A pipeline tank leakage identification method and system based on machine vision

Abstract

The present invention provides a pipeline tank leakage identification method and system based on machine vision, which relates to the field of data identification technology. The method includes: controlling a compensation acquisition unit according to basic distribution data, generating video acquisition control data from the compensation acquisition result, exporting video acquisition data, performing auxiliary angle background information acquisition through the compensation acquisition unit, generating background noise features from the acquisition results and then performing frame subtraction processing on the video acquisition data, performing feature recognition on the frame-subtracted video acquisition data through an abnormal feature recognition database, performing leakage identification verification after obtaining recognition results of data in different formats, and outputting the identification verification results. The present invention solves the technical problem in the prior art of low recognition accuracy when a pipeline tank leaks due to the limitations of manual recognition and detection of pipeline tank leaks, realizes accurate recognition of pipeline tank leaks based on machine vision, and thereby improves the recognition accuracy when a pipeline tank leaks.

Description

Technical Field

[0001] The present invention relates to the technical field of data recognition, and in particular to a pipeline tank leakage recognition method and system based on machine vision. Background Art

[0002] With the development of chemical plants, especially the existing pipelines and tanks in chemical plants, leakage in liquid transport pipelines has become a major problem in large chemical plants. Pipeline damage not only affects the normal operation of the plant, but also increases maintenance costs. In addition, it can threaten the safety of operators. Therefore, the detection and location of pipeline leaks are key tasks in maintenance and condition monitoring.

[0003] However, due to the limitations of manual detection of pipeline and tank leaks in the prior art, there is a technical problem of low accuracy in identifying leaks in pipeline and tank bodies. Summary of the Invention

[0004] The present application provides a pipeline tank leakage identification method and system based on machine vision, which is used to solve the technical problem in the prior art of low recognition accuracy when pipeline tank leakage exists due to the limitations of manual pipeline tank leakage identification and detection.

[0005] In view of the above problems, the present application provides a pipeline tank leakage identification method and system based on machine vision.

[0006] In the first aspect, the present application provides a method for identifying pipeline tank leakage based on machine vision, the method comprising: a basic data set of a connected pipeline tank, and interactive basic distribution data; controlling the compensation acquisition unit to perform compensation acquisition of pipeline data according to the basic distribution data, and generating video acquisition control data according to the compensation acquisition result; controlling the video acquisition unit to perform video acquisition based on the video acquisition control data, and exporting video acquisition data, wherein the video acquisition data includes at least two formats of data; performing auxiliary angle background information acquisition through the compensation acquisition unit, and generating background noise features based on the acquisition results; performing frame subtraction processing on the video acquisition data through the background noise features, and performing feature recognition on the video acquisition data after frame subtraction through an abnormal feature recognition database; obtaining recognition results of data in different formats, and performing leakage recognition verification based on the recognition results, and outputting recognition verification results.

[0007] In the second aspect, the present application provides a pipeline tank leakage identification system based on machine vision, the system comprising: a data acquisition module, the data acquisition module being used to connect the basic data set of the pipeline tank and exchange basic distribution data; a compensation acquisition module, the compensation acquisition module being used to control the compensation acquisition unit to perform compensation acquisition of pipeline data according to the basic distribution data, and generate video acquisition control data according to the compensation acquisition result; a video acquisition module, the video acquisition module being used to control the video acquisition unit to perform video acquisition based on the video acquisition control data, and export video acquisition data, wherein the video acquisition data includes at least two formats of data; an information acquisition module, the information acquisition module being used to perform auxiliary angle background information acquisition through the compensation acquisition unit, and generate background noise features according to the acquisition results; a feature recognition module, the feature recognition module being used to perform frame subtraction processing of the video acquisition data through the background noise features, and perform feature recognition on the video acquisition data after frame subtraction through an abnormal feature recognition database; a leakage identification verification module, the leakage identification verification module being used to obtain recognition results of data in different formats, and perform leakage identification verification based on the recognition results, and output recognition verification results.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The present application provides a pipeline tank leakage identification method and system based on machine vision, which relates to the field of data recognition technology. It solves the technical problem in the prior art of low recognition accuracy when pipeline tank leakage exists due to the limitations of manual recognition and detection of pipeline tank leakage, and realizes accurate recognition of pipeline tank leakage based on machine vision, thereby improving the recognition accuracy when pipeline tank leakage exists. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flow chart of a pipeline tank leakage identification method based on machine vision is provided for this application;

