movement, and dwell time by recording the temporal changes in personnel position coordinates. Preset restricted area boundaries are defined according to port safety regulations, including the geometric boundaries of hazardous areas such as crane operating radii, cargo swing range, and equipment running tracks. Real-time comparative detection employs a continuous monitoring method, acquiring the current position coordinates of personnel at fixed time intervals and determining their spatial relationship with the restricted area boundaries. The start time of the violation is recorded when the personnel's position exceeds the safety boundary, and the end time is recorded when the personnel return to the safe area. The duration of the violation is calculated using the time difference, forming the violation detection result.

[0061] Weighted fusion calculation combines the swing boundary crossing risk level from the collision trajectory prediction results with the personnel intrusion risk level from the violation detection results. Weighted fusion involves assigning weight coefficients based on the importance of different data sources, and then linearly combining the values ​​from each data source according to their weights. The swing boundary crossing risk level is quantified based on factors such as the number of intersections between the trajectory and the boundary, the location of the intersections, and the collision intensity. The personnel intrusion risk level is quantified based on factors such as the type of violation area, the duration of the violation, and the number of personnel. The weight coefficients are set considering the actual risk characteristics of port operations; typically, the weight of swing collision risk is higher because it involves equipment damage and personnel injury, while the weight of personnel violation risk is relatively lower but still requires close attention. Threshold range division classifies the fused risk values ​​according to preset numerical ranges, including low risk, medium risk, high risk, and extremely high risk. The level labeling process assigns a corresponding identifier and processing priority to each risk level, forming a dual-assessment hazard level.

[0062] Prioritization is based on the numerical distribution characteristics of the dual-assessment hazard levels to rank the importance of different risk types. These numerical distribution characteristics include statistical features such as the frequency, spatial, and temporal distribution of risk levels. Risk types are categorized according to their source and scope of impact, including equipment sway risk, personnel violation risk, and compound risks. The priority ranking algorithm calculates a priority score for each risk based on a combination of factors, including risk level value, scope of impact, and urgency. Risks with higher scores are ranked higher. Hierarchical labeling assigns visual identifiers such as color codes, graphic symbols, and text labels to risks of different priorities. Hierarchical classification storage uses a tree structure to organize risk data. The top-level node represents the overall risk assessment result, the middle-level nodes represent risk classifications of different types and levels, and the bottom-level nodes contain specific risk event records. Risk urgency is assessed based on time sensitivity and the severity of consequences. Risks with high urgency require immediate action, while risks with low urgency can be addressed later, resulting in a hierarchical composite risk assessment.

[0063] In one specific embodiment, based on the boom swing angle parameter and cargo center of gravity offset parameter in the composite risk characteristic data of loading and unloading operations, physical dynamics modeling and calculation are performed on the motion trajectory of the boom and cargo. The calculated trajectory coordinates are then spatially intersected with the boundary of the port safety operation area to obtain the collision trajectory prediction result, including:

[0064] Based on the swing amplitude and swing frequency values ​​in the swing angle parameters of the boom, the periodic motion of the boom end is modeled and calculated using a sine function. The boom length and swing angle are input into the kinematic model to solve for the trajectory coordinates, and the boom motion trajectory coordinate sequence is obtained.

[0065] Based on the offset distance and offset velocity vector in the cargo center of gravity offset parameters, damped vibration modeling calculation is performed on the inertial swing of the suspended cargo. The cargo mass and wind load coefficient are input into the dynamic model to predict the displacement and obtain the cargo swing trajectory coordinate sequence.

[0066] The coordinate system is transformed into the coordinate system of the boom motion trajectory coordinate sequence and the cargo swing trajectory coordinate sequence. The intersection of the unified trajectory coordinates with the geometric boundary line of the port safety operation area is calculated to obtain the set of trajectory boundary intersection point coordinates.

[0067] Based on the spatial distribution characteristics of the coordinate set of the intersection points of the trajectory boundary, the collision time and collision intensity at the intersection points are quantified and calculated. The quantification results are then classified and labeled according to the degree of collision danger to obtain the collision trajectory prediction results.

