Hazardous chemical substance loading and unloading operation process safety monitoring method, device, equipment and medium

Abstract

The invention provides a hazardous chemical loading and unloading operation process safety monitoring method, device and equipment and a medium, and belongs to the technical field of industrial safety detection and early warning. The method comprises the following steps: acquiring images (including camera numbers and acquisition moments) of various operation processes of loading and unloading the hazardous chemical substances, and identifying alarm types of the hazardous chemical substances; performing intra-process association according to the camera number, the alarm type and the acquisition time to form a multi-class image set with the same alarm type; the process is matched with a predefined operation process alarm type combination table, and a due alarm type combination of the process is obtained; and if the alarm type of any image set in the process belongs to the combination, outputting the alarm types of all the image sets. According to the method, image sources are divided according to the operation process, repeated entries are reduced through alarm type grouping, irrelevant alarm types are dynamically filtered based on the process stage to improve correlation, low-code management of complex correlation logic is achieved by means of a configurable rule engine, and the maintenance cost is remarkably reduced.

Description

Technical Field

[0001] This invention relates to the field of industrial safety detection and early warning technology, specifically to a method for safety monitoring of hazardous chemical loading and unloading operations, a device for safety monitoring of hazardous chemical loading and unloading operations, an electronic device, and a computer-readable storage medium. Background Technology

[0002] Hazardous chemical loading and unloading operations involve flammable, explosive, and toxic substances. Their multi-stage nature (such as vehicle guidance, loading arm docking, oil transfer, oil and gas recovery, and vehicle departure) makes them inherently risky and complex. Traditional manual monitoring and single-point sensor monitoring suffer from blind spots, poor real-time performance, and difficulty in handling dynamic, multi-stage scenarios.

[0003] Currently, intelligent monitoring systems based primarily on visual recognition have been introduced. By deploying cameras at multiple points and using image analysis algorithms, these systems automatically identify violations (such as not wearing anti-static clothing, leaks, etc.) and generate alarms, significantly improving the monitoring range and automation level.

[0004] However, the existing visual recognition intelligent monitoring system has the following problems: (1) When the existing system monitors the complex process of loading and unloading hazardous chemicals with multiple cameras, the lack of a multi-source image alarm association mechanism based on the operation process leads to the fragmentation and dispersion of risk monitoring of key operation nodes, and it is impossible to effectively aggregate multi-view and multi-type risk information under the same node to ensure the safety of different process nodes; (2) Alarm entries are usually generated based on a single image, which leads to a large number of repeated alarms of the same alarm type (such as multiple people not wearing safety helmets) in a short period of time, which significantly increases the information processing burden; (3) The system cannot automatically identify the alarm type that should be paid attention to in the current operation process (such as electrostatic alarms should not be paid attention to in the oil and gas recovery process), which leads to a large number of irrelevant alarm interferences and reduces the effective information concentration; (4) The existing system needs to adjust the alarm rules by modifying the code. For example, when updating the alarm type of the operation process mapping, it relies on professional development support, which increases the complexity of business logic, maintenance costs and expansion difficulty, and cannot flexibly adapt to changes in on-site specifications. Summary of the Invention

[0005] The purpose of this invention is to provide a safety monitoring method for hazardous chemical loading and unloading operations to solve the above-mentioned problems.

[0006] To achieve the above objectives, embodiments of the present invention provide a safety monitoring method for hazardous chemical loading and unloading operations, comprising: Acquire multiple images from various operational processes during the loading and unloading of hazardous chemicals, and determine the alarm types for these images. Each image is labeled with its corresponding camera number and acquisition time. Based on the camera number, alarm type, and acquisition time of multiple images under each operational process of hazardous chemical loading and unloading, image association processing is performed on the multiple images under each operational process of hazardous chemical loading and unloading to obtain multiple image sets under each operational process of hazardous chemical loading and unloading; among them, multiple images in each image set have the same alarm type. The alarm type combination of each operation process of hazardous chemical loading and unloading is matched with the pre-built operation process alarm type combination table to obtain the alarm type combination of each operation process of hazardous chemical loading and unloading; wherein, the pre-built operation process alarm type is used to characterize the mapping relationship between different operation processes of hazardous chemical loading and unloading and the corresponding alarm type combination. If the alarm type of at least one image set under each operation process of hazardous chemical loading and unloading is a component of the alarm type combination of each operation process of hazardous chemical loading and unloading, then the output displays the alarm types of multiple image sets under each operation process of hazardous chemical loading and unloading.

[0007] Optionally, determine the alarm types for multiple images across multiple operational processes during the handling of hazardous chemicals, including: Multiple images from each operational process of hazardous chemical loading and unloading are input into the alarm type analysis model to obtain the alarm types of multiple images from each operational process of hazardous chemical loading and unloading. The alarm type analysis model is trained on a pre-built neural network model by the alarm types corresponding to multiple images from each operational process of hazardous chemical loading and unloading.

[0008] Optionally, the multiple operational processes for loading and unloading hazardous chemicals include: pre-operation preparation process, loading and unloading operation phase process, and post-operation cleanup process; image association processing includes: combination processing, aggregation processing, and clustering processing; Based on the camera IDs, alarm types, and acquisition times of multiple images from various operational processes in hazardous chemical loading and unloading, image association processing is performed on these images for each operational process, resulting in multiple image sets for each operational process in hazardous chemical loading and unloading, including: Based on the camera number, alarm type, and acquisition time of multiple images in the pre-operation preparation process of hazardous chemical loading and unloading, the multiple images in the pre-operation preparation process of hazardous chemical loading and unloading are combined and processed to obtain a multi-type image set in the pre-operation preparation process of hazardous chemical loading and unloading. Based on the camera number, alarm type, and acquisition time of multiple images in the loading and unloading operation process of hazardous chemicals, the multiple images in the loading and unloading operation process of hazardous chemicals are grouped and clustered to obtain a multi-type image set in the loading and unloading operation process of hazardous chemicals. Based on the camera number, alarm type, and acquisition time of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, the multiple images from the post-operation cleanup process of hazardous chemical loading and unloading are aggregated to obtain a multi-type image set of the post-operation cleanup process of hazardous chemical loading and unloading.