[0011] Figure 2 This application provides a schematic diagram of the frame subtraction processing flow in a pipeline tank leakage identification method based on machine vision;

[0012] Figure 3 This application provides a flow chart of abnormal verification results in a pipeline tank leakage identification method based on machine vision;

[0013] Figure 4 This application provides a schematic diagram of an abnormality recognition optimization process in a pipeline tank leakage recognition method based on machine vision;

[0014] Figure 5A schematic diagram of the structure of a pipeline tank leakage identification system based on machine vision is provided for this application.

[0015] Explanation of the accompanying symbols: data acquisition module 1, compensation acquisition module 2, video acquisition module 3, information acquisition module 4, feature recognition module 5, leakage recognition and verification module 6. Implementation Method

[0016] The present application provides a pipeline tank leakage identification method and system based on machine vision to solve the technical problem in the prior art of low recognition accuracy when pipeline tank leakage exists due to the limitations of manual pipeline tank leakage identification and detection. Example

[0017] like Figure 1 As shown, an embodiment of the present application provides a method for identifying pipeline tank leakage based on machine vision. The method is applied to an intelligent recognition system, and the intelligent recognition system is communicatively connected with a video acquisition unit and a compensation acquisition unit. The method includes:

[0018] Step S100: basic data set of connected pipeline tanks and interactive basic distribution data;

[0019] Specifically, a machine vision-based pipeline tank leakage identification method provided in an embodiment of the present application is applied to an intelligent identification system, which is communicatively connected to a video acquisition unit and a compensation acquisition unit, and the video acquisition unit and the compensation acquisition unit are used to collect environmental parameters.

[0020] In order to ensure the accuracy of leakage caused by damage to the pipeline tank in the later stage, the intelligent identification system first needs to connect with the target pipeline tank and extract the basic data of the connected pipeline tank. The basic data of the pipeline tank may include basic data such as tank volume, tank material, tank pressure, and tank connection. On this basis, the data is summarized and recorded as the basic data set of the pipeline tank at the same time. According to the interactive distribution between the various data in the basic data set, for example, when the target pipeline tank is connected with other pipeline tanks, the data of the tank connection between the two need to be interacted, and then the basic distribution data of the pipeline tank is obtained, which serves as an important reference for the later identification of pipeline tank leakage.

[0021] Step S200: controlling the compensation acquisition unit to perform compensation acquisition of pipeline data according to the basic distribution data, and generating video acquisition control data according to the compensation acquisition result;

[0022] Specifically, since only the basic information of the pipeline tank body is not sufficient to obtain various factors when the pipeline tank body leaks, it is necessary to use the above-mentioned basic distribution data of the pipeline tank body as a benchmark to perform compensation acquisition control on the compensation acquisition unit connected to the intelligent identification system for communication, that is, all data other than the basic information of the pipeline tank body are compensated and acquired through the compensation acquisition unit. The results of the compensated acquisition can include the length of the pipeline tank body, the diameter of the tank body, the integrity of the tank body, etc. Further, based on the obtained compensation acquisition results, video acquisition control data is generated. The video acquisition data can perform video acquisition control on the pipeline tank body according to the tank body length, tank body diameter and tank body integrity that need to be collected in the compensation acquisition results, thereby ensuring the identification of pipeline tank body leakage.

[0023] Step S300: controlling the video acquisition unit to perform video acquisition based on the video acquisition control data, and exporting video acquisition data, wherein the video acquisition data includes data in at least two formats;

[0024] Specifically, the video acquisition control data obtained above is used to perform video acquisition of the pipeline tank by the video acquisition unit to which the intelligent identification system communicates. The video acquisition data includes at least two formats of data. Among them, an infrared camera can be used in the video acquisition unit, and the infrared camera can capture liquid leakage with a temperature higher (or lower) than the ambient temperature. The infrared camera can be used to detect the image reflected by the target pipeline tank when there is light, the image reflected by infrared rays when there is no light on the target pipeline tank, and when the target pipeline tank is damaged and leaking, the tank leakage can also be judged based on the temperature difference collected around the pipeline tank in the image. Furthermore, the images reflected in the two cases are exported from the video acquisition unit and recorded as video acquisition data at the same time, laying a solid foundation for the subsequent identification of pipeline tank leakage.

