CN110766035A - Intelligent detection method for single fault of bearing - Google Patents
Intelligent detection method for single fault of bearing Download PDFInfo
- Publication number
- CN110766035A CN110766035A CN201910780429.XA CN201910780429A CN110766035A CN 110766035 A CN110766035 A CN 110766035A CN 201910780429 A CN201910780429 A CN 201910780429A CN 110766035 A CN110766035 A CN 110766035A
- Authority
- CN
- China
- Prior art keywords
- evidence
- bpa
- bearing
- detection method
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses an intelligent detection method for single fault of a bearing. The method comprises the following steps: a. signal acquisition: collecting bearing multi-sensor data and constructing a data set; b. data preprocessing: randomly sampling from the data set to form a training set and a testing set; c. feature extraction: extracting the features of the training set and the test set based on an SAE algorithm to obtain a feature data set; d. training a classifier: carrying out classification training on the feature data in the feature data set to obtain a plurality of classifiers; ids fusion: performing fusion calculation on the classification results of all classifiers through an improved D-S evidence theory to obtain a fusion result; the improved D-S evidence theory is obtained by establishing an identification framework and recalculating a weight distribution function; f. the accuracy is obtained: and comparing the actual label with the fusion result to obtain the precision of the method. The invention has high detection precision and strong applicability.
Description
Technical Field
The invention relates to a bearing fault diagnosis method, in particular to a bearing single fault intelligent detection method.
Background
According to incomplete statistics, about 30% of faults in the rotary machine are caused by bearing damage, and therefore, the research of bearing fault diagnosis technology has important economic value and social benefit. At present, the research on the multi-fault diagnosis of the bearing mainly focuses on the use of a signal decomposition method and the analysis, transformation and separation of coupling signals manually. These investigated methods generally focus on the diagnosis of three or two combinational failure points without analyzing all of the multi-failure combinational failure cases throughout, nor do they consider using an automatic feature extraction method to identify a single failure from among the different multi-failures.
Disclosure of Invention
The invention aims to provide an intelligent detection method for single fault of a bearing. The invention has high detection precision and strong applicability.
The technical scheme of the invention is as follows: the intelligent detection method for the single fault of the bearing comprises the following steps:
a. signal acquisition: collecting bearing multi-sensor data and constructing a data set;
b. data preprocessing: randomly sampling from the data set to form a training set and a testing set;
c. feature extraction: extracting the features of the training set and the test set based on an SAE algorithm to obtain a feature data set;
d. training a classifier: carrying out classification training on the feature data in the feature data set to obtain a plurality of classifiers;
ids fusion: performing fusion calculation on the classification results of all classifiers through an improved D-S evidence theory to obtain a fusion result; the improved D-S evidence theory is obtained by establishing an identification framework and recalculating a weight distribution function;
f. the accuracy is obtained: and comparing the actual label with the fusion result to obtain the precision of the method.
In step d of the intelligent detection method for single fault of bearing, the classifier is an SVM classifier.
In step e of the intelligent detection method for single fault of bearing, the improved D-S evidence theory is obtained by the following steps:
e1. each classifier is used as an evidence body, and a judgment probability matrix called BPA is obtained after a plurality of classifiers obtained by training are given to the same test sample;
e2. calculating the weight of an evidence body according to the Pearson correlation coefficient, and constructing a correlation matrix;
e3. distributing the trust weight of the evidence body to obtain the trust of the evidence body;
e4. correcting the original evidence body BPA by using the evidence body credibility to obtain a new BPA matrix;
e5.0 factor correction: correcting the item with the probability of 0 in the new BPA with the correction size of 0.01 to obtain the corrected BPA;
e6. calculating the corrected BPA according to a D-S calculation rule to obtain a credibility vector;
e7. adding the credibility vector into the original BPA to form a new basic probability distribution matrix;
e8. repeating the factor 0 modification of step e 5;
e9. and e, repeating the D-S calculation rule calculation of the step e6 to obtain a final fusion result.
