CN116823816A - Detection equipment and detection method based on security monitoring static memory - Google Patents
Detection equipment and detection method based on security monitoring static memory Download PDFInfo
- Publication number
- CN116823816A CN116823816A CN202311083045.5A CN202311083045A CN116823816A CN 116823816 A CN116823816 A CN 116823816A CN 202311083045 A CN202311083045 A CN 202311083045A CN 116823816 A CN116823816 A CN 116823816A
- Authority
- CN
- China
- Prior art keywords
- module
- hash
- dimensional
- data
- static memory
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 117
- 230000003068 static effect Effects 0.000 title claims abstract description 64
- 238000012544 monitoring process Methods 0.000 title claims abstract description 38
- 230000003993 interaction Effects 0.000 claims abstract description 59
- 238000004458 analytical method Methods 0.000 claims abstract description 26
- 230000000007 visual effect Effects 0.000 claims abstract description 8
- 230000002159 abnormal effect Effects 0.000 claims description 61
- 239000013598 vector Substances 0.000 claims description 41
- 238000000034 method Methods 0.000 claims description 40
- 238000004422 calculation algorithm Methods 0.000 claims description 36
- 239000011159 matrix material Substances 0.000 claims description 36
- 238000004364 calculation method Methods 0.000 claims description 33
- 230000006870 function Effects 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 20
- 230000005856 abnormality Effects 0.000 claims description 17
- 238000010276 construction Methods 0.000 claims description 16
- 230000004927 fusion Effects 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 16
- 230000002452 interceptive effect Effects 0.000 claims description 13
- 238000009826 distribution Methods 0.000 claims description 10
- 238000004519 manufacturing process Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 10
- 230000007547 defect Effects 0.000 claims description 9
- 238000011084 recovery Methods 0.000 claims description 9
- 238000010845 search algorithm Methods 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 238000005286 illumination Methods 0.000 claims description 7
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 230000000877 morphologic effect Effects 0.000 claims description 5
- 238000013519 translation Methods 0.000 claims description 5
- 239000000463 material Substances 0.000 claims description 3
- 239000004065 semiconductor Substances 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 238000003745 diagnosis Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 238000012806 monitoring device Methods 0.000 description 3
- 238000013079 data visualisation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007654 immersion Methods 0.000 description 2
- 238000002347 injection Methods 0.000 description 2
- 239000007924 injection Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005670 electromagnetic radiation Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 238000005316 response function Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses detection equipment and a detection method based on a security monitoring static memory, which are used in the field of semiconductors and comprise a detector body, wherein a real-time interaction module is arranged at the top of the detector body, a camera module is arranged at the top of the front surface of the detector body and positioned at one side of the real-time interaction module, and a three-dimensional reconstruction module, a local sensitive hash forest anomaly detection module, an anomaly analysis module and a k-nearest neighbor search module are sequentially arranged in the detector body. According to the invention, the three-dimensional model can reflect the real scene more truly by integrating the multi-source image data, so that a user can obtain higher-quality and more realistic visual experience in the virtual interaction environment, and the multi-source image data can help to improve the accuracy and detail display of the three-dimensional model.
Description
Technical Field
The invention relates to the field of semiconductors, in particular to detection equipment and detection method based on a security monitoring static memory.
Background
A static memory (Static Random Access Memory, SRAM for short) is an electronic device that stores data and can hold data without requiring refreshing, and is therefore also referred to as a "static" memory. Static memory is commonly used in high performance computers, graphics processors, routers, switches, and the like applications because of its fast read and write speeds, low power consumption, and ease of integration.
Static memory is an important component in computers, and its reliability and stability are critical to the operation of the computer system. The following are several common static memory detection methods:
1. dynamic fault injection method: the method simulates the fault condition of the memory by injecting external interference such as voltage pulse or electromagnetic radiation into the circuit of the static memory, thereby testing and evaluating
2. Equivalent discriminant diagnosis method: the method replaces the actual memory circuit by establishing an equivalent model, and performs fault detection and diagnosis for the model.
3. Failure mode analysis: the method determines all fault types possibly occurring in the memory by analyzing various fault modes possibly occurring in the memory, and designs a detection method according to the fault types.
In the prior art, an abnormality detection method and an abnormality detection system for security equipment exist:
the Chinese patent with publication number of CN114627079B discloses an anomaly detection method and system of security monitoring equipment based on artificial intelligence; specifically disclosed is: the method comprises the steps of determining a target monitoring device, and acquiring a standard mixed Gaussian model of each pixel position in an image acquired by the target monitoring device; for each pixel in each frame of image, inputting the pixel value of the pixel into a corresponding standard mixed Gaussian model to calculate a probability value, and calculating a probability deviation value based on the probability value and the maximum probability value of the corresponding standard mixed Gaussian model; counting the ratio of the illumination intensity of each pixel in the background area of each frame of image to the probability deviation value to obtain a standard ratio; the real-time ratio of the illumination intensity corresponding to each pixel in the background area of the real-time image acquired by the target monitoring equipment to the probability deviation value is calculated and compared with the standard ratio to obtain an abnormal value corresponding to each pixel in the background area, and the abnormal detection of the target monitoring equipment is carried out based on the abnormal value, so that the effective operation of the security system can be ensured.
However, in the prior art, detection is performed based on a probability deviation value, a certain probability error exists, fault detection is difficult to achieve with high precision and high accuracy, and particularly, the prior art only detects the illumination intensity and cannot learn the fault condition of the equipment when the color appears.
The Chinese patent with publication number CN112712021B discloses a grain surface abnormal state identification method based on a perceptual hash and connected domain analysis algorithm; specifically disclosed is: the method comprises the steps of acquiring images of two sides in a granary at regular time and in real time, automatically storing video streams, comparing the current grain surface state with the original grain surface state through a perception hash algorithm to obtain similarity, comparing the similarity with a preset threshold value to judge whether the granary state is abnormal or not, and carrying out recognition calculation of an abnormal region through a connected domain algorithm when the grain surface state in the granary is abnormal.
However, this conventional technique is directed to abnormality detection of a monitoring object, and the detection criterion is that the device itself is not required to fail, and when an abnormal image is found, it cannot be recognized whether the abnormal image originates from the monitoring object or from the monitoring device.
Chinese patent publication No. CN115641538A discloses a method, apparatus, computer device and storage medium for detecting device abnormality; specifically disclosed is: acquiring video information shot by the monitoring equipment in at least two historical time periods, and acquiring video information shot by the monitoring equipment in a current time period; respectively extracting the background in the video information shot in at least two historical time periods and the background in the video information shot in the current time period to obtain background images corresponding to the video information shot in at least two historical time periods and background images corresponding to the video information shot in the current time period; performing image fusion processing on background images corresponding to video information shot in at least two historical time periods to obtain background reference images; according to the background reference image and the background image corresponding to the video information shot in the current period, whether the monitoring equipment is abnormal or not is determined, the implementation cost of monitoring equipment abnormality detection is reduced, the influence of environmental interference on monitoring equipment abnormality detection is reduced, and false detection is avoided.