[0025] Step S400: collecting auxiliary angle background information through the compensation collection unit, and generating background noise characteristics according to the collection results;

[0026] Specifically, in order to improve the accuracy of identifying pipeline tank leakage, on this basis, the compensation collection unit communicated with the intelligent identification system is used to collect pipeline tank background information at multiple other auxiliary angles on the basis of the original collection angle, wherein the multiple other auxiliary angles do not include the original collection angle of the compensation collection unit. The collection results obtained by collecting background information of the pipeline tank at multiple other auxiliary angles refer to all interferences that are not related to the existence or non-existence of the pipeline tank in the occurrence, inspection, measurement or recording of the intelligent identification system, that is, the surrounding environmental noise outside the pipeline tank, and it is recorded as the background noise feature of the pipeline tank, which has a promoting effect on the identification of pipeline tank leakage.

[0027] Step S500: performing frame subtraction processing on the video acquisition data using the background noise feature, and performing feature recognition on the video acquisition data after frame subtraction using an abnormal feature recognition database;

[0028] Specifically, based on the background noise features collected by the auxiliary angle, the video acquisition data collected by the video acquisition unit is subjected to frame subtraction processing. Frame subtraction processing refers to first identifying the acquisition time nodes of different video frames in the video acquisition data, and then associating the background noise features with the frame subtraction nodes in the video acquisition data. Frame subtraction is to subtract two frames of images of the same part in the video acquisition data to highlight the difference between the two frames of images. The noise features in the background noise features are restored according to the auxiliary angle and the video acquisition control data. Furthermore, the frame subtraction processing of the video acquisition data is completed according to the node association and feature restoration processing results. At the same time, feature recognition is performed on the video acquisition data after frame subtraction processing through the abnormal feature recognition database. The abnormal feature recognition database refers to the integration and construction of the abnormal features of the pipeline and tank contained in the big data, and the video acquisition data after frame subtraction processing is matched with the abnormal features in the abnormal feature recognition database, and the successfully matched data is output as the final feature recognition result, so as to be used as reference data for the subsequent identification of pipeline and tank leaks.

[0029] Step S600: obtaining recognition results of data in different formats, performing leakage recognition verification based on the recognition results, and outputting recognition verification results.

[0030] Specifically, after performing feature recognition on the video acquisition data after frame subtraction through the abnormal feature recognition database, the recognition results of data in different formats are obtained. The recognition results of data in different formats can be MPEG format, AVI format, WMV format, F4V format, RMVB format, etc., and the target pipeline tank is identified and verified whether there is a leak based on the recognition results corresponding to different formats. This means that since the video clarity, video frame, and video resolution under different formats are different, the same position of the target pipeline tank is compared based on the videos under different formats, so as to better identify and verify the target pipeline tank with a leak, and then output the recognition verification results, realizing the accurate identification of pipeline tank leaks based on machine vision, thereby improving the recognition accuracy when the pipeline tank has a leak.

[0031] Furthermore, if Figure 2 As shown, step S500 of this application also includes:

[0032] Step S510: Marking the acquisition time node of the video acquisition data to obtain time node data;

[0033] Step S520: Associating the background noise feature with the frame subtraction node of the video acquisition data through the time node data;

[0034] Step S530: generating background correction data according to the auxiliary angle and the video acquisition control data;

[0035] Step S540: performing feature restoration processing of the background noise feature using the background correction data;

[0036] Step S550: completing the frame subtraction processing according to the node association and feature restoration processing results.