[0068] Specifically, sine function modeling refers to using the periodicity of trigonometric functions to describe the reciprocating motion of an object, including key parameters such as amplitude, frequency, and phase. The swing amplitude corresponds to the amplitude parameter of the sine function, representing the maximum angle at which the boom deviates from its vertical position. The swing frequency corresponds to the angular frequency parameter of the sine function, representing the number of swing cycles completed per unit time. In establishing the kinematic model, the boom length is used as the radius parameter, and the swing angle as the angle parameter. The spatial position of the boom's end at different times is calculated using the polar coordinate to rectangular coordinate conversion relationship. The trajectory coordinate solution employs a time discretization method, recursively calculating from the initial moment according to a fixed time step. At each time point, the corresponding angle value is calculated based on the sine function value, and then combined with the boom length to calculate the rectangular coordinate position of the end, forming a sequence of boom motion trajectory coordinates arranged in chronological order.

[0069] Damped vibration modeling and calculation are based on the offset distance and offset velocity vector in the cargo's center of gravity offset parameters to construct a mathematical model of the vibration system affected by damping. Damped vibration refers to the vibrational motion of an object under the combined action of elastic restoring force and damping force. The elastic restoring force tends to pull the object back to its equilibrium position, while the damping force hinders the object's motion and dissipates kinetic energy. The offset distance corresponds to the vibration displacement parameter, representing the degree of deviation of the cargo's center of gravity from its equilibrium position, and the offset velocity vector corresponds to the vibration velocity parameter, containing information on the magnitude and direction of the velocity. During the establishment of the dynamic model, the cargo mass, as an inertial parameter, affects the vibration period and decay characteristics, while the wind load coefficient, as an external force parameter, affects the excitation and damping effects of the vibration. Displacement prediction calculation uses a numerical integration method. Based on the current displacement and velocity state, combined with parameters such as mass, elastic coefficient, and damping coefficient, the motion state at the next moment is calculated. By progressively recursively obtaining the cargo's position coordinates at various future time points, a sequence of cargo swing trajectory coordinates is formed.

[0070] The coordinate system transformation converts the coordinate sequences of the boom's motion trajectory and the cargo's swing trajectory to the same reference coordinate system for unified processing. This transformation refers to the mathematical transformation process of converting coordinate data from different coordinate systems to a unified reference coordinate system, including geometric transformations such as translation, rotation, and scaling. The boom's motion trajectory is typically established in a polar coordinate system with the boom base as the origin, while the cargo's swing trajectory is typically established in a rectangular coordinate system with the suspension point as the origin. The unified transformation process requires aligning the origins, coordinate axis directions, and scale units of the two coordinate systems. The line segment intersection judgment calculation uses a line segment intersection algorithm from computational geometry. Adjacent points in the trajectory coordinate sequence are connected to form trajectory line segments. The boundary of the port safety operation area is represented as a set of polygonal boundary line segments. Intersections between trajectory line segments and boundary line segments are calculated one by one to determine whether an intersection has occurred. The intersection judgment process includes two stages: a rapid rejection test and a crossover test. The rapid rejection test quickly eliminates obviously non-intersecting line segment pairs by comparing whether the circumscribed rectangles of the line segments overlap. The crossover test accurately judges whether the line segments intersect by calculating the positional relationship between the points and the lines, and obtains the set of coordinates of the intersection points of the trajectory boundaries.

[0071] Risk quantification calculation uses the spatial distribution characteristics of the coordinate set of trajectory boundary intersection points to numerically assess collision risk. These spatial distribution characteristics include the number, location, density, and other geometric and statistical properties of the intersection points. Collision time calculation determines the time corresponding to the intersection point using time stamps on the trajectory coordinate sequence; that is, the time value when the trajectory reaches the intersection point. Collision intensity calculation considers the momentum of the moving object and the stiffness characteristics of the colliding object. Momentum is determined by the object's mass and velocity at the moment of collision, while stiffness characteristics are determined by the material properties and geometry of the colliding object. The product of these two factors reflects the magnitude of the impact force generated by the collision. Grading and labeling processing classifies the risk into different levels based on the collision intensity value, including minor risk, moderate risk, severe risk, and extremely severe risk. Each level corresponds to a different color label and processing priority, forming the collision trajectory prediction result.

