Development and Application Methods and Systems of Intelligent Empty Container Inspection Robot System for Ports

By using quadruped robots for autonomous inspection and image processing technology, the problem of missing empty container inspections at ports due to manual inspection has been solved, and highly accurate risk assessment and identification of disease vectors has been achieved.

CN123113522APending Publication Date: 2026-11-17ZHEJIANG LANCHENG EMERGENCY INFORMATION TECHNOLOGY CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611287223.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Current technology relies on manual inspection of empty containers at ports, which leads to staff fatigue, makes it easy for inspections to be missed, and poses a safety hazard of cross-border transmission of pathogens.

Method used

A quadruped robot is used to autonomously inspect empty containers. Through image acquisition, credibility assessment and target selection, the terminal completes risk data segmentation and analysis, generates vector risk warnings, and implements airflow injection and image correction when the target credibility is insufficient, and updates the selected data.

Benefits of technology

It has improved the accuracy of empty container inspection at ports, reduced omissions in identification, and enhanced the accuracy of identifying disease vector debris and assessing the risk of pathogen transmission.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113522A_ABST
    Figure CN123113522A_ABST
Patent Text Reader

Abstract

This invention relates to the development and application method and system of a port intelligent empty container inspection robot system, specifically in the field of port empty container inspection. The system includes: collecting multi-source data; determining the port route of the quadruped robot within the port based on the multi-source data and generating movement instructions along the route; retrieving the real-time positioning of the quadruped robot from the multi-source data according to the movement instructions; generating and sending an empty container scanning instruction when the real-time positioning falls within the positioning range; collecting image information and identifying suspicious targets based on the image information; determining the target credibility based on the suspicious targets; when the target credibility is greater than a preset credibility threshold, determining bounding box data based on the suspicious targets and image information; determining transmission data based on the bounding box data and generating a data transmission instruction; and receiving and analyzing the transmission data at a terminal to determine risk data and generate a vector risk warning. This invention achieves accurate inspection of empty containers in ports.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of port empty container inspection, and in particular to the development and application methods and systems of intelligent port empty container inspection robot systems. Background Technology

[0002] Port empty container inspection is a specialized quarantine and inspection technique for empty containers arriving at the port. It is used to screen for pathogens inside and outside the container in order to prevent the cross-border spread of pathogens through empty containers.

[0003] Current technology mainly relies on manual on-site operations, where staff enter empty containers and use flashlights to visually inspect various areas and details inside the container in order to identify and screen for various disease-carrying organisms hidden inside, thus completing the quarantine and inspection work.

[0004] As the port's container throughput continues to climb, long-term, high-intensity manual inspection work can easily lead to staff fatigue, resulting in oversights and potentially causing cross-border transmission and spread of pathogens, posing a significant safety hazard. Summary of the Invention

[0005] To accurately inspect empty containers at ports, this invention provides a method and system for developing and applying an intelligent empty container inspection robot system for ports.

[0006] In a first aspect, the present invention provides a method for developing and applying a port intelligent empty container inspection robot system, which adopts the following technical solution: A method for developing and applying a port intelligent empty container inspection robot system, comprising: Step 100: Collect data from multiple sources; Step 101: Determine the port route of the quadruped robot in the port based on the multi-source data and the preset positioning interval, and send movement instructions along the route in response to the generation of the port route. Step 102: Retrieve the real-time positioning of the quadruped robot from the multi-source data according to the motion instructions along the line; Step 103: When the real-time positioning falls within the positioning range, respond to the real-time positioning by generating and sending an empty box scanning command; Step 104: Acquire image information according to the empty box scanning command, and determine suspicious targets based on the image information; Step 105: Determine the credibility of the target based on the suspected target; Step 106: When the credibility of the target is greater than the preset credibility threshold, determine the bounding box data based on the suspicious target and image information; Step 107: Determine the transmission data based on the selected data, and generate a data transmission instruction according to the transmission data; Step 108: The terminal receives the transmitted data and determines the segmented data based on the transmitted data; Step 109: Determine risk data according to the segmented data, and generate vector risk warnings based on the risk data.

[0007] By adopting the above technical solution, a quadruped robot is used to autonomously inspect empty containers in the port. When it reaches the container, it triggers a scanning command and transmits the collected images to the terminal after credibility assessment and suspicious target selection. The terminal then completes the segmentation and analysis of risk data and finally automatically generates a vector risk warning, thereby improving the accuracy of empty container inspection in the port.

[0008] Optionally, the method for determining the selected data further includes: Step 110: When the target's credibility is lower than a preset credibility threshold, determine the jet coordinates based on the suspected target and image information; Step 111: Generate an airflow injection command according to the jet coordinates, and acquire a post-jet image in response to the airflow injection command; Step 112: Determine the target displacement based on the post-spray image, image information, and suspected target; Step 113: When the target displacement is greater than a preset displacement threshold, determine the fragment target based on the target displacement, and determine the adhesion index by combining the target displacement and the fragment target; Step 114: Determine the corrected targets according to the adhesion index and the suspected targets, and update the selected data according to the corrected targets.

[0009] By adopting the above technical solution, when the credibility of the target is insufficient, airflow is sprayed towards the suspicious location and images are collected. The displacement of the suspicious target and the debris adhesion index are used to correct the original suspicious target, thereby updating the bounding box data, reducing identification omissions, and improving the accuracy of port empty container inspection data.

[0010] Optionally, the method for correcting the vector risk warning further includes: Step 200: Determine the fragment coordinates based on the vector risk warning, and determine the air blowing coordinates according to the fragment coordinates; Step 201: Determine the surface area of ​​the debris based on the debris coordinates and vector risk warning, and determine the blowing force based on the debris surface area; Step 202: Determine the fragment detection command by combining the blowing coordinates and blowing force; Step 203: In response to the fragment detection command, determine the blown image of the fragment, and determine the cross-sectional gloss of the fragment based on the blown image and image information; Step 204: Determine the freshness according to the cross-sectional gloss, and determine the degree of danger based on the freshness; Step 205: Update the vector risk warning in response to the stated level of danger.

[0011] By adopting the above technical solution, the fresher the disease vector debris, the greater the possibility of spreading pathogens. At this time, a directional blowing operation is carried out on the identified debris, and the freshness of the disease vector residue is inferred by analyzing the cross-sectional gloss of the debris. The hazard level is adjusted to reflect the hazard level of the disease vector debris, thereby improving the accuracy of port empty container inspection.

[0012] Optionally, the method for determining the freshness includes: Step 206: In response to the blowing force and the preset monitoring area, determine the real-time movement image of the fragment under the action of continuous airflow; Step 207: Determine the real-time movement trajectory of the fragment based on the moving image; Step 208: Determine the response characteristics according to the real-time movement trajectory; Step 209: Retrieve the fragment material from the vector risk warning, and determine the risk factor based on the fragment material and response characteristics; Step 210: Update the freshness based on the risk factor.

[0013] By adopting the above technical solution, the blowing may cause the fragments of disease vectors to move continuously. At this time, real-time images of the fragments are acquired and the response characteristics of the fragment movement are determined. The risk factor is determined in combination with the fragment material, thereby correcting the freshness of the fragments and improving the accuracy of port empty container inspection.

[0014] Optionally, the method for determining the freshness further includes: Step 211: When the freshness is greater than the preset freshness threshold, determine the movement speed according to the real-time movement trajectory, and retrieve the fragment volume from the vector risk warning; Step 212: Determine the mass of the fragment according to its material and volume; Step 213: Determine the kinetic energy of motion based on the motion speed and fragment mass, and determine the energy density during vibration according to the kinetic energy of motion and fragment volume; Step 214: Determine the vibration frequency by combining the energy density and the fragment mass, and determine the volume-frequency ratio based on the vibration frequency and the fragment volume; Step 215: Determine the frequency coefficient based on the volume-frequency ratio, and update the freshness according to the frequency coefficient.