[0037] Specifically, in the process of performing frame subtraction processing on the video acquisition data under the condition of the obtained background noise characteristics, it is first necessary to identify the video acquisition time nodes corresponding to different video frames in the video acquisition data, and further, associate the noise characteristics in the background noise characteristics with the corresponding frame subtraction nodes in the video acquisition data through the obtained event node data, which refers to the quantitative analysis of the specific intrinsic relationship embodied by the correlation characteristics between the noise characteristics and the corresponding frame subtraction nodes. At the same time, the characteristics of the background noise characteristics are restored through the background correction data, which means that the background correction data refers to the standard data of the background noise appearing around the target pipeline tank. On this basis, the characteristics of the background noise characteristics are used to restore the background noise to its original state. Finally, the node association and feature restoration processing are used as the processing basis for frame subtraction of the video acquisition data, thereby completing the frame subtraction processing of the video acquisition data of the pipeline tank, and improving the accuracy of the later recognition of pipeline tank leakage.

[0038] Furthermore, step S540 of this application includes:

[0039] Step S541: determining a restoration reference point for the background noise feature according to the auxiliary angle and the video acquisition control data;

[0040] Step S542: extracting the distance between the video acquisition data and the reference point through the restored reference point, and generating a restoration ratio based on the extraction result and the distance ratio of the restored reference point;

[0041] Step S543: performing feature size restoration on the background noise feature using the restoration ratio, and obtaining the feature restoration processing result based on the size restoration result.

[0042] Specifically, the restoration reference point of the background noise feature is determined according to the auxiliary angle and the video acquisition control data. The restoration reference point of the background noise feature refers to the reference feature used to assist in establishing other background noises, and can also assist in defining the position of the background noise. Furthermore, the point distance of the same reference point contained in the video acquisition data is extracted through the determined restoration reference point, which means calculating the distance difference between the two points of the restoration reference point and the reference point at the same position in the video acquisition data. Furthermore, the feature size of the background noise feature is restored in the same proportion through the restoration ratio of the restoration reference point. Finally, the feature size result of the restored background noise feature is recorded as the feature restoration processing result, thereby achieving the technical effect of providing an important basis for the later identification of pipeline tank leakage.

[0043] Furthermore, step S543 of this application includes:

[0044] Step S5431: generating distortion correction data using the auxiliary angle and the restoration reference point;

[0045] Step S5432: performing distortion restoration on the size restoration result based on the distortion correction data;

[0046] Step S5433: Obtain the feature restoration processing result through the distortion restoration result.

[0047] Specifically, in order to ensure the accuracy of the feature restoration processing results, based on the original collection angle, the background information of the pipeline and tank is collected at multiple other auxiliary angles. The geometric correction of the distortion phenomenon caused by the background noise characteristics caused by the auxiliary angles and the restoration reference points during the feature size restoration is recorded as distortion correction data. Furthermore, the size restoration result is subjected to distortion restoration based on the distortion correction data, which means that the distorted data existing in the size restoration result is restored. Finally, the feature restoration processing result is improved according to the restored distortion restoration result, so as to ensure better identification of pipeline and tank leakage in the later stage.

[0048] Furthermore, if Figure 3 As shown, step S600 of this application also includes:

[0049] Step S610: determining whether the identification verification result contains a single abnormal verification result;

[0050] Step S620: when the single abnormal verification result appears at any node, a non-abnormal frame image of the corresponding node is extracted through the video acquisition data;

[0051] Step S630: authenticating and identifying the non-abnormal frame image through the abnormal feature recognition database;

[0052] Step S640: If the authentication and identification is passed, the single abnormal verification result is corrected to an abnormal verification result with re-inspection passed.

[0053] Specifically, in order to ensure the accuracy of the identification verification results output by the leakage identification verification based on the identification results, it is necessary to judge whether there is a single abnormal verification result in the output identification verification result. The single abnormal verification result refers to the situation where only one format of data is abnormal in the identification results of different format data obtained. Furthermore, when a single abnormal verification result appears at any node, the non-abnormal frame image of the node corresponding to the single abnormal verification result is extracted in the video acquisition data, and the non-abnormal frame image extracted from the video acquisition data is authenticated and identified by the constructed abnormal feature recognition database. The authentication and identification means that the non-abnormal situation in the non-abnormal frame image needs to be matched and compared with the abnormal situation in the single abnormal verification result. If the match is successful, it is regarded as the authentication and identification passed. When the authentication and identification passes, the single abnormal verification result is updated and corrected, and the updated and corrected result is recorded as an abnormal verification result with re-inspection and output, so as to achieve more accurate identification of pipeline tank leakage based on the abnormal verification result with re-inspection.