[0009] Optionally, based on the camera number, alarm type, and acquisition time of multiple images from the pre-operation preparation process for hazardous chemical loading and unloading, the multiple images from the pre-operation preparation process for hazardous chemical loading and unloading are combined to obtain a multi-type image set from the pre-operation preparation process for hazardous chemical loading and unloading, including: Images belonging to the same alarm type and whose acquisition time is less than the first preset acquisition time under the pre-operation preparation process for hazardous chemical loading and unloading are identified as the same image set, so as to obtain multiple image sets under the pre-operation preparation process for hazardous chemical loading and unloading.

[0010] Optionally, based on the camera number, alarm type, and acquisition time of multiple images from the loading and unloading operation phase of hazardous chemicals, the multiple images from the loading and unloading operation phase of hazardous chemicals are grouped and clustered to obtain multiple image sets from the loading and unloading operation phase of hazardous chemicals, including: Multiple images belonging to the same alarm type and whose difference between the collection time and the preset collection time is less than the first preset time are identified as the same image set after the completion process of hazardous chemical loading and unloading operations. Based on the acquisition time of multiple images in various image sets, the multiple images in various image sets are sorted from first to last; Multiple images whose acquisition time is less than the acquisition time of adjacent images in each image set are identified as target images for each image set. Target images belonging to multiple image sets with the same camera number are identified as the same image set to obtain multiple image sets under the post-operation cleanup process for hazardous chemical loading and unloading.

[0011] Optionally, based on the camera number, alarm type, and acquisition time of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, the multiple images from the post-operation cleanup process of hazardous chemical loading and unloading are aggregated to obtain multiple image sets from the post-operation cleanup process of hazardous chemical loading and unloading, including: Images belonging to the same alarm type and the same camera number under the loading and unloading operation process of hazardous chemicals are identified as the same image set; Based on the acquisition time of multiple images in various image sets, the multiple images in various image sets are sorted from first to last; Images whose acquisition time is less than the acquisition time of adjacent images are identified as target images for each image set, so that the target images can be used as multiple image sets in the loading and unloading operation process of hazardous chemicals.

[0012] In a second aspect of the present invention, a safety monitoring device for hazardous chemical loading and unloading operations is provided, comprising: The data acquisition module is used to acquire multiple images under multiple operational processes of hazardous chemical loading and unloading, and to determine the alarm type of multiple images under multiple operational processes of hazardous chemical loading and unloading; each image is marked with a corresponding camera number and acquisition time; The image association module is used to perform image association processing on multiple images under each operation process of hazardous chemical loading and unloading based on the camera number, alarm type and acquisition time of multiple images under each operation process of hazardous chemical loading and unloading, to obtain multiple image sets under each operation process of hazardous chemical loading and unloading; among them, multiple images in each image set have the same alarm type. The combination matching module is used to match each operation process of hazardous chemical loading and unloading with a pre-built operation process alarm type combination table to obtain the alarm type combination of each operation process of hazardous chemical loading and unloading; wherein, the pre-built operation process alarm type is used to characterize the mapping relationship between different operation processes of hazardous chemical loading and unloading and the corresponding alarm type combination. The alarm verification module is used to output and display the alarm types of multiple image sets under each operation process of hazardous chemical loading and unloading when the alarm type of at least one image set under each operation process of hazardous chemical loading and unloading belongs to the combination of alarm types under each operation process of hazardous chemical loading and unloading.

[0013] Optionally, the multiple operational processes for loading and unloading hazardous chemicals include: pre-operation preparation process, loading and unloading operation phase process, and post-operation cleanup process; image association processing includes: combination processing, aggregation processing, and clustering processing; The image association module is specifically used for: Based on the camera number, alarm type, and acquisition time of multiple images in the pre-operation preparation process of hazardous chemical loading and unloading, the multiple images in the pre-operation preparation process of hazardous chemical loading and unloading are combined and processed to obtain a multi-type image set in the pre-operation preparation process of hazardous chemical loading and unloading. Based on the camera number, alarm type, and acquisition time of multiple images in the loading and unloading operation process of hazardous chemicals, the multiple images in the loading and unloading operation process of hazardous chemicals are grouped and clustered to obtain a multi-type image set in the loading and unloading operation process of hazardous chemicals. Based on the camera number, alarm type, and acquisition time of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, the multiple images from the post-operation cleanup process of hazardous chemical loading and unloading are aggregated to obtain a multi-type image set of the post-operation cleanup process of hazardous chemical loading and unloading.

[0014] In a third aspect of the present invention, an electronic device is provided, comprising: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and the machine-readable instructions, when executed by the processor, perform the above-described safety monitoring method for hazardous chemical loading and unloading operations.

[0015] In a fourth aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions that, when executed on a computer, cause the computer to perform the above-described safety monitoring method for hazardous chemical loading and unloading operations.

[0016] The beneficial effects of this invention are: (1) By dividing the image sources according to the work process and performing correlation processing based on camera number, alarm type and acquisition time, the safety monitoring needs of the complex and multi-stage loading and unloading of hazardous chemicals can be effectively met, ensuring coverage of key operation nodes.

[0017] (2) By grouping images of the same alarm type into a set of images, the granularity of alarm display is increased from a single image to an alarm type group, thereby reducing the number of duplicate alarm entries.