However, the prior art also relies on Gaussian algorithm, the mathematical basis of which is normal distribution statistics, which are found inThere is an error of about 1% outside the interval of (a), and the error rate is amplified when it is faced with high bit amount data such as image files, resulting in low recognition accuracy.
However, in the detection process of the method, the dynamic fault injection method needs to perform external interference on the memory, which may cause that a fault which does not occur when the memory actually works is simulated, thereby misjudging the reliability of the memory. The equivalent model established by the equivalent discriminant diagnosis method may be different from the actual memory, thereby affecting the accuracy of the diagnosis result. Fault pattern analysis requires analysis and identification of different fault types and thus requires extensive knowledge of the memory circuitry. In addition, failure mode analysis cannot detect some abnormal conditions, such as single bit errors in memory.
For the problems in the related art, no effective solution has been proposed at present.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide detection equipment and a detection method based on a security monitoring static memory, which have the advantages of high detection efficiency and high precision and aim to solve the problem that some common detection methods in the prior art have defects.
In order to realize the advantages of high detection efficiency and high precision, the invention adopts the following specific technical scheme:
according to one aspect of the invention, the detection equipment based on the security monitoring static memory is improved, and comprises a detector body, wherein a real-time interaction module is arranged at the top of the detector body, a camera module is arranged at the top of the front surface of the detector body and positioned at one side of the real-time interaction module, and a three-dimensional reconstruction module, a local sensitive hash forest anomaly detection module, an anomaly analysis module and a k-nearest neighbor search module are sequentially arranged in the detector body;
The camera module is used for capturing static memory images on the surface of the wafer at multiple angles by using the panoramic camera to obtain multi-source image data;
the three-dimensional reconstruction module is used for constructing a three-dimensional model by utilizing the multi-source image data and fusing the three-dimensional model into a virtual interaction environment;
the local sensitive hash forest anomaly detection module is used for carrying out anomaly detection on the static memory image in the virtual interactive scene by utilizing a local sensitive hash isolated forest algorithm;
the abnormality analysis module is used for analyzing the area with the defect in the static memory image in the virtual interaction scene according to the abnormality detection result;
the k-neighbor searching module is used for inquiring the position where the data abnormality occurs by adopting a k-neighbor searching algorithm;
the real-time interaction module is used for highlighting the abnormal area by utilizing the real-time interaction module and providing an interaction function.
Optionally, the three-dimensional reconstruction module comprises a feature extraction module, an image matching module, a motion recovery and three-dimensional point cloud production module, a global optimization module, a point cloud fusion and surface reconstruction module, a texture mapping module and a virtual interaction environment construction module;
the feature extraction module is used for acquiring multi-source image data and extracting feature descriptors of the multi-source image data;
The image matching module is used for calculating the similarity between images according to the feature descriptors and finding out feature point matching pairs existing between the images;
the motion recovery and three-dimensional point cloud production module is used for solving an intrinsic matrix between two adjacent images according to the feature point matching pairs, recovering a motion relation between cameras, calculating three-dimensional coordinates of the matching feature points in the real world by using a triangulation method, and generating an initial three-dimensional point cloud;
the global optimization module is used for integrating the camera pose of the multi-source image data and the three-dimensional point cloud into a unified coordinate system and performing global optimization;
the point cloud fusion and surface reconstruction module is used for fusing the three-dimensional point cloud and extracting the surface information of the three-dimensional model from the point cloud data through a surface reconstruction algorithm;
the texture mapping module is used for mapping the color information in the original image onto the surface of the three-dimensional model to obtain a reconstructed three-dimensional model;
the virtual interactive environment construction module is used for importing the reconstructed three-dimensional model into a virtual reality platform, setting the parallel illumination and material properties, and constructing a virtual interactive environment.
Optionally, the motion restoration and three-dimensional point cloud production module comprises a feature descriptor extraction module, a feature point matching module, an essential matrix calculation module, a camera motion restoration module, a triangulation module and an initial three-dimensional point cloud generation module;
The feature descriptor extraction module is used for extracting a group of stable feature points from two adjacent pictures and calculating feature descriptors of each feature point;
the characteristic point matching module is used for finding out characteristic point pairs matched with each other in two adjacent pictures;
the essential matrix calculation module is used for calculating an essential matrix between two cameras under the condition that the internal parameters of the cameras are known;
the camera motion recovery module is used for obtaining a rotation matrix and a translation vector through the decomposition of the essential matrix, and further calculating the motion between the cameras;
the triangulation module is used for calculating three-dimensional coordinates of the feature point pairs in the real world by using a triangulation method according to the feature point pairs and the essence matrix;
the initial three-dimensional point cloud generation module is used for connecting all three-dimensional coordinates to generate an initial three-dimensional point cloud.
Optionally, the point cloud fusion and surface reconstruction module comprises a three-dimensional point cloud fusion module, a direction field calculation module and a triangular grid generation module;
the three-dimensional point cloud fusion module is used for fusing three-dimensional point clouds with different visual angles so that the three-dimensional point clouds are aligned in the same coordinate system;
The direction field calculation module is used for calculating the normal vector of each three-dimensional point cloud in the point cloud, projecting the point cloud onto a regular grid by using known normal vector information, calculating the direction field of each grid, and solving a poisson equation on the whole grid to obtain the height value of each grid unit;
the triangular grid generating module is used for constructing a continuous triangular grid according to the solved height value and outputting the triangular grid into a common three-dimensional model file format.
Optionally, the local sensitive hash forest anomaly detection module comprises a feature vector extraction module, a local sensitive hash module and an isolated forest anomaly detection module;
the feature vector extraction module is used for extracting feature vectors of the static memory image data and converting the static memory image data into numerical vectors with fixed lengths;
the local sensitive hash module is used for mapping similar data points in the numerical vector into similar barrels by utilizing a local sensitive hash algorithm to obtain a hash table;
the isolated forest anomaly detection module is used for taking the hash table as input, and performing anomaly detection by utilizing an isolated forest algorithm to obtain an anomaly detection result.