[0054] Furthermore, step S640 of the present application includes:

[0055] Step S641: Setting the verification association time interval;

[0056] Step S642: When the single abnormal verification result appears at any node, extended authentication of the corresponding node is performed according to the verification associated time interval;

[0057] Step S643: When the extended authentication result meets the preset threshold, the single abnormal verification result is corrected to an abnormal verification result with re-verification passing.

[0058] Specifically, since it is necessary to set a time limit within a certain range when verifying an anomaly to avoid a timeout in the anomaly verification, a verification-related time interval is set. The verification-related time interval can be set to 2 hours. If it is within 2 hours, the verification is deemed valid. If it exceeds 2 hours, the verification is deemed invalid. When a single-body abnormal verification result appears at any time node, the corresponding node is extended and authenticated according to the set verification-related time interval. For example, when the single-body abnormal verification result is preceded and followed by non-single-body authentication, or there are multiple consecutive non-single-body authentications in one direction, etc., it is deemed that the current single-body abnormal verification result meets the preset threshold, wherein the preset threshold is preset by relevant technical personnel based on the current amount of single-body and non-single-body data. Furthermore, when the extended authentication result meets the preset threshold, the single-body abnormal verification result is updated to an abnormal verification result with re-inspection for corresponding output to ensure high efficiency in identifying pipeline tank leaks.

[0059] Furthermore, if Figure 4 As shown, step S700 of this application also includes:

[0060] Step S710: generating a multi-level warning mark according to the identification and verification result;

[0061] Step S720: issuing an early warning prompt through the multi-level early warning identification and receiving feedback verification data, wherein the feedback verification data includes key features of the manual identification;

[0062] Step S730: Optimizing abnormality recognition of the abnormality feature recognition database using the key features.

[0063] Specifically, based on the recognition verification results output by the leakage recognition verification based on the recognition results of data in different formats, a multi-level warning mark is generated for the recognition verification results, that is, the recognition verification of data in different formats corresponds to different levels of warning marks. The more accurate the video reflected by the format data, the lower the warning mark. Furthermore, a warning prompt is given for pipeline tank leakage through a multi-level warning mark, and feedback verification data is received at the same time, and the feedback verification data contains key features of manual identification. If the warning level of the target pipeline tank is high, an warning prompt is given for the leakage of the pipeline tank, and a corresponding verification is made for the leakage of the target pipeline tank in combination with manual verification, so that the pipeline tank abnormality existing in the abnormal feature representation database is improved and optimized based on the key features in the manual verification. For example, if the target pipeline tank is only damaged but not leaking, and there is no corresponding abnormality in the abnormal feature recognition database, when performing manual identification verification, it can be found that it is abnormal and it is judged that there may be leakage in the future, then the corresponding abnormal data in the abnormal feature recognition database is updated, and the improvement and optimization of the abnormal feature recognition database is completed on this basis.

[0064] To sum up, the embodiment of the present application provides a pipeline tank leakage identification method based on machine vision, which includes at least the following technical effects, realizing accurate identification of pipeline tank leakage based on machine vision, thereby improving the identification accuracy when there is leakage in the pipeline tank. Example

[0065] Based on the same inventive concept as the method for identifying pipeline tank leakage based on machine vision in the aforementioned embodiment, Figure 5 As shown, the present application provides a pipeline tank leakage identification system based on machine vision, the system comprising:

[0066] Data acquisition module 1, the data acquisition module 1 is used to connect the basic data set of the pipeline tank and the interactive basic distribution data;

[0067] Compensation acquisition module 2, the compensation acquisition module 2 is used to control the compensation acquisition unit to perform compensation acquisition of pipeline data according to the basic distribution data, and generate video acquisition control data according to the compensation acquisition result;

[0068] A video acquisition module 3, configured to control the video acquisition unit to perform video acquisition based on the video acquisition control data, and to export video acquisition data, wherein the video acquisition data includes data in at least two formats;

[0069] An information acquisition module 4 is configured to acquire auxiliary angle background information through the compensation acquisition unit and generate background noise characteristics based on the acquisition results;

[0070] A feature recognition module 5 is configured to perform frame subtraction processing on the video acquisition data using the background noise feature, and perform feature recognition on the video acquisition data after frame subtraction using an abnormal feature recognition database;

[0071] The leakage identification and verification module 6 is used to obtain identification results of data in different formats, perform leakage identification and verification based on the identification results, and output the identification and verification results.