[0018] (3) By matching the work process with the predefined alarm type combination, alarm types that are not related to the current work process are automatically filtered out (such as ignoring electrostatic alarms during the oil transfer stage), which greatly reduces invalid alarm interference and improves the relevance of alarm information.

[0019] (4) By using the pre-built "job process-alarm type combination" mapping rule engine, the complex multi-source alarm association logic is transformed into rule configuration. Users can define or adjust these mapping rules according to the actual job specifications. No deep code development is required, which significantly reduces the complexity of business logic, maintenance costs and subsequent expansion difficulty.

[0020] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the safety monitoring method for hazardous chemical loading and unloading operations provided in this embodiment of the invention. Figure 2 This is a schematic diagram illustrating the principle of image aggregation processing provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the safety monitoring device for hazardous chemical loading and unloading operations provided in an embodiment of the present invention. Detailed Implementation

[0022] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this application.

[0024] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0025] Example 1 Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the safety monitoring method for hazardous chemical loading and unloading operations provided in this embodiment of the invention. The method includes the following steps: S100: Acquire multiple images under multiple operational processes of hazardous chemical loading and unloading, and determine the alarm type of multiple images under multiple operational processes of hazardous chemical loading and unloading; wherein, each image is marked with the corresponding camera number and acquisition time; This patented system adopts a cloud-edge architecture, with the platform and edge working collaboratively. The edge is responsible for real-time data acquisition and preliminary analysis. It acquires video streams through remote cameras and uses a convolutional neural network (CNN) AI algorithm deployed on that side for image recognition and analysis. When an event meeting the alarm criteria is identified, the edge's AI analyzer generates alarm data and sends the alarm information to the platform via MQTT message middleware. The platform handles the core processing and display functions. Its backend service (based on Java EE architecture) receives the alarm data transmitted via MQTT, first performing data cleaning, combination, and aggregation operations. Then, it associates and matches the processed data with the business rule model, triggering and executing the corresponding process rules for processing and flow. Finally, the alarm data that has completed the process processing is persistently stored in a non-volatile storage unit. Meanwhile, the platform's frontend (web interface) is responsible for displaying the processed alarm information and analysis results to the user.

[0026] In one embodiment, the multiple operational processes for loading and unloading hazardous chemicals include a pre-operation preparation process, a loading and unloading operation phase process, and a post-operation cleanup process.

[0027] It should be noted that cameras were installed in multiple operational scenarios during the handling of hazardous chemicals. These cameras captured images of each operational process. The cameras installed in each operational process have different numbers, and the images were captured at different times. Therefore, each image is labeled with its corresponding camera number and capture time.

[0028] It should be noted that multiple images were collected during the various operational processes of hazardous chemical loading and unloading, and each image was marked with an alarm type, meaning that each operational process included multiple alarm types.

[0029] The alarm types in the pre-operation preparation process include: personnel safety distance alarm, missing protective equipment alarm, vehicle not grounded alarm, missing fire-fighting equipment alarm, abnormal status of hose / arming hose alarm, vehicle parking beyond the boundary alarm, and environmental leakage warning alarm. The alarm types during the loading and unloading operation phase include: liquid leakage alarm, personnel leaving their post alarm, vehicle displacement alarm, static grounding disconnection alarm, abnormal pressure alarm, and external intrusion alarm. The alarm types for the post-operation cleanup process include: residual liquid in hoses / arming arms, vehicle not disconnected, incomplete site cleanup, tank lid not closed, personnel remaining, and abnormal seal.

[0030] In one embodiment, determining the alarm types of multiple images under multiple operational processes in the handling of hazardous chemicals includes: Multiple images from each operational process of hazardous chemical loading and unloading are input into the alarm type analysis model to obtain the alarm types of multiple images from each operational process of hazardous chemical loading and unloading. The alarm type analysis model is trained on a pre-built neural network model by the alarm types corresponding to multiple images from each operational process of hazardous chemical loading and unloading.

[0031] It should be noted that the neural network model can be deployed at the edge. When the camera captures an image, it is uploaded to the alarm type analysis model deployed at the edge, which is trained by the convolutional neural network model, for image recognition, and thus outputs the alarm type.

[0032] In one embodiment, the neural network model can be a convolutional neural network, a deep learning neural network, a YOLO model, etc.

[0033] The training process for the alarm type analysis model is as follows: Step 1 (Data Collection and Acquisition): This forms the basis of the training. A large number of real-world images or video frames from hazardous materials loading and unloading operations need to be collected. These images must cover all critical processes of the loading and unloading operation, such as vehicle / ship positioning checks, PPE donning checks, static grounding connection verification, equipment and pipeline connection checks, leak monitoring, work permit verification, fire extinguisher inspection, warning zone setup, actual operation monitoring, and post-operation checks.

[0034] The images are primarily taken by surveillance cameras installed on-site at different times, angles, and lighting conditions (daytime, nighttime, sunny / cloudy). A large amount of data is required, typically thousands or even hundreds of thousands of high-quality images, to ensure the model learns sufficiently rich features and scene variations.

[0035] Step 2 (Data Labeling): The raw images collected are meaningless to the model and must be manually annotated by domain experts (usually safety engineers familiar with the safety procedures for handling hazardous chemicals) or highly trained professional annotators.

[0036] The core task of labeling is to "tag" each image. This tag represents the type of alarm the image signifies. For example: if an image clearly shows an operator not wearing a safety helmet, it is labeled "Helmet Not Wearing"; if an image shows an unconnected or loosely connected static grounding clamp, it is labeled "Static Grounding Not Connected" or "Poor Static Grounding"; if an image shows oil stains or chemical residue on the ground, it is labeled "Liquid Leak"; if an image shows a vehicle without wheel chocks, it is labeled "Wheel chocks Not Placed"; if no safety issues are found, it is labeled "Normal" or "No Alarm".

