Chemical production site on-line inspection method and system based on closed-circuit television system

Abstract

Embodiments of the present specification provide a closed-circuit television system-based online inspection method, device and system for a chemical production site. In the closed-circuit television system-based online inspection method for the chemical production site, a target image for shooting each to-be-collected image area of the chemical production site is acquired from a closed-circuit television system; each target image is stored in a first path with a respective name according to a first name designation rule and a first path designation rule; an inspection score for each to-be-collected image area is generated according to a comparison result between each target image and a matched comparison reference image, wherein the matched comparison reference image is determined according to the name of the target image; and each generated inspection score is stored in a file in a second path accessible by a client with a respective name according to a second name designation rule and a second path designation rule.

Description

Technical Field

[0001] The embodiments in this specification generally relate to the field of image monitoring and processing technology in chemical production scenarios, and particularly to online inspection methods, devices and systems for chemical production sites based on closed-circuit television systems. Background Technology

[0002] In the chemical production field, traditional inspection methods typically rely on regular manual inspections, which are often inefficient and unable to conduct high-frequency, comprehensive, and detailed checks on large-scale chemical production sites, leading to the failure to detect potential problems in a timely manner. Furthermore, human judgment is susceptible to subjective factors, resulting in insufficient objectivity in inspection results. Other methods involve deploying various sensors around chemical production equipment to collect relevant data and transmit it to a monitoring center for analysis and judgment to determine whether the equipment's operating status and environmental conditions are normal. Closed-circuit television systems often require personnel to be stationed at the control center to constantly monitor images transmitted from the production site. However, with the expansion of chemical production scale and the advancement of automation technology, a more efficient, objective, and labor-saving online inspection solution is needed. Summary of the Invention

[0003] In view of the above, embodiments of this specification provide an online inspection method, apparatus, and system for chemical production sites based on a closed-circuit television system. Using this online inspection scheme, target images acquired on demand are first stored in an orderly manner according to a designed first name specification rule and a first path specification rule. Then, an inspection score is generated for the target images using comparison reference images determined based on the target image names. The inspection score is then stored in a file accessible to the client using a second name specification rule and a second path specification rule. This not only enables rapid, objective, and timely inspection of the image areas to be acquired, but also provides a technical foundation for remote monitoring terminals or other related systems to obtain inspection result data in a timely manner. It facilitates data sharing and collaborative work among multiple systems, lays the foundation for improving the overall management level and emergency response capabilities of chemical production sites, and ensures the safe and stable operation and production efficiency of chemical production.

[0004] According to one aspect of the embodiments of this specification, an online inspection method for a chemical production site based on a closed-circuit television system is provided, comprising: acquiring target images captured from the closed-circuit television system for each image area to be acquired at the chemical production site; storing each target image with its own name in a first path according to a first name specification rule and a first path specification rule; generating an inspection score for each image area to be acquired based on the comparison results between each target image and a matching comparison reference image, wherein the matching comparison reference image is determined according to the name of the target image; and storing each generated inspection score with its own name in a file accessible to a client in a second path according to a second name specification rule and a second path specification rule.

[0005] According to another aspect of the embodiments of this specification, an online inspection device for a chemical production site based on a closed-circuit television system is provided, comprising: an image acquisition module configured to acquire target images captured from the closed-circuit television system for each image area to be collected at the chemical production site; an image storage module configured to store each target image with its own name in a first path according to a first name specification rule and a first path specification rule; a score generation module configured to generate an inspection score for each image area to be collected based on the comparison result between each target image and a matching comparison reference image, wherein the matching comparison reference image is determined according to the name of the target image; and a file storage module configured to store each generated inspection score with its own name in a file accessible to a client in a second path according to a second name specification rule and a second path specification rule.

[0006] According to another aspect of the embodiments of this specification, an online inspection system for a chemical production site based on a closed-circuit television system is provided, comprising: a closed-circuit television system for acquiring and storing images captured for various image areas to be acquired at the chemical production site; a server for executing the online inspection method for a chemical production site based on the closed-circuit television system as described above; and a client for extracting corresponding inspection scores from various files under a second path; determining area identifiers corresponding to each extracted inspection score according to a name-area identifier lookup table matching the second name specification rule; and presenting the online inspection results in the form of area identifier-inspection score.

[0007] According to another aspect of the embodiments of this specification, an online inspection device for a chemical production site based on a closed-circuit television system is provided, comprising: at least one processor, and a memory coupled to the at least one processor, the memory storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the online inspection method for a chemical production site based on a closed-circuit television system as described above. Attached Figure Description

[0008] A further understanding of the nature and advantages of this specification can be achieved by referring to the following figures. In the figures, similar components or features may have the same reference numerals.

[0009] Figure 1 An exemplary architecture of a method, apparatus, and system for online inspection of chemical production sites based on a closed-circuit television system, according to embodiments of this specification, is shown.

[0010] Figure 2 A flowchart illustrating an example of an online inspection method for a chemical production site based on a closed-circuit television system, according to an embodiment of this specification, is shown.

[0011] Figure 3 A flowchart illustrating an example of the process for generating inspection scores according to an embodiment of this specification is shown.

[0012] Figure 4 A flowchart illustrating yet another example of an online inspection method for a chemical production site based on a closed-circuit television system, according to an embodiment of this specification, is shown.

[0013] Figure 5 A flowchart illustrating an example of a similarity determination process according to embodiments of this specification is shown.

[0014] Figure 6 A flowchart illustrating an example of a similarity determination process according to embodiments of this specification is shown.

[0015] Figure 7 A block diagram of an example of an online inspection device for a chemical production site based on a closed-circuit television system, according to an embodiment, is shown.

[0016] Figure 8 A signaling diagram illustrating an example of interaction between various devices in an online inspection system for a chemical production site according to an embodiment of this specification is shown.

[0017] Figure 9 A schematic diagram of an example of an online inspection device for a chemical production site based on a closed-circuit television system, as described in this specification, is shown. Detailed Implementation

[0018] The subject matter described herein will be discussed below with reference to exemplary embodiments. It should be understood that these embodiments are discussed merely to enable those skilled in the art to better understand and implement the subject matter described herein, and are not intended to limit the scope, applicability, or examples set forth in the claims. The function and arrangement of the elements discussed may be changed without departing from the scope of the embodiments described herein. Various processes or components may be omitted, substituted, or added as needed in the various examples. Furthermore, features described in some examples may be combined in other examples.