Optionally, the local sensitive hash module comprises a local sensitive hash function family module, a hash value calculation module and a similar data point query module;
the local sensitive hash function family module is used for selecting a local sensitive hash function family and determining the number of hash functions to be used and the number of bits allocated to each hash bucket;
the hash value calculation module is used for traversing the input numerical value vector, calculating hash values of each data point under all hash functions, and combining the calculated hash values into a hash signature;
the similar data point query module is used for mapping the hash signature to a corresponding hash bucket, giving a query data point, calculating the hash signature of the given query data point, finding the corresponding bucket in the hash table, traversing all the data points in the bucket, calculating the similarity with the query point, and returning the data point with the highest similarity as an output result to obtain the hash table.
Optionally, the isolated forest anomaly detection module comprises a hash table conversion module, an isolated forest construction model, a path length calculation module and an anomaly point judgment module;
the hash table conversion module is used for converting the hash table into a characteristic vector set of data points, creating vectors with the length equal to the number of hash buckets for each data point of the characteristic vector set, and filling vector elements according to the distribution of the data points in the hash table;
The isolated forest construction model is used for creating and constructing an isolated forest model according to parameter setting;
the path length calculation module is used for traversing each tree in the isolated forest and calculating the path length of each data point on the tree;
the abnormal point judging module is used for converting the average path length of each data point on all trees into abnormal scores and setting a threshold value to judge whether the data point is abnormal or not.
Optionally, the k-nearest neighbor searching module comprises an abnormal data point inquiring module, an abnormal point distance calculating module, a k-adjacent selecting module and a k-adjacent analyzing module;
the abnormal data point query module is used for determining abnormal data points to be queried according to an abnormal detection result;
the abnormal point distance calculation module is used for traversing each data point in the data set and calculating the distance between the abnormal data points by using the morphological similarity distance;
the k-adjacent selection module is used for sequencing the calculated distances and selecting k data points with the smallest distance as k-adjacent of the abnormal data points;
and the k-neighbor analysis module is used for analyzing the positions of the abnormal data points and other nearby data points according to the search result of the k-neighbor.
Optionally, the calculating the distance between outlier data points using morphology similarity distancesThe formula of (2) is:
in (1) the->The Euclidean distance between abnormal points;
is the Manhattan distance between outliers;
is the absolute value of the sum of the dimensional differences between outliers.
According to one aspect of the invention, a detection method based on security monitoring of a static memory is improved, the detection method comprising the following steps:
s1, capturing a static memory image of the surface of a wafer by using a panoramic camera at multiple angles to obtain multi-source image data;
s2, constructing a three-dimensional model by utilizing multi-source image data, and fusing the three-dimensional model into a virtual interaction environment;
s3, performing anomaly detection on the static memory image in the virtual interaction scene by using a local sensitive hash isolated forest algorithm;
s4, analyzing the area with the defects in the static memory image in the virtual interaction scene according to the abnormal detection result;
s5, inquiring the position where the data abnormality occurs by using a k-nearest neighbor search algorithm;
s6, highlighting the abnormal area by using the real-time interaction module and providing an interaction function.
Compared with the prior art, the invention provides the detection equipment and the detection method based on the security monitoring static memory, which have the following beneficial effects:
(1) According to the invention, the panoramic camera is used for capturing the static memory image on the surface of the wafer at multiple angles, so that abundant data information can be obtained, the analysis accuracy is improved, the data visualization is more visual through the construction of the three-dimensional model and the virtual interaction environment, the observation and analysis are convenient, the anomaly detection and positioning are performed by combining a plurality of algorithms such as a local sensitive hash isolated forest algorithm, a k-nearest neighbor search algorithm and the like, and the detection efficiency and the detection accuracy are improved.
(2) According to the invention, the three-dimensional model can reflect a real scene more truly by integrating the multi-source image data, a user can obtain higher-quality and more realistic visual experience in a virtual interaction environment, the multi-source image data can help to improve the accuracy and detail display of the three-dimensional model, the mode can reduce errors or missing information caused by a single data source, so that a more comprehensive and high-quality model is provided, the virtual interaction environment can provide multiple interaction modes for the user, such as clicking, dragging, rotating, zooming and the like, the rich interaction functions can enable the user to explore the three-dimensional model more freely, the immersion and participation are enhanced, the three-dimensional model is constructed by utilizing the multi-source image data and fused into the virtual interaction environment, the virtual interaction environment can be widely applied to various industries, remote collaborative work can be realized, and the user can operate and participate in projects at the same time in different regions.
(3) According to the method, the isolated forest algorithm is used for realizing efficient anomaly detection by constructing a plurality of trees and randomly selecting the attribute for segmentation, compared with other anomaly detection algorithms based on distance or density, the isolated forest has higher speed in the training and detection processes, different types and distribution data sets can be dealt with, whether the data sets are uniform or not and whether the data sets are distributed or not is balanced or not, the method can find out an anomaly sample, and the isolated forest has higher detection accuracy on anomaly points due to the fact that the random attribute selection and segmentation strategy is adopted in the construction process, and meanwhile, the similarity measurement among data points is reserved in the dimension reduction process by the local sensitive hash technology, so that the anomaly positioning accuracy is improved.
(4) The k-neighbor search algorithm is simple in principle, easy to understand, relatively simple to realize, only needs to calculate the distance between the point to be queried and other data points, then selects the nearest k neighbors for analysis, has strong robustness to abnormal values and noise, can reduce the influence of individual abnormal values on results by considering the information of the k nearest neighbors, has good adaptability to nonlinear and changeable data distribution, does not need to assume that the data obeys specific distribution or has a certain structure, and can still perform well in complex data environments.
Drawings
The above features, features and advantages of the present invention, as well as the manner of attaining them and method of attaining them, will become more apparent and the invention will be better understood by reference to the following description of embodiments, taken in conjunction with the accompanying drawings. Here shown in schematic diagram:
FIG. 1 is a schematic structural diagram of a detection device based on a security monitoring static memory according to an embodiment of the present invention;
FIG. 2 is a cross-sectional view of a security monitoring static memory-based detection device in accordance with an embodiment of the present invention;
FIG. 3 is a functional block diagram of a detection device based on security monitoring static memory according to an embodiment of the present invention;
FIG. 4 is a schematic block diagram of a three-dimensional reconstruction module in a schematic block diagram of a detection device based on security monitoring static memory according to an embodiment of the present invention;
FIG. 5 is a schematic block diagram of a locally sensitive hash forest anomaly detection module in a schematic block diagram of a detection device based on security monitoring static memory according to an embodiment of the present invention;
fig. 6 is a schematic block diagram of a k-nearest neighbor search module in a schematic block diagram of a detection device based on a security monitoring static memory according to an embodiment of the present invention.
