Engine abnormal state detection method and device
阅读说明：本技术 一种发动机异常状态检测方法和装置 (Engine abnormal state detection method and device ) 是由 姜艺林 于乾坤 金忠孝 于 2021-07-09 设计创作，主要内容包括：本发明提供一种发动机异常状态检测方法和装置,方案在实施过程中,首先采用特征抽取取方式获取存储的视频数据中的特征数据,然后再对所述特征数据进行分类,将所述特征数据分为不同的簇,计算每个簇的簇中心的值,计算下一周期获取到的视频数据对应的特征数据的值与对应的所述簇中心的值的差值,将计算得到的差值与报警阈值进行对比,如果特征数据中与所述簇中心的最小差值大于报警阈值时,更新报警状态为第一预设状态,从而实现了基于图像数据对发动机进行监测,在发动机异常时进行提醒,提高了发动机监测结果的可靠性。(The invention provides a method and a device for detecting an abnormal state of an engine, wherein in the implementation process of the scheme, firstly, a characteristic data in stored video data is obtained by adopting a characteristic extraction mode, then the characteristic data is classified, the characteristic data is divided into different clusters, the value of the cluster center of each cluster is calculated, the difference value between the value of the characteristic data corresponding to the video data obtained in the next period and the value of the corresponding cluster center is calculated, the calculated difference value is compared with an alarm threshold value, and if the minimum difference value between the characteristic data and the cluster center is greater than the alarm threshold value, the alarm state is updated to be a first preset state, so that the engine is monitored based on image data, the alarm is given when the engine is abnormal, and the reliability of the monitoring result of the engine is improved.)
1. An engine abnormal state detection method characterized by comprising:
adopting image acquisition equipment to carry out video acquisition on a target area in an engine in the working process;
storing video data with a preset length;
acquiring feature data in the stored video data by adopting a feature extraction mode;
classifying the characteristic data, and dividing the characteristic data into different clusters;
calculating to obtain a cluster center value corresponding to each type of feature data based on the classification result;
calculating the difference value between the value of the characteristic data calculated in the current period and the value of the cluster center corresponding to the value of the characteristic data;
when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is larger than an alarm threshold value, updating the alarm state of the image area corresponding to the feature data into a first preset state;
and when the minimum difference value between each characteristic data and the cluster center is smaller than an alarm threshold value, updating the updated alarm state of the image area corresponding to the characteristic data to a second preset state.
2. The engine abnormal state detection method according to claim 1, further comprising, when a minimum difference between the feature data calculated in the current cycle and the cluster center is smaller than an alarm threshold:
judging whether the minimum difference value between the characteristic data and the cluster center is smaller than a preset updating threshold value or not;
and when the value is smaller than the preset updating threshold value, updating the value of the cluster center corresponding to the minimum difference value based on the feature data acquired in the current period by adopting a linear data processing method.
3. The engine abnormal state detection method according to claim 2, characterized in that the method further comprises:
configuring a cluster center score for each cluster center;
after updating the value of the cluster center corresponding to the minimum difference value, the method further includes:
increasing the updated cluster center score corresponding to the cluster center;
and when the number of cluster centers corresponding to the same feature data is larger than a preset value, removing the cluster center with the lowest cluster center score.
4. The engine abnormal state detection method according to claim 3, further comprising, after updating the value of the cluster center corresponding to the minimum difference value:
and increasing the cluster center fraction of the cluster center corresponding to the minimum difference value.
5. The engine abnormal state detection method according to claim 3, further comprising:
and when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than the preset updating threshold value, calculating to obtain a corresponding cluster center based on the feature data obtained in the current period, configuring a cluster center score for the corresponding cluster center obtained by calculation, and reducing the cluster center scores of other cluster centers corresponding to the feature data.