[0019] As used herein, the term "comprising" and its variations are open terms meaning "including but not limited to". The term "based on" means "at least partially based on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. Unless explicitly indicated by the context, the definition of a term shall remain consistent throughout the specification.

[0020] The flowcharts used in this specification illustrate operations implemented according to some embodiments of this specification. It should be clearly understood that the operations in the flowcharts may not be implemented in a sequential order. Instead, the operations may be implemented in reverse order or simultaneously. Furthermore, one or more additional operations may be added to the flowcharts. One or more operations may be removed from the flowcharts.

[0021] In view of this, the embodiments of this specification propose an online inspection scheme for chemical production sites based on a closed-circuit television system. Using this online inspection scheme, target images acquired on demand are first stored in an orderly manner according to a designed first name specification rule and a first path specification rule. Then, an inspection score is generated for the target images using comparison reference images determined by the target image names. The inspection score is then stored in a file accessible to the client using a second name specification rule and a second path specification rule. This not only enables rapid, objective, and timely inspection of the image areas to be acquired, but also provides a technical foundation for remote monitoring terminals or other related systems to obtain inspection result data in a timely manner. It facilitates data sharing and collaborative work among multiple systems, lays the foundation for improving the overall management level and emergency response capabilities of chemical production sites, and ensures the safe and stable operation and production efficiency of chemical production.

[0022] The following will describe in detail, with reference to the accompanying drawings, the online inspection method, apparatus and system for chemical production sites based on closed-circuit television systems according to embodiments of this specification.

[0023] Figure 1An exemplary architecture 100 of a chemical production site online inspection method, apparatus, and system based on a closed-circuit television system, according to an embodiment of this specification, is shown.

[0024] exist Figure 1 In this context, network 110 is used to interconnect the closed-circuit television (CCTV) system 120 and the application server 130.

[0025] Network 110 can be any type of network capable of interconnecting network entities. Network 110 can be a single network or a combination of various networks. In terms of coverage, network 110 can be a local area network (LAN), a wide area network (WAN), etc. In terms of the carrying medium, network 110 can be a wired network, a wireless network, etc. In terms of data switching technology, network 110 can be a circuit-switched network, a packet-switched network, etc.

[0026] A closed-circuit television (CCTV) system 120 can refer to an image communication system capable of transmitting video over a specific area. In some examples, the CCTV system 120 may include several components such as image acquisition equipment, image transmission equipment, image control equipment, and image processing, display, and recording equipment. In some examples, the image acquisition equipment may be a camera. Different CCTV cameras can be set up in designated locations according to the layout of the chemical plant, the production process, and the functional characteristics of different areas. In one example, in a production workshop, high-definition cameras can be set up in key areas such as critical production equipment, material storage areas, and operating passages. For example, cameras with high resolution (such as 4K), wide dynamic range, and low-light capabilities can be selected to adapt to the complex lighting environment of chemical plants (such as production workshops with strong light and dimly lit warehouse corners), thereby ensuring image clarity and preserving image details. Alternatively, cameras with high stability and durability can be selected to operate stably for extended periods in the special environment of chemical plants (such as the presence of chemical gases and dust).

[0027] Server 130 can access CCTV system 120 via network 110, thereby acquiring images collected by CCTV system 120. Furthermore, server 130 can generate corresponding inspection scores based on the acquired images and store the inspection scores in a corresponding file under a specified path.

[0028] In one implementation, the terminal device used by the user ( Figure 1(Not shown) can access server 130 and read file contents under the specified path. In some examples, the terminal device can be any type of electronic computing device capable of accessing server 130, processing data or signals, etc. For example, terminal device 120 can be a laptop, tablet, smartphone, etc.

[0029] It should be understood that Figure 1 All network entities shown are exemplary, and any other network entities may be involved in Architecture 100 depending on the specific application requirements.

[0030] Figure 2 A flowchart illustrating an example of an online inspection method 200 for a chemical production site based on a closed-circuit television system, according to an embodiment of this specification, is shown.

[0031] like Figure 2 As shown in Figure 210, target images are captured from the closed-circuit television system for each image area to be acquired at the chemical production site.

[0032] In this embodiment, the image area to be acquired typically needs to include key areas of concern in the chemical production site. For example, key areas of concern may include areas containing important production equipment, material storage areas, workstations, finished product packaging areas, emergency exit areas, etc. In some examples, the closed-circuit television system can acquire real-time images of each image area to be acquired in the chemical production site. In some examples, the closed-circuit television system can store the acquired images of each image area to be acquired in the chemical production site, along with the corresponding acquisition time. In some examples, the acquired images can be obtained by remotely accessing the closed-circuit television system.

[0033] In some examples, images captured from a closed-circuit television system can be sequentially acquired for different areas of the chemical production site based on various methods, such as predetermined time intervals (e.g., every 15 minutes), predetermined event triggers (e.g., detecting personnel or vehicles entering a designated area), or predetermined commands (e.g., receiving a command from a client to initiate online inspection). For example, the acquired images can be directly used as the target images for the corresponding areas. Alternatively, the acquired images can be pre-processed (e.g., denoising, contrast enhancement), and the pre-processed images can be used as the target images for the corresponding areas.

[0034] In some implementations, simulated mouse operations can be used to sequentially retrieve and capture images of various areas within a chemical production site from a closed-circuit television (CCTV) system. Mouse operations can include at least one of the following: movement, clicking, and scrolling. In some examples, automation scripting software can be used to program tasks to simulate mouse movement, clicking, and scrolling. For instance, mouse wheel operations can be simulated to switch cameras, and clicking the screenshot button can be simulated to capture the image currently being captured by the camera. This allows for automated acquisition of target images from the CCTV system.

[0035] In step 220, each target image is stored in the first path with its own name, according to the first name specification rule and the first path specification rule.

[0036] In this embodiment, the first name specification rule can be set according to actual needs. For example, it can be named according to the camera number or the area identifier of the target image area. The area identifier can be, for example, an area code. Alternatively, the name of the target image can be determined based on information such as the image acquisition time, location, and area number, for example, in the format "Acquisition Time_Location Number_Area Number.jpg". In some examples, the first path can be a designated server folder or local disk path specifically used to store the target image.

[0037] In step 230, based on the comparison results between each target image and the matching reference image, an inspection score is generated for each image region to be acquired.