[0072] In one specific embodiment, an adaptive region growing algorithm based on spatiotemporal correlation is used to dynamically track the risk propagation range of the hierarchical composite risk assessment results. This algorithm, combined with the superposition effect of swing collision risk and personnel injury risk, identifies comprehensive safety hazards in loading and unloading operations, resulting in port loading and unloading composite risk early warning information, including:

[0073] An adaptive region growth algorithm based on spatiotemporal correlation selects seed points for high-risk areas in the hierarchical composite risk assessment results. Pixels with risk level values ​​exceeding the safety threshold are set as initial growth seeds. The spatiotemporal neighborhood correlation of the seed points is calculated and analyzed to obtain a set of risk propagation seed points.

[0074] Based on the spatial distribution characteristics of the risk propagation seed point set, a time-series change analysis of the risk correlation between adjacent seed points is performed. Regions with correlation exceeding the propagation threshold are adaptively expanded and grown to obtain the risk propagation diffusion region.

[0075] The combined effect of the swing collision risk value and the personnel injury risk value in the risk propagation and diffusion area is calculated. The combined comprehensive risk intensity is then spatially weighted and fused to obtain a comprehensive safety hazard distribution map.

[0076] Based on the risk density distribution characteristics of the comprehensive safety hazard distribution map, the early warning level of high-density risk areas is marked and the time urgency is assessed. The assessment results are classified, organized and stored according to the early warning priority to obtain port loading and unloading composite risk early warning information.

[0077] Specifically, a seed point selection mechanism identifies high-risk areas in the graded composite risk assessment results. Spatiotemporal correlation refers to the relationship between risks in spatial location and time series, including the propagation characteristics of risks between adjacent areas and the evolution of risks over time. The seed point selection process compares risk level values ​​with preset safety thresholds. When the risk level value of a pixel exceeds the safety threshold, that pixel is marked as an initial growth seed. The safety threshold is set according to the port's risk tolerance and safety regulations. Spatiotemporal neighborhood correlation calculation analyzes the risk distribution within the spatial neighborhood of the seed point and the risk change trend within the temporal neighborhood. The spatial neighborhood includes eight connected pixels around the seed point, and the temporal neighborhood includes the risk change sequence of the seed point within several time frames before and after it. The correlation calculation uses the correlation coefficient method, quantifying the correlation strength by calculating the statistical correlation between the risk values ​​of the seed point and its neighboring pixels. A higher correlation coefficient value indicates a stronger correlation in risk propagation, forming a risk propagation seed point set.

[0078] The temporal variation analysis of risk correlation assesses the risk propagation relationship between seed points based on the spatial distribution characteristics of the seed point set. These spatial distribution characteristics include geometric and statistical properties such as seed point density distribution, clustering degree, and distribution direction. Adjacent seed points are identified using a distance threshold method; when the Euclidean distance between two seed points is less than a preset threshold, they are considered spatially adjacent. Risk correlation calculation considers multiple factors, including distance, risk level similarity, and time synchronization. Seed points that are closer, have more similar risk levels, and are more synchronized in time have a higher correlation. The temporal variation analysis assesses the dynamic characteristics of risk propagation by changing correlation over continuous time frames. An increasing correlation over time indicates that the risk is spreading. An adaptive radius expansion growth mechanism dynamically adjusts the growth radius based on the correlation value. When the correlation exceeds the propagation threshold, the radius expands around the seed point. The expansion radius is proportional to the correlation value; the higher the correlation, the larger the expansion radius, forming a risk propagation and diffusion area.