[0037] The annotation process requires extremely high accuracy and consistency. For a single image, it may be necessary to annotate it with multiple alarm types (if multiple issues exist simultaneously) or label it as "normal." Annotation results are typically stored as structured data (e.g., one or more alarm type codes per image).

[0038] Step 3 (Data Preprocessing and Augmentation): Image normalization: The original images obtained from different cameras are adjusted to a uniform size required by the model (e.g., 224 pixels wide by 224 pixels high), and the pixel values ​​are normalized (e.g., scaled to between 0 and 1 or normalized) so that all input data are within a uniform numerical range.

[0039] Data augmentation: This is a core technique for improving model generalization ability and preventing overfitting (i.e., the model only remembers the training data and cannot adapt to new data). It artificially expands the dataset by performing a series of random transformations on the original training images, generating new and diverse "virtual" samples. Common augmentation methods include geometric transformations, color transformations, noise injection, and random cropping.

[0040] Step 4 (Model Architecture Selection and Construction): Alarm type analysis is essentially an image classification task (and may also combine object detection). Therefore, choosing convolutional neural networks as the underlying architecture is the mainstream approach.

[0041] Typically, a mature CNN model pre-trained on a large image dataset (such as ImageNet) is chosen as the starting point (called a "pre-trained model" or "backbone network"), such as ResNet (e.g., ResNet50), VGGNet, MobileNet, EfficientNet, etc. These models already possess powerful general image feature extraction capabilities.

[0042] Key modification: Remove the original classifier at the top layer of the pre-trained model (usually a 1000-class classifier for ImageNet) and replace it with a new fully connected network adapted to our task requirements. The last layer of this new network is a Softmax classification layer, with the number of nodes equal to the total number of alarm types we need to identify (e.g., if we want to identify 20 different types of alarms and 1 normal state, this layer would have 21 nodes). The bottom convolutional layers of the pre-trained model are responsible for extracting general low- and mid-level features of the image (edges, textures, shapes, etc.), while our newly added top layer is responsible for learning to map these general features to our specific alarm types.

[0043] Step 5 (Model Training): Prepare the training environment: Prepare the processed training dataset (images + labels), validation dataset, model architecture, and set the initial training parameters (hyperparameters), such as learning rate, batch size, optimizer type (commonly Adam or stochastic gradient descent with momentum SGD), training epochs, etc.

[0044] Forward propagation: A small batch of preprocessed training images is fed into the model. The model performs convolutions and calculations layer by layer, and finally produces a predicted probability for each alarm type at the output layer (e.g., the probability of predicting an image as "not wearing a helmet" is 80%, and the probability of predicting it as "normal" is 15%, etc.).

[0045] Loss calculation: The model's predictions (probability distribution) are compared with the image's true labels (i.e., manually labeled alert types). A loss function is used to quantify the difference between the predictions and the true values. For multi-class classification tasks, the cross-entropy loss function is most commonly used. The larger the loss value, the less accurate the model's predictions are.

[0046] Backpropagation: The loss value is a key signal guiding model learning. The backpropagation algorithm calculates the gradient of the loss function relative to all adjustable parameters within the model (primarily convolutional kernel weights and fully connected layer weights). The gradient indicates how each parameter needs to be adjusted (increased or decreased) to reduce the loss value.

[0047] Parameter optimization: Using an optimizer, the model parameters are updated according to the calculated gradients and a set learning rate (which controls the size of each adjustment step). The goal is to find a set of parameters that minimizes the model's average loss across all training data.

[0048] Iterative loop: Repeat the above process (forward propagation -> loss calculation -> back propagation -> parameter optimization). Update using a batch of data each time. A complete traversal of all training data is called an "epoch". Training typically requires dozens or even hundreds of epochs.

[0049] Learning rate adjustment: During training, a fixed learning rate is not typically used. A common strategy is to use a slightly larger learning rate initially and then rapidly decrease it, followed by gradual reduction of the learning rate as training progresses (e.g., learning rate decay) to finely adjust the parameters and approximate the optimal solution.

[0050] Step 6 (Model Validation and Evaluation): During training, a model's performance cannot be judged solely by its performance on the training data (as this can easily lead to overfitting). Therefore, it is necessary to reserve a portion of data that has never been used during training (the validation set) and periodically evaluate the model's performance on the validation set.

[0051] Evaluation indicators: Accuracy: The percentage of images correctly predicted by the model out of the total number of images. This is the most intuitive metric.

[0052] Confusion matrix: A table that details the model's predictions for each class (True Positives TP, False Positives FP, True Negatives TN, False Negatives FN). It reveals more about the problems (e.g., which classes are easily confused) than accuracy.

[0053] Accuracy: What percentage of images predicted as a certain alarm type are actually alarms of that type? High accuracy means fewer false alarms.

[0054] Recall: For a given alarm type that actually exists, what percentage of the model successfully identifies it? A high recall rate means fewer false negatives and higher security.

[0055] F1 score: The harmonic mean of precision and recall, is a commonly used metric to comprehensively measure a model’s performance on a given class (especially when classes are imbalanced).

[0056] Step 7 (Model Testing and Final Evaluation): After training and validation tuning are completed, the final selected model is evaluated using a completely independent test set (data that has never been involved in the training or validation process). Performance metrics on the test set (accuracy, precision, recall, F1 score, etc.) best reflect the model's generalization ability on real unknown data and are the ultimate standard for evaluating the model's practical application value.

[0057] Step 8 (Model Deployment and Continuous Iteration): After testing confirms that the performance meets the requirements, the model can be deployed to a real production environment (such as a monitoring center server or edge computing device). The deployed model receives real-time image streams from on-site cameras, analyzes them, and outputs alarm types and confidence levels.