In the figure:
1. a detector body; 101. a real-time interaction module; 102. a camera module; 103. a three-dimensional reconstruction module; 1031. a feature extraction module; 1032. an image matching module; 1033. a motion recovery and three-dimensional point cloud production module; 1034. a global optimization module; 1035. a point cloud fusion and surface reconstruction module; 1036. a texture mapping module; 1037. the virtual interaction environment construction module; 104. a local sensitive hash forest anomaly detection module; 1041. a feature vector extraction module; 1042. a locally sensitive hash module; 1043. an isolated forest anomaly detection module; 105. an anomaly analysis module; 106. k-neighbor search module; 1061. an abnormal data point query module; 1062. an outlier distance calculation module; 1063. k-a proximity selection module; 1064. a k-proximity analysis module; 107. a light source module; 108. an alarm module; 109. and operating the button.
Detailed Description
In order to make the present application better understood by those skilled in the art, the following description will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
According to the embodiment of the application, a detection device and a detection method based on security monitoring static memory are provided.
The application is further described with reference to the drawings and the specific embodiments, as shown in fig. 1-6, according to an embodiment of the application, a detection device based on a security monitoring static memory is provided, and the detection device comprises a detector body 1, wherein a real-time interaction module 101 is arranged at the top of the detector body 1, a camera module 102 is arranged at the top of the front surface of the detector body 1 and positioned at one side of the real-time interaction module 101, and a three-dimensional reconstruction module 103, a local sensitive hash forest anomaly detection module 104, an anomaly analysis module 105 and a k-nearest neighbor search module 106 are sequentially arranged in the detector body 1.
In addition, the two ends of one side of the camera module 102 are provided with light source modules 107, the top of the front surface of the detector body 1 and the side with the real-time interaction module 101 are provided with alarm modules 108, and one side of each alarm module 108 is provided with a plurality of control buttons 109 positioned on the surface of the detector body 1.
In particular, in order to facilitate better understanding of those skilled in the art, related embodiments of the present application will now be explained with reference to technical terms or partial terms that may be involved in the present application:
description of: refers to a method of mathematically describing the characteristics of an object or image. In the fields of computer vision, pattern recognition, image processing and the like, descriptors are used for describing characteristics of objects or images, and can be used for performing tasks such as classification, retrieval, matching and the like. Common descriptors include SIFT, SURF, ORB, which generally have the characteristics of rotation invariance, scale invariance, illumination invariance and the like, and can effectively extract and match characteristics in multiple scenes.
And (3) a barrel: in the locality sensitive hashing algorithm (LSH algorithm), mapping data points into similar "buckets" means that packets of data points with similar characteristics are stored in the same "bucket". A "bucket" herein may be understood as a storage structure for gathering similar data points together for faster similarity matching and querying. Briefly, "bucket" is a container of a set of data points, where the data points have some common characteristics or similar attributes. Efficient similarity searching and querying can be achieved by distributively mapping similar data points into similar "buckets" using a locality sensitive hashing algorithm.
The camera module 102 is configured to capture a static memory image of a wafer surface at multiple angles by using a panoramic camera, so as to obtain multi-source image data.
Specifically, because the panoramic camera can capture images of a plurality of angles on the surface of the wafer, more comprehensive and detailed information can be provided, and therefore, the accuracy of static memory testing is improved, and the panoramic camera can be used for rapidly capturing images of a plurality of angles and combining the images into a complete image, so that time and labor cost are saved.
The three-dimensional reconstruction module 103 is configured to construct a three-dimensional model using multi-source image data, and fuse the three-dimensional model into a virtual interactive environment.
Preferably, the three-dimensional reconstruction module 103 includes a feature extraction module 1031, an image matching module 1032, a motion recovery and three-dimensional point cloud production module 1033, a global optimization module 1034, a point cloud fusion and surface reconstruction module 1035, a texture mapping module 1036, and a virtual interaction environment construction module 1037;
the feature extraction module 1031 is configured to obtain multi-source image data, and extract feature descriptors of the multi-source image data;
the image matching module 1032 is configured to calculate similarities between the images according to the feature descriptors, and find feature point matching pairs that exist between the images;
The motion restoration and three-dimensional point cloud production module 1033 is configured to solve an eigenvector between two adjacent images according to a feature point matching pair, recover a motion relationship between cameras, and calculate three-dimensional coordinates of the matching feature point in the real world by using a triangulation method to generate an initial three-dimensional point cloud;
the global optimization module 1034 is configured to integrate the camera pose of the multi-source image data and the three-dimensional point cloud into a unified coordinate system, and perform global optimization;
the point cloud fusion and surface reconstruction module 1035 is configured to fuse three-dimensional point clouds, and extract surface information of a three-dimensional model from the point cloud data through a surface reconstruction algorithm;
the texture mapping module 1036 is configured to map color information in an original image onto a surface of a three-dimensional model, so as to obtain a reconstructed three-dimensional model;
the virtual interactive environment construction module 1037 is configured to import the reconstructed three-dimensional model into a virtual reality platform, set the juxtaposition illumination and material properties, and construct a virtual interactive environment.
Specifically, the feature point matching pair refers to finding a group of feature point pairs matched with each other in two adjacent images through feature point detection and descriptor matching. The relation between the matched characteristic points is key information required by the steps of calculating an eigenvatrix based on the matched pairs, recovering the camera motion relation, calculating three-dimensional coordinates by a triangulation method and the like. Specifically, when the eigenvalue matrix is calculated, the eigenvalue matching pairs are needed to be used for calculating the eigenvalue matrix, so as to calculate the camera motion parameters; in performing triangulation, three-dimensional coordinates of feature points in the real world need to be calculated using pixel coordinates of matching feature points and camera internal parameters. Therefore, the relationship between the feature point matching pair and the matching feature point is closely related, and is an important basis for calculating the pose of the camera and generating the three-dimensional point cloud. The matching feature points, the eigenvatrices and the feature point matching pairs are all important concepts for solving camera motion relations and three-dimensional reconstruction in computer vision, and close relations exist between the two important concepts. Firstly, the feature point matching pairs are that feature point pairs matched with each other are found in two pictures, the feature points represent projections of the same object or scene under different view angles, and geometric information between images can be provided. By means of the feature point matching pairs, we can estimate the motion relationship between camera poses by solving the intrinsic matrix, a process called the computation of the intrinsic matrix. The eigenvector can be computed by at least 8 pairs of eigenvalue matches, and the 8 degrees of freedom of the eigenvector corresponds to exactly 8 pairs of eigenvalue matches. Second, the eigen matrix is an important tool to describe the motion relationship between two cameras. The eigen matrix may transform the matching feature points from one camera coordinate system to another and may resolve the rotation and translation matrices. The motion parameters of the camera obtained by the calculation of the eigenvector matrix can further recover the motion trail of the camera in the three-dimensional space. Finally, the matching feature points are the information necessary to calculate the eigen matrix and recover the camera pose. Through the matched characteristic point pairs, an essential matrix between cameras can be calculated, so that a rotation matrix and a translation matrix between camera postures are obtained, three-dimensional coordinates of the matched characteristic points in the real world are calculated by using a triangulation method, and a three-dimensional point cloud is generated. Thus, matching feature points, eigenvectors, and feature point matching pairs are closely related concepts in computer vision that together form the basis for solving camera motion relationships and three-dimensional reconstruction.