[0079] The superposition effect calculation comprehensively assesses the swing collision risk and personnel injury risk values ​​within the risk propagation and diffusion area. The superposition effect refers to the combined impact of multiple risk factors existing simultaneously, and its overall risk level is typically higher than the simple sum of the individual risks. The swing collision risk value is derived from the collision prediction results of the boom and cargo movement trajectories, reflecting the physical impact threat posed by equipment swinging to surrounding objects and personnel. The personnel injury risk value is derived from the detection results of worker violations, reflecting the probability of personnel being injured in a hazardous environment. The superposition effect calculation employs a nonlinear fusion method, considering the interaction and amplification effects between different risk types. When swing collision risk and personnel injury risk coexist in the same area, their combined risk intensity exhibits an exponential growth characteristic. Spatial weighted fusion processing assigns different weight coefficients based on the spatial location of the risk source; areas closer to the risk source have higher weights. The weight allocation uses an inverse distance-proportional relationship. A weighted average is used to calculate the comprehensive risk intensity value within the area, forming a comprehensive safety hazard distribution map.

[0080] The early warning level classification is based on the risk density distribution characteristics of the comprehensive safety hazard distribution map to classify different areas into risk levels. Risk density distribution characteristics refer to the statistical distribution of risk intensity per unit area, including statistical parameters such as the average, maximum, and variance of risk intensity. High-density risk area identification uses a risk density threshold screening method; when the risk density of an area exceeds the high-density threshold, that area is marked as a high-density risk area. Early warning level classification divides high-density risk areas into different early warning levels based on risk density values, including yellow, orange, and red warnings, each corresponding to different emergency response measures and processing time limits. Time urgency assessment uses risk evolution trend analysis to determine the time urgency of risk development. Evolution trend analysis is based on the risk density change curve of continuous time frames; when the risk density shows a rapid upward trend, it indicates high time urgency, requiring immediate emergency measures. The assessment results are hierarchically organized according to early warning priority. High-priority early warning information is ranked first and marked as urgent, while low-priority early warning information is ranked last and marked as routine, forming a composite risk early warning information system for port loading and unloading.

[0081] In one specific embodiment, the oscillating collision risk value and the personnel injury risk value in the risk propagation and diffusion area are superimposed to calculate the effect. The combined risk intensity after superposition is then spatially weighted and fused to obtain a comprehensive safety hazard distribution map, including:

[0082] Based on the spatial coordinate distribution of the risk propagation and diffusion area, the hazard intensity of the swing collision risk value in the area is quantified. The collision speed, collision area and equipment mass parameters are numerically processed to obtain the quantified swing collision risk value.

[0083] Based on the population distribution density in the risk transmission and diffusion area, the exposure risk of personnel injury in the area is assessed and calculated. The number of personnel, exposure time, and protection level parameters are quantified to obtain a quantitative value of personnel injury risk.

[0084] The quantitative values ​​of swing collision risk and personnel injury risk are nonlinearly superimposed and calculated. The composite risk value after superposition is weighted according to the spatial distance weight to obtain the spatial weighted comprehensive risk intensity value.

[0085] Based on the distribution characteristics of spatially weighted comprehensive risk intensity values, color coding and contour line drawing are performed on different intensity intervals. The processed risk intensity data is then spatially mapped and stored according to geographic coordinates to obtain a comprehensive safety hazard distribution map.

[0086] Specifically, spatial coordinate distribution refers to the geographic coordinate information of various locations within the risk propagation and diffusion area, including the latitude and longitude coordinates and relative positional relationships of each risk point. The swing collision risk value is calculated by comprehensively considering three key parameters: collision speed, collision area, and equipment mass. Collision speed refers to the velocity of the moving object at the moment of collision, calculated by analyzing the trajectory of the boom and cargo. Collision area refers to the geometric area of ​​the collision contact surface, determined by the geometry and contact method of the colliding objects. Equipment mass refers to the total mass of the equipment involved in the collision, including the total mass of components such as the boom, cargo, and lifting gear. The numerical processing converts these three physical parameters into a dimensionless risk score. The conversion method uses normalization, dividing each parameter value by its corresponding maximum possible value, and then performing a weighted sum according to preset weights. Collision speed has a higher weight because speed directly affects the impact force; collision area has a medium weight because area affects the damage range; and equipment mass has a relatively low weight, but inertial effects still need to be considered. The final quantified value of the swing collision risk is obtained.