[0058] S200: Based on the camera number, alarm type and acquisition time of multiple images under each operation process of hazardous chemical loading and unloading, perform image association processing corresponding to each operation process to obtain multiple image sets under each operation process of hazardous chemical loading and unloading; among them, multiple images in each image set have the same alarm type. Understandably, the required level of safety monitoring varies depending on the different operational processes involved in the loading and unloading of hazardous chemicals, and different levels of safety monitoring correspond to different image association processing modes. Therefore, this embodiment provides different image association processing modes for different operational processes.

[0059] Specifically, the multiple operational processes for loading and unloading hazardous chemicals include: pre-operation preparation, loading and unloading operation phases, and post-operation cleanup. Image association processing includes: combination processing, aggregation processing, and clustering processing; In one embodiment, step S200 includes: S210, based on the camera number, alarm type and acquisition time of multiple images under the pre-operation preparation process of hazardous chemical loading and unloading, the multiple images under the pre-operation preparation process of hazardous chemical loading and unloading are combined and processed to obtain a multi-type image set under the pre-operation preparation process of hazardous chemical loading and unloading. Specifically, step S210 includes: Images belonging to the same alarm type and whose acquisition time is less than the first preset acquisition time under the pre-operation preparation process for hazardous chemical loading and unloading are identified as the same image set, so as to obtain multiple image sets under the pre-operation preparation process for hazardous chemical loading and unloading.

[0060] For ease of understanding, a specific implementation method is given below: Configuration is performed via a web-based interface. This interface maintains a configuration table for image set generation rules. The configuration table includes information such as camera number, target alarm type, first preset duration Δt, and the effective time period of the configuration.

[0061] The core function of the configuration table is to determine which images collected at specific times should be grouped into the same image set when a specific alarm type occurs on a specific camera (based on the fact that the difference between the collection time and the preset collection time is less than Δt).

[0062] The program flow for the image set generation unit is as follows: Step 1 (Query the configuration table): Based on the camera number and alarm type in the received alarm message, query the "Image Set Generation Rule Configuration Table" to obtain the corresponding configuration information (especially the time difference threshold Δt).

[0063] Step 2 (Determine if processing is needed): Determine if this alarm type is configured to require generating an image set. If not, the process ends. If so, proceed to the next step.

[0064] Step 3 (Insert alarm / image information records into the cache (set TTL)): Insert the key information of this alarm (alarm type, precise alarm occurrence time, associated image identifier or metadata) into the Cache device in an overwrite or append manner (logic needs to be designed, see description), and set the TTL parameter (TTL should be greater than or equal to the configured Δt to ensure that the relevant records are valid within the time window).

[0065] Step 4 (Query recent records of the same alarm type in the cache): Search the Cache for all alarm / image records of the same camera and alarm type where the difference between the alarm occurrence time and the current alarm time is less than the configured time difference threshold Δt.

[0066] Step 5 (Determine if a set of records matching the criteria has been found): If the query result is empty (i.e., the current alarm is the first of its kind within this time window), then only the current record is cached and the process ends.

[0067] If one or more records that meet the criteria are found (including current alerts), proceed to the next step.

[0068] Step 6 (Merge records, generate / update image set, clear cache related records): All images associated with the retrieved alarm records (including the one triggered this time) are merged into a single image set.

[0069] After generating or updating the information for the image set, clear all associated alarm records belonging to that image set from the cache.

[0070] Step 7 (Output Image Set): The platform outputs or records the final generated image set information (including all associated image identifiers or metadata, alarm types, time windows, etc.).

[0071] An expiration time caching middleware technology (such as Redis, ttlcache) is employed. This technology is used for temporary data storage and to improve data access speed. TTL (Time To Live) is the core parameter controlling the validity period of cached data. By setting a TTL for each alarm record (usually equal to or slightly greater than the configured Δt), expired records (i.e., records exceeding the current alarm time Δt range) can be automatically cleaned up, ensuring that the records queried in step 4 always meet the time difference condition. This mechanism efficiently implements the image filtering logic that "the difference between the collection time and the preset collection time is less than the first preset duration".

[0072] In this embodiment, by integrating image information from multiple cameras with the same alarm type under the same work process, the blind spots or false alarm interference of a single camera are overcome, and the authenticity of the alarm is comprehensively judged, thereby significantly improving the accuracy of security alarms.

[0073] S220: Based on the camera number, alarm type and acquisition time of multiple images in the loading and unloading operation process of hazardous chemicals, perform grouping and clustering processing on multiple images in the loading and unloading operation process of hazardous chemicals to obtain a multi-type image set in the loading and unloading operation process of hazardous chemicals. Specifically, step S220 includes: S221, determine that multiple images belonging to the same alarm type and whose difference between the collection time and the preset collection time is less than the first preset time are the same type of image set under the closing process of hazardous chemical loading and unloading operations; S222, based on the acquisition time of multiple images in various image sets, sort the multiple images in various image sets from first to last; S223, determine the target images of each image set whose acquisition time is less than the acquisition time of adjacent images and the acquisition time of each image set is less than the second preset time. S224, determine the target images of multiple image sets belonging to the same camera number as the same image set, so as to obtain multiple image sets under the post-operation closing process of hazardous chemical loading and unloading.

[0074] In this embodiment, by focusing on the temporal continuity (based on a second preset duration for selecting target images) and alarm type consistency under a single camera's perspective, the temporal characteristics of key operational nodes during the oil unloading process are accurately captured. This method effectively eliminates interference caused by differences in cross-camera perspectives and filters out discontinuous or isolated instantaneous false alarm images, significantly improving the accuracy of alarm authenticity judgment for the specific process stage of oil unloading completion.