Step e2 of the foregoing intelligent detection method for single bearing fault specifically includes: calculating the weight of the evidence body through the Pearson correlation coefficient, and defining the evidence body miAnd mjIs a correlation coefficient SijForming a correlation matrix Sij，
Defining a body of evidence miThe support degree of (m) is Supi) The evidence body m can be calculated by the correlation between the evidence bodiesiAll correlation values of (a):
step e3 of the foregoing intelligent detection method for single bearing fault specifically includes: defining a body of evidence miConfidence level of (m) is credi)，cred(mi)∈[0,1]And is
The calculation method is as follows:
then, the confidence of the evidence of the overall negative linear correlation is set as an absolute value, and the others are kept unchanged.
Step e4 of the foregoing intelligent detection method for single bearing fault specifically includes: correcting the original evidence body BPA by using the evidence body credibility to obtain a new BPA matrix mi *(X)，
Credibility vector m in step e6 of the intelligent bearing single-fault detection methodi #(X) is calculated according to the following formula:
new basic probability distribution matrix M in step e7 of the intelligent bearing single fault detection methodnewThe method specifically comprises the following steps:
compared with the prior art, the invention has the following beneficial effects:
the invention realizes the intelligent detection of the automatic extraction of the single fault characteristics of the bearing;
the method can fuse decision results of a plurality of sensors, greatly improves the detection precision, and has better fusion effect of the improved D-S evidence theory compared with the traditional D-S theory;
the invention can achieve higher detection accuracy in various faults of various types, and compared with the traditional signal analysis which depends on artificial knowledge, the method has stronger applicability.
In order to verify the beneficial effects of the method of the invention, the invention was used to analyze the mixed multiple fault condition of the rolling bearing. Seven sets of experiments were performed on different code sizes using data from three directional sensors. The experimental results are shown in fig. 4 to 11. The results demonstrate that the idea of intelligent single fault detection from multiple faults is meaningful and effective.
Furthermore, the inventors have also conducted the following experimental studies:
effectiveness analysis of experiment 1SAE and D-S evidence theory on single-fault detection method
The coding dimension reflects the number of extracted features, so the coding dimension directly influences whether the output of the hidden layer can completely reserve the required fault features or not, and directly influences the classification precision. In order to research the relation between the characteristic number and the detection precision, the invention selects different coding dimensions with gradient change to carry out experimental analysis. As the code size increases, the detection accuracy also increases, as shown in fig. 4, 5, and 6.
In fig. 4, 5 and 6, there are various fault detection details, whether internal, spherical or external. From the proposed results, the research of intelligent SFD (single fault detection) is of great significance, and the SAE method proves its effectiveness for single fault feature extraction in multi-fault bearing data. First, as the feature size changes, the accuracy rate trends to be as expected, the accuracy rises rapidly at the beginning and remains stable when the feature dimension reaches around 150 to 200, but as the number of feature sizes increases, the computation amount extraction function also increases. As can be seen from the results of the three pictures, the sensors at different positions are shown to have significant differences of different detection accuracies for different faults. Typically, the data of the X-axis sensor has the best detection results for all three faults, except for a few features with dimensions of 25 to 100 on the inner and outer ring faults. It is clear that the data of the Y-axis sensor is more difficult to detect faults than the other two, with an accuracy always below 0.85. To compare the three sensor detection capabilities for three faults, we averaged the three fault accuracy rates for each sensor, resulting in the results of fig. 7 and 8. As can be seen from fig. 7, regardless of the increase in the code size, the accuracy of the three sensors is in order of X, Z, and Y from large to small, in addition to the code size of 25, and the difference in detection accuracy is also becoming large as the feature size increases. This indicates that the location of the sensor has a large impact on fault detection. The results in fig. 8 indicate that there are differences in the difficulty of detecting different fault types, and that it is relatively difficult to identify a sphere fault compared to other fault types in the experiments of the present invention. Thus, it can be concluded that the X-axis signal is advantageous for fault detection. To take advantage of the difference between the sensor and the fault, a method of fusing a plurality of direction sensors is considered.