[0015] By adopting the above technical solution, the vibration frequency of disease vector debris with different freshness levels under the action of wind is different. At this time, the motion speed, kinetic energy and vibration frequency are determined from the swing trajectory of the debris under the action of airflow, so as to calculate the volume frequency ratio of the debris, thereby reflecting the change of the internal structure of the debris with freshness, and thus correcting the freshness of the debris, improving the accuracy of port empty container inspection.

[0016] Optionally, the method for determining the freshness further includes: Step 300: When the freshness is greater than a preset freshness threshold, acquire multi-light source images according to the freshness, and determine the scattering parameters of the fragment surface based on the multi-light source images; Step 301: Determine the surface roughness of the fragment according to the scattering parameters; Step 302: Determine the theoretical force by combining the roughness and blowing force, and determine the theoretical speed based on the vector risk warning and the theoretical force; Step 303: Determine the speed deviation by combining the theoretical speed and the actual speed; Step 304: When the speed deviation is greater than a preset deviation threshold, determine the detachment mass based on the speed deviation; Step 305: Determine the dust correction coefficient according to the detachment mass and fragment surface area, and update the freshness based on the dust correction coefficient.

[0017] By adopting the above technical solution, the longer the fragment has been present, the less fresh the fragment is. Therefore, by collecting multi-source images to determine the scattering parameters and roughness of the fragment surface, and comparing them with the theoretical movement speed to obtain the speed deviation, the influence of dust shedding can be determined. Then, the freshness can be updated using the dust correction coefficient, thereby improving the accuracy of empty container inspection at the port.

[0018] Optionally, the method for determining the freshness further includes: Step 306: When the dust correction coefficient is less than the preset correction threshold, determine the pre-blowing image in response to the fragment detection command; Step 307: Determine the region image of the scattering area based on the post-blowing image; Step 308: Determine the dust dispersion concentration by combining the area image and the pre-blowing image, and determine the deposition thickness based on the dispersion concentration; Step 309: Determine the sealing coefficient based on the deposition thickness; Step 310: Determine the real-time speed according to the real-time movement trajectory, and determine the rate of change of speed based on the real-time speed; Step 311: When the rate of change of speed is greater than a preset threshold, determine the detachment time based on the rate of change of speed, and determine the degree of sealing based on the detachment time; Step 312: Determine the enhancement coefficient by combining the sealing coefficient and the sealing degree, and update the freshness according to the enhancement coefficient.

[0019] By adopting the above technical solution, blowing air will cause dust to be stirred up. At this time, by comparing the images before and after blowing air, the dust dispersion concentration can be obtained to determine the deposition thickness. Then, combined with the dust shedding time, the degree of sealing can be determined, thereby updating the freshness and improving the accuracy of port empty container inspection.

[0020] Optionally, the method for determining the freshness further includes: Step 313: Determine the scattering boundary according to the region image, and determine the diffusion gradient per unit distance based on the scattering boundary; Step 314: Determine the dust settling rate by combining the diffusion gradient and dispersion concentration; Step 315: Determine the suspension time based on the dust settling rate; Step 316: When the suspension time exceeds a preset time threshold, determine the particle size of the dust in response to the suspension time; Step 317: Determine the shielding capability of the redeposited dust based on the deposition thickness and particle size; Step 318: Determine the descent coefficient based on the shielding capability, and update the freshness according to the descent coefficient.

[0021] By adopting the above technical solution, dust can shield the spread of pathogens, resulting in a higher risk of pathogen transmission from stale pathogen debris covered with dust. In this case, the suspension time of dust is determined based on the image of dust in the air, and the shielding ability after deposition is determined in combination with the deposition thickness assessment, so as to update the freshness and improve the accuracy of port empty container inspection.

[0022] Optionally, the method for updating the vector risk warning further includes: Step 319: When the decrease coefficient is less than a preset coefficient threshold, determine the dispersion concentration curve based on the dispersion concentration; Step 320: Determine the dust mass curve in the air according to the dispersion concentration curve; Step 321: Determine the wind force coefficient according to the blowing force, and determine the vibration coefficient according to the vibration frequency; Step 322: Determine the coverage quality by combining the vibration coefficient, wind force coefficient, and dust mass; Step 323: Determine the density curve based on the coverage quality and fragment surface area; Step 324: Determine the dangerous period according to the density curve and the preset density threshold, and update the vector risk warning according to the dangerous period.

[0023] By adopting the above technical solution, the wind force coefficient and vibration coefficient are determined based on the wind force during air blowing and the vibration frequency of the disease vector debris under forced vibration. This determines the mass of dust covering the debris before air blowing, and then determines the density curve to update the disease vector risk warning, thereby improving the accuracy of port empty container inspection.

[0024] Secondly, this invention provides a development and application system for a port intelligent empty container inspection robot system, which adopts the following technical solution: A port intelligent empty container inspection robot system development and application system, comprising: The data acquisition module is used to collect data from multiple sources. A memory for storing the program of the development and application method of the port intelligent empty container inspection robot system as described in any one of the claims; The processor is the unit of memory that allows programs to be loaded and executed by the processor.

[0025] By adopting the above technical solution, a quadruped robot is used to autonomously inspect empty containers in the port. When it reaches the container, it triggers a scanning command and transmits the collected images to the terminal after credibility assessment and suspicious target selection. The terminal then completes the segmentation and analysis of risk data and finally automatically generates a vector risk warning, thereby improving the accuracy of empty container inspection in the port.

[0026] In summary, the present invention has at least one of the following beneficial technical effects: A quadruped robot is used to autonomously inspect empty containers in the port. When it reaches a container, it triggers a scanning command and transmits the collected images to the terminal after credibility assessment and suspicious target selection. The terminal then completes the segmentation and analysis of risk data and finally automatically generates a vector risk warning, which improves the accuracy of empty container inspection in the port. The fresher the disease vector debris, the greater the possibility of it spreading pathogens. In this case, a directional blowing operation is carried out on the identified debris, and the freshness of the disease vector residue is inferred by analyzing the cross-sectional gloss of the debris. The hazard level is then adjusted to reflect the hazard level of the disease vector debris, thus improving the accuracy of port empty container inspection. The longer the fragment has been present, the less fresh it is. Therefore, by collecting multi-source images to determine the scattering parameters and roughness of the fragment surface, and comparing them with the theoretical motion speed to obtain the velocity deviation, the influence of dust shedding can be determined. Then, the freshness can be updated using a dust correction coefficient, thereby improving the accuracy of empty container inspection at the port. Attached Figure Description

[0027] Figure 1 This is a schematic diagram illustrating a scenario for the development and application of a port intelligent empty container inspection robot system. Figure 2This is a flowchart illustrating the development and application method of a port intelligent empty container inspection robot system; Figure 3 This is a flowchart of the method for revising vector risk warnings; Figure 4 This is a flowchart of the method for determining freshness. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0029] This invention discloses a method for developing and applying a port intelligent empty container inspection robot system.

[0030] Reference Figure 1 and Figure 2 The development and application methods of intelligent empty container inspection robot systems in ports include: Step 100: Collect multi-source data.

[0031] Multi-source data refers to the real-time data collection from various sensors mounted on the quadruped robot, including point cloud data collected by LiDAR, pose data collected by IMU inertial sensors, positioning data collected by GNSS / RTK modules, and visual image data collected by high-definition cameras. In this embodiment, the point cloud data is collected at a frequency of 10Hz, and the positioning data is updated at a frequency of 1Hz.

[0032] Step 101: Determine the port route of the quadruped robot in the port based on the multi-source data and the preset positioning range, and send movement instructions along the route in response to the generation of the port route.

[0033] The positioning range refers to the location of the container that each quadruped robot needs to inspect. The positioning range can be pre-entered into the system by the staff.