[0038] In this embodiment, the comparison reference image can be an image of the image area to be acquired when it meets the inspection standards, such as an image with a perfect inspection score. In some examples, the comparison reference image can correspond one-to-one with the image area to be acquired, that is, different image areas to be acquired can correspond to different comparison reference images. In some examples, different comparison reference images can be distinguished by region identifiers. In some examples, for key areas in a chemical production site, such as a large equipment area, there can be a corresponding set of comparison reference images. Each set of comparison reference images can contain multiple comparison reference images taken from different angles. In some examples, the required number of target images and viewing angles can be determined based on the importance and complexity of the image area to be acquired.

[0039] In this embodiment, comparison reference images corresponding to each image region to be acquired can be pre-stored. After acquiring the target image, the matching comparison reference image can be determined based on the name of the target image. For example, if the name of the target image contains a region number, the comparison reference image corresponding to the image region to be acquired indicated by that region number can be determined as the matching comparison reference image.

[0040] In this embodiment, the degree of difference between the target image and the matching comparison reference image can be determined in various ways, and an inspection score can be generated based on the degree of difference. In some examples, the target image and the matching comparison reference image can be compared pixel by pixel, and the inspection score can be determined by combining the comparison results of each pixel. In some examples, the target image and the matching comparison reference image can be preprocessed, such as normalization or grayscale processing. Then, the structural similarity index (SSIM) between the preprocessed target image and the matching comparison reference image can be determined, and the inspection score can be determined accordingly. It can be understood that, generally, the smaller the degree of difference between the target image and the matching comparison reference image, the higher the inspection score.

[0041] In step 240, according to the second name specification rule and the second path specification rule, each generated inspection score is stored with its own name in a file on a second path that can be accessed by the client.

[0042] In this embodiment, the second naming rule can adopt a naming convention similar to that of the target image to facilitate association and querying. For example, the name of the inspection score can be in the format of "area identifier.txt". Another example is that the name of the inspection score can be in the format of "area number_inspection time.txt" or "area number_inspection time_score value.txt". The inspection scores can be written to the corresponding file. In some examples, the second path can be specified as a network shared folder or a specific table field in a database, so that clients can easily access and query the inspection scores.

[0043] It should be understood that all steps and their order in process 200 are exemplary, and embodiments of this disclosure will also cover modifications to process 200 in any manner. For example, although Figure 2 Step 230 shown is executed after step 220, but it can also be executed before step 220.

[0044] The following is for reference. Figure 3 , Figure 3 A flowchart illustrating an example of an inspection score generation process 300 according to an embodiment of this specification is shown. Process 300 may be... Figure 2An exemplary implementation of step 230 in the example.

[0045] like Figure 3 As shown in 310, the similarity between each target image and the matched comparison reference image is determined.

[0046] In this embodiment, various image similarity algorithms can be used to determine the similarity between the target image and the matching reference image. For example, methods based on color feature histograms, differences in corresponding pixel values, or structural similarity indices can be used to determine the similarity between the target image and the matching reference image. Alternatively, a similarity calculation method based on extracted feature vectors can be used to determine the similarity between the target image and the matching reference image. In one example, feature vectors corresponding to the similarity between the target image and the matching reference image can be extracted separately. For instance, algorithms such as SURF (Speeded Up Robust Feature), SIFT (Scale-invariant feature transform), and ORB (Oriented FAST and Rotated BRIEF) can be used to extract feature points and construct feature vectors corresponding to the target image and the matching reference image, respectively. Then, the similarity between the target image and the matching reference image is determined based on the similarity metric between the two constructed feature vectors (e.g., Euclidean distance, cosine similarity, etc.).

[0047] In step 320, the determined similarity is calibrated using calibration coefficients corresponding to each target image to obtain the calibrated similarity score.

[0048] In this embodiment, the calibration coefficients can be determined by fitting the similarity to historical images and the corresponding benchmark inspection scores. For example, historical scoring data corresponding to each comparison reference image can be collected in advance. For each comparison reference image, each historical scoring data can consist of the similarity determined for the comparison reference image and the corresponding historical image, and the corresponding manually annotated scoring data. For example, for image region 1 to be collected, images collected regularly every day over the past month can be selected as historical images. For each historical image and the comparison reference image of image region 1 to be collected, the similarity can be determined and manually annotated scoring data can be obtained. In some examples, the determined similarity can be fitted with the corresponding scoring data using the collected historical scoring data. For example, each similarity can be denoted as x, the corresponding scoring data as y, and the regression equation y = ax + b can be fitted using the least squares method to determine the values ​​of the calibration coefficients a and b for the comparison reference image. Then, the values ​​of a and b of the calibration system can be used to convert the newly obtained similarity of the target image corresponding to the comparison reference image into a calibrated similarity score.

[0049] In some examples, the binning principle can be used to divide the determined similarity range into several intervals (bins), and then the mean or median of the score data within each interval can be calculated as the calibration coefficient of the comparison reference image in each interval. For example, the similarity range determined by comparing each historical image of the image region 1 to the image region 1 can be divided into [0,10), [10,20), ..., [90,100). Taking the interval [60,70) as an example, the average (or median) of the manually labeled score data corresponding to the historical images with similarity in this interval can be calculated. For example, if the average is 67, then the calibration coefficient for this interval can be 67. Similarly, the calibration coefficients for each interval can be obtained. Thus, when the similarity of a newly obtained target image corresponding to the comparison reference image falls into the corresponding interval, the calibration coefficient of that interval can be determined as the calibrated similarity score.

[0050] It is understandable that, based on the above method, the calibration coefficient corresponding to each comparison reference image can be determined. Once the target image is determined, the corresponding calibration coefficient can be determined based on the matched comparison reference image.

[0051] At 330, a corresponding inspection score is generated based on each similarity score.

[0052] In this embodiment, the inspection score corresponding to each similarity score can be determined according to a pre-defined correspondence table between similarity scores and inspection scores. For example, the similarity scores can be divided into several intervals, each interval corresponding to a specific inspection score. In some examples, other factors can also be combined to adjust the inspection score. For example, if specific abnormal features are detected in the target image, such as the presence of smoke or obstacles, the inspection score can be reduced accordingly. In some examples, a pre-defined mapping function can be used to normalize the inspection score, for example, the inspection score can be mapped to the [0, 100] interval.