6. An engine abnormal state detection device characterized by comprising:
the acquisition unit is used for acquiring a video of a target area in the engine in the working process by adopting image acquisition equipment;
the characteristic data extraction unit is used for storing video data with preset length and acquiring characteristic data in the stored video data by adopting a characteristic extraction mode;
the clustering unit is used for classifying the characteristic data and dividing the characteristic data into different clusters;
the data processing unit is used for calculating and obtaining a cluster center value corresponding to each type of feature data based on the classification result; calculating the difference value between the value of the characteristic data calculated in the current period and the value of the cluster center corresponding to the value of the characteristic data;
the early warning unit is used for updating the warning state of the image area corresponding to the feature data to be a first preset state when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than a warning threshold value; and when the minimum difference value between each characteristic data and the cluster center is smaller than an alarm threshold value, updating the updated alarm state of the image area corresponding to the characteristic data to a second preset state.
7. The engine abnormal state detection device according to claim 6, further comprising:
the data updating unit is used for judging whether the minimum difference value between the feature data and the cluster center is smaller than a preset updating threshold value or not when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is smaller than the alarm threshold value; and when the value is smaller than the preset updating threshold value, updating the value of the cluster center corresponding to the minimum difference value based on the feature data acquired in the current period by adopting a linear data processing method.
8. The engine abnormal state detection device according to claim 7, wherein the data update unit is further configured to:
configuring a cluster center score for each cluster center;
after updating the value of the cluster center corresponding to the minimum difference value, the data updating unit is further configured to:
increasing the updated cluster center score corresponding to the cluster center;
and when the number of cluster centers corresponding to the same feature data is larger than a preset value, removing the cluster center with the lowest cluster center score.
9. The engine abnormal state detection device according to claim 8, wherein after updating the value of the cluster center corresponding to the minimum difference value, the data updating unit is further configured to:
and increasing the cluster center fraction of the cluster center corresponding to the minimum difference value.
10. The engine abnormal state detection method according to claim 7, wherein the data update unit is further configured to:
and when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than the preset updating threshold value, calculating to obtain a corresponding cluster center based on the feature data obtained in the current period, configuring a cluster center score for the corresponding cluster center obtained by calculation, and reducing the cluster center scores of other cluster centers corresponding to the feature data.
Technical Field
The invention relates to the technical field of power equipment, in particular to a method and a device for detecting an abnormal state of an engine.
Background
The engine performance test is an important link of an engine test, in the prior art, the working condition of the engine is mainly detected through the running state of the engine, but the applicant finds that along with the progress of the engine test, a considerable number of moving parts such as a cable tie, a pipeline and the like are influenced by vibration and ventilation in the test to generate approximately periodic movement, along with the progress of the engine test, parts such as a cylinder body and the like can be influenced by high temperature, the surface color can slowly change, the position and the posture of other moving parts can slowly deviate, the change of the states is difficult to distinguish from abnormal states by adopting a manual design characteristic mode, and therefore false positives can occur at a high frequency when the abnormal states are judged.
Disclosure of Invention
In view of this, embodiments of the present invention provide a method and an apparatus for detecting an abnormal state of an engine, so as to improve reliability of a determination result of the abnormal state of the engine.
In order to achieve the above purpose, the embodiments of the present invention provide the following technical solutions:
an engine abnormal state detection method includes:
adopting image acquisition equipment to acquire an image of a target area in an engine in the working process;
storing video data with a preset length;
acquiring feature data in the stored video data by adopting a feature extraction mode;
classifying the characteristic data, and dividing the characteristic data into different clusters;
calculating to obtain a cluster center value corresponding to each type of feature data based on the classification result;
calculating the difference value between the value of the characteristic data calculated in the current period and the value of the cluster center corresponding to the value of the characteristic data;
when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is larger than an alarm threshold value, updating the alarm state of the image area corresponding to the feature data into a first preset state;
and when the minimum difference value between each characteristic data and the cluster center is smaller than an alarm threshold value, updating the updated alarm state of the image area corresponding to the characteristic data to a second preset state.
Optionally, in the method for detecting an abnormal state of an engine, when a minimum difference between the feature data obtained by calculation in the current period and the cluster center is smaller than an alarm threshold, the method further includes:
judging whether the minimum difference value between the characteristic data and the cluster center is smaller than a preset updating threshold value or not;
and when the value is smaller than the preset updating threshold value, updating the value of the cluster center corresponding to the minimum difference value based on the feature data acquired in the current period by adopting a linear data processing method.