[0075] S230: Based on the camera number, alarm type, and acquisition time of multiple images in the post-operation cleanup process of hazardous chemical loading and unloading, aggregate the multiple images in the post-operation cleanup process of hazardous chemical loading and unloading to obtain a multi-type image set in the post-operation cleanup process of hazardous chemical loading and unloading.

[0076] Specifically, step S230 includes: S231, determine that images belonging to the same alarm type and the same camera number under the loading and unloading operation process of hazardous chemicals are the same image set; S232, based on the acquisition time of multiple images in various image sets, sort the multiple images in various image sets from first to last; S233, determine the images whose acquisition time of each type of image set is less than the acquisition time of adjacent images and the difference is less than the second preset time as the target images of each type of image set, so as to use the target images as multiple image sets under the loading and unloading operation process of hazardous chemicals loading and unloading.

[0077] For ease of understanding, a specific implementation method is given below: Step 1 (Configuration Management): The configuration table for aggregated alarm information is managed through the web-based configuration function. Key information in the table includes: camera number, alarm type, and aggregation time threshold t.

[0078] Step 2 (Alarm Reception and Grouping): The system receives alarm information for each image in real time.

[0079] Each alarm carries an alarm type / AI tag and a camera number.

[0080] Grouping rules: The system assigns alarms to unique alarm groups based on a combination of alarm type and camera number (i.e., group identifier).

[0081] Step 3 (Alarm Sequence Processing within Groups): For image alarms in each alarm group: (1) Time sequence sorting: Based on the time of alarm acquisition, all images in the group are sorted from first to last.

[0082] (2) Continuity determination and target image set generation (e.g.) Figure 2 (as shown) Traverse the sorted image sequence and calculate the time difference Δt between the acquisition time of the current image and the next adjacent image.

[0083] If Δt <= the configured time threshold t, then the two images are considered to belong to the same continuous abnormal event segment, and the current image is included in the target image set (or continuous alarm set).

[0084] If Δt > t, it marks the end of the current continuous event segment. The system outputs the current target image set that has been constructed (this set contains all continuous images that satisfy the condition Δt <= t), and uses the next image as the starting point for the new target image set.

[0085] Step 4 (Output target image set): When a continuous set of target images is constructed (i.e., when Δt>t is encountered or the buffer times out), the set is output.

[0086] Step 5 (Cache Management): Used to temporarily store the target image set being built in each alarm group (i.e., the latest state of the current consecutive segments).

[0087] Set a TTL (Time To Live) for each cached segment to ensure that the current segment is forcibly output if no new alarms arrive for a long time.

[0088] There is a cache expiration clearing mechanism that periodically cleans up expired cache fragments (TTL expires).

[0089] In this embodiment, by integrating image information from the same camera and with the same alarm type, it is automatically aggregated into a "class image set" and only a single alarm entry is output, eliminating the "alarm storm" caused by a massive number of duplicate alarms, avoiding the flooding of critical risk information, and thus improving alarm efficiency.

[0090] In this embodiment, by combining the advantages of multi-camera perspective integration and time series analysis, the blind spots of a single camera's perspective and the interference of false alarms are effectively overcome. At the same time, by intelligently aggregating alarm entries, the "alarm storm" caused by a large number of duplicate alarms is avoided. Thus, while ensuring the authenticity of alarms, the accuracy of security alarms and the efficiency of alarm processing are significantly improved.

[0091] S300, match each operation process of hazardous chemical loading and unloading with a pre-built operation process alarm type combination table to obtain the alarm type combination of each operation process of hazardous chemical loading and unloading; wherein, the pre-built operation process alarm type is used to characterize the mapping relationship between different operation processes of hazardous chemical loading and unloading and the corresponding alarm type combination. For ease of understanding, the following are examples of pre-built combinations of job workflow alarm types, as shown in Table 1 below: Table 1. Alarm Type Combination Table for Workflow

[0092] S400, if the alarm type of at least one type of image set under each operation process of hazardous chemical loading and unloading belongs to the alarm type combination of each operation process of hazardous chemical loading and unloading, then the alarm types of multiple types of image sets under each operation process of hazardous chemical loading and unloading are displayed.

[0093] To make it easier to understand, the following examples are provided: If the hazardous chemical loading and unloading operation process is a pre-operation preparation process, there are three types of image sets under this process. The alarm types of these three types of image sets are personnel safety distance alarm, protective equipment missing alarm, and environmental leakage warning alarm. If the alarm type combination of each operation process of hazardous chemical loading and unloading is personnel safety distance alarm + vehicle not grounded alarm + hose / arming arm abnormal status alarm + environmental leakage warning alarm, it means that there are alarm types of two types of image sets in the alarm type combination of each operation process of hazardous chemical loading and unloading. Therefore, the output displays the alarm types of the three types of image sets (personnel safety distance alarm, protective equipment missing alarm, and environmental leakage warning alarm) under the pre-operation preparation process of hazardous chemical loading and unloading.

[0094] The beneficial effects of this invention are: (1) By dividing the image sources according to the work process and performing correlation processing based on camera number, alarm type and acquisition time, the safety monitoring needs of the complex and multi-stage loading and unloading of hazardous chemicals can be effectively met, ensuring coverage of key operation nodes.

[0095] (2) By grouping images of the same alarm type into a set of images, the granularity of alarm display is increased from a single image to an alarm type group, thereby reducing the number of duplicate alarm entries.

[0096] (3) By matching the work process with the predefined alarm type combination, alarm types that are not related to the current work process are automatically filtered out (such as ignoring electrostatic alarms during the oil transfer stage), which greatly reduces invalid alarm interference and improves the relevance of alarm information.

[0097] (4) By using the pre-built "job process-alarm type combination" mapping rule engine, the complex multi-source alarm association logic is transformed into rule configuration. Users can define or adjust these mapping rules according to the actual job specifications. No deep code development is required, which significantly reduces the complexity of business logic, maintenance costs and subsequent expansion difficulty.