In addition, in fig. 4, 5 and 6, the detection accuracy of three faults of the D-S method is always higher than that of a single sensor, and thus it can be seen that combining the results of different sensors will improve the detection accuracy of various faults. In fig. 9, similar to the comparison of fig. 4, fig. 5 and fig. 6 for a single fault, the fusion result is always higher than the average accuracy of the three sensors. As the feature size increases, the detection accuracy of the internal fault becomes higher than the other two, and this phenomenon never occurs on average. This indicates that the fusion method has a higher effect on inner ring failure than outer ring failure, which also demonstrates the benefit of combining different sensors together to detect the results. Overall, the advantages of the fusion theory can be seen from this figure. Conventional D-S exhibits good multi-sensor fusion capability, which can achieve high recognition accuracy when sized around 150 to 200 a and held stable.
In this study, the data also showed that the fused results were the best results for each fault case, since the feature has a dimension of 150, and the D-S method results are 0.1209,0.1133, and 0.1242 higher than the results for the internal, external, and external unfused results. And (4) detecting the ball fault.
Due to the D-S evidence problem described previously, an experimental analysis of the bearing data is shown in FIGS. 10 and 11 for the improved method. In fig. 10, the higher accuracy effect of IDS can be seen from the figure compared to D-S. To clearly see the effect of the improved method, the difference between IDS and D-S is shown in FIG. 11. As feature size increases, the difference increases and then decreases in IF detection, remains relatively stable and high in OF detection, and remains increasing and decreasing slowly in BF detection. OF the maximum differences, IDS is 1.41%, 1.78% and 2.33% more accurate than D-S in the IF, OF and BF detections. From the results compared with the D-S evidence theory, it can be seen that the PCC can effectively realize the weight calculation of different sensors, and the better results can prove the influence of different sensor data on the fault identification again. Considering the overall performance of the three failure types, it is most apparent that IDS exhibits better performance than D-S in the 150 and 200 dimensions, and the results of IDS and D-S in the 150 and 200 characteristic dimensions are shown in Table 1.
As shown in table 1, the accuracies OF IF, OF, and BF are 0.987, 0.991, and 0.977 in the feature size OF 150; the dimension is 200 with accuracies of 0.991, 0.993 and 0.985, respectively, which are 0.42%, 0.21% and 0.79% higher than the dimension of 150, respectively. The improved fusion method has more obvious improvement on BF detection, and the detection precision is 2.33 percent and 2.06 percent higher than that of D-S. Finally, it can be concluded in the bearing failure detection data that PCC plays an important role in improving D-S by redistributing the weight of the sensor.
TABLE 1 accuracy comparison between D-S and IDS methods
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a single fault classification training rule;
FIG. 3 is a schematic flow diagram of a modified D-S;
FIG. 4 is a comparison of fused and unfused inner ring faults;
FIG. 5 is a table comparing fused and unfused outer ring faults;
FIG. 6 is a table comparing failure of fused and unfused spheres;
FIG. 7 is the average detection accuracy for three types of faults;
FIG. 8 is the average accuracy of three sensors;
FIG. 9 compares the results of D-S with no fusion;
FIG. 10 compares the results of D-S and IDS;
fig. 11 is the magnitude of the difference between D-S and IDS.
Detailed Description
The invention is further illustrated by the following figures and examples, which are not to be construed as limiting the invention.
Example 1. An intelligent detection method for single fault of bearing, referring to fig. 1, is carried out according to the following steps:
a. signal acquisition: collecting bearing multi-sensor data and constructing a data set;
b. data preprocessing: randomly sampling from the data set to form a training set and a testing set;
c. feature extraction: extracting the features of the training set and the test set based on an SAE algorithm to obtain a feature data set;
d. training a classifier: carrying out classification training on the feature data in the feature data set to obtain a plurality of classifiers;
ids fusion: performing fusion calculation on the classification results of all classifiers through an improved D-S evidence theory to obtain a fusion result; the improved D-S evidence theory is obtained by establishing an identification framework and recalculating a weight distribution function;
f. the accuracy is obtained: and comparing the actual label with the fusion result to obtain the precision of the method.