Optionally, in the method for detecting an abnormal state of an engine, the method further includes:
configuring a cluster center score for each cluster center;
after updating the value of the cluster center corresponding to the minimum difference value, the method further includes:
increasing the updated cluster center score corresponding to the cluster center;
and when the number of cluster centers corresponding to the same feature data is larger than a preset value, removing the cluster center with the lowest cluster center score.
Optionally, in the method for detecting an abnormal state of an engine, after updating the value of the cluster center corresponding to the minimum difference, the method further includes:
and increasing the cluster center fraction of the cluster center corresponding to the minimum difference value.
Optionally, the method for detecting an abnormal state of an engine further includes:
and when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than the preset updating threshold value, calculating to obtain a corresponding cluster center based on the feature data obtained in the current period, configuring a cluster center score for the corresponding cluster center obtained by calculation, and reducing the cluster center scores of other cluster centers corresponding to the feature data.
An engine abnormal state detection device comprising:
the acquisition unit is used for acquiring a video of a target area in the engine in the working process by adopting image acquisition equipment;
the characteristic data extraction unit is used for storing video data with preset length and acquiring characteristic data in the stored video data by adopting a characteristic extraction mode;
the clustering unit is used for classifying the characteristic data and dividing the characteristic data into different clusters;
the data processing unit is used for calculating and obtaining a cluster center value corresponding to each type of feature data based on the classification result; calculating the difference value between the value of the characteristic data calculated in the current period and the value of the cluster center corresponding to the value of the characteristic data;
the early warning unit is used for updating the warning state of the image area corresponding to the feature data to be a first preset state when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than a warning threshold value; and when the minimum difference value between each characteristic data and the cluster center is smaller than an alarm threshold value, updating the updated alarm state of the image area corresponding to the characteristic data to a second preset state.
Optionally, the engine abnormal state detection device further includes:
the data updating unit is used for judging whether the minimum difference value between the feature data and the cluster center is smaller than a preset updating threshold value or not when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is smaller than the alarm threshold value; and when the value is smaller than the preset updating threshold value, updating the value of the cluster center corresponding to the minimum difference value based on the feature data acquired in the current period by adopting a linear data processing method.
Optionally, in the above apparatus for detecting an abnormal state of an engine, the data updating unit is further configured to:
configuring a cluster center score for each cluster center;
after updating the value of the cluster center corresponding to the minimum difference value, the data updating unit is further configured to:
increasing the updated cluster center score corresponding to the cluster center;
and when the number of cluster centers corresponding to the same feature data is larger than a preset value, removing the cluster center with the lowest cluster center score.
Optionally, in the above apparatus for detecting an abnormal state of an engine, after updating the value of the cluster center corresponding to the minimum difference, the data updating unit is further configured to:
and increasing the cluster center fraction of the cluster center corresponding to the minimum difference value.
Optionally, in the above apparatus for detecting an abnormal state of an engine, the data updating unit is further configured to:
and when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than the preset updating threshold value, calculating to obtain a corresponding cluster center based on the feature data obtained in the current period, configuring a cluster center score for the corresponding cluster center obtained by calculation, and reducing the cluster center scores of other cluster centers corresponding to the feature data.