[0034] The port route refers to the route taken by the robot from its current location to the container that needs to be inspected. It can integrate multi-source localization and perception information such as GNSS / RTK, LiDAR, and IMU, and combine SLAM mapping and localization, Kalman filter fusion, attitude calculation, obstacle detection, and path tracking control to determine the port route. In this embodiment, extended Kalman filter is used to tightly couple and fuse LiDAR, IMU, and RTK data to construct a grid occupancy map, and static obstacles such as container yard boundaries and container door orientations are semantically labeled. Then, an improved hybrid A* algorithm is used for path search, which introduces a risk heat map based on historical inspection data into the cost function to avoid high-risk areas. Finally, the path is discretized into a sequence of path points containing coordinates, heading angle, and velocity as the port route.

[0035] The motion commands along the route refer to the set of motion control commands that control the quadruped robot to move along the port route and avoid obstacles. These commands include the travel speed, turning angle, and stopping trigger conditions. In this embodiment, the robot's travel speed is 1.0 m / s and the safe obstacle avoidance distance is 0.5 m.

[0036] Step 102: Retrieve the real-time positioning of the quadruped robot from the multi-source data according to the motion instructions along the line.

[0037] Real-time positioning refers to the process by which a quadruped robot, while executing movement commands along the route, collects its current coordinates and orientation in the port coordinate system in real time using its onboard GNSS / RTK module. The latest positioning data can be read from multiple sources as real-time positioning data.

[0038] Step 103: When the real-time positioning falls within the positioning range, an empty box scanning command is sent in response to the real-time positioning.

[0039] The empty container scanning command is a command that triggers the quadruped robot to enter the interior of an empty container and start the inspection program. When the real-time positioning falls within the positioning range, the quadruped robot arrives at the container to be inspected. At this time, the empty container scanning command is generated, which includes the quadruped robot's walking and shooting parameters inside the container.

[0040] Step 104: Collect image information according to the empty box scanning command, and determine suspicious targets based on the image information.

[0041] Image information refers to the visual data captured by the quadruped robot after it enters the empty box. After entering the empty box, the quadruped robot walks along a preset route and stops automatically when it reaches the designated position. At each position, a wide-angle camera first captures the overall picture inside the box, and then an adjustable focus gimbal camera zooms in and captures key areas such as floor gaps, box corners, and door seals one by one in the order from far to near, from top to bottom, and from left to right to obtain image information. When collecting image information, a ring light is used for supplementary lighting. The ring light consists of multiple independently controllable LED beads.

[0042] Suspicious targets refer to areas in image information that, as identified by the target detection algorithm, may pose a risk of disease vector transmission. In this embodiment, the target detection algorithm uses an improved small target detection network as its base model. The overall structure is based on a single-stage detection framework, such as the YOLOv10 structure. Multi-scale feature fusion and high-resolution feature channels are added to this framework to focus on magnifying and enhancing small-area targets. By adjusting the loss function, higher weights are assigned to targets with smaller areas and lower contrast, allowing the model to pay more attention to scattered suspected targets in areas such as floors and corners during training, thus identifying suspicious targets. In this embodiment, the detection network uses an improved YOLOv10-nan network. The architecture incorporates an additional detection head with an input of 160*160*64 and an output of 160*160*128, following the second C2f module in the Neck section. This head is specifically designed for small object detection. Furthermore, the output channel count of the third Conv module in the original Backbone is increased from 256 to 512 to enhance the perception of fine textures. During the training phase, 13,000 images of empty containers at the port were collected, of which 8,000 contained disease vectors or their fragments. These images were cross-labeled by two quarantine personnel. Training employed the Adam optimizer with an initial learning rate of 0.001, which decayed to 0.1 times the original rate every 30 epochs. The batch size was 16, and training lasted for 200 epochs. The loss function consists of three parts: classification loss, localization loss, and confidence loss. The localization loss assigns a double penalty weight to small objects.

[0043] Step 105: Determine the credibility of the target based on the suspected target.

[0044] Target credibility refers to a quantitative indicator for assessing the authenticity of a suspicious target. The target image corresponding to the suspicious target in the image information can be input into the verification model to obtain the target credibility. The target image refers to the image cropped from the image information with the minimum bounding rectangle of the suspicious target as the size. In this embodiment, an image dataset containing positive and negative samples can be pre-constructed. The positive samples are patches of various disease vectors and their traces in a real empty box environment, and the negative samples are pseudo-targets with similar morphology. Then, a lightweight convolutional neural network is selected as the backbone, the original classification head is removed and replaced with a fully connected layer with a one-dimensional scalar output and a sigmoid activation function. Finally, supervised training is performed with a binary cross-entropy loss function so that the model learns to directly output a value between 0 and 1 for the input suspicious target patch. This value represents the target credibility.

[0045] Step 106: When the credibility of the target is greater than the preset credibility threshold, determine the bounding box data based on the suspicious target and image information.

[0046] The credibility threshold is a critical value for judging whether a suspicious target is credible. When the credibility of a target is greater than the preset credibility threshold, it means that the suspicious target is a disease vector. At this time, the relevant image data needs to be transmitted to the terminal for further verification. In this embodiment, the credibility threshold is 0.7.

[0047] The frame selection data refers to the set of local image data extracted from the image information surrounding the suspicious target, which is the target image mentioned above.

[0048] Step 107: Determine the transmission data based on the selected data, and generate a data transmission instruction according to the transmission data.

[0049] Transmitted data refers to the data set sent from the quadruped robot to the terminal. Transmitted data includes frame selection data, real-time positioning, and acquisition time. The transmitted data is compressed and encoded before being encapsulated into transmission frames.

[0050] Data transmission instructions are instructions that control the robot to send data to the terminal server, used to upload the actually collected data to the terminal.

[0051] Step 108: The terminal receives the transmitted data and determines the segmented data based on the transmitted data.

[0052] The terminal refers to the data processing server located in the port monitoring center. After receiving the transmitted data, the terminal decompresses it to obtain the image data to be segmented.

[0053] Segmentation data refers to image data that separates suspicious target regions from the background in image information. The image data to be segmented can be used as prompts for the segmentation network. Based on these prompts, the segmentation network automatically generates a set of pixel-level masks in the corresponding regions. At the same time, combined with multi-scale feature extraction, the shape, texture and color distribution of the target are statistically analyzed and encoded to form a low-dimensional but complete feature vector as segmentation data. In this embodiment, the segmentation network is the DeepLabV3+ network.

[0054] Step 109: Determine risk data according to the segmented data, and generate vector risk warnings based on the risk data.

[0055] Risk data refers to risk assessment data obtained after the terminal analyzes segmented data. Category names and textual descriptions such as flying vectors, crawling vectors, insect eggs / egg masses, vector debris, rust spots, stains, and cracks can be input into a semantic coding network to obtain a stable set of category semantic vectors. On the other hand, segmented data can be input into the same feature space for encoding to obtain target semantic vectors. By calculating the similarity between the target vector and each category vector, the system can complete the category classification judgment within a unified semantic space. On this basis, quantity, location, and clustering information are superimposed to form a box-level risk assessment result, which serves as risk data.

[0056] Vector risk warning refers to a warning that reminds staff of the risk of pathogen spread inside the container. The vector risk warning includes the type of vector organism, size of the vector organism, volume of the vector organism, coordinates of the vector organism, material of any remaining vector organism debris, and the corresponding location range of the container. The vector risk warning is then sent to the staff's handheld terminal.

[0057] In this embodiment, for the same container to be inspected, the terminal will dispatch both robots to the location of the container, and the two robots will independently enter the container to complete a complete image acquisition and candidate target detection process. The two robots will use a unified coordinate system for spatial positioning and path coordination to ensure that they do not interfere with each other during dispatching and movement.

[0058] The terminal receives images and metadata from two robots, and sequentially initiates procedures such as candidate target verification, fine segmentation, semantic recognition, and risk assessment to generate structured inspection results. After the inspection of each container is completed, the terminal system writes the results to the database and interfaces with the port's existing business system, supporting intuitive viewing on the inspection terminal or management platform. The terminal is also responsible for the unified management of the versions of various algorithm programs and knowledge bases, regularly optimizing them based on operational feedback, and distributing the upgraded programs to the edge through the management platform to achieve online updates and rollbacks of models and rules.