In the foregoing step d, the classifier is an SVM classifier. And for the feature data set obtained by SAE extraction, training single fault classification according to the partition rule of FIG. 2. In fig. 2, each letter represents a fault type, a plurality of letters represent coupling situations of a plurality of faults, and the classifier a represents a classifier obtained after training, which can be used to detect whether a fault is included, and so on.
In the foregoing step e, the improved process of the improved D-S evidence theory is shown in fig. 3, each classifier is regarded as an evidence body, and a judgment probability matrix, called basic probability distribution (BPA), is obtained by training the obtained multiple classifiers for the same test sample. The 'one-vote rejection' means that if the probability is 0 in the basic probability distribution, the corresponding proposition of the probability is completely denied according to the traditional D-S calculation method, and the rule of the correction of the method is that 0.01 is borrowed from the maximum value of the probability in the same evidence body to the item with the probability of 0, so that the correction purpose is achieved under the condition that the integral probability and the integral trend are not changed. The improvement is specifically carried out by the following steps:
e1. each classifier is used as an evidence body, and a judgment probability matrix called BPA (basic probability distribution) is obtained after a plurality of classifiers obtained by training are given to the same test sample;
e2. calculating the weight of an evidence body according to the Pearson correlation coefficient, and constructing a correlation matrix;
e3. distributing the trust weight of the evidence body to obtain the trust of the evidence body;
e4. correcting the original evidence body BPA by using the evidence body credibility to obtain a new BPA matrix;
e5.0 factor correction: correcting the item with the probability of 0 in the new BPA with the correction size of 0.01 to obtain the corrected BPA;
e6. calculating the corrected BPA according to a D-S calculation rule to obtain a credibility vector;
e7. adding the credibility vector into the original BPA to form a new basic probability distribution matrix;
e8. repeating the factor 0 modification of step e 5;
e9. and e, repeating the D-S calculation rule calculation of the step e6 to obtain a final fusion result.
The step e2 is specifically as follows: calculating the weight of the evidence body through the Pearson correlation coefficient, and defining the evidence body miAnd mjIs a correlation coefficient SijForming a correlation matrix Sij，
Defining a body of evidence miThe support degree of (m) is Supi) The evidence body m can be calculated by the correlation between the evidence bodiesiAll correlation values of (a):
the step e3 is specifically as follows: defining a body of evidence miConfidence level of (m) is credi)，cred(mi)∈[0,1]And is
The calculation method is as follows:
then, the confidence of the evidence of the overall negative linear correlation is set as an absolute value, and the others are kept unchanged.
The step e4 is specifically as follows: correcting the original evidence body BPA by using the evidence body credibility to obtain a new BPA matrix mi *(X)，
The aforementioned confidence vector m in step e6i #(X) is calculated according to the following formula:
the new basic probability distribution matrix M in step e7 described abovenewThe method specifically comprises the following steps:
Claims (8)
1. the intelligent detection method for the single fault of the bearing is characterized by comprising the following steps of:
a. signal acquisition: collecting bearing multi-sensor data and constructing a data set;
b. data preprocessing: randomly sampling from the data set to form a training set and a testing set;
c. feature extraction: extracting the features of the training set and the test set based on an SAE algorithm to obtain a feature data set;
d. training a classifier: carrying out classification training on the feature data in the feature data set to obtain a plurality of classifiers;
ids fusion: performing fusion calculation on the classification results of all classifiers through an improved D-S evidence theory to obtain a fusion result; the improved D-S evidence theory is obtained by establishing an identification framework and recalculating a weight distribution function;
f. the accuracy is obtained: and comparing the actual label with the fusion result to obtain the precision of the method.
2. The intelligent bearing single fault detection method according to claim 1, wherein in the step d, the classifier is an SVM classifier.