Based on the technical scheme, in the scheme provided by the embodiment of the invention, the characteristic data in the stored video data is obtained by adopting a characteristic extraction mode, then the characteristic data is classified, the characteristic data is divided into different clusters, the value of the cluster center of each cluster is calculated, the difference value between the value of the characteristic data corresponding to the video data obtained in the next period and the value of the corresponding cluster center is calculated, the calculated difference value is compared with the alarm threshold value, and if the minimum difference value between the characteristic data and the cluster center is greater than the alarm threshold value, the alarm state is updated to be the first preset state, so that the monitoring of the engine based on the image data is realized, the reminding is performed when the engine is abnormal, and the reliability of the engine monitoring result is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for detecting an abnormal engine condition according to an embodiment of the present disclosure;
FIG. 2 is a schematic flow chart diagram illustrating a method for detecting an abnormal engine condition according to another embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of an engine abnormal state detection device disclosed in the embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to improve the reliability of the engine test result, the application discloses an engine abnormal state detection method and an engine abnormal state detection device, referring to fig. 1, the engine abnormal state detection method disclosed in the embodiment of the application may include:
step S101: adopting image acquisition equipment to carry out video acquisition on a target area in an engine in the working process;
in the scheme, an image acquisition module is adopted to acquire videos of a target area in an engine in a working state in the experimental process, and then a video stream callback module is adopted to process an original video stream acquired by the image acquisition module so as to obtain a video frame in the working process of the engine;
step S102: storing video data with a preset length;
in this step, when an initialization instruction given by a user is obtained, an initialization module is triggered to store a video frame with a preset length n, wherein n is a positive integer, the value of n can be set according to the user requirement, and when video data with the preset length is stored, a window is slid, and the video data can be stored through the sliding window;
step S103: acquiring feature data in the stored video data by adopting a feature extraction mode;
in this step, the feature data may include any one or more of data of some channels sensitive to smoke and light in different color spaces in a video frame, optical flow added with time sequence, and other information, and these feature data may be obtained by performing data extraction on the video frame through different feature extraction modules (filters), and each feature extraction module extracts a type of feature data;
specifically, in the feature extraction module, a single-frame video frame is divided into w × h image blocks b (i, j) with the same size and without overlapping from top to bottom and from left to right, i is greater than or equal to 0 and is less than or equal to w-1, j is greater than or equal to 0 and is less than or equal to h-1, a feature vector with a length of k can be extracted by manual design, data dimension reduction or other methods for each single image block, and the feature vector is used for representing whether the engine has abnormal conditions such as smoke/fire and the like, and is the feature data in the step.
Specifically, when calculating the feature vector corresponding to each image block, the following method is specifically used to implement:
for a single image block divided by a single video frame, p is designed as the number of neighborhoods of the block, the neighborhoods refer to areas adjacent to the single image block, the feature vector finally corresponding to the single image block is determined by the original features of the image block and the features of the p neighborhoods of the image block, the length K of the feature vector is set to be K (K, p), the original K-dimensional feature vector can also be understood as a degradation form under the condition of no neighborhoods, n frames of videos are processed by the method, and finally, n multiplied by w multiplied by h multiplied by K feature vectors are obtained.
Step S104: classifying the characteristic data, and dividing the characteristic data into different clusters;
in this step, cluster initialization is started for the feature vectors, the feature vectors of the regions corresponding to the w × h image blocks corresponding to each group of video frames are clustered in the time dimension, and m K-dimensional cluster sets corresponding to the positions corresponding to each image block, that is, the feature vectors of all the image blocks corresponding to the positions of each image block in a group of video frames are clustered in the time dimension to obtain m cluster sets.
Step S105: calculating to obtain a cluster center value corresponding to each type of feature data based on the classification result;
in this step, the feature vector in each cluster set is calculated to obtain a value of a cluster center corresponding to each cluster set, in this step, the value of the cluster center may be a vector value, the cluster centers of the initial m K-dimensional cluster sets are obtained by a local area (an area within an image block) of each image block, and are used as cluster centers (which may also be in a reference State) State (i, j, n _ center) under the area, i is greater than or equal to 0 and less than or equal to w-1, j is greater than or equal to 0 and less than or equal to h-1, n _ center is greater than or equal to 0 and less than or equal to m-1, the maximum distance between the cluster centers is denoted as d _ ref (i, j), i is greater than or equal to 0 and less than or equal to w-1, and j is greater than or equal to 0 and less than or equal to h-1.
Step S106: calculating the difference value between the value of the characteristic data calculated in the current period and the value of the cluster center corresponding to the value of the characteristic data;
in this step, step S103 is executed in each period, the feature data of each image block corresponding to each video frame in the video data of the current period is obtained, the value of the feature data of each image block calculated in the current period is compared with the recorded value of the cluster center matched with the image block area, and the distance between the value of the feature data of each image block of each video frame calculated in the current period and the recorded value of the cluster center is calculated.