[0059] In terms of key technology implementation, the cloud-edge collaborative process adopts a unified data format and communication protocol, standardizing and encapsulating information such as images, candidate regions, and recognition results to ensure interoperability between different devices and systems. The edge terminal is equipped with local caching and breakpoint resume mechanisms. When the on-site network quality is unstable, the detection results can be temporarily stored locally and uploaded in batches after the network recovers, ensuring no data loss. The terminal uses a task queue and scheduling module to queue and process tasks from multiple robots in parallel, avoiding performance fluctuations caused by concentrated tasks in a single period. Simultaneously, the system is designed with operational status monitoring and alarm functions at both the terminal and edge terminals, monitoring the online status of devices, task execution status, and key indicators such as latency and failure rate in real time. In case of an anomaly, it can automatically retry or switch to a backup process, ensuring the stability and reliability of the overall operation chain.

[0060] A quadruped robot is used to autonomously inspect empty containers in the port. When it reaches a container, it triggers a scanning command and transmits the collected images to the terminal after credibility assessment and suspicious target selection. The terminal then completes the segmentation and analysis of risk data and automatically generates a vector risk warning, which improves the accuracy of empty container inspection in the port.

[0061] Methods for determining the selected data also include: Step 110: When the target's credibility is lower than a preset credibility threshold, determine the jet coordinates based on the suspected target and image information.

[0062] The jet coordinates refer to the spatial position of the airflow jetting operation. The center point of the box selection data corresponding to the suspected target can be mapped to the three-dimensional coordinates under the port coordinate system through the spatial coordinate transformation algorithm. In this embodiment, the port coordinate system is a spatial rectangular coordinate system with the port area preset benchmark point as the origin, the direction along the wharf shoreline as the X-axis, the direction perpendicular to the shoreline towards the sea as the Y-axis, and the direction vertically upward based on the local normal height system as the Z-axis. The coordinate unit is meters.

[0063] Step 111: Generate an airflow injection command according to the jet coordinates, and acquire a post-jet image in response to the airflow injection command.

[0064] The airflow injection command refers to the command that controls the jet module on the quadruped robot to inject air at the jet coordinate. The command includes the jet coordinate, the jet pressure and the jet duration. In this embodiment, the jet pressure is 5 kPa and the jet duration is 1 second.

[0065] Post-spray images refer to visual image data obtained by a high-definition camera re-capturing the suspicious target after the airflow spray command has been executed.

[0066] Step 112: Determine the target displacement based on the post-spray image, image information, and suspected target.

[0067] Target displacement refers to the change in position of a suspected target before and after the action of airflow. The original suspected target in the image information is used as the reference, and the area corresponding to the suspected target in the post-spray image is used as the comparison image. The pixel-level offset field between the two is calculated using the optical flow method. The average displacement of the top 10% of pixels with the largest offset is taken as the initial pixel displacement. Then, the pixel displacement is converted into the actual physical displacement in the port coordinate system, which is the target displacement.

[0068] Step 113: When the target displacement is greater than the preset displacement threshold, determine the fragment target based on the target displacement, and determine the adhesion index by combining the target displacement and the fragment target.

[0069] The displacement threshold refers to the critical distance at which a suspicious target can be blown away by airflow. When the target displacement is greater than the displacement threshold, it means that the suspicious target is a fragment that is not firmly attached. In this embodiment, the displacement threshold is 2mm.

[0070] Fragment targets refer to fragments that are re-identified from post-spray images after displacement, corresponding to the original suspicious targets; these are the suspicious targets that have moved.

[0071] The adhesion index is an indicator for assessing the firmness of the adhesion of debris to the surface of the container. The larger the target displacement, the easier the debris is to move, and the weaker the adhesion to the container surface, resulting in a smaller adhesion index. The adhesion index corresponding to the target displacement can be found in the adhesion index correspondence table. The adhesion index correspondence table is a data table that records different target displacements and their corresponding adhesion indices. Six different materials of container surface test plates can be selected, and 50 samples of disease vector debris with known adhesion strength can be placed on the test plates. The adhesion strength can be measured by a tensile tester. A jet test is conducted under the conditions of jet pressure of 5 kPa and nozzle distance of 10 cm. The target displacement of each sample is recorded, and the maximum adhesion strength is normalized to 1 and zero adhesion strength is normalized to 0. Then, based on the negative correlation scatter trend between displacement and adhesion strength, an exponential decay or piecewise linear model is used to fit the mapping formula. Then, the continuous displacement range is discretized into several levels and substituted into the formula to calculate the corresponding adhesion index, forming a reference table that can be directly looked up. Finally, in actual operation, the fitting coefficient is continuously corrected by manual verification or sensor feedback, so that the table dynamically adapts to different container materials and pollution conditions.

[0072] In this embodiment, the tables corresponding to various parameters such as air blowing force, freshness, danger level, response characteristics, danger coefficient, density, frequency, frequency coefficient, roughness, theoretical stress, shedding mass, dust correction coefficient, dispersion concentration, deposition thickness, sealing coefficient, shedding time, sealing degree, dust settling rate, particle size, shielding capability, descent coefficient, particle density, wind force coefficient, and vibration coefficient can all be initially obtained by selecting typical samples and testing them under standard conditions to fit an initial mapping formula, forming an initial reference table that can be directly queried. Then, in actual operation, data is continuously collected and the fitting coefficient is dynamically corrected through manual verification or sensor feedback, so that the table can adapt to complex environments such as different box materials and pollution conditions.

[0073] Step 114: Determine the corrected targets according to the adhesion index and the suspected targets, and update the selected data according to the corrected targets.

[0074] The corrected target refers to the target whose credibility has been corrected by combining the adhesion index. The smaller the adhesion index, the easier the target is to be blown away, which is more in line with the characteristics of light weight and weak adhesion of disease vector biological debris. The quotient of the original target credibility and the adhesion index can be used as the corrected credibility. When the corrected credibility is greater than the credibility threshold, the box selection data is obtained according to the corresponding suspicious target.

[0075] When the credibility of the target is insufficient, airflow is sprayed towards the suspicious location and images are collected. The displacement of the suspicious target and the debris adhesion index are used to correct the original suspicious target, thereby updating the bounding box data, reducing identification omissions, and improving the accuracy of port empty container inspection data.

[0076] Reference Figure 3 Methods for revising vector risk warnings also include: Step 200: Determine the fragment coordinates based on the vector risk warning, and determine the air blowing coordinates according to the fragment coordinates.

[0077] The debris coordinates refer to the three-dimensional coordinates of the disease vector debris after it has been blown by the wind in the port coordinate system. The disease vector coordinates can be read from the disease vector risk warning as the debris coordinates.

[0078] The blowing coordinates refer to the position where the directional blowing operation is performed, which is also the coordinates of the fragment.

[0079] Step 201: Determine the surface area of ​​the fragment based on the fragment coordinates and the vector risk warning, and determine the blowing force based on the fragment surface area.

[0080] The surface area of ​​the debris refers to the surface area of ​​the disease vector debris. It can be obtained by reading the size of the disease vector corresponding to the coordinates of the debris from the disease vector risk warning and calculating the corresponding area as the surface area of ​​the debris.

[0081] The blowing force refers to the air pressure value used to break up disease vector debris. The larger the surface area of ​​the debris, the greater the blowing force required. The blowing force corresponding to the surface area of ​​the debris can be found in the blowing force correspondence table, which is a data table that records different debris surface areas and their corresponding blowing forces.

[0082] Step 202: Determine the fragment detection command by combining the blowing coordinates and blowing force.

[0083] The fragment detection command refers to the instruction that controls the quadruped robot to blow air onto the identified fragments in order to calculate the freshness of the fragments.

[0084] Step 203: In response to the fragment detection command, determine the blown image of the fragment, and determine the cross-sectional gloss of the fragment based on the blown image and image information.

[0085] Post-blowing images refer to visual images re-acquired after the fragment detection command is executed. These images are captured by a high-definition camera and are used to reflect the fragment's breakage state.