Preferably, the motion restoration and three-dimensional point cloud production module 1033 includes a feature descriptor extraction module, a feature point matching module, an essential matrix calculation module, a camera motion restoration module, a triangulation module, and an initial three-dimensional point cloud generation module;
the feature descriptor extraction module is used for extracting a group of stable feature points from two adjacent pictures and calculating feature descriptors of each feature point;
the characteristic point matching module is used for finding out characteristic point pairs matched with each other in two adjacent pictures;
the essential matrix calculation module is used for calculating an essential matrix between two cameras under the condition that the internal parameters of the cameras are known;
the camera motion recovery module is used for obtaining a rotation matrix and a translation vector through the decomposition of the essential matrix, and further calculating the motion between the cameras;
the triangulation module is used for calculating three-dimensional coordinates of the feature point pairs in the real world by using a triangulation method according to the feature point pairs and the essence matrix;
the initial three-dimensional point cloud generation module is used for connecting all three-dimensional coordinates to generate an initial three-dimensional point cloud.
Specifically, the three-dimensional coordinates of the same object in the two images need to be matched by calculating the three-dimensional coordinates in the real world by using a triangulation method. Feature point detection and matching can be generally performed by using SIFT, SURF and other algorithms, and meanwhile, in order to obtain parameters of internal parameters, external parameters and the like of a camera, calibration of the camera is required. The calibration process may be performed by shooting a checkerboard or other known three-dimensional model, and after matching a sufficient number of feature point pairs, an essential matrix may be estimated using an algorithm such as RANSAC, and three-dimensional coordinates thereof in the real world may be calculated by using a triangulation method based on the feature point pairs and the essential matrix. The process mainly uses the principle of similar triangle and vector operation to calculate, and different cameras may adopt different coordinate systems, so that the three-dimensional coordinates obtained by calculation need to be subjected to coordinate system conversion so as to adapt to different application scenes.
Preferably, the point cloud fusion and surface reconstruction module 1035 includes a three-dimensional point cloud fusion module, a direction field calculation module, and a triangular mesh generation module;
the three-dimensional point cloud fusion module is used for fusing three-dimensional point clouds with different visual angles so that the three-dimensional point clouds are aligned in the same coordinate system;
the direction field calculation module is used for calculating the normal vector of each three-dimensional point cloud in the point cloud, projecting the point cloud onto a regular grid by using known normal vector information, calculating the direction field of each grid, and solving a poisson equation on the whole grid to obtain the height value of each grid unit;
the triangular grid generating module is used for constructing a continuous triangular grid according to the solved height value and outputting the triangular grid into a common three-dimensional model file format.
Specifically, the step of solving the poisson equation on the whole grid to obtain the height value of each grid cell includes: it is necessary to determine the boundary conditions, i.e. at which points a fixed height is required. Typically, these points correspond to points on the model surface, and a coefficient matrix needs to be constructed. Each row of the matrix corresponds to a grid cell and the coefficients need to be set according to the relationship between the cell and the adjacent cells, and then the coefficient matrix and the boundary conditions are substituted into poisson's equation to form a linear equation set. The system of equations may be solved using iterative methods (e.g., jacobi, gauss-Seidel) or direct methods (e.g., LU decomposition, cholesky decomposition), and finally, the height value of each grid cell is calculated. The height values of all points inside each grid cell can be solved by weighted averaging the height values of the four vertices of the cell.
The local sensitive hash forest anomaly detection module 104 is configured to perform anomaly detection on the static memory image in the virtual interactive scene by using a local sensitive hash isolated forest algorithm.
Preferably, the local sensitive hash forest anomaly detection module 104 includes a feature vector extraction module 1041, a local sensitive hash module 1042, and an isolated forest anomaly detection module 1043;
the feature vector extraction module 1041 is configured to extract a feature vector of the static memory image data, and convert the static memory image data into a numerical vector with a fixed length;
the local sensitive hash module 1042 is configured to map similar data points in the numerical vector to similar buckets by using a local sensitive hash algorithm to obtain a hash table;
the isolated forest anomaly detection module 1043 is configured to take the hash table as input, and perform anomaly detection by using an isolated forest algorithm to obtain an anomaly detection result.
Preferably, the local sensitive hash module 1042 includes a local sensitive hash function family module, a hash value calculation module, and a similar data point query module;
the local sensitive hash function family module is used for selecting a local sensitive hash function family and determining the number of hash functions to be used and the number of bits allocated to each hash bucket;
The hash value calculation module is used for traversing the input numerical value vector, calculating hash values of each data point under all hash functions, and combining the calculated hash values into a hash signature;
the similar data point query module is used for mapping the hash signature to a corresponding hash bucket, giving a query data point, calculating the hash signature of the given query data point, finding the corresponding bucket in the hash table, traversing all the data points in the bucket, calculating the similarity with the query point, and returning the data point with the highest similarity as an output result to obtain the hash table.
Preferably, the isolated forest anomaly detection module 1043 includes a hash table conversion module, an isolated forest construction model, a path length calculation module, and an anomaly point judgment module;
the hash table conversion module is used for converting the hash table into a characteristic vector set of data points, creating vectors with the length equal to the number of hash buckets for each data point of the characteristic vector set, and filling vector elements according to the distribution of the data points in the hash table;
the isolated forest construction model is used for creating and constructing an isolated forest model according to parameter settings (such as the number of trees, the maximum depth of the trees and the like);
The path length calculation module is used for traversing each tree in the isolated forest and calculating the path length of each data point on the tree;
furthermore, path length represents the distance from the root node to a leaf node, a shorter path representing a possible outlier;
the abnormal point judging module is used for converting the average path length of each data point on all trees into abnormal scores and setting a threshold value to judge whether the data point is abnormal or not.
The anomaly score may be normalized to a range of 0 to 1. The closer the score is to 1, the more likely the data point is an outlier.
The anomaly analysis module 105 is configured to analyze, according to the anomaly detection result, an area in the static memory image in the virtual interactive scene where a defect exists.