[0087] Exposure assessment is based on the quantitative analysis of personnel injury risk through personnel distribution density within the risk propagation area. Personnel distribution density refers to the distribution of the number of workers per unit area. This density is calculated by statistically analyzing the number of workers within the area using image recognition technology and dividing by the area. Personnel injury risk assessment comprehensively considers three core parameters: number of workers, exposure time, and protection level. The number of workers refers to the total number of workers simultaneously exposed to the risk area; a higher number indicates a higher risk. Exposure time refers to the duration of time workers remain in the hazardous area, calculated through continuous monitoring of personnel location changes. Protection level refers to the level of safety equipment worn by workers, including the comprehensive protection capability rating of safety helmets, protective clothing, safety belts, and other protective equipment. Numerical processing employs a risk matrix method, classifying the number of workers, exposure time, and protection level according to their number, duration, and protective capability. Then, a lookup table method is used to determine the corresponding risk coefficients. Multiplying these three risk coefficients yields the quantitative value of the personnel injury risk.

[0088] Nonlinear superposition calculation combines the quantified values ​​of swing collision risk and personnel injury risk for a composite risk assessment. Nonlinear superposition means that the combined effect of the two risks is not a simple linear addition of their individual risks, but rather exhibits a nonlinear relationship of mutual amplification or synergistic effect. The superposition calculation uses an exponential amplification model; when both risks coexist, the composite risk value grows exponentially. The growth exponent is determined by the interaction strength of the risk types. The interaction between swing collision risk and personnel injury risk is strong because physical collisions directly threaten personnel safety. Spatial distance weighting is based on the spatial distance relationship between the risk source and the assessment point. A distance attenuation model is used: the closer the distance, the greater the weight; the farther the distance, the smaller the weight. The attenuation function uses an inverse proportional relationship, meaning the weight is inversely proportional to the square of the distance. The weighting coefficient allocation process multiplies the superimposed composite risk value by the corresponding distance weight coefficient. The distance weight coefficient reflects the spatial attenuation characteristics of the risk impact, ultimately yielding a spatially weighted comprehensive risk intensity value.

[0089] Color coding and contour plotting are based on the spatially weighted comprehensive risk intensity value distribution characteristics for visualization. Color coding assigns corresponding color labels to different risk intensity intervals, using a heatmap color scheme: low-risk areas are represented by green, medium-risk areas by yellow, high-risk areas by orange, and extremely high-risk areas by red. The intensity of the color is directly proportional to the risk intensity value. Contour plotting connects points with the same risk intensity value to form a contour map. Contour plotting uses a linear interpolation method, performing numerical interpolation between discrete points with known risk intensity values ​​to generate continuous contour outlines. Contour density reflects the magnitude of the risk gradient; denser contour lines indicate more drastic risk changes. Spatial mapping storage establishes a correspondence between the processed risk intensity data and geographic coordinate information. Geographic coordinates use the actual coordinate system of the port operation area, including latitude and longitude coordinates and relative coordinates. The storage format adopts a raster data structure, dividing the port operation area into regular grids. Each grid cell stores the corresponding location's risk intensity value and color coding information, forming a comprehensive safety hazard distribution map.

[0090] The above describes the port loading and unloading risk identification method based on image processing in the embodiments of this application. The following describes the port loading and unloading risk identification system based on image processing in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 3 One embodiment of the port loading and unloading risk identification system based on image processing in this application includes:

[0091] The correction module is used to acquire the original image sequence of the port loading and unloading operation site through multispectral image acquisition, and to perform composite environmental correction on the original image sequence using sea breeze disturbance compensation and metal surface reflection suppression processing to obtain standardized image data.

[0092] The extraction module is used to extract multi-dimensional risk features of crane boom swing angle, cargo center of gravity shift and personnel violations based on the target recognition results of crane boom, suspended cargo and operators in the standardized image data, so as to obtain composite risk feature data of loading and unloading operations.

[0093] The detection module is used to predict the collision trajectory and detect violations by comparing the composite risk characteristic data of the loading and unloading operation with the boundary of the port safety operation area. The prediction results are used to determine the danger level through dual assessment of swinging over the boundary and personnel intrusion, so as to obtain the graded composite risk assessment results.