[0086] Cross-sectional gloss refers to the reflective properties of the fresh fracture surface of a fragment after it has been broken by a directional airflow. The cross-sectional gloss is determined by the specular reflection intensity of the fragment's cross-section in the post-blowing image using image recognition technology. In this embodiment, the pixel mask of the fresh fracture cross-section is extracted using the Otsu algorithm combined with highlight and area constraints. Then, the average normalized brightness and the proportion of highlight pixels in the brightness component of the post-blowing image are statistically analyzed within this mask. The weighted sum of the two is the cross-sectional gloss. The weight of the average normalized brightness is 0.6, and the weight of the proportion of highlight pixels is 0.4.

[0087] Step 204: Determine the freshness according to the cross-sectional gloss, and determine the degree of danger based on the freshness.

[0088] Freshness refers to the length of time since the vector organism fragments were separated from the organism or died. When the vector organism species are the same, the greater the cross-sectional gloss, the drier and more fragile the fragments are. The longer the time, the lower the freshness. The freshness corresponding to the cross-sectional gloss and the vector organism species can be found in the freshness correspondence table. The freshness correspondence table is a data table that records different cross-sectional gloss ranges, vector organism species and their corresponding freshness.

[0089] The risk level refers to the revised level of the risk of pathogen transmission from the fragment. The fresher the fragment, the higher the probability of pathogen survival and the risk of transmission. The risk level corresponding to the freshness can be found in the risk level correspondence table. The risk level correspondence table is a data table that records different freshness levels and their corresponding risk level revision weights, and the risk level is added to the vector risk warning.

[0090] Step 205: Update the vector risk warning in response to the stated level of danger.

[0091] The fresher the disease vector debris, the greater the likelihood of it spreading pathogens. In this case, targeted air blowing is performed on the identified debris, and the freshness of the disease vector residue is inferred by analyzing the cross-sectional gloss of the debris. The hazard level is then adjusted to reflect the degree of danger of the disease vector debris, thus improving the accuracy of empty container inspection at the port.

[0092] Methods for determining freshness include: Step 206: In response to the blowing force and the preset monitoring area, determine the real-time movement image of the fragment under the action of continuous airflow.

[0093] Moving images refer to a sequence of images captured continuously at a high frame rate by a high-definition camera under the influence of a continuous airflow. In this embodiment, the shooting frame rate is 120fps.

[0094] The monitoring area refers to a circular area with a radius of 2m centered on the air blowing coordinates, used to define the spatial range for dynamic observation of the debris.

[0095] Step 207: Determine the real-time movement trajectory of the fragment based on the moving image.

[0096] The real-time motion trajectory refers to the path curve of the debris moving in a continuous airflow. The coordinates of the center point of the debris in each frame of the moving image can be extracted by the target tracking algorithm and connected in time sequence to fit a continuous curve as the real-time motion trajectory.

[0097] Step 208: Determine the response characteristics according to the real-time movement trajectory.

[0098] Response characteristics refer to the dynamic behavior of debris under the action of airflow. The movement trajectory of debris can be converted into a quantifiable motion pattern through image tracking algorithms. Response characteristics include suspension, rolling, sliding, jumping, breaking and scattering.

[0099] Step 209: Retrieve the fragment material from the vector risk warning and determine the risk factor based on the fragment material and response characteristics.

[0100] The material of the fragments refers to the material type of the vector-borne organism fragments. The material of the residual vector-borne organism fragments can be read from the vector risk warning as the fragment material.

[0101] The risk factor refers to the risk coefficient of pathogen dissemination under different response characteristics of different materials. The risk factor corresponding to the material and response characteristics of the fragment can be found in the risk factor correspondence table. The risk factor correspondence table is a data table that records different fragment materials and response characteristics and their corresponding risk factors. The product of the original freshness and the risk factor is used as the new freshness. For example, when the response characteristic is suspension, light insect wings are easy to suspend in the airflow and drift over long distances, with a very high risk of spreading pathogens. At this time, the risk factor is 0.9.

[0102] Step 210: Update the freshness based on the risk factor.

[0103] Inhaling air may cause the debris of disease vectors to move continuously. In this case, real-time images of the debris are acquired and the response characteristics of the debris movement are determined. Combined with the material of the debris, the risk factor is determined, thereby correcting the freshness of the debris and improving the accuracy of empty container inspection at the port.

[0104] Methods for determining freshness also include: Step 211: When the freshness is greater than the preset freshness threshold, determine the movement speed according to the real-time movement trajectory, and retrieve the fragment volume from the vector risk warning.

[0105] The freshness threshold is a critical value for judging whether a fragment is fresh enough. When the freshness is greater than the freshness threshold, it means that the fragment is relatively fresh. At this time, the risk of the spread of pathogens carried on the surface of the fragment is relatively high, and further analysis is required.

[0106] Motion speed refers to the average speed of the fragment moving along the real-time trajectory. It can be obtained by differentiating the real-time trajectory to obtain the instantaneous velocity sequence, and then taking the average value of the time sequence as the motion speed.

[0107] The fragment volume refers to the estimated three-dimensional size of the vector organism fragment, which can be read from the vector risk warning as the fragment volume.

[0108] Step 212: Determine the mass of the fragment according to the fragment material and fragment volume.

[0109] The mass of a fragment refers to the total amount of material contained in the fragment. The mass of a fragment can be the product of its density and volume. The density of a fragment refers to the mass of a unit volume of fragment. The density of a fragment can be found in a density correspondence table, which is a data table that records different fragment materials and their corresponding densities.

[0110] Step 213: Determine the kinetic energy of motion based on the motion speed and fragment mass, and determine the energy density during vibration according to the kinetic energy of motion and fragment volume.

[0111] Kinetic energy refers to the kinetic energy possessed by a fragment when it moves under the influence of airflow. It can be calculated using the formula: Kinetic Energy = 0.5 * Fragment Mass * Velocity².

[0112] Energy density refers to the average kinetic energy contained in a unit volume of a fragment. The quotient of the kinetic energy and the volume of the fragment can be used as the energy density.

[0113] Step 214: Determine the vibration frequency by combining the energy density and the fragment mass, and determine the volume frequency ratio based on the vibration frequency and the fragment volume.

[0114] Vibration frequency refers to the frequency of forced vibration of a fragment. When the fragment material is the same, the higher the energy density, the stronger the external excitation, the faster the fragment is pushed, and the faster the vibration. The smaller the fragment mass, the smaller the inertia of the object itself, the faster the fragment is pushed, and the higher the vibration frequency. The vibration frequency corresponding to the fragment material, energy density, and fragment mass can be found in the frequency correspondence table. The frequency correspondence table is a data table that records different fragment materials, energy densities, fragment masses, and their corresponding vibration frequencies.

[0115] The volume-frequency ratio is the ratio of the vibration frequency to the volume of the fragment, and it is used to normalize and evaluate the internal structural stiffness of fragments of different sizes.

[0116] Step 215: Determine the frequency coefficient based on the volume-frequency ratio, and update the freshness according to the frequency coefficient.

[0117] The frequency coefficient refers to the correction amount of the volume frequency ratio to freshness. The larger the volume frequency ratio, the denser and stiffer the structure, the lower the freshness, and the smaller the frequency coefficient. The frequency coefficient corresponding to the volume frequency ratio can be found in the frequency coefficient correspondence table. The frequency coefficient correspondence table is a data table that records different volume frequency ratios and their corresponding frequency coefficients, and the product of the frequency coefficient and the original freshness is used as the new freshness.

[0118] The vibration frequencies of disease vector debris with different freshness levels differ under wind action. In this case, the motion speed, kinetic energy, and vibration frequency are determined from the swing trajectory of the debris under the action of airflow, so as to calculate the volume frequency ratio of the debris. This reflects the changes in the internal structure of the debris with freshness, and thus corrects the freshness of the debris, improving the accuracy of port empty container inspection.

[0119] Reference Figure 4 Methods for determining freshness also include: Step 300: When the freshness is greater than a preset freshness threshold, acquire multi-source images according to the freshness, and determine the scattering parameters of the fragment surface based on the multi-source images.