[0053] Through the above methods, this scheme introduces calibration coefficients to make the inspection score fit the manually labeled standard to a certain extent. Compared with the supervised training process of machine learning models, it is simpler, more intuitive and easier to implement. This helps to make the inspection score more objectively reflect the relationship between the similarity score of the image and the actual production situation, and provides strong support for the safety management of chemical production.

[0054] Continue to refer to Figure 4 , Figure 4 A flowchart is shown as yet another example of an online inspection method 400 for a chemical production site based on a closed-circuit television system, according to an embodiment of this specification.

[0055] like Figure 4 As shown, at 410, target images are captured from the closed-circuit television system for each image area to be acquired at the chemical production site.

[0056] In some implementations, the comparison reference images may have corresponding capture times. In some examples, the capture time of the comparison reference images can serve as an important reference factor in determining when to trigger the acquisition of the target image. Acquiring a target image of a region to be imaged at a chemical production site can be performed in response to the current time matching the capture time of the matching comparison reference image. In some examples, when the system detects that the current time matches the capture time of a comparison reference image, it can automatically trigger the acquisition of the target image from the closed-circuit television system. For example, if the comparison reference image was captured within a specific time period (e.g., 20:00-21:00), the target image can be acquired within the same time period to ensure the accuracy of the comparison. In some examples, a certain time tolerance can also be set, meaning that if the current time is close to the capture time of the comparison reference image within a certain range, it is also considered a match, and the acquisition of the target image is performed. For example, the time tolerance can be set to ±10 minutes.

[0057] In some implementations, acquiring a target image of a specific area within a chemical production site can be performed in response to sensor data indicating an anomaly in that area. In these implementations, chemical production sites often have multiple sensors used to collect data such as temperature, pressure, liquid level, and gas leaks. In some examples, these sensors can be integrated with a closed-circuit television (CCTV) system and equipment for performing an online inspection method for the chemical production site based on the CCTV system (e.g., [missing information]). Figure 1 The server 130 in the middle establishes a data connection. In some examples, the closed-circuit television system can acquire sensor data at the corresponding time while acquiring images, and can synchronize and associate the image data of each image area to be acquired with the sensor data of that image area in time.

[0058] In some examples, when a sensor detects abnormal data, it can immediately trigger the acquisition of a target image of the corresponding image area from the closed-circuit television system to promptly understand the situation on site. For example, if a temperature sensor detects an excessively high temperature, a pressure sensor detects an abnormal pressure change, or a waste liquid tank detects that the liquid level is approaching a warning threshold, the aforementioned target image acquisition operation can be automatically triggered. In some examples, different target image acquisition priorities can be preset for different levels of sensor anomalies. For example, for severe anomalies, target images are acquired first for emergency handling; for general anomalies, target images can be acquired within a certain time range (e.g., 10 minutes). In this case, step 420 can be executed first, followed by step 410.

[0059] At 420, for each target image, sensor data from each sensor in the corresponding image area to be acquired is collected at the time the target image is captured.

[0060] In some examples, current sensor data can be acquired directly at the moment the target image was captured. In some examples, sensor data can be monitored and recorded in real time. In some examples, the capture time can be determined based on the timestamp of the target image, and the corresponding sensor data can be extracted accordingly.

[0061] At 430, each target image is stored in the first path with its own name, according to the first name specification rule and the first path specification rule.

[0062] At 440, the similarity between each target image and the matched reference image is determined.

[0063] At 450, the determined similarity is calibrated using calibration coefficients corresponding to each target image to obtain the calibrated similarity score.

[0064] At 460, a corresponding inspection score is generated based on each similarity score and the corresponding sensor data.

[0065] In this embodiment, similarity scores and sensor data can be combined in various ways to generate inspection scores. In some examples, different weights can be pre-assigned to similarity scores and different types of sensor data, and then a weighted sum can be calculated as the inspection score. In some examples, the inspection score can be dynamically adjusted based on the changing trends of the sensor data. For example, if sensor data shows that the condition of a certain area is gradually deteriorating, a certain score can be reduced accordingly based on the similarity score to obtain the inspection score for that area. In some examples, a machine learning model can be pre-trained, and the inspection score corresponding to the target image can be obtained by inputting the similarity scores and corresponding sensor data into the machine learning model.

[0066] At 470, in accordance with the second name specification rule and the second path specification rule, each generated inspection score is stored with its own name in a file on a second path that can be accessed by the client.

[0067] It should be noted that steps 410 and 430-450 above can be referred to in the previous sections. Figure 2 Steps 210-220 in the embodiment and Figure 3 The corresponding descriptions of steps 310-320 in the embodiments will not be repeated here. It should also be noted that step 420 can be executed before steps 430-450, after any of steps 430-450, or before step 410.

[0068] By using the above methods, this solution can more accurately assess the on-site conditions and improve the safety and reliability of chemical production by determining the similarity between the target image obtained from the closed-circuit television system and the comparison reference image, and by combining sensor data for comprehensive analysis.

[0069] The following is for reference. Figure 5 , Figure 5 A flowchart illustrating an example of a similarity determination process 500 according to embodiments of this specification is provided. Process 500 may be... Figure 3 An exemplary implementation of step 310 in the above. Process 500 can be performed for each target image.

[0070] like Figure 5 As shown in Figure 510, local texture features are extracted for each part of the target image.

[0071] In this embodiment, various texture feature extraction algorithms can be used, such as Gray-Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP). For example, GLCM can be used to extract texture roughness, contrast, and other features from different regions in various target images. In some examples, different weights can be pre-set according to the importance of different local regions in the target image. For example, for local images involving critical equipment or high-risk areas, a larger weight coefficient can be applied when extracting texture features. In some examples, the extracted local texture features can also be preprocessed, such as normalization, to eliminate scale differences between different localities. For example, the extracted texture feature values ​​can be mapped to specific intervals to make the texture features of different localities comparable. By combining the GLCM and LBP methods, typical scenarios such as the texture state of the production equipment surface (e.g., whether there is wear, scratches, etc.) and the cleanliness texture of the ground (e.g., whether there are stains, debris accumulation, or other special textures) can be effectively identified.

[0072] In step 520, the extracted local texture features are fused together to obtain the texture feature vector of the target image.