That is, in this step, p is designed as the number of neighborhoods of a single image block divided by the latest video frame (a video frame acquired in the current period), and finally, i is greater than or equal to 0 and less than or equal to w-1, j is greater than or equal to 0 and less than or equal to h-1 is determined by the original features of the block and the features of the p neighborhoods thereof, and the length of the Feature vector Feature (i, j) is K (K, p). That is, in this step, the Feature vector Feature (i, j) of the image block of the video frame calculated in the current period is compared with the m cluster center states (i, j, n _ center) matching the image block position, and the distance between the two is calculated.
Step S107: when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is larger than an alarm threshold value, updating the alarm state of the image area corresponding to the feature data into a first preset state;
in this step, comparing the Feature vector Feature (i, j) of an image block b (i, j) of a video frame calculated in the current period with the values (i, j, n _ center) of m cluster centers matched with the position of the image block, and if the minimum difference between the Feature vector Feature (i, j) calculated in the previous period and the values State (i, j, n _ center) of the m cluster centers is greater than a preset alarm threshold, updating the alarm State of the position corresponding to the image block b (i, j) to a first preset State.
Specifically, if the minimum difference is greater than the alarm threshold Talert(i, j), i is more than or equal to 0 and less than or equal to w-1, j is more than or equal to 0 and less than or equal to h-1, and the alarm state of the corresponding position of the image block b (i, j) in the alarm module is set as the Boolean value True.
Step S108: when the minimum difference value between each feature data and the cluster center is smaller than an alarm threshold value, updating the updated alarm state of the image area corresponding to the feature data to a second preset state;
when the minimum difference between the feature data of a certain image block of the video frame measured in the current period and the m cluster centers matched with the positions of the image blocks b (i, j) is smaller than an alarm threshold TalertAnd (i, j), setting the alarm state of the position corresponding to the image block b (i, j) in the alarm module to be a Boolean value False (a second preset state). The system can either alarm or not alarm at the time of selection according to the marking result of the block b (i, j) in the alarm module.
According to the scheme, the image data of the running state of the engine in the working process is obtained, the feature data in the stored video data are obtained in a feature extraction mode, then the feature data are classified, the feature data are divided into different clusters, the value of the cluster center of each cluster is calculated, the difference value between the value of each feature data corresponding to the video data obtained in the next period and the value of the corresponding cluster center is calculated, the calculated difference value is compared with an alarm threshold value, and if the minimum difference value between the feature data and the cluster center is larger than the alarm threshold value, the alarm state is updated to be the first preset state, so that the engine is monitored based on the image data, the warning is carried out when the engine is abnormal, and the reliability of the engine monitoring result is improved.
In a technical solution disclosed in another embodiment of the present application, the value of the cluster center is used to represent a characteristic value corresponding to the cluster under a normal operating condition of the engine, and in order to further ensure reliability of the value of the cluster center, the value of the cluster center may be dynamically updated in a monitoring process in the present solution. Specifically, in the scheme:
referring to fig. 2, when the minimum difference between the feature data calculated in the current period and the cluster center is smaller than the alarm threshold, the method further includes:
step S201: judging whether the minimum difference value between the characteristic data and the cluster center is smaller than a preset updating threshold value or not;
step S202: when the current cycle is smaller than the preset updating threshold, updating the value of the cluster center corresponding to the characteristic data based on the characteristic data acquired in the current cycle by adopting a linear data processing method;
when detecting that the minimum difference value between the characteristic value corresponding to a certain image block b (i, j) in the current period and the centers of the m clusters is smaller than an alarm threshold Talert(i, j), judging whether the minimum difference value is still less than a preset updating threshold value Tupdate(i, j), i is more than or equal to 0 and less than or equal to w-1, j is more than or equal to 0 and less than or equal to h-1, and the alarm state of the corresponding position of the image block b (i, j) in the alarm module is set as a Boolean value False.