[0072] Furthermore, the system also includes:

[0073] A time node identification module, the time node identification module is used to identify the acquisition time node of the video acquisition data to obtain time node data;

[0074] a frame subtraction node association module, configured to associate the background noise feature with the frame subtraction node of the video acquisition data using the time node data;

[0075] A background correction data module, the background correction data module is used to generate background correction data according to the auxiliary angle and the video acquisition control data;

[0076] A feature restoration processing module, configured to perform feature restoration processing on the background noise feature using the background correction data;

[0077] A frame subtraction processing module is used to complete the frame subtraction processing according to the node association and feature restoration processing results.

[0078] Furthermore, the system also includes:

[0079] A reference point restoration module, configured to determine a restoration reference point for the background noise feature based on the auxiliary angle and the video acquisition control data;

[0080] a point distance extraction module, configured to extract the point distance between the video acquisition data and the reference point through the restored reference point, and generate a restoration ratio based on the extraction result and the distance ratio of the restored reference point;

[0081] A feature size restoration module is used to restore the feature size of the background noise feature according to the restoration ratio, and obtain the feature restoration processing result based on the size restoration result.

[0082] Furthermore, the system also includes:

[0083] a distortion correction module, configured to generate distortion correction data using the auxiliary angle and the restoration reference point;

[0084] a distortion restoration module, configured to perform distortion restoration on the size restoration result based on the distortion correction data;

[0085] A feature restoration module is used to obtain the feature restoration processing result through the distortion restoration result.

[0086] Furthermore, the system also includes:

[0087] A first judgment module, the first judgment module is used to judge whether there is a single abnormal verification result in the identification verification result;

[0088] A second judgment module, wherein when any node has the single abnormal verification result, the second judgment module is used to extract a non-abnormal frame image of the corresponding node through the video acquisition data;

[0089] An authentication and identification module, configured to authenticate and identify the non-abnormal frame image using the abnormal feature recognition database;

[0090] The first correction module is used to correct the single abnormal verification result to an abnormal verification result with re-inspection passing if the authentication identification passes.

[0091] Furthermore, the system also includes:

[0092] An interval setting module, the interval setting module is used to set a verification association time interval;

[0093] An extended authentication module, configured to perform extended authentication of the corresponding node according to the verification-related time interval when the single abnormal verification result occurs at any node;

[0094] The second correction module is used to correct the single abnormal verification result to an abnormal verification result with re-inspection passing when the expanded authentication result meets the preset threshold.

[0095] Furthermore, the system also includes:

[0096] A multi-level warning identification module, the multi-level warning identification module is used to generate a multi-level warning identification according to the identification and verification results;

[0097] A feedback verification module, the feedback verification module is used to provide early warning prompts through the multi-level early warning identification and receive feedback verification data, wherein the feedback verification data includes key features of manual identification;

[0098] An optimization module is used to optimize the abnormality recognition of the abnormality feature recognition database through the key features.