[0098] Example 2 Based on the same inventive concept, such as Figure 3 As shown, this embodiment of the invention also provides a safety monitoring device 200 for hazardous chemical loading and unloading operations, comprising: The data acquisition module 210 is used to acquire multiple images under multiple operational processes of hazardous chemical loading and unloading, and to determine the alarm type of multiple images under multiple operational processes of hazardous chemical loading and unloading; wherein, each image is marked with a corresponding camera number and acquisition time; The image association module 220 is used to perform image association processing on multiple images under each operation process of hazardous chemical loading and unloading based on the camera number, alarm type and acquisition time of multiple images under each operation process of hazardous chemical loading and unloading, so as to obtain multiple image sets under each operation process of hazardous chemical loading and unloading; wherein, multiple images in each image set have the same alarm type. The combination matching module 230 is used to match each operation process of hazardous chemical loading and unloading with a pre-built operation process alarm type combination table to obtain the alarm type combination of each operation process of hazardous chemical loading and unloading; wherein, the pre-built operation process alarm type is used to characterize the mapping relationship between different operation processes of hazardous chemical loading and unloading and the corresponding alarm type combination. The alarm verification module 240 is used to output and display the alarm types of multiple image sets under each operation process of hazardous chemical loading and unloading when the alarm type of at least one image set under each operation process of hazardous chemical loading and unloading belongs to the combination of alarm types under each operation process of hazardous chemical loading and unloading.

[0099] Optionally, the multiple operational processes for loading and unloading hazardous chemicals include: pre-operation preparation process, loading and unloading operation phase process, and post-operation cleanup process; image association processing includes: combination processing, aggregation processing, and clustering processing; The image association module 220 is specifically used for: Based on the camera number, alarm type, and acquisition time of multiple images in the pre-operation preparation process of hazardous chemical loading and unloading, the multiple images in the pre-operation preparation process of hazardous chemical loading and unloading are combined and processed to obtain a multi-type image set in the pre-operation preparation process of hazardous chemical loading and unloading. Based on the camera number, alarm type, and acquisition time of multiple images in the loading and unloading operation process of hazardous chemicals, the multiple images in the loading and unloading operation process of hazardous chemicals are grouped and clustered to obtain a multi-type image set in the loading and unloading operation process of hazardous chemicals. Based on the camera number, alarm type, and acquisition time of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, the multiple images from the post-operation cleanup process of hazardous chemical loading and unloading are aggregated to obtain a multi-type image set of the post-operation cleanup process of hazardous chemical loading and unloading.

[0100] It should be understood that this device corresponds to the above-described safety monitoring method embodiment for hazardous chemical loading and unloading operations, and is capable of executing the various steps involved in the above method embodiment. The specific functions of this device can be found in the description above, and detailed descriptions are omitted here to avoid repetition. The device includes at least one software functional module that can be stored in memory or embedded in the device's operating system (OS) in the form of software or firmware.

[0101] Example 3 Based on the same inventive concept, embodiments of the present invention also provide an electronic device, including: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and when the machine-readable instructions are executed by the processor, the above-described safety monitoring method for hazardous chemical loading and unloading operations is performed.

[0102] In a typical configuration, an electronic device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0103] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0104] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0105] Example 4 Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing computer instructions, which, when executed on a computer, cause the computer to perform the aforementioned safety monitoring method for hazardous chemical loading and unloading operations.

[0106] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0111] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0112] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0113] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for safety monitoring of hazardous chemical loading and unloading operations, characterized in that, include: Acquire multiple images from various operational processes during the loading and unloading of hazardous chemicals, and determine the alarm types for these images. Each image is labeled with its corresponding camera number and acquisition time. Based on the camera number, alarm type, and acquisition time of multiple images under each operational process of hazardous chemical loading and unloading, image association processing is performed on the multiple images under each operational process of hazardous chemical loading and unloading to obtain multiple image sets under each operational process of hazardous chemical loading and unloading; among them, multiple images in each image set have the same alarm type. The alarm type combination of each operation process of hazardous chemical loading and unloading is matched with the pre-built operation process alarm type combination table to obtain the alarm type combination of each operation process of hazardous chemical loading and unloading; wherein, the pre-built operation process alarm type is used to characterize the mapping relationship between different operation processes of hazardous chemical loading and unloading and the corresponding alarm type combination. If the alarm type of at least one image set under each operation process of hazardous chemical loading and unloading is a component of the alarm type combination of each operation process of hazardous chemical loading and unloading, then the output displays the alarm types of multiple image sets under each operation process of hazardous chemical loading and unloading.

2. The safety monitoring method for hazardous chemical loading and unloading operations according to claim 1, characterized in that, Identify the alarm types for multiple images across various operational processes during the handling of hazardous chemicals, including: Multiple images from each operational process of hazardous chemical loading and unloading are input into the alarm type analysis model to obtain the alarm types of multiple images from each operational process of hazardous chemical loading and unloading. The alarm type analysis model is trained on a pre-built neural network model by the alarm types corresponding to multiple images from each operational process of hazardous chemical loading and unloading.