When the minimum difference between the feature data corresponding to the image block b (i, j) and all cluster centers corresponding to the position of the image block b (i, j) is smaller than a preset updating threshold, the feature data corresponding to the image block b (i, j) is used for updating the value of the cluster center, at this time, the updated value of the cluster center refers to the value of the cluster center corresponding to the feature data corresponding to the image block b (i, j) with the minimum difference, and during updating, a linear data processing method can be used.
For example, a simple linear formula can be used:
wherein the w (i, j) is a weight value calculated based on a Score (i, j, n _ center ') value, the Score (i, j, n _ center ') is a value of a cluster center corresponding to a cluster center having a smallest difference in feature data corresponding to the image block b (i, j), and the Score (i, j, n _ center ') is a value of a cluster center corresponding to a cluster center having a smallest difference in feature data corresponding to the image block b (i, j)initIs the initial value of the cluster center;
the value of the cluster center is updated by this weight w (i, j), and, in particular,
State′(i,j,n′center)＝w(i,j)*State(i,j,n′center) + (1-w (i, j)). Feature (i, j) is Feature data calculated from image blocks b (i, j) of the video frame of the current period, and State '(i, j, n'center) I.e. the updated cluster center value.
In the updating process of the value of the cluster center, the value State (i, j, n _ center') of the cluster center is moved to the Feature vector Feature (i, j) corresponding to the image block b (i, j) by the weight w (i, j), so that the updating of the cluster center is completed, and the updated value of the cluster center is more accurate and reliable.
Step S203: when the current period is larger than the preset updating threshold value, calculating to obtain a cluster center based on the feature data acquired in the current period;
when the minimum difference between the feature data corresponding to the image block b (i, j) and all cluster centers corresponding to the position of the image block b (i, j) is greater than the preset update threshold, it indicates that the setting reliability of the value of the cluster center (where the cluster center refers to the value of the cluster center having the minimum difference between the feature data corresponding to the image block b (i, j)) is low, and therefore, in this scheme, the feature data corresponding to the image block b (i, j) needs to be used as a new cluster center corresponding to the position of the image block b (i, j), and at this time, the position of the image block b (i, j) corresponds to m +1 cluster centers.
In the technical solution disclosed in the embodiment of the present application, the alarm threshold T is set asalert(i, j) and an update threshold Tupdate(i, j) is determined by the maximum distance d _ ref (i, j) between the centers of the corresponding clusters and an empirical preset value, and different alarm thresholds T corresponding to d _ ref (i, j)alert(i, j) and an update threshold Tupdate(i, j) different, the corresponding alarm threshold value and the updating threshold value can be obtained from the preset mapping relation through the maximum distance d _ ref (i, j) and the empirical preset value.
In the above scheme, when the feature data and the cluster center are both represented by vectors, the "minimum difference value from the cluster center in the feature data" refers to a minimum distance from the cluster center to the feature data.
In the technical solution disclosed in the embodiment of the present application, as can be seen from the description of the above specific embodiment, each piece of feature data corresponds to a plurality of (at least m) cluster centers, where the "minimum difference between the feature data and the cluster centers" refers to comparing the feature data with the plurality of cluster centers respectively, obtaining differences between the feature data and the plurality of cluster centers respectively, at this time, obtaining a plurality of differences, and extracting a minimum value of the plurality of differences as a minimum value of the "minimum difference between the feature data and the cluster centers", and in a subsequent updating process, updating the cluster center corresponding to the "minimum difference between the feature data and the cluster centers".
In the technical solution disclosed in another embodiment of the present application, a corresponding cluster center score value may be created for each cluster center, where the cluster center score value may be increased or decreased based on a comparison result between a minimum difference value between the minimum difference value and the cluster center in the feature data of each image block in the current period and a preset update threshold and an alarm threshold, and specifically, in the foregoing solution, when each cluster center is created, an initial cluster center score is configured for each cluster center;
in step S202, after updating the value of the cluster center corresponding to the feature data, the method further includes: increasing the updated cluster center score corresponding to the cluster center; and reducing the score value of the other cluster centers corresponding to the characteristic data. That is, after the value of the cluster center corresponding to the feature data is updated, the cluster center Score (i, j, n _ center') corresponding to the updated cluster center is awarded (i.e., the value of the cluster center Score is increased), and the cluster center scores Score (i, j, n _ center) corresponding to the other cluster centers corresponding to the same feature data are penalized (i.e., the cluster center scores corresponding to the cluster centers are decreased), thereby completing the update of the cluster center scores.