[0094] The tracking module is used to dynamically track the risk propagation range of the hierarchical composite risk assessment results using an adaptive region growth algorithm based on spatiotemporal correlation. It combines the superposition effect of swing collision risk and personnel injury risk to identify comprehensive safety hazards in loading and unloading operations and obtain port loading and unloading composite risk early warning information.

[0095] above Figure 3 The image processing-based port loading and unloading risk identification system in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The image processing-based port loading and unloading risk identification device in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0096] Reference Figure 4 This invention also provides a port loading and unloading risk identification device based on image processing. This device can be a server, and its internal structure can be as follows: Figure 4 As shown, the image processing-based port loading and unloading risk identification device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computing and control capabilities. The memory of the image processing-based port loading and unloading risk identification device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the image processing-based port loading and unloading risk identification device stores the data corresponding to this embodiment. The network interface of the image processing-based port loading and unloading risk identification device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0097] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the image processing-based port loading and unloading risk identification device to which the present invention is applied.

[0098] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the image processing-based port loading and unloading risk identification method.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a port loading and unloading risk identification device based on image processing (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A port loading and unloading risk identification method based on image processing, characterized in that, The method includes: The original image sequence of the port loading and unloading operation site was acquired by multispectral image acquisition. The original image sequence was then subjected to composite environmental correction using sea breeze disturbance compensation and metal surface reflection suppression processing to obtain standardized image data. Based on the target recognition results of crane boom, suspended cargo and operators in the standardized image data, multi-dimensional risk features are extracted from boom swing angle, cargo center of gravity shift and personnel violations to obtain composite risk feature data of loading and unloading operations. The composite risk characteristic data of loading and unloading operations are used to predict collision trajectories and detect violations by the boundary of the port safety operation area. The prediction results are then used to determine the hazard level by dual assessment of swinging over the boundary and personnel intrusion, resulting in a graded composite risk assessment result. An adaptive region growing algorithm based on spatiotemporal correlation is used to dynamically track the risk propagation range of the hierarchical composite risk assessment results. Combined with the superposition effect of swing collision risk and personnel injury risk, comprehensive safety hazards in loading and unloading operations are identified, and port loading and unloading composite risk early warning information is obtained.

2. The port loading and unloading risk identification method based on image processing according to claim 1, characterized in that, The process involves acquiring raw image sequences of the port loading and unloading operation site through multispectral image acquisition, and then performing composite environmental correction on the raw image sequences using sea breeze disturbance compensation and metal surface reflectivity suppression processing to obtain standardized image data, including: Simultaneous image acquisition using a binocular stereo vision camera and an infrared thermal imaging camera; time-stamping and aligning visible light and thermal infrared images of the port loading and unloading site to obtain a multispectral raw image sequence. Based on the pixel grayscale value change gradient in the multispectral original image sequence, the sea breeze disturbance amplitude is calculated for the image displacement deviation between consecutive frames. The calculation results are then processed for image stabilization to obtain a sea breeze compensation and correction image. Based on the pixel brightness distribution characteristics of the metal surface area in the sea breeze compensation and correction image, reflective areas are identified and intensity attenuation is performed on high-brightness pixels. The attenuated pixel values ​​are then smoothly fused with the surrounding normal areas to obtain a reflection suppression processed image. The image with the reflection suppression processing is subjected to global contrast adjustment, and the adjusted image is then subjected to noise filtering and edge sharpening processing to obtain standardized image data.