3. The intelligent detection method for the single fault of the bearing according to claim 1, wherein in the step e, the improved D-S evidence theory is obtained by improving the following steps:
e1. each classifier is used as an evidence body, and a judgment probability matrix called BPA is obtained after a plurality of classifiers obtained by training are given to the same test sample;
e2. calculating the weight of an evidence body according to the Pearson correlation coefficient, and constructing a correlation matrix;
e3. distributing the trust weight of the evidence body to obtain the trust of the evidence body;
e4. correcting the original evidence body BPA by using the evidence body credibility to obtain a new BPA matrix;
e5.0 factor correction: correcting the item with the probability of 0 in the new BPA with the correction size of 0.01 to obtain the corrected BPA;
e6. calculating the corrected BPA according to a D-S calculation rule to obtain a credibility vector;
e7. adding the credibility vector into the original BPA to form a new basic probability distribution matrix;
e8. repeating the factor 0 modification of step e 5;
e9. and e, repeating the D-S calculation rule calculation of the step e6 to obtain a final fusion result.
4. The intelligent detection method for the single fault of the bearing according to claim 3, wherein the step e2 is specifically as follows: calculating the weight of the evidence body through the Pearson correlation coefficient, and defining the evidence body miAnd mjIs a correlation coefficient SijForming a correlation matrix Sij，
Defining a body of evidence miThe support degree of (m) is Supi) The evidence body m can be calculated by the correlation between the evidence bodiesiAll correlation values of (a):
5. the intelligent detection method for the single fault of the bearing according to claim 4, wherein the step e3 is specifically as follows: defining a body of evidence miConfidence level of (m) is credi)，cred(mi)∈[0,1]And isThe calculation method is as follows:
then, the confidence of the evidence of the overall negative linear correlation is set as an absolute value, and the others are kept unchanged.
6. The intelligent detection method for the single fault of the bearing according to claim 5, wherein the step e4 is specifically as follows: correcting the original evidence body BPA by using the evidence body credibility to obtain a new BPA matrix mi *(X)，
7. The intelligent detection method for single fault of bearing according to claim 6, wherein the confidence vector m in step e6i #(X) is calculated according to the following formula:
8. the intelligent bearing single fault detection method as claimed in claim 7, wherein said new fundamental probability of step e7Distribution matrix MnewThe method specifically comprises the following steps:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910780429.XA CN110766035A (en) | 2019-08-22 | 2019-08-22 | Intelligent detection method for single fault of bearing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910780429.XA CN110766035A (en) | 2019-08-22 | 2019-08-22 | Intelligent detection method for single fault of bearing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110766035A true CN110766035A (en) | 2020-02-07 |
Family
ID=69329854
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910780429.XA Pending CN110766035A (en) | 2019-08-22 | 2019-08-22 | Intelligent detection method for single fault of bearing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110766035A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101161A (en) * | 2020-09-04 | 2020-12-18 | 西安交通大学 | Evidence theory fault state identification method based on correlation coefficient distance and iteration improvement |
| CN112633493A (en) * | 2020-12-01 | 2021-04-09 | 北京理工大学 | Fault diagnosis method and system for industrial equipment data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160327452A1 (en) * | 2015-05-04 | 2016-11-10 | Deere & Company | Fault detection for bearings |
| CN107101829A (en) * | 2017-04-11 | 2017-08-29 | 西北工业大学 | A kind of intelligent diagnosing method of aero-engine structure class failure |
| CN108332970A (en) * | 2017-11-17 | 2018-07-27 | 中国铁路总公司 | A kind of Method for Bearing Fault Diagnosis based on LS-SVM and D-S evidence theory |