[0120] Multi-source images refer to a sequence of images of the fragment captured under illumination from different angles. The multi-source images are acquired after the airflow jet command. In this embodiment, the light source angles include 0°, 90°, 180° and 270°. The LED beads at the corresponding positions are lit sequentially according to the preset angle order. Only one direction of the light source is lit at a time, while the other directions remain off. One image is acquired under each angle, for a total of 4 images. Each time the angle is switched, the high-definition camera acquires one image, obtaining a total of 4 images under different illumination angles, which are used for subsequent calculation of the scattering parameters of the fragment surface.

[0121] Scattering parameters refer to the scattering parameters of light incident on the surface of the fragment. The brightness changes of each pixel under different light sources can be extracted from multi-source images, and the approximate parameters of the bidirectional reflection distribution function can be calculated as scattering parameters. Scattering parameters include diffuse reflection coefficient and specular reflection coefficient.

[0122] Step 301: Determine the surface roughness of the fragment according to the scattering parameters.

[0123] Roughness refers to the unevenness of the surface of the fragment. The larger the diffuse reflection coefficient and the smaller the specular reflection coefficient, the rougher the surface. The roughness corresponding to the scattering parameter can be found in the roughness correspondence table, which is a data table that records different scattering parameters and their corresponding roughness levels.

[0124] Step 302: Determine the theoretical force by combining the roughness and blowing force, and determine the theoretical speed based on the vector risk warning and the theoretical force.

[0125] The theoretical force refers to the force exerted by the airflow on an ideal fragment with that roughness. The theoretical force corresponding to the surface roughness index and the blowing force can be found in the theoretical force correspondence table. The theoretical force correspondence table is a data table that records different roughnesses, blowing forces and their corresponding theoretical forces.

[0126] Theoretical speed refers to the ideal motion speed calculated under theoretical force, combining the mass of the fragment and the area of ​​force application. It can be calculated according to Newton's second law and the aerodynamic drag force formula.

[0127] Step 303: Determine the speed deviation by combining the theoretical speed and the actual speed.

[0128] Speed ​​deviation refers to the difference between theoretical speed and actual speed. The absolute value of the difference between theoretical speed and actual speed can be used as speed deviation.

[0129] Step 304: When the speed deviation is greater than a preset deviation threshold, determine the shedding mass based on the speed deviation.

[0130] The deviation threshold is a critical value that determines whether surface dust or other adhering substances have a significant impact on motion. When the velocity deviation is greater than the deviation threshold, it means that dust adheres to the surface of the fragment, resulting in a large difference between the velocity and the theoretical value. In this embodiment, the deviation threshold is 0.05 m / s.

[0131] The shedding mass refers to the mass of dust that falls off the surface of the fragment. The larger the velocity deviation, the greater the velocity at which the fragment decreases, the more obvious the change in fragment mass, and the greater the shedding mass. The shedding mass corresponding to the velocity deviation can be found in the shedding mass correspondence table, which is a data table that records different velocity deviations and their corresponding shedding masses.

[0132] Step 305: Determine the dust correction coefficient according to the detachment mass and fragment surface area, and update the freshness based on the dust correction coefficient.

[0133] The dust correction factor is an adjustment factor used to correct for freshness. The smaller the surface area of ​​the fragment, the greater the shedding mass, which means that there is more dust covering the surface of the fragment, the longer the fragment has been there, the less fresh it is, and the smaller the dust correction factor. The dust correction factor corresponding to the shedding mass and the fragment surface area can be found in the dust correction factor correspondence table. The dust correction factor correspondence table is a data table that records different shedding masses and fragment surface areas and their corresponding dust correction factors, and the product of the original freshness and the dust correction factor is used as the new freshness.

[0134] The longer the fragment has been present, the less fresh it is. Therefore, by collecting multi-source images to determine the scattering parameters and roughness of the fragment surface, and comparing them with the theoretical motion speed to obtain the velocity deviation, the influence of dust shedding can be determined. Then, the freshness can be updated using a dust correction coefficient, thereby improving the accuracy of empty container inspection at the port.

[0135] Methods for determining freshness also include: Step 306: When the dust correction coefficient is less than the preset correction threshold, determine the pre-blowing image in response to the fragment detection command.

[0136] The correction threshold is the critical value used to determine whether further dust analysis is needed. When the dust correction coefficient is less than the correction threshold, the surface of the fragment is covered with a large amount of dust, and further analysis is required.

[0137] The pre-blowing image refers to the original image before the fragment detection command is executed, and it can be retrieved from the image information.

[0138] Step 307: Determine the regional image of the scattering area based on the blown image.

[0139] The dispersion area refers to the spatial area formed by the diffusion of dust in the air after the blowing operation. The image after blowing and the image before blowing can be differentiated, and the difference image can be binarized. The set of continuous pixels with gray values ​​greater than a preset threshold is taken as the dispersion area of ​​dust. In this embodiment, the gray value threshold is 30.

[0140] A region image is an image of a scattered area, which can be extracted from a blown image to form a region image.

[0141] Step 308: Determine the dust dispersion concentration by combining the area image and the pre-blowing image, and determine the deposition thickness based on the dispersion concentration.

[0142] Scattering concentration refers to the number of dust particles per unit volume of space. The method involves differentiating the region image from the image before blowing, binarizing the difference image, and calculating the average grayscale value of each pixel in the binarized difference image as the average grayscale. The scattering concentration corresponding to the average grayscale value is then retrieved from a scattering concentration correspondence table, which records different average grayscale values ​​and their corresponding scattering concentrations.

[0143] Deposition thickness refers to the thickness of airborne dust that settles onto the surface of the fragment. The higher the dispersion concentration, the greater the amount of airborne dust, and the thicker the dust layer accumulated on the surface of the fragment after complete settling. The deposition thickness can be found in the deposition thickness correspondence table, which records data on different dispersion concentrations and their corresponding deposition thicknesses.

[0144] Step 309: Determine the sealing coefficient based on the deposition thickness.

[0145] The sealing coefficient refers to the physical barrier coefficient of dust against pathogens and debris. The greater the deposition thickness, the more complete the dust coverage and the better the sealing effect, and the larger the sealing coefficient. The sealing coefficient corresponding to the deposition thickness can be found in the sealing coefficient correspondence table, which is a data table that records different deposition thicknesses and their corresponding sealing coefficients.

[0146] Step 310: Determine the real-time speed according to the real-time movement trajectory, and determine the rate of change of speed based on the real-time speed.

[0147] Real-time velocity refers to the instantaneous velocity of the debris at each moment in the real-time moving trajectory. It can be obtained by performing frame-by-frame differential calculation on the real-time moving trajectory, which is the sequence of instantaneous velocities at each moment.

[0148] The rate of change of velocity refers to how quickly the velocity of the fragment changes with time. It can be calculated as the absolute value of the derivative of the real-time velocity sequence with respect to time.

[0149] Step 311: When the rate of change of speed is greater than the preset threshold, determine the detachment time based on the rate of change of speed, and determine the degree of sealing based on the detachment time.

[0150] The change threshold is the critical value used to determine whether dust is falling off the surface of the fragment. When the rate of change of speed is greater than the change threshold, it means that the fragment's mass is continuously decreasing and its speed is continuously increasing due to the continuous shedding of dust.

[0151] The shedding time refers to the time elapsed from the start of blowing until the dust has basically been shed. The greater the rate of change of speed, the faster the dust sheds and the shorter the shedding time. The shedding time corresponding to the rate of change of speed can be found in the shedding time correspondence table.

[0152] The degree of containment refers to the ability of the dust on the surface of the fragment to contain pathogens before blowing. The longer the shedding time, the slower the dust falls off and the denser the original dust layer, the greater the degree of containment. The degree of containment corresponding to the shedding time can be found in the degree of containment correspondence table, which is a data table that records different shedding times and their corresponding degree of containment.

[0153] Step 312: Determine the enhancement coefficient by combining the sealing coefficient and the sealing degree, and update the freshness according to the enhancement coefficient.