3. The port loading and unloading risk identification method based on image processing according to claim 1, characterized in that, Based on the target identification results of the crane boom, suspended cargo, and operators in the standardized image data, multi-dimensional risk features are extracted from the boom swing angle, cargo center of gravity shift, and personnel violations to obtain composite risk feature data for loading and unloading operations, including: Based on the contour features in the standardized image data, target classification and recognition are performed. The steel structure contour of the crane boom, the geometric shape of the suspended goods, and the human body contour of the workers are segmented and labeled to obtain the classification target recognition results. Based on the spatial position coordinates of the crane boom in the classification target recognition results, the angle between the boom and the vertical baseline is calculated geometrically. The angle change values ​​between consecutive frames are statistically analyzed for swing amplitude and swing frequency to obtain the boom swing angle parameters. Based on the trajectory of the change of the center of gravity of the suspended cargo in the classification target identification results, the offset distance of the cargo's center of gravity relative to the lifting point is measured and calculated in real time. The offset direction and offset speed are processed by vector analysis to obtain the cargo's center of gravity offset parameters. The spatial relationship between the location coordinates of the workers in the classification target identification results and the boundary of the preset safety area is determined. The violation type and duration of the personnel entering the dangerous area are identified and statistically analyzed. The crane swing angle parameter, cargo center of gravity offset parameter and violation behavior statistical results are fused and processed to obtain composite risk characteristic data of loading and unloading operations.

4. The port loading and unloading risk identification method based on image processing according to claim 1, characterized in that, The process involves predicting collision trajectories and detecting violations by comparing the composite risk characteristic data of loading and unloading operations with the boundary of the port's safe operating area. The prediction results are then used to determine the hazard level through a dual assessment of boundary crossing and personnel intrusion, resulting in a graded composite risk assessment, including: Based on the boom swing angle parameters and cargo center of gravity offset parameters in the composite risk characteristic data of loading and unloading operations, physical dynamics modeling and calculation of the motion trajectory of the boom and cargo are performed. The calculated trajectory coordinates are spatially intersected with the boundary of the port safety operation area to obtain the collision trajectory prediction results. Based on the statistical results of violations in the composite risk characteristic data of loading and unloading operations, a time series analysis is performed on the behavior patterns of workers entering dangerous areas. The personnel location coordinates are compared and detected in real time with the boundaries of the preset prohibited areas to obtain the violation detection results. The risk level of swinging out of bounds in the collision trajectory prediction result and the risk level of personnel intrusion in the violation detection result are weighted and fused together. The fused risk value is then divided into threshold ranges and graded to obtain a dual-assessment hazard level. Based on the numerical distribution characteristics of the dual-assessment risk levels, different risk types are prioritized and classified, and the ranking results are hierarchically classified and stored according to the degree of risk urgency to obtain the graded composite risk assessment results.

5. The port loading and unloading risk identification method based on image processing according to claim 4, characterized in that, Based on the boom swing angle parameters and cargo center of gravity offset parameters in the composite risk characteristic data of the loading and unloading operation, the motion trajectory of the boom and cargo is modeled and calculated using physical dynamics. The calculated trajectory coordinates are then spatially intersected with the boundary of the port safety operation area to obtain collision trajectory prediction results, including: Based on the swing amplitude and swing frequency values ​​in the swing angle parameters of the boom, the periodic motion of the boom end is modeled and calculated using a sine function. The boom length and swing angle are input into the kinematic model to solve for the trajectory coordinates, and the boom motion trajectory coordinate sequence is obtained. Based on the offset distance and offset velocity vector in the cargo center of gravity offset parameters, damped vibration modeling calculation is performed on the inertial swing of the suspended cargo. The cargo mass and wind load coefficient are input into the dynamic model to predict the displacement and obtain the cargo swing trajectory coordinate sequence. The coordinate system of the boom motion trajectory coordinate sequence and the cargo swing trajectory coordinate sequence is transformed into a single coordinate system. The intersection of the unified trajectory coordinates with the geometric boundary line of the port safety operation area is calculated to obtain the set of trajectory boundary intersection point coordinates. Based on the spatial distribution characteristics of the coordinate set of the intersection points of the trajectory boundaries, the collision time and collision intensity at the intersection points are quantified and calculated. The quantification results are then classified and labeled according to the degree of collision danger to obtain the collision trajectory prediction results.