| CN109540520A (en) * | 2018-11-29 | 2019-03-29 | 中国船舶重工集团海装风电股份有限公司 | A kind of rolling bearing fault fusion diagnosis method based on improvement D-S evidence theory |
| CN109932179A (en) * | 2019-04-09 | 2019-06-25 | 东南大学 | A kind of rolling bearing fault testing method based on the reconstruct of DS Adaptive spectra |
-
2019
- 2019-08-22 CN CN201910780429.XA patent/CN110766035A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160327452A1 (en) * | 2015-05-04 | 2016-11-10 | Deere & Company | Fault detection for bearings |
| CN107101829A (en) * | 2017-04-11 | 2017-08-29 | 西北工业大学 | A kind of intelligent diagnosing method of aero-engine structure class failure |
| CN108332970A (en) * | 2017-11-17 | 2018-07-27 | 中国铁路总公司 | A kind of Method for Bearing Fault Diagnosis based on LS-SVM and D-S evidence theory |
| CN109540520A (en) * | 2018-11-29 | 2019-03-29 | 中国船舶重工集团海装风电股份有限公司 | A kind of rolling bearing fault fusion diagnosis method based on improvement D-S evidence theory |
| CN109932179A (en) * | 2019-04-09 | 2019-06-25 | 东南大学 | A kind of rolling bearing fault testing method based on the reconstruct of DS Adaptive spectra |
Non-Patent Citations (3)
| Title |
|---|
| DUY TANG HOANG 等: "Deep Belief Network and Dempster-Shafer Evidence Theory for Bearing Fault Diagnosis", 《ISIE 2018》 * |
| JIANGUANG LU 等: "A Novel Method for Intelligent Single Fault Detection of Bearings Using SAE and Improved D–S Evidence Theory", 《ENTROPY》 * |
| 梅检民 等: "基于多分类支持向量机和 D-S证据理论的轴承故障诊断", 《汽车工程》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101161A (en) * | 2020-09-04 | 2020-12-18 | 西安交通大学 | Evidence theory fault state identification method based on correlation coefficient distance and iteration improvement |
| CN112633493A (en) * | 2020-12-01 | 2021-04-09 | 北京理工大学 | Fault diagnosis method and system for industrial equipment data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gao et al. | Automatic change detection in synthetic aperture radar images based on PCANet | |
| CN109050535B (en) | Rapid terrain condition identification method based on vehicle attitude | |
| CN113255793B (en) | Fine-grained ship identification method based on contrast learning | |
| CN109241901B (en) | A kind of detection and recognition methods to the three-dimensional point cloud with hole | |
| CN113838054B (en) | Mechanical part surface damage detection method based on artificial intelligence | |
| CN103839065A (en) | Extraction method for dynamic crowd gathering characteristics | |
| CN103279738B (en) | Automatic identification method and system for vehicle logo | |
| CN111914911B (en) | Vehicle re-identification method based on improved depth relative distance learning model | |
| CN105989597B (en) | Hyperspectral image abnormal target detection method based on pixel selection process | |
| CN110766035A (en) | Intelligent detection method for single fault of bearing | |
| CN102385592A (en) | Image concept detection method and device | |
| CN111965183A (en) | Titanium alloy microstructure detection method based on deep learning | |
| CN117495891B (en) | Point cloud edge detection method and device and electronic equipment | |
| CN105868778B (en) | A kind of median filtering detection method based on frequency domain fluctuation characteristic | |
| CN111914930A (en) | Density peak value clustering method based on self-adaptive micro-cluster fusion | |
| CN104462826B (en) | The detection of multisensor evidences conflict and measure based on Singular Value Decomposition Using | |
| CN111144280A (en) | Monitoring video leaf occlusion detection method | |
| CN107622667B (en) | Method and system for detecting altered license plate number | |
| Li et al. | MA 2-FPN for Tiny Object Detection from Remote Sensing Images | |
| CN112014821B (en) | Unknown vehicle target identification method based on radar broadband characteristics | |
| CN113554036A (en) | Characteristic point extraction and matching method for improving ORB algorithm | |
| CN112037193A (en) | Power line feature marking method and device | |
| Zhu et al. | OGNet: Salient object detection with output-guided attention module | |
| Pan et al. | The Crime Scene Tools Identification Algorithm Based on GVF-Harris-SIFT and KNN. | |
| Xu et al. | Abnormal behavior detection in crowd scenes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200207 |