However, it is necessary to detect an abnormality in the static memory image by using a method such as machine learning or deep learning, identify a region where a defect may exist, divide the detected abnormal region, and determine the boundary and size of each region. Image segmentation may be performed using a threshold-based approach (e.g., otsu algorithm), an edge-based approach (e.g., canny algorithm), or a Watershed-based approach (e.g., watershed algorithm), etc., and some feature information may need to be extracted for each anomaly region for subsequent analysis. The features may include information on color, texture, shape, etc., and the abnormal region is classified according to the extracted feature information. The region classification may be performed using a machine learning method such as a Support Vector Machine (SVM), decision tree, etc.
The k-nearest neighbor searching module 106 is configured to query a location where the data anomaly occurs by using a k-nearest neighbor searching algorithm.
Preferably, the k-nearest neighbor search module 106 includes an outlier query module 1061, an outlier distance calculation module 1062, a k-nearest neighbor selection module 1063, and a k-nearest neighbor analysis module 1064;
the abnormal data point query module 1061 is configured to determine abnormal data points to be queried according to an abnormal detection result;
the outlier distance calculation module 1062 is configured to traverse each data point in the dataset and calculate a distance between outlier data points using the morphological similarity distance;
the k-neighbor selection module 1063 is configured to rank the calculated distances, and select k data points with the smallest distances as k-neighbors of the outlier data points;
the k-neighbor analysis module 1064 is configured to analyze the position of the abnormal data point and other data points nearby according to the search result of the k-neighbor.
Preferably, the method uses morphological similarity distance to calculate distance between abnormal data pointsThe formula of (2) is:
in (1) the->The Euclidean distance between abnormal points;
is the Manhattan distance between outliers;
is the absolute value of the sum of the dimensional differences between outliers.
The euclidean distance refers to a linear distance between two points in a two-dimensional or three-dimensional space. Manhattan distance, also known as city block distance, refers to the distance between two points in two dimensions that travel along grid lines. Outliers refer to points in the dataset that differ significantly from other points.
The meaning of this formula is therefore to calculate the absolute value of the sum of the differences of two data points in each dimension, which is used to measure their degree of similarity or degree of difference in the various dimensions. In the field of abnormal point detection and the like, the formula can be used for judging whether a certain point is greatly different from other points, so that abnormal points are identified.
The real-time interaction module 101 is configured to highlight the abnormal area by using the real-time interaction module and provide an interaction function.
For highlighting the abnormal region by the real-time interactive module, the abnormal region may be highlighted by the real-time interactive module, and the highlighted region may be represented by a shape such as a rectangle, a circle, or a polygon, and may be implemented by adding an event listener or a response function. For example, when a user clicks on a highlight region, detailed information about the region may be displayed, and the highlighting color, transparency, and border width properties may be altered to enhance the visualization, integrate the real-time interaction module into the application, and ensure that it is connected to a data source (e.g., database).
According to another embodiment of the present invention, there is also provided a detection method based on security monitoring of a static memory, the detection method including the steps of:
s1, capturing a static memory image of the surface of a wafer by using a panoramic camera at multiple angles to obtain multi-source image data;
s2, constructing a three-dimensional model by utilizing multi-source image data, and fusing the three-dimensional model into a virtual interaction environment;
s3, performing anomaly detection on the static memory image in the virtual interaction scene by using a local sensitive hash isolated forest algorithm;
s4, analyzing the area with the defects in the static memory image in the virtual interaction scene according to the abnormal detection result;
s5, inquiring the position where the data abnormality occurs by using a k-nearest neighbor search algorithm;
s6, highlighting the abnormal area by using the real-time interaction module and providing an interaction function.
In summary, by means of the technical scheme, the panoramic camera is used for capturing the static memory image of the wafer surface at multiple angles, so that abundant data information can be obtained, analysis accuracy is improved, data visualization is more visual through construction of a three-dimensional model and a virtual interaction environment, observation and analysis are facilitated, and abnormality detection and positioning are performed by combining multiple algorithms such as a local sensitive hash isolated forest algorithm, a k-nearest neighbor search algorithm and the like, so that detection efficiency and accuracy are improved; according to the method, a three-dimensional model can reflect a real scene more truly by integrating multi-source image data, a user can acquire visual experience with higher quality and stronger sense of reality in a virtual interaction environment, the multi-source image data can help to improve accuracy and detail display of the three-dimensional model, the mode can reduce errors or missing information caused by a single data source, so that a more comprehensive and high-quality model is provided, the virtual interaction environment can provide multiple interaction modes for the user, such as clicking, dragging, rotating, scaling and the like, the rich interaction functions can enable the user to explore the three-dimensional model more freely, immersion sense and participation degree are enhanced, the three-dimensional model is constructed by utilizing the multi-source image data and fused into the virtual interaction environment, remote collaborative work can be realized by the virtual interaction environment, the user can operate and participate in projects simultaneously in different regions, compared with other anomaly detection algorithms based on distance or density, a forest has higher interaction modes in a training process and a detection process, the advantages of being capable of accurately balancing the difference between different types of data points in a random distribution and the anomaly, and the anomaly detection process can be well balanced, and the anomaly detection policy is kept, and the anomaly is more accurate in the situation is well distributed because of the anomaly is distributed in the different technologies; the k-neighbor search algorithm is simple in principle, easy to understand, relatively simple to realize, only needs to calculate the distance between the point to be queried and other data points, then selects the nearest k neighbors for analysis, has strong robustness to abnormal values and noise, can reduce the influence of individual abnormal values on results by considering the information of the k nearest neighbors, has good adaptability to nonlinear and changeable data distribution, does not need to assume that the data obeys specific distribution or has a certain structure, and can still perform well in complex data environments.
Although the invention has been described with respect to the preferred embodiments, the embodiments are for illustrative purposes only and are not intended to limit the invention, as those skilled in the art will appreciate that various modifications can be made without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. The detection equipment based on the security monitoring static memory is characterized by comprising a detector body (1), wherein a real-time interaction module (101) is arranged at the top of the detector body (1), a camera module (102) is arranged at the top of the front surface of the detector body (1) and positioned at one side of the real-time interaction module (101), and a three-dimensional reconstruction module (103), a local sensitive hash forest anomaly detection module (104), an anomaly analysis module (105) and a k-nearest neighbor search module (106) are sequentially arranged in the detector body (1);
the camera module (102) is used for capturing static memory images of the surface of the wafer at multiple angles by using the panoramic camera to obtain multi-source image data;
the three-dimensional reconstruction module (103) is used for constructing a three-dimensional model by utilizing multi-source image data and fusing the three-dimensional model into a virtual interaction environment;
The local sensitive hash forest anomaly detection module (104) is used for carrying out anomaly detection on the static memory image in the virtual interactive scene by utilizing a local sensitive hash isolated forest algorithm;
the abnormality analysis module (105) is used for analyzing the area with the defect in the static memory image in the virtual interaction scene according to the abnormality detection result;
the k-neighbor search module (106) is used for inquiring the position where the data abnormality occurs by adopting a k-neighbor search algorithm;
the real-time interaction module (101) is used for highlighting the abnormal area by utilizing the real-time interaction module and providing interaction functions.