6. The port loading and unloading risk identification method based on image processing according to claim 1, characterized in that, The method employs an adaptive region growing algorithm based on spatiotemporal correlation to dynamically track the risk propagation range of the hierarchical composite risk assessment results. It combines the superimposed effects of swing collision risk and personnel injury risk to identify comprehensive safety hazards in loading and unloading operations, thereby obtaining port loading and unloading composite risk early warning information, including: An adaptive region growth algorithm based on spatiotemporal correlation is used to select seed points for high-risk regions in the hierarchical composite risk assessment results. Pixels with risk level values ​​exceeding the safety threshold are set as initial growth seeds. The spatiotemporal neighborhood correlation of the seed points is calculated and analyzed to obtain a set of risk propagation seed points. Based on the spatial distribution characteristics of the risk propagation seed point set, a time-series change analysis of the risk correlation between adjacent seed points is performed, and the region with a correlation exceeding the propagation threshold is adaptively expanded and grown to obtain the risk propagation diffusion region. The oscillation collision risk value and the personnel injury risk value in the risk propagation and diffusion area are superimposed to calculate the effect, and the spatial weighted fusion processing of the superimposed comprehensive risk intensity is performed to obtain a comprehensive safety hazard distribution map. Based on the risk density distribution characteristics of the comprehensive safety hazard distribution map, the early warning level of high-density risk areas is marked and the time urgency is assessed. The assessment results are classified, organized and stored according to the early warning priority to obtain port loading and unloading composite risk early warning information.

7. The port loading and unloading risk identification method based on image processing according to claim 6, characterized in that, The calculation of the superposition effect of the swing collision risk value and the personnel injury risk value in the risk propagation and diffusion area, and the spatial weighted fusion processing of the superimposed comprehensive risk intensity, yields a comprehensive safety hazard distribution map, including: Based on the spatial coordinate distribution of the risk propagation and diffusion area, the hazard intensity of the swing collision risk value in the area is quantitatively calculated. The collision speed, collision area and equipment mass parameters are numerically processed to obtain the quantified swing collision risk value. Based on the population distribution density in the risk transmission and diffusion area, the exposure risk of personnel injury in the area is assessed and calculated. The number of personnel, exposure time, and protection level parameters are quantified to obtain a quantitative value of personnel injury risk. The quantified values ​​of swing collision risk and personnel injury risk are nonlinearly superimposed and calculated. The superimposed composite risk value is then weighted according to spatial distance weight to obtain a spatially weighted comprehensive risk intensity value. Based on the distribution characteristics of the spatially weighted comprehensive risk intensity values, color coding and contour line drawing are performed on different intensity intervals. The processed risk intensity data is then spatially mapped and stored according to geographical coordinates to obtain a comprehensive safety hazard distribution map.

8. A port loading and unloading risk identification system based on image processing, characterized in that, For implementing the port loading and unloading risk identification method based on image processing as described in any one of claims 1-7, the port loading and unloading risk identification system based on image processing comprises: The correction module is used to acquire the original image sequence of the port loading and unloading operation site through multispectral image acquisition, and to perform composite environmental correction on the original image sequence using sea breeze disturbance compensation and metal surface reflection suppression processing to obtain standardized image data. The extraction module is used to extract multi-dimensional risk features of crane boom swing angle, cargo center of gravity shift and personnel violations based on the target recognition results of crane boom, suspended cargo and operators in the standardized image data, so as to obtain composite risk feature data of loading and unloading operations. The detection module is used to predict the collision trajectory and detect violations by comparing the composite risk characteristic data of the loading and unloading operation with the boundary of the port safety operation area. The prediction results are used to determine the danger level through dual assessment of swinging over the boundary and personnel intrusion, so as to obtain the graded composite risk assessment results. The tracking module is used to dynamically track the risk propagation range of the hierarchical composite risk assessment results using an adaptive region growth algorithm based on spatiotemporal correlation. It combines the superposition effect of swing collision risk and personnel injury risk to identify comprehensive safety hazards in loading and unloading operations and obtain port loading and unloading composite risk early warning information.

9. A port loading and unloading risk identification device based on image processing, characterized in that, The method includes a memory and a processor, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the port loading and unloading risk identification method based on image processing as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to perform the port loading and unloading risk identification method based on image processing as described in any one of claims 1 to 7.

Images