[0099] Through the above detailed description of a pipeline tank leakage identification method based on machine vision in this specification, those skilled in the art can clearly understand a pipeline tank leakage identification system based on machine vision in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0100] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A pipeline tank leakage identification method based on machine vision, characterized in that: The method is applied to an intelligent recognition system, wherein the intelligent recognition system is communicatively connected with a video acquisition unit and a compensation acquisition unit, and the method includes: Basic data sets of connected pipelines and tanks, and interactive basic distribution data; Controlling the compensation acquisition unit to perform compensation acquisition of pipeline data according to the basic distribution data, and generating video acquisition control data according to the compensation acquisition result; Controlling the video acquisition unit to perform video acquisition based on the video acquisition control data, and exporting video acquisition data, wherein the video acquisition data includes data in at least two formats; Perform auxiliary angle background information collection through the compensation collection unit, and generate background noise characteristics based on the collection results; Performing frame subtraction processing on the video acquisition data using the background noise feature, and performing feature recognition on the video acquisition data after frame subtraction using an abnormal feature recognition database; Obtaining recognition results of data in different formats, performing leakage recognition verification based on the recognition results, and outputting recognition verification results; The method further comprises: Determine whether there is a single abnormal verification result in the identification verification result; When the single abnormal verification result appears at any node, a non-abnormal frame image of the corresponding node is extracted through the video acquisition data; Authentication and identification of the non-abnormal frame image is performed using the abnormal feature recognition database; If the authentication and identification is passed, the single abnormal verification result is corrected to an abnormal verification result with re-inspection passed.

2. The method according to claim 1, wherein The method further comprises: Marking the acquisition time nodes of the video acquisition data to obtain time node data; Associating the background noise feature with the frame subtraction node of the video acquisition data through the time node data; generating background correction data according to the auxiliary angle and the video acquisition control data; Performing feature restoration processing of the background noise feature using the background correction data; The frame subtraction process is completed according to the node association and feature restoration process results.

3. The method according to claim 2, wherein The method further comprises: Determining a restoration reference point of the background noise feature according to the auxiliary angle and the video acquisition control data; Extracting the distance between the video acquisition data and the reference point through the restored reference point, and generating a restoration ratio based on the extraction result and the distance ratio of the restored reference point; The feature size of the background noise feature is restored according to the restoration ratio, and the feature restoration processing result is obtained based on the size restoration result.

4. The method according to claim 3, wherein The method further comprises: generating distortion correction data using the auxiliary angle and the restored reference point; Performing distortion restoration on the size restoration result based on the distortion correction data; The feature restoration processing result is obtained through the distortion restoration result.

5. The method according to claim 1, wherein The method further comprises: Set the verification association time interval; When any node has the said single abnormal verification result, the extended authentication of the said corresponding node is performed according to the said verification associated time interval; When the expanded authentication result meets the preset threshold, the single abnormal verification result is corrected to an abnormal verification result with re-verification passing.

6. The method according to claim 1, wherein The method further comprises: Generate a multi-level warning sign based on the identification and verification results; Providing early warning prompts through the multi-level early warning identification and receiving feedback verification data, wherein the feedback verification data includes key features of the manual identification; The abnormality recognition optimization of the abnormality feature recognition database is performed using the key features.

7. A pipeline tank leakage identification system based on machine vision, characterized in that: The system is communicatively connected with the video acquisition unit and the compensation acquisition unit, and the system includes: A data acquisition module, which is used to communicate with the basic data set of the pipeline tank and exchange basic distribution data; a compensation acquisition module, configured to control the compensation acquisition unit to perform compensation acquisition of pipeline data according to the basic distribution data, and generate video acquisition control data according to the compensation acquisition result; A video acquisition module, configured to control the video acquisition unit to perform video acquisition based on the video acquisition control data, and to export video acquisition data, wherein the video acquisition data includes data in at least two formats; An information acquisition module, configured to acquire auxiliary angle background information through the compensation acquisition unit and generate background noise characteristics based on the acquisition results; a feature recognition module, the feature recognition module being configured to perform frame subtraction processing on the video acquisition data using the background noise feature, and perform feature recognition on the video acquisition data after frame subtraction using an abnormal feature recognition database; A leakage identification and verification module, the leakage identification and verification module is used to obtain identification results of data in different formats, perform leakage identification and verification based on the identification results, and output the identification and verification results; A first judgment module, the first judgment module is used to judge whether there is a single abnormal verification result in the identification verification result; A second judgment module, wherein when any node has the single abnormal verification result, the second judgment module is used to extract a non-abnormal frame image of the corresponding node through the video acquisition data; An authentication and identification module, configured to authenticate and identify the non-abnormal frame image using the abnormal feature recognition database; The first correction module is used to correct the single abnormal verification result to an abnormal verification result with re-inspection passing if the authentication identification passes.

Images