[0073] In this embodiment, methods such as weighted averaging and principal component analysis (PCA) can be used to fuse the extracted local texture features, thereby reducing the dimensionality of multiple local texture feature vectors and fusing them into a more representative texture feature vector. In some examples, the fusion method can be dynamically adjusted based on the reliability of different local texture features. For example, if there is significant noise interference (i.e., low reliability) during the extraction of certain local texture features, their weight in the fusion can be reduced. In some examples, the fused texture feature vector can be further optimized, such as removing redundant features. For example, feature selection algorithms can be used to select the most valuable features for similarity judgment from the fused texture feature vector.

[0074] At 530, a layout feature vector of the target image is generated based on the geometry detected from the target image.

[0075] In this embodiment, edge detection algorithms (such as Canny edge detection) and Hough transform can be used to detect geometric shapes in the image. For example, edge detection algorithms can be used to extract the contours of objects in the image, and then Hough transform can be used to detect geometric shapes such as straight lines and circles (such as circular chemical containers, rectangular shelves, etc.), thereby constructing a layout feature vector. In some examples, different weights can be assigned to different types of geometric shapes to highlight important layout features. For example, regular-shaped equipment in a chemical production site can be given higher weights, while irregularly shaped background areas can be given lower weights. In some examples, the geometric parameters (such as area, perimeter, aspect ratio, etc.) of the detected geometric shapes can be further calculated, and the calculated geometric parameters can be used as elements in the constructed layout feature vector.

[0076] At 540, extract the color features of the target image.

[0077] In this embodiment, color features can be extracted using methods such as color histograms and color moments. In some examples, a feature histogram extraction method based on the HSV color space can be used, which divides the hue, saturation, and brightness of the image into multiple intervals and counts the distribution of the number of pixels in each interval, thereby forming a color feature histogram.

[0078] At 550, the extracted texture feature vector, layout feature vector, and color feature are compared with the corresponding features of the matching reference image to obtain the texture feature comparison result, layout feature comparison result, and color feature comparison result.

[0079] In this embodiment, methods such as distance metrics and similarity functions can be used for comparison. In some examples, the Euclidean distance between the texture feature vectors of the target image and the comparison reference image can be used as the texture feature comparison result. Alternatively, the texture feature comparison result can be obtained based on the difference between the intersection distance of the LBP histograms and the GLCM feature parameters. In some examples, the Euclidean distance between the layout feature vectors of the target image and the comparison reference image can be used as the layout feature comparison result. Furthermore, the geometric parameter differences (such as area difference, perimeter difference, etc.) between the extracted shape and the corresponding shape extracted from the comparison reference image can be calculated to obtain a layout feature comparison result that incorporates the geometric parameter differences. In some examples, the Bhattacharyya distance between the color feature histograms of the target image and the comparison reference image can be used as the color feature comparison result.

[0080] In some examples, the results of texture feature matching, layout feature matching, and color feature matching can be normalized. For example, the results of texture feature matching, layout feature matching, and color feature matching can be mapped to the [0,1] interval, with the closer to 1 representing greater similarity.

[0081] In step 560, based on the fusion result between the obtained texture feature comparison results, layout feature comparison results, and color feature comparison results, the similarity between the target image and the matched comparison reference image is determined.

[0082] In this embodiment, methods such as weighted summation and machine learning models can be used for fusion. For example, a neural network model for outputting a comprehensive similarity score can be pre-trained, taking texture feature matching results, layout feature matching results, and color feature matching results as input, and outputting the similarity between the target image and the matching reference image. In some examples, the aforementioned neural network model may include an adaptive weight module based on an attention mechanism, which can dynamically generate weights for various feature matching results.

[0083] By employing the methods described above, this solution can comprehensively utilize multi-dimensional features of the target image, including texture, layout, and color characteristics. This allows for a more comprehensive and accurate assessment of the similarity between the target image and the comparison reference image. It is particularly suitable for monitoring equipment operating status in complex environments at chemical production sites, checking production process compliance, and implementing 5S (Sort, Set in order, Shine, Standardize, and Sustain) management. For example, changes in the texture of equipment may indicate wear, corrosion, or material adhesion. It can also be used to identify the cleanliness and smoothness of equipment surfaces to determine whether sorting and cleaning work is adequate. Furthermore, differences in layout features can visually demonstrate whether the placement of equipment, tools, and materials meets the requirements of sorting, whether space utilization is reasonable, and can reflect key information such as whether equipment has been moved or whether pipe connections are normal. Color features help distinguish the functional positioning of different areas and the completeness of warning signs, examining the implementation of cleaning and sustaining practices. Changes in color features may also indicate potential hazards such as chemical leaks or discoloration caused by abnormal temperatures.

[0084] Continue to refer to Figure 6 , Figure 6 A flowchart illustrating an example of a similarity determination process 600 according to an embodiment of this specification is provided. Inspection scoring may include 5S management scoring. Process 600 may be... Figure 5Another exemplary implementation of step 310 in the above. Procedure 600 is applicable to target image groups corresponding to various image regions of a predetermined type. Procedure 600 can be performed for each target image group.

[0085] In this embodiment, for a predetermined type of image region to be acquired, target images captured simultaneously from different perspectives can be obtained to form a corresponding target image group. For example, for a key area in a chemical production site, such as a large equipment area, target images can be acquired simultaneously from multiple cameras positioned at different angles. In some examples, the required number of target images and perspectives can be determined based on the importance and complexity of the image region to be acquired. In some examples, the target image group can also be labeled and classified. For example, the target image group can be classified and stored according to information such as the image's shooting angle and time, facilitating quick retrieval and use.

[0086] like Figure 6 As shown in Figure 610, a three-dimensional model is constructed for the corresponding image region to be acquired based on the target image group.

[0087] In this embodiment, 3D reconstruction algorithms (such as structured light and stereo vision) can be used to fuse images from different perspectives in a target image set to construct a 3D model. For example, stereo vision can be used to recover the 3D shape and spatial position of an object by calculating the disparity of corresponding points in different images. In some examples, the constructed 3D model can be optimized and simplified to improve processing efficiency and visualization effects. For example, unnecessary details and noise in the 3D model can be removed, while retaining key structures and features. In some examples, the 3D model can also be rendered and annotated according to actual needs for better display and analysis. For example, key parts of the device can be annotated, or different areas can be color-coded to highlight specific information.

[0088] At 620, determine the spatial position of the target entity in the 3D model.