3. The safety monitoring method for hazardous chemical loading and unloading operations according to claim 1, characterized in that, The multiple operational processes for loading and unloading hazardous chemicals include: pre-operation preparation process, loading and unloading operation phase process, and post-operation cleanup process; image association processing includes: combination processing, aggregation processing, and clustering processing. Based on the camera IDs, alarm types, and acquisition times of multiple images from various operational processes in hazardous chemical loading and unloading, image association processing is performed on these images for each operational process, resulting in multiple image sets for each operational process in hazardous chemical loading and unloading, including: Based on the camera number, alarm type, and acquisition time of multiple images in the pre-operation preparation process of hazardous chemical loading and unloading, the multiple images in the pre-operation preparation process of hazardous chemical loading and unloading are combined and processed to obtain a multi-type image set in the pre-operation preparation process of hazardous chemical loading and unloading. Based on the camera number, alarm type, and acquisition time of multiple images in the loading and unloading operation process of hazardous chemicals, the multiple images in the loading and unloading operation process of hazardous chemicals are grouped and clustered to obtain a multi-type image set in the loading and unloading operation process of hazardous chemicals. Based on the camera number, alarm type, and acquisition time of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, the multiple images from the post-operation cleanup process of hazardous chemical loading and unloading are aggregated to obtain a multi-type image set of the post-operation cleanup process of hazardous chemical loading and unloading.

4. The safety monitoring method for hazardous chemical loading and unloading operations according to claim 3, characterized in that, Based on the camera IDs, alarm types, and acquisition times of multiple images from the pre-operation preparation process for hazardous chemical loading and unloading, these images are combined to obtain multiple image sets for the pre-operation preparation process for hazardous chemical loading and unloading, including: Images belonging to the same alarm type and whose acquisition time is less than the first preset acquisition time under the pre-operation preparation process for hazardous chemical loading and unloading are identified as the same image set, so as to obtain multiple image sets under the pre-operation preparation process for hazardous chemical loading and unloading.

5. The safety monitoring method for hazardous chemical loading and unloading operations according to claim 3, characterized in that, Based on the camera IDs, alarm types, and acquisition times of multiple images from the loading and unloading operation phases of hazardous chemicals, clustering processing is performed on these images to obtain multiple image sets for the hazardous chemicals loading and unloading operation phases, including: Multiple images belonging to the same alarm type and whose difference between the collection time and the preset collection time is less than the first preset time are identified as the same image set after the completion process of hazardous chemical loading and unloading operations. Based on the acquisition time of multiple images in various image sets, the multiple images in various image sets are sorted from first to last; Multiple images whose acquisition time is less than the acquisition time of adjacent images in each image set are identified as target images for each image set. Target images belonging to multiple image sets with the same camera number are identified as the same image set to obtain multiple image sets under the post-operation cleanup process for hazardous chemical loading and unloading.

6. The safety monitoring method for hazardous chemical loading and unloading operations according to claim 3, characterized in that, Based on the camera IDs, alarm types, and acquisition times of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, these images are aggregated to obtain multiple image sets for the post-operation cleanup process of hazardous chemical loading and unloading, including: Images belonging to the same alarm type and the same camera number under the loading and unloading operation process of hazardous chemicals are identified as the same image set; Based on the acquisition time of multiple images in various image sets, the multiple images in various image sets are sorted from first to last; Images whose acquisition time is less than the acquisition time of adjacent images are identified as target images for each image set, so that the target images can be used as multiple image sets in the loading and unloading operation process of hazardous chemicals.

7. A safety monitoring device for hazardous chemical loading and unloading operations, characterized in that, include: The data acquisition module is used to acquire multiple images under multiple operational processes of hazardous chemical loading and unloading, and to determine the alarm type of multiple images under multiple operational processes of hazardous chemical loading and unloading; each image is marked with a corresponding camera number and acquisition time; The image association module is used to perform image association processing on multiple images under each operation process of hazardous chemical loading and unloading based on the camera number, alarm type and acquisition time of multiple images under each operation process of hazardous chemical loading and unloading, to obtain multiple image sets under each operation process of hazardous chemical loading and unloading; among them, multiple images in each image set have the same alarm type. The combination matching module is used to match each operation process of hazardous chemical loading and unloading with a pre-built operation process alarm type combination table to obtain the alarm type combination of each operation process of hazardous chemical loading and unloading; wherein, the pre-built operation process alarm type is used to characterize the mapping relationship between different operation processes of hazardous chemical loading and unloading and the corresponding alarm type combination. The alarm verification module is used to output and display the alarm types of multiple image sets under each operation process of hazardous chemical loading and unloading when the alarm type of at least one image set under each operation process of hazardous chemical loading and unloading belongs to the combination of alarm types under each operation process of hazardous chemical loading and unloading.

8. The safety monitoring device for hazardous chemical loading and unloading operations according to claim 7, characterized in that, The multiple operational processes for loading and unloading hazardous chemicals include: pre-operation preparation process, loading and unloading operation phase process, and post-operation cleanup process; image association processing includes: combination processing, aggregation processing, and clustering processing. The image association module is specifically used for: Based on the camera number, alarm type, and acquisition time of multiple images in the pre-operation preparation process of hazardous chemical loading and unloading, the multiple images in the pre-operation preparation process of hazardous chemical loading and unloading are combined and processed to obtain a multi-type image set in the pre-operation preparation process of hazardous chemical loading and unloading. Based on the camera number, alarm type, and acquisition time of multiple images in the loading and unloading operation process of hazardous chemicals, the multiple images in the loading and unloading operation process of hazardous chemicals are grouped and clustered to obtain a multi-type image set in the loading and unloading operation process of hazardous chemicals. Based on the camera number, alarm type, and acquisition time of multiple images from the post-operation cleanup process of hazardous chemical loading and unloading, the multiple images from the post-operation cleanup process of hazardous chemical loading and unloading are aggregated to obtain a multi-type image set of the post-operation cleanup process of hazardous chemical loading and unloading.

9. An electronic device, characterized in that, include: A processor and a memory, the memory storing machine-readable instructions executable by the processor, which, when executed by the processor, perform the safety monitoring method for hazardous chemical loading and unloading operations as described in any one of claims 1-6.

10. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed on the computer, the computer performs the safety monitoring method for hazardous chemical loading and unloading operations as described in any one of claims 1-6.

Images