2. The detection device based on a security monitoring static memory according to claim 1, wherein the three-dimensional reconstruction module (103) comprises a feature extraction module (1031), an image matching module (1032), a motion recovery and three-dimensional point cloud production module (1033), a global optimization module (1034), a point cloud fusion and surface reconstruction module (1035), a texture mapping module (1036) and a virtual interaction environment construction module (1037);
the feature extraction module (1031) is used for acquiring multi-source image data and extracting feature descriptors of the multi-source image data;
The image matching module (1032) is used for calculating the similarity between the images according to the feature descriptors and finding out feature point matching pairs existing between the images;
the motion recovery and three-dimensional point cloud production module (1033) is used for solving an intrinsic matrix between two adjacent images according to the characteristic point matching pair, recovering a motion relation between cameras, calculating three-dimensional coordinates of the matching characteristic points in the real world by using a triangulation method, and generating an initial three-dimensional point cloud;
the global optimization module (1034) is used for integrating the camera pose of the multi-source image data and the three-dimensional point cloud into a unified coordinate system and performing global optimization;
the point cloud fusion and surface reconstruction module (1035) is used for fusing the three-dimensional point cloud and extracting surface information of the three-dimensional model from the point cloud data through a surface reconstruction algorithm;
the texture mapping module (1036) is used for mapping the color information in the original image onto the surface of the three-dimensional model to obtain a reconstructed three-dimensional model;
the virtual interactive environment construction module (1037) is used for importing the reconstructed three-dimensional model into a virtual reality platform, setting the juxtaposition illumination and material properties, and constructing a virtual interactive environment.
3. The detection device based on a security monitoring static memory according to claim 2, wherein the motion restoration and three-dimensional point cloud production module (1033) comprises a feature description sub-extraction module, a feature point matching module, an essential matrix calculation module, a restoration camera motion module, a triangulation module and an initial three-dimensional point cloud generation module;
the feature descriptor extraction module is used for extracting a group of stable feature points from two adjacent pictures and calculating feature descriptors of each feature point;
the characteristic point matching module is used for finding out characteristic point pairs matched with each other in two adjacent pictures;
the essential matrix calculation module is used for calculating an essential matrix between two cameras under the condition that the internal parameters of the cameras are known;
the camera motion recovery module is used for obtaining a rotation matrix and a translation vector through the decomposition of the essential matrix, and further calculating the motion between the cameras;
the triangulation module is used for calculating three-dimensional coordinates of the feature point pairs in the real world by using a triangulation method according to the feature point pairs and the essence matrix;
the initial three-dimensional point cloud generation module is used for connecting all three-dimensional coordinates to generate an initial three-dimensional point cloud.
4. A detection device based on a security monitoring static memory according to claim 3, wherein the point cloud fusion and surface reconstruction module (1035) comprises a three-dimensional point cloud fusion module, a direction field calculation module and a triangular grid generation module;
the three-dimensional point cloud fusion module is used for fusing three-dimensional point clouds with different visual angles so that the three-dimensional point clouds are aligned in the same coordinate system;
the direction field calculation module is used for calculating the normal vector of each three-dimensional point cloud in the point cloud, projecting the point cloud onto a regular grid by using known normal vector information, calculating the direction field of each grid, and solving a poisson equation on the whole grid to obtain the height value of each grid unit;
the triangular grid generating module is used for constructing a continuous triangular grid according to the solved height value and outputting the triangular grid into a common three-dimensional model file format.
5. The detection device based on a security monitoring static memory according to claim 4, wherein the locally sensitive hash forest anomaly detection module (104) comprises a feature vector extraction module (1041), a locally sensitive hash module (1042) and an isolated forest anomaly detection module (1043);
The feature vector extraction module (1041) is configured to extract a feature vector of the static memory image data, and convert the static memory image data into a numerical vector with a fixed length;
the local sensitive hash module (1042) is used for mapping similar data points in the numerical vector into similar barrels by utilizing a local sensitive hash algorithm to obtain a hash table;
the isolated forest anomaly detection module (1043) is used for taking the hash table as input, and performing anomaly detection by utilizing an isolated forest algorithm to obtain an anomaly detection result.
6. The security monitoring static memory-based detection device of claim 5, wherein the locality sensitive hashing module (1042) comprises a locality sensitive hashing function family module, a hash value calculation module, a similar data point query module;
the local sensitive hash function family module is used for selecting a local sensitive hash function family and determining the number of hash functions to be used and the number of bits allocated to each hash bucket;
the hash value calculation module is used for traversing the input numerical value vector, calculating hash values of each data point under all hash functions, and combining the calculated hash values into a hash signature;
The similar data point query module is used for mapping the hash signature to a corresponding hash bucket, giving a query data point, calculating the hash signature of the given query data point, finding the corresponding bucket in the hash table, traversing all the data points in the bucket, calculating the similarity with the query point, and returning the data point with the highest similarity as an output result to obtain the hash table.
7. The detection device based on a security monitoring static memory according to claim 6, wherein the isolated forest anomaly detection module (1043) comprises a hash table conversion module, an isolated forest construction model, a path length calculation module and an anomaly point judgment module;
the hash table conversion module is used for converting the hash table into a characteristic vector set of data points, creating vectors with the length equal to the number of hash buckets for each data point of the characteristic vector set, and filling vector elements according to the distribution of the data points in the hash table;
the isolated forest construction model is used for creating and constructing an isolated forest model according to parameter setting;
the path length calculation module is used for traversing each tree in the isolated forest and calculating the path length of each data point on the tree;
The abnormal point judging module is used for converting the average path length of each data point on all trees into abnormal scores and setting a threshold value to judge whether the data point is abnormal or not.
8. The security monitoring static memory-based detection device according to claim 7, wherein the k-nearest neighbor search module (106) comprises an outlier query module (1061), an outlier distance calculation module (1062), a k-nearest neighbor selection module (1063), and a k-nearest neighbor analysis module (1064);
the abnormal data point query module (1061) is configured to determine an abnormal data point to be queried according to an abnormal detection result;
the outlier distance calculation module (1062) is configured to traverse each data point in the data set and calculate a distance between outlier data points using the morphological similarity distance;
the k-neighbor selection module (1063) is configured to rank the calculated distances, and select k data points with the smallest distances as k-neighbors of the outlier data points;
the k-neighbor analysis module (1064) is configured to analyze the position of the abnormal data point and other data points nearby according to the search result of the k-neighbor.