[0154] The enhancement factor is a coefficient used to correct the freshness. It can be calculated according to the formula: enhancement factor = 1 + storage degree * (1 - storage factor). The product of the original freshness and the enhancement factor is taken as the new freshness.

[0155] Blowing air causes dust to be stirred up. By comparing images before and after blowing air, the dust dispersion concentration can be obtained to determine the deposition thickness. This, combined with the dust shedding time, determines the degree of sealing and thus updates the freshness, improving the accuracy of empty container inspection at the port.

[0156] Methods for determining freshness also include: Step 313: Determine the scattering boundary according to the region image, and determine the diffusion gradient per unit distance based on the scattering boundary.

[0157] The scattering boundary refers to the outline of the dust scattering area, and the scattering boundary can be extracted from the area image using edge detection algorithms.

[0158] The diffusion gradient refers to the rate of change of dust particle concentration along the radial direction at the dispersion boundary. Taking a point on the dispersion boundary as the starting point, a unit distance is taken inward along its normal direction, and the concentration difference between the two ends of the line segment is calculated as the diffusion gradient at that point. The average diffusion gradient of all boundary points is taken as the diffusion gradient per unit distance. In this embodiment, the unit distance is 1 mm.

[0159] Step 314: Determine the dust settling rate by combining the diffusion gradient and dispersion concentration.

[0160] Dust settling rate refers to the speed at which dust falls and settles in still air. The greater the diffusion gradient and the higher the dispersion concentration, the more large dust particles there are and the faster the settling rate. The dust settling rate corresponding to the diffusion gradient and dispersion concentration can be found in the dust settling rate correspondence table. The dust settling rate correspondence table is a data table that records different diffusion gradients and dispersion concentrations and their corresponding dust settling rates.

[0161] Step 315: Determine the suspension time based on the dust settling rate.

[0162] Suspension time refers to the time required from the end of blowing to the completion of most dust settling. The suspension time can be the quotient of the center height of the scattered area and the dust settling rate. The blowing coordinates can be used as the object distance. By using the similar triangle ratio between the vertical pixel span of the dust area in the image and the camera focal length, the actual physical height of the dust can be deduced and half of it can be taken as the center height.

[0163] Step 316: When the suspension duration exceeds a preset duration threshold, determine the particle size of the dust in response to the suspension duration.

[0164] The duration threshold refers to the critical suspension time for determining whether dust particles have the ability to remain suspended for a long time. When the suspension duration is greater than the duration threshold, it indicates that the dust particles are relatively small and require further analysis. In this embodiment, the duration threshold is 10 seconds.

[0165] Particle size refers to the diameter of dust particles. According to Stokes' law of settling, the longer the suspension time, the smaller the dust particles. The particle size corresponding to the suspension time can be found in the particle size correspondence table, which is a data table that records different suspension times and their corresponding particle sizes.

[0166] Step 317: Determine the shielding capability of the redeposited dust based on the deposition thickness and particle size.

[0167] Shielding capability refers to the ability of the covering layer that forms again on the surface of the debris after dust settles to prevent pathogens from escaping from the debris. The smaller the particles, the thicker the deposition thickness, and the denser the covering layer formed after deposition, the stronger the shielding capability. The shielding capability corresponding to the deposition thickness and particle size can be found in the shielding capability correspondence table. The shielding capability correspondence table is a data table that records different deposition thicknesses and particle sizes and their corresponding shielding capabilities.

[0168] Step 318: Determine the descent coefficient based on the shielding capability, and update the freshness according to the descent coefficient.

[0169] The reduction factor is a correction factor that should be adjusted down in terms of freshness due to the shielding effect of redeposited dust. The greater the shielding ability, the greater the obstruction of pathogen transmission by the dust-deposited cover layer. The lower the risk of pathogen transmission, the smaller the reduction factor. The reduction factor corresponding to the shielding ability can be found in the reduction factor correspondence table. The reduction factor correspondence table is a data table that records different shielding abilities and their corresponding reduction factors. The product of the original freshness and the reduction factor is taken as the new freshness.

[0170] Dust can act as a shield against the spread of pathogens, resulting in a higher risk of pathogen transmission from stale pathogen debris covered in dust. In this case, the suspension time of dust is determined by analyzing images of dust in the air, and the shielding ability after deposition is assessed in conjunction with the deposition thickness to update the freshness and improve the accuracy of empty container inspection at ports.

[0171] The methods for updating vector risk warnings also include: Step 319: When the decrease coefficient is less than the preset coefficient threshold, determine the dispersion concentration curve based on the dispersion concentration.

[0172] The coefficient threshold is a critical value used to determine whether further dust quality analysis is needed. When the decreasing coefficient is less than the coefficient threshold, it indicates that the dust shielding effect is significant and further analysis is required. In this embodiment, the coefficient threshold is 0.8.

[0173] The dispersion concentration curve refers to the curve of dispersion concentration changing with the blowing time. According to the acquisition sequence of the moving images, the dispersion concentration corresponding to each frame of the image can be determined to obtain the dispersion concentration value at different times. Then, with time as the horizontal axis and dispersion concentration as the vertical axis, the data points at each time are placed at the corresponding positions in the coordinate system, and connected sequentially with a smooth curve to fit the dispersion concentration curve.

[0174] Step 320: Determine the dust mass curve in the air according to the dispersion concentration curve.

[0175] The dust mass curve refers to the curve showing the change of the total mass of suspended dust in the air over time. For each moment on the dispersion concentration curve, the dust mass corresponding to each dispersion concentration can be calculated using the formula: Dust Mass = Dispersion Concentration * Dispersion Area Volume * Dust Particle Density. These values ​​are then connected sequentially with a smooth curve to fit the dust mass curve. The particle density corresponding to each particle size can be looked up in a particle density correspondence table, which records data on different particle sizes and their corresponding particle densities. Using a pixel area mapping algorithm, the dispersion area is determined based on the number of pixels in the dispersion area mask image and the actual area of ​​a single pixel. The dispersion height is then obtained by combining the vertical pixel span method or infrared thickness calibration method, and the product of the dispersion height and dispersion area is taken as the dispersion area volume.

[0176] Step 321: Determine the wind force coefficient according to the blowing force, and determine the vibration coefficient according to the vibration frequency.

[0177] The wind force coefficient refers to the effective dust-carrying capacity of the blowing airflow. It is positively correlated with the blowing force. The wind force coefficient corresponding to the blowing force can be found in the wind force coefficient correspondence table, which is a data table that records different blowing forces and their corresponding wind force coefficients.

[0178] The vibration coefficient refers to the effect of fragment vibration on dust shedding. The higher the vibration frequency, the easier it is for the attached dust to be shaken off. The vibration coefficient corresponding to the vibration frequency can be found in the vibration coefficient correspondence table, which is a data table that records different vibration frequencies and their corresponding vibration coefficients.

[0179] Step 322: Determine the coverage quality by combining the vibration coefficient, wind force coefficient and dust mass.

[0180] Coverage mass refers to the mass of dust covering the fragments before blowing. The mass is calculated using the formula: Coverage mass = Dust mass / (Wind force coefficient + Vibration coefficient).

[0181] Step 323: Determine the density curve based on the coverage quality and the surface area of ​​the fragment.

[0182] The density curve is a curve that shows how the density changes over time. The density can be obtained by taking the quotient of the covering mass and the surface area of ​​the fragment as the density, and then fitting the density curve with time as the horizontal axis and density as the vertical axis.

[0183] Step 324: Determine the dangerous period according to the density curve and the preset density threshold, and update the vector risk warning according to the dangerous period.

[0184] The density threshold refers to the critical density value for determining whether the dust covering layer is sufficient to effectively seal pathogens. In this embodiment, the density threshold is 0.5 mg / mm².

[0185] The danger period refers to the time period during which the density remains below the density threshold. The time interval during which the density is continuously below the density threshold can be read from the density curve as the danger period. The danger period includes the start and end of the time interval and is added to the vector risk warning.