In order to prevent the number of cluster centers corresponding to the same feature data from being too large and increase the workload of the system, in the present scheme, the number of cluster centers corresponding to the same feature data may be further controlled, so that the value of the number of cluster centers corresponding to the same feature data is not greater than a preset value, that is, when a certain feature data is detected or the number of cluster centers corresponding to a certain image block is greater than the preset value, the cluster center with the lowest cluster center score in the cluster centers corresponding to the feature data is removed.
Corresponding to the above method, the detailed operation content of each unit in the engine abnormal state detection device in the embodiment of the present application refers to the content of the above method embodiment, and the engine abnormal state detection device provided in the embodiment of the present invention is described below, and the engine abnormal state detection device described below and the engine abnormal state detection method described above may be referred to correspondingly.
Specifically, referring to fig. 3, the present application discloses an engine abnormal state detection device including:
the acquisition unit 100 corresponds to the step S101 in the method, and is used for performing video acquisition on a target area in the engine in the working process by adopting image acquisition equipment;
a feature data extraction unit 200, corresponding to the above method, in steps S102-S103, configured to store video data with a preset length, and obtain feature data in the stored video data by using a feature extraction manner;
a clustering unit 300, corresponding to step S104 in the above method, for classifying the feature data and dividing the feature data into different clusters;
a data processing unit 400, configured to calculate a cluster center value corresponding to each type of feature data based on the classification result; calculating the difference value between the value of the characteristic data calculated in the current period and the value of the cluster center corresponding to the value of the characteristic data;
the early warning unit 500 corresponds to the steps S105 to S108 in the method, and is configured to update the warning state of the image area corresponding to the feature data to a first preset state when the minimum difference between the feature data obtained by the calculation in the current period and the cluster center is greater than a warning threshold; and when the minimum difference value between each characteristic data and the cluster center is smaller than an alarm threshold value, updating the updated alarm state of the image area corresponding to the characteristic data to a second preset state.
Corresponding to the method, the device may further include a data updating unit, configured to determine whether a minimum difference between the feature data obtained by the current period calculation and the cluster center is smaller than a preset updating threshold when the minimum difference between the feature data and the cluster center is smaller than an alarm threshold; and when the value is smaller than the preset updating threshold value, updating the value of the cluster center corresponding to the minimum difference value based on the feature data acquired in the current period by adopting a linear data processing method.
Corresponding to the method, in the above apparatus, the data updating unit is further configured to:
configuring a cluster center score for each cluster center;
after updating the value of the cluster center corresponding to the minimum difference value, the data updating unit is further configured to:
increasing the updated cluster center score corresponding to the cluster center;
and when the number of cluster centers corresponding to the same feature data is larger than a preset value, removing the cluster center with the lowest cluster center score.
Corresponding to the method, in the apparatus, after updating the value of the cluster center corresponding to the feature data, the data updating unit is further configured to:
and increasing the cluster center fraction of the cluster center corresponding to the minimum difference value.
Corresponding to the method, in the apparatus, the data updating unit is further configured to:
and when the minimum difference value between the feature data obtained by calculation in the current period and the cluster center is greater than the preset updating threshold value, calculating to obtain a corresponding cluster center based on the feature data obtained in the current period, configuring a cluster center score for the corresponding cluster center obtained by calculation, and reducing the cluster center scores of other cluster centers corresponding to the feature data.
For convenience of description, the above system is described with the functions divided into various modules, which are described separately. Of course, the functionality of the various modules may be implemented in the same one or more software and/or hardware implementations of the invention.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
It is further noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.