[0089] In this embodiment, the spatial position of a target entity can be determined by analyzing information such as the object's coordinates and orientation in the 3D model. In some examples, the spatial position may include the position coordinates and tilt angle of chemical equipment in 3D space. In other examples, historical data can be used to analyze trends in the spatial position of the target entity. For example, by comparing 3D models at different points in time, the changing trends of the spatial position of chemical equipment can be analyzed, thereby predicting potential problems in advance.

[0090] At 630, based on whether the determined spatial location status conforms to the corresponding 5S management rules, a spatial score is generated for the target image group.

[0091] In this embodiment, the rules of 5S management, such as sorting, straightening, sweeping, cleaning, and sustaining, can be pre-converted into specific spatial location status indicators. Then, a spatial score is generated based on the degree to which the actual spatial location of the target entity conforms to these indicators. For example, if chemical equipment is neatly positioned and oriented in a 3D model, meeting the requirements of straightening, a higher spatial score can be awarded. In some examples, different weights can be set according to different 5S management rules to reflect their importance in the spatial score. For example, for areas with high safety requirements in chemical production sites, the weights of cleaning and sustaining rules can be relatively high. In some examples, specific analysis and feedback can be provided for spatial location states that do not conform to 5S management rules, so that corresponding improvement measures can be taken. For example, if a certain area has a low cleaning score, the specific dirty and messy areas and the problems can be marked.

[0092] At 640, a feature extraction process is performed for each target image in the target image group.

[0093] In this embodiment, the above feature extraction process can refer to the corresponding description of steps 510-540 in the previous embodiment, and will not be repeated here.

[0094] At 650, for each target image in the target image group, the extracted texture feature vector, layout feature vector, and color feature are compared with the corresponding features of the matching reference image to obtain the texture feature comparison result, layout feature comparison result, and color feature comparison result of each target image.

[0095] The above step 650 can be referred to the corresponding description of step 550 in the foregoing embodiments, and will not be repeated here.

[0096] At 660, based on the fusion result between the texture feature comparison results, layout feature comparison results, and color feature comparison results of each target image in the target image group, the feature comprehensive comparison result of the target image group is determined.

[0097] In this embodiment, the texture feature comparison results, layout feature comparison results, and color feature comparison results of each target image in the target image group can be fused in various ways (such as weighted summation, using machine learning models, etc.) to finally obtain a feature comprehensive comparison result, which is used to evaluate the comprehensive similarity between the target image group and the corresponding comparison reference image group in terms of the extracted features.

[0098] At 670, based on the determined feature integration comparison results and the generated spatial score, the similarity between the target image group and the matched comparison reference image group is determined.

[0099] In this embodiment, the feature integration comparison results and spatial scores can be weighted and summed to obtain the similarity between the target image group and the comparison reference image group. In some examples, the weights of the feature integration comparison results and spatial scores can be adjusted according to the needs of the actual application scenario. For example, in the visual inspection of a chemical production site, the weight of the feature integration comparison results can be increased; while in the evaluation of site layout and space management, the weight of the spatial score can be increased.

[0100] Using the methods described above, this solution can combine the construction of a 3D model of a multi-view target image group, feature extraction, and spatial score evaluation to analyze the situation at the chemical production site more comprehensively and accurately.

[0101] use Figures 1-6 The publicly disclosed online inspection method for chemical production sites based on closed-circuit television (CCTV) systems provides fully automated operations for image acquisition, storage, comparison, and scoring, enabling real-time and efficient inspection of chemical production sites. This allows for the timely detection of potential problems and safety hazards, improving the safety, standardization, and management efficiency of chemical production. Furthermore, by comprehensively considering image texture, layout, color features, and spatial location, and combining calibration systems and other technical means, the online inspection method for chemical production sites can be further optimized and improved.

[0102] Figure 7 A block diagram of an example of an online inspection device 700 for a chemical production site based on a closed-circuit television system, according to an embodiment, is shown.

[0103] The online inspection device 700 may include: an image acquisition module 710, configured to acquire target images of various image areas to be collected at a chemical production site from a closed-circuit television system; an image storage module 720, configured to store each target image with its own name in a first path according to a first name specification rule and a first path specification rule; a score generation module 730, configured to generate an inspection score for each image area to be collected based on the comparison results between each target image and a matching comparison reference image, wherein the matching comparison reference image is determined according to the name of the target image; and a file storage module 740, configured to store each generated inspection score with its own name in a file accessible to a client in a second path according to a second name specification rule and a second path specification rule.

[0104] In addition, the online inspection device 700 may also include any other modules configured to perform any operation of the online inspection method for chemical production sites based on a closed-circuit television system according to the embodiments of the present disclosure described above.

[0105] Reference above Figures 1 to 7 This specification describes embodiments of the online inspection method and apparatus for chemical production sites based on closed-circuit television systems.

[0106] Figure 8 A signaling diagram illustrating an example of interaction between various devices in an online inspection system 800 for a chemical production site according to an embodiment of this specification is shown.

[0107] like Figure 8 As shown, in 810, the closed-circuit television system acquires and stores images captured from various areas of the chemical production site.

[0108] At 820, the server acquires target images from the closed-circuit television system for each area of ​​the chemical production site to be captured.

[0109] At 830, the server stores each target image in the first path with its own name, according to the first name specification rule and the first path specification rule.

[0110] In step 840, the server generates an inspection score for each image region to be acquired based on the comparison results between each target image and the matching reference image.

[0111] In 850, the server stores each generated inspection score with its own name in a file accessible to the client in a second path, according to the second name specification rule and the second path specification rule.

[0112] It should be noted that steps 810-850 above can be referred to the aforementioned... Figures 1-6 The corresponding steps of the online inspection method for chemical production sites based on closed-circuit television systems described in the embodiments will not be repeated here.

[0113] In 860, the client extracts the corresponding inspection scores from the various files under the second path.

[0114] In this embodiment, the client can establish a communication connection with the server storing the inspection scores through a pre-configured network connection and corresponding permission verification mechanism. This allows direct access to files in the second path, and extraction of various inspection score data according to the file format specifications of the second path. In some examples, the client can perform preliminary filtering and organization of the extracted inspection score data based on actual needs. For example, only inspection scores corresponding to specific time periods or specific types of image areas to be collected can be extracted for more efficient subsequent analysis and display.