[0186] The wind force coefficient and vibration coefficient are determined by the wind force during air blowing and the vibration frequency of the disease vector debris under forced vibration. This determines the mass of dust covering the debris before air blowing, and then determines the density curve to update the disease vector risk warning, thus improving the accuracy of port empty container inspection.

[0187] Based on the same inventive concept, embodiments of the present invention provide a development and application system for a port intelligent empty container inspection robot system, including: The data acquisition module is used to collect data from multiple sources. A memory for storing the program of the development and application method of the port intelligent empty container inspection robot system as described in any one of the claims; The processor is the unit of memory that allows programs to be loaded and executed by the processor.

[0188] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0189] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for developing and applying a port intelligent empty container inspection robot system, characterized in that, include: Step 100: Collect data from multiple sources; Step 101: Determine the port route of the quadruped robot in the port based on the multi-source data and the preset positioning interval, and send movement instructions along the route in response to the generation of the port route. Step 102: Retrieve the real-time positioning of the quadruped robot from the multi-source data according to the motion instructions along the line; Step 103: When the real-time positioning falls within the positioning range, respond to the real-time positioning by generating and sending an empty box scanning command; Step 104: Acquire image information according to the empty box scanning command, and determine suspicious targets based on the image information; Step 105: Determine the credibility of the target based on the suspected target; Step 106: When the credibility of the target is greater than the preset credibility threshold, determine the bounding box data based on the suspicious target and image information; Step 107: Determine the transmission data based on the selected data, and generate a data transmission instruction according to the transmission data; Step 108: The terminal receives the transmitted data and determines the segmented data based on the transmitted data; Step 109: Determine risk data according to the segmented data, and generate vector risk warnings based on the risk data.

2. The method for developing and applying the port intelligent empty container inspection robot system according to claim 1, characterized in that, The method for determining the selected data also includes: Step 110: When the target's credibility is lower than a preset credibility threshold, determine the jet coordinates based on the suspected target and image information; Step 111: Generate an airflow injection command according to the jet coordinates, and acquire a post-jet image in response to the airflow injection command; Step 112: Determine the target displacement based on the post-spray image, image information, and suspected target; Step 113: When the target displacement is greater than a preset displacement threshold, determine the fragment target based on the target displacement, and determine the adhesion index by combining the target displacement and the fragment target; Step 114: Determine the corrected targets according to the adhesion index and the suspected targets, and update the selected data according to the corrected targets.

3. The method for developing and applying the port intelligent empty container inspection robot system according to claim 1, characterized in that, The method for correcting the vector risk warning also includes: Step 200: Determine the fragment coordinates based on the vector risk warning, and determine the air blowing coordinates according to the fragment coordinates; Step 201: Determine the surface area of ​​the debris based on the debris coordinates and vector risk warning, and determine the blowing force based on the debris surface area; Step 202: Determine the fragment detection command by combining the blowing coordinates and blowing force; Step 203: In response to the fragment detection command, determine the blown image of the fragment, and determine the cross-sectional gloss of the fragment based on the blown image and image information; Step 204: Determine the degree of freshness according to the gloss of the cross-section, and determine the degree of danger based on the degree of freshness; Step 205: Update the vector risk warning in response to the stated level of danger.

4. The method for developing and applying the port intelligent empty container inspection robot system according to claim 3, characterized in that, The methods for determining the freshness include: Step 206: In response to the blowing force and the preset monitoring area, determine the real-time movement image of the fragment under the action of continuous airflow; Step 207: Determine the real-time movement trajectory of the fragment based on the moving image; Step 208: Determine the response characteristics according to the real-time movement trajectory; Step 209: Retrieve the fragment material from the vector risk warning, and determine the risk factor based on the fragment material and response characteristics; Step 210: Update the freshness based on the risk factor.

5. The method for developing and applying the port intelligent empty container inspection robot system according to claim 4, characterized in that, The method for determining freshness also includes: Step 211: When the freshness is greater than the preset freshness threshold, determine the movement speed according to the real-time movement trajectory, and retrieve the fragment volume from the vector risk warning; Step 212: Determine the mass of the fragment according to its material and volume; Step 213: Determine the kinetic energy of motion based on the motion speed and the mass of the fragment, and determine the energy density during vibration according to the kinetic energy of motion and the volume of the fragment; Step 214: Determine the vibration frequency by combining the energy density and the fragment mass, and determine the volume-frequency ratio based on the vibration frequency and the fragment volume; Step 215: Determine the frequency coefficient based on the volume-frequency ratio, and update the freshness according to the frequency coefficient.

6. The method for developing and applying the port intelligent empty container inspection robot system according to claim 5, characterized in that, The method for determining freshness also includes: Step 300: When the freshness is greater than a preset freshness threshold, acquire multi-light source images according to the freshness, and determine the scattering parameters of the fragment surface based on the multi-light source images; Step 301: Determine the surface roughness of the fragment according to the scattering parameters; Step 302: Determine the theoretical force by combining the roughness and blowing force, and determine the theoretical speed based on the vector risk warning and the theoretical force; Step 303: Determine the speed deviation by combining the theoretical speed and the actual speed; Step 304: When the speed deviation is greater than a preset deviation threshold, determine the detachment mass based on the speed deviation; Step 305: Determine the dust correction coefficient according to the detachment mass and fragment surface area, and update the freshness based on the dust correction coefficient.

7. The method for developing and applying the port intelligent empty container inspection robot system according to claim 6, characterized in that, The method for determining freshness also includes: Step 306: When the dust correction coefficient is less than the preset correction threshold, determine the pre-blowing image in response to the fragment detection command; Step 307: Determine the region image of the scattering area based on the post-blowing image; Step 308: Determine the dust dispersion concentration by combining the area image and the pre-blowing image, and determine the deposition thickness based on the dispersion concentration; Step 309: Determine the sealing coefficient based on the deposition thickness; Step 310: Determine the real-time speed according to the real-time movement trajectory, and determine the rate of change of speed based on the real-time speed; Step 311: When the rate of change of speed is greater than a preset threshold, determine the detachment time based on the rate of change of speed, and determine the degree of sealing based on the detachment time; Step 312: Determine the enhancement coefficient by combining the sealing coefficient and the sealing degree, and update the freshness according to the enhancement coefficient.

8. The method for developing and applying the port intelligent empty container inspection robot system according to claim 7, characterized in that, The method for determining freshness also includes: Step 313: Determine the scattering boundary according to the region image, and determine the diffusion gradient per unit distance based on the scattering boundary; Step 314: Determine the dust settling rate by combining the diffusion gradient and dispersion concentration; Step 315: Determine the suspension time based on the dust settling rate; Step 316: When the suspension duration exceeds a preset duration threshold, determine the particle size of the dust in response to the suspension duration; Step 317: Determine the shielding capability of the redeposited dust based on the deposition thickness and particle size; Step 318: Determine the descent coefficient based on the shielding capability, and update the freshness according to the descent coefficient.

9. The method for developing and applying the port intelligent empty container inspection robot system according to claim 8, characterized in that, The method for updating vector risk warnings also includes: Step 319: When the decrease coefficient is less than a preset coefficient threshold, determine the dispersion concentration curve based on the dispersion concentration; Step 320: Determine the dust mass curve in the air according to the dispersion concentration curve; Step 321: Determine the wind force coefficient according to the blowing force, and determine the vibration coefficient according to the vibration frequency; Step 322: Determine the coverage quality by combining the vibration coefficient, wind force coefficient, and dust mass; Step 323: Determine the density curve based on the coverage quality and fragment surface area; Step 324: Determine the dangerous period according to the density curve and the preset density threshold, and update the vector risk warning according to the dangerous period.

10. A development and application system for a port intelligent empty container inspection robot system, characterized in that, include: The data acquisition module is used to collect data from multiple sources. A memory for storing the program of the development and application method of the port intelligent empty container inspection robot system as described in any one of claims 1 to 9; The processor is the unit of memory that allows programs to be loaded and executed by the processor.