[0115] In 870, the client determines the area identifier corresponding to each extracted inspection score based on a name-area identifier lookup table that matches the second name specification rule.

[0116] In this embodiment, a name-region identifier lookup table can be maintained internally on the client side. In some examples, this name-region identifier lookup table can be synchronously obtained from the server during system initialization and updated periodically. The region identifier matching the filename corresponding to each extracted inspection score can be determined based on the name-region identifier lookup table. For example, the name of the inspection score file may contain a specific coded field, and the client can determine the corresponding region identifier based on the correspondence between the coded field and the region identifier in the name-region identifier lookup table. In some examples, the name-region identifier lookup table can also have a corresponding index. For example, an index can be created according to rules such as the first letter of the name or a range of numbers to accelerate the search process, especially when the data volume is large, which can significantly improve the speed of determining the region identifier.

[0117] On the 880, the client displays online inspection results in the form of area identifier-inspection score.

[0118] In this embodiment, the client can use an intuitive and clear visual interface to display the online inspection results. For example, the area identifiers and their corresponding inspection scores can be listed in a table, and different colors can be used to indicate the level of the inspection scores, making it easy for managers to quickly understand the situation of each area. For example, green can indicate a high score and good condition, while red can indicate a low score and potential problems.

[0119] In some examples, the client can also provide chart display functionality. For instance, a bar chart can be used to compare inspection scores in different areas, or a line chart can be used to show the changing trend of inspection scores in a specific area over different time periods, allowing managers to more intuitively understand the dynamic changes and overall status of various areas in the chemical production site.

[0120] In some examples, the client can implement interactive functions. When an administrator clicks on a certain area icon, more detailed image information and historical inspection data of that area can be displayed, which facilitates in-depth analysis of the problems in that area or observation of the improvement effect.

[0121] Through the methods described above, this solution provides a complete approach to automating, informatizing, and visualizing the on-site inspection process in chemical production. By leveraging the image acquisition capabilities of a closed-circuit television system and the image storage, comparison, and analysis functions of a server, inspection scores can be generated efficiently and accurately. Furthermore, through a series of data extraction, identifier matching, and result presentation operations on the client side, managers can conveniently and intuitively grasp the actual situation in various areas of the chemical production site, promptly identify potential problem areas, and make targeted decisions. This effectively improves the safety, standardization, and overall management efficiency of chemical production, ensuring its stable and orderly operation.

[0122] Reference above Figures 1 to 8 This specification describes embodiments of a method, apparatus, and system for online inspection of chemical production sites based on closed-circuit television systems.

[0123] The online inspection device for chemical production sites based on a closed-circuit television system, as described in the embodiments of this specification, can be implemented in hardware, software, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of its host device reading the corresponding computer program instructions from the memory into the main memory and executing them. In the embodiments of this specification, the online inspection device for chemical production sites based on a closed-circuit television system can, for example, be implemented using electronic equipment.

[0124] Figure 9 A schematic diagram of an example of an online inspection device 900 for a chemical production site based on a closed-circuit television system, according to an embodiment of this specification, is shown.

[0125] like Figure 9 As shown, the online inspection device 900 for chemical production sites based on a closed-circuit television system may include at least one processor 910, a memory (e.g., non-volatile memory) 920, a main memory 930, and a communication interface 940, and the at least one processor 910, memory 920, main memory 930, and communication interface 940 are connected together via a bus 950. The at least one processor 910 executes at least one computer-readable instruction (i.e., the elements implemented in software described above) stored or encoded in the memory.

[0126] In one embodiment, computer-executable instructions are stored in memory, which, when executed, cause at least one processor 910 to: acquire target images of various image areas to be acquired at a chemical production site from a closed-circuit television system; store each target image with its respective name in a first path according to a first naming rule and a first path naming rule; generate inspection scores for each image area to be acquired based on comparison results between each target image and a matching comparison reference image, wherein the matching comparison reference image is determined according to the name of the target image; and store each generated inspection score with its respective name in a file accessible to a client in a second path according to a second naming rule and a second path naming rule.

[0127] It should be understood that the computer-executable instructions stored in memory, when executed, cause at least one processor 910 to perform the above-described combinations in the various embodiments of this specification. Figures 1-6 The description includes various operations and functions.

[0128] Specifically, a system or apparatus equipped with a readable storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer or processor of the system or apparatus can read and execute the instructions stored in the readable storage medium.

[0129] In this case, the program code itself, which can be read from a readable medium, can perform the functions of any of the above embodiments. Therefore, the machine-readable code and the readable storage medium storing the machine-readable code constitute a part of the present invention.

[0130] The computer program code required for the operation of each part of this manual can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB, .NET, and Python; conventional procedural programming languages ​​such as C, Visual Basic 2003, Perl, COBOL 2002, PHP, and ABAP; dynamic programming languages ​​such as Python, Ruby, and Groovy; or other programming languages. This program code can run on the user's computer, or as a standalone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer via any network, such as a local area network (LAN) or wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service, such as Software as a Service (SaaS).

[0131] Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer or the cloud via a communication network.

[0132] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0133] Not all steps and units in the above process and system structure diagrams are mandatory; some steps or units can be omitted as needed. The execution order of each step is not fixed and can be determined as required. The device structure described in the above embodiments can be a physical structure or a logical structure. That is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities, or they may be jointly implemented by certain components in multiple independent devices.

[0134] The term "exemplary" as used throughout this specification means "serving as an example, instance, or illustration" and does not imply that it is "preferred" or "advantageous" over other embodiments. Detailed descriptions are included for the purpose of providing an understanding of the described techniques. However, these techniques may be practiced without these detailed descriptions. In some instances, well-known structures and apparatuses are shown in block diagram form to avoid obscuring the concepts of the described embodiments.

[0135] The optional embodiments of the present specification have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present specification are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present specification, various simple modifications can be made to the technical solutions of the embodiments of the present specification, and these simple modifications all fall within the protection scope of the embodiments of the present specification.

[0136] The foregoing description of this specification is provided to enable any person skilled in the art to implement or use the content of this specification. Various modifications to the content of this specification will be apparent to those skilled in the art, and the general principles defined herein can be applied to other variations without departing from the scope of protection of this specification. Therefore, this specification is not limited to the examples and designs described herein, but is consistent with the widest scope of the principles and novel features disclosed herein.

Claims

1. A method for online inspection of chemical production sites based on a closed-circuit television system, comprising: Acquire target images of various areas to be captured at the chemical production site from the closed-circuit television system; According to the first name specification rule and the first path specification rule, each target image is stored in the first path with its own name; Based on the comparison results between each target image and the matched comparison reference image, an inspection score is generated for each image region to be acquired. The matched comparison reference image is determined according to the name of the target image. According to the second naming rule and the second path rule, each generated inspection score is stored with its own name in a file accessible to the client on a second path. The inspection score includes a 5S management score, and the generation of inspection scores for each image area to be acquired includes: Determine the similarity between each target image and the matched comparison reference image; and generate the corresponding inspection score based on the similarity. The target image includes target image groups corresponding to various image regions of a predetermined type. Each target image group includes target images captured at the same time from different perspectives targeting the image regions of the predetermined type. Determining the similarity between each target image and the matching comparison reference image includes: Based on each target image group, construct a 3D model for each corresponding image region to be acquired; For each 3D model, determine the spatial position state of the target entity in that 3D model; Based on whether the determined spatial location status conforms to the corresponding 5S management rules, a spatial score is generated for the target image group. The spatial score is determined based on the degree of conformity between the spatial location status of the target entity and the spatial location status index transformed according to the 5S management rules. For each target image in each target image group, a feature extraction process is performed to obtain the texture feature vector, layout feature vector, and color features of each target image. The layout feature vector is generated based on the geometric shape detected from the target image. The extracted texture feature vector, layout feature vector, and color features are compared with the corresponding features of the matching reference image to obtain the texture feature comparison result, layout feature comparison result, and color feature comparison result of each target image. For each target image group, the feature comprehensive comparison result of the target image group is determined based on the fusion result between the texture feature comparison result, layout feature comparison result and color feature comparison result of each target image in the target image group; and the similarity between the target image group and the matching comparison reference image group is determined based on the determined feature comprehensive comparison result and the generated spatial score.

2. The online inspection method as described in claim 1, wherein, The acquisition of target images from the closed-circuit television system for each area of ​​the chemical production site includes: The closed-circuit television system sequentially retrieves images of various areas to be captured at the chemical production site by simulating mouse operations, and then performs image cropping. The mouse operations include at least one of the following: moving, clicking, and scrolling.

3. The online inspection method as described in claim 1, wherein, The step of generating the corresponding inspection score based on the similarity includes: The determined similarity is calibrated using calibration coefficients corresponding to each target image to obtain a calibrated similarity score. The calibration coefficients are determined by fitting the similarity scores to historical images and their corresponding benchmark inspection scores. Based on the similarity scores, the corresponding inspection score is generated.

4. The online inspection method as described in claim 3, wherein, The online inspection method also includes: For each target image, sensor data from each sensor in the corresponding image region at the time the target image was captured is collected. The process of generating corresponding inspection scores based on each similarity score includes: Based on the similarity scores and the corresponding sensor data, a corresponding inspection score is generated.

5. The online inspection method as described in claim 4, wherein, The comparison reference image has a corresponding shooting time. Acquiring the target image of the area to be captured at the chemical production site is performed in response to the current time matching the shooting time of the corresponding comparison reference image. The process of acquiring target images of a region in a chemical production site is executed in response to sensor data indicating an anomaly in that region.

6. The online inspection method as described in any one of claims 3 to 5, wherein, The feature extraction process includes: For each target image, Local texture features are extracted for each part of the target image; The extracted local texture features are fused together to obtain the texture feature vector of the target image; Based on the detected geometry from the target image, a layout feature vector of the target image is generated; and Extract the color features of the target image.

7. An online inspection device for chemical production sites based on a closed-circuit television system, comprising: The image acquisition module is configured to acquire target images from the closed-circuit television system for each image area to be captured at the chemical production site. The image storage module is configured to store each target image in the first path with its own name according to the first name specification rule and the first path specification rule; The scoring generation module is configured to generate inspection scores for each image region to be acquired based on the comparison results between each target image and a matching comparison reference image, wherein the matching comparison reference image is determined according to the name of the target image; and The file storage module is configured to store each generated inspection score, with its own name, in a file accessible to the client, according to a second naming rule and a second path rule. The inspection score includes a 5S management score, and the score generation module is further configured to: determine the similarity between each target image and the matched comparison reference image; and generate a corresponding inspection score based on the similarity. The target image includes target image groups corresponding to various image regions of a predetermined type. Each target image group includes target images captured at the same time from different perspectives targeting the image regions of the predetermined type. Determining the similarity between each target image and the matching comparison reference image includes: Based on each target image group, construct a 3D model for each corresponding image region to be acquired; For each 3D model, determine the spatial position state of the target entity in that 3D model; Based on whether the determined spatial location status conforms to the corresponding 5S management rules, a spatial score is generated for the target image group. The spatial score is determined based on the degree of conformity between the spatial location status of the target entity and the spatial location status index transformed according to the 5S management rules. For each target image in each target image group, a feature extraction process is performed to obtain the texture feature vector, layout feature vector, and color features of each target image. The layout feature vector is generated based on the geometric shape detected from the target image. The extracted texture feature vector, layout feature vector, and color features are compared with the corresponding features of the matching reference image to obtain the texture feature comparison result, layout feature comparison result, and color feature comparison result of each target image. For each target image group, the feature comprehensive comparison result of the target image group is determined based on the fusion result between the texture feature comparison result, layout feature comparison result and color feature comparison result of each target image in the target image group; and the similarity between the target image group and the matching comparison reference image group is determined based on the determined feature comprehensive comparison result and the generated spatial score.

8. An online inspection system for chemical production sites, comprising: Closed-circuit television systems are used to acquire and store images captured from various areas of a chemical production site. A server is used to execute the online inspection method for chemical production sites based on a closed-circuit television system as described in any one of claims 1 to 6. as well as The client is used to extract the corresponding inspection scores from the files in the second path. Based on the name-region identifier lookup table that matches the second name specification rule, determine the region identifier corresponding to each extracted inspection score; and present the online inspection results in the form of region identifier-inspection score.

9. An online inspection device for chemical production sites based on a closed-circuit television system, comprising: At least one processor, a memory coupled to the at least one processor, and a computer program stored on the memory, wherein the at least one processor executes the computer program to implement the online inspection method for chemical production sites based on a closed-circuit television system as described in any one of claims 1 to 6.

Images