9. The security monitoring static memory-based detection device according to claim 8, wherein the distances between outlier data points are calculated using morphological similarity distances The formula of (2) is:
in the middle of，/>The Euclidean distance between abnormal points;
is the Manhattan distance between outliers;
is the absolute value of the sum of the dimensional differences between outliers.
10. A detection method based on a security monitoring static memory, for implementing detection of the static memory by a detection device based on the security monitoring static memory according to any one of claims 1 to 9, characterized in that the detection method comprises the following steps:
s1, capturing a static memory image of the surface of a wafer by using a panoramic camera at multiple angles to obtain multi-source image data;
s2, constructing a three-dimensional model by utilizing multi-source image data, and fusing the three-dimensional model into a virtual interaction environment;
s3, performing anomaly detection on the static memory image in the virtual interaction scene by using a local sensitive hash isolated forest algorithm;
s4, analyzing the area with the defects in the static memory image in the virtual interaction scene according to the abnormal detection result;
s5, inquiring the position where the data abnormality occurs by using a k-nearest neighbor search algorithm;
s6, highlighting the abnormal area by using the real-time interaction module and providing an interaction function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311083045.5A CN116823816B (en) | 2023-08-28 | 2023-08-28 | Detection equipment and detection method based on security monitoring static memory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311083045.5A CN116823816B (en) | 2023-08-28 | 2023-08-28 | Detection equipment and detection method based on security monitoring static memory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116823816A true CN116823816A (en) | 2023-09-29 |
| CN116823816B CN116823816B (en) | 2023-11-21 |
Family
ID=88122438
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311083045.5A Active CN116823816B (en) | 2023-08-28 | 2023-08-28 | Detection equipment and detection method based on security monitoring static memory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116823816B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117195607A (en) * | 2023-11-08 | 2023-12-08 | 苏州长光华芯光电技术股份有限公司 | Reconstruction method of micro-nano air hole three-dimensional model |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH08123946A (en) * | 1994-10-27 | 1996-05-17 | Toshiba Corp | Plant monitoring device |
| US20130020391A1 (en) * | 2011-01-24 | 2013-01-24 | Datalogic ADC, Inc. | Exception detection and handling in automated optical code reading systems |
| CN107944502A (en) * | 2017-12-06 | 2018-04-20 | 南京航空航天大学 | A kind of Outlier Detection Algorithm based on random Harsh |
| CN108961390A (en) * | 2018-06-08 | 2018-12-07 | 华中科技大学 | Real-time three-dimensional method for reconstructing based on depth map |
| US20200374720A1 (en) * | 2018-06-04 | 2020-11-26 | Jiangnan University | Method for Detecting Abnormal Data in Sensor Network |
| CN112053391A (en) * | 2020-09-11 | 2020-12-08 | 中德（珠海）人工智能研究院有限公司 | Monitoring and early warning method and system based on dynamic three-dimensional model and storage medium |
| CN114201374A (en) * | 2021-12-07 | 2022-03-18 | 华融融通（北京）科技有限公司 | Operation and maintenance time sequence data anomaly detection method and system based on hybrid machine learning |
| WO2022142855A1 (en) * | 2020-12-31 | 2022-07-07 | 深圳市优必选科技股份有限公司 | Loop closure detection method and apparatus, terminal device, and readable storage medium |
| CN115187753A (en) * | 2022-05-27 | 2022-10-14 | 国网冀北电力有限公司张家口供电公司 | Equipment abnormality detection method and device, storage medium and computer equipment |
| CN115266757A (en) * | 2022-08-26 | 2022-11-01 | 英特尔产品（成都）有限公司 | Defect detection apparatus and method |
| CN115641538A (en) * | 2022-10-31 | 2023-01-24 | 杭州励飞软件技术有限公司 | Equipment abnormity detection method and device, computer equipment and storage medium |
| CN115658940A (en) * | 2022-10-24 | 2023-01-31 | 国网上海市电力公司 | Method, device and storage medium for constructing and updating heterogeneous detection feature database |
| WO2023050620A1 (en) * | 2021-09-29 | 2023-04-06 | 西安交通大学 | Method and system for monitoring abnormal user behavior in distributed blockchain system |
| CN116291724A (en) * | 2022-12-29 | 2023-06-23 | 北京航空航天大学云南创新研究院 | Real-time early warning detection method and system for highway tunnel construction |
| CN116612243A (en) * | 2023-07-21 | 2023-08-18 | 武汉国遥新天地信息技术有限公司 | Method for inhibiting and processing abnormal points of three-dimensional track of optical motion capture system |
-
2023
- 2023-08-28 CN CN202311083045.5A patent/CN116823816B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH08123946A (en) * | 1994-10-27 | 1996-05-17 | Toshiba Corp | Plant monitoring device |
| US20130020391A1 (en) * | 2011-01-24 | 2013-01-24 | Datalogic ADC, Inc. | Exception detection and handling in automated optical code reading systems |
| CN107944502A (en) * | 2017-12-06 | 2018-04-20 | 南京航空航天大学 | A kind of Outlier Detection Algorithm based on random Harsh |
| US20200374720A1 (en) * | 2018-06-04 | 2020-11-26 | Jiangnan University | Method for Detecting Abnormal Data in Sensor Network |
| CN108961390A (en) * | 2018-06-08 | 2018-12-07 | 华中科技大学 | Real-time three-dimensional method for reconstructing based on depth map |
| CN112053391A (en) * | 2020-09-11 | 2020-12-08 | 中德（珠海）人工智能研究院有限公司 | Monitoring and early warning method and system based on dynamic three-dimensional model and storage medium |
| WO2022142855A1 (en) * | 2020-12-31 | 2022-07-07 | 深圳市优必选科技股份有限公司 | Loop closure detection method and apparatus, terminal device, and readable storage medium |
| WO2023050620A1 (en) * | 2021-09-29 | 2023-04-06 | 西安交通大学 | Method and system for monitoring abnormal user behavior in distributed blockchain system |
| CN114201374A (en) * | 2021-12-07 | 2022-03-18 | 华融融通（北京）科技有限公司 | Operation and maintenance time sequence data anomaly detection method and system based on hybrid machine learning |
| CN115187753A (en) * | 2022-05-27 | 2022-10-14 | 国网冀北电力有限公司张家口供电公司 | Equipment abnormality detection method and device, storage medium and computer equipment |
| CN115266757A (en) * | 2022-08-26 | 2022-11-01 | 英特尔产品（成都）有限公司 | Defect detection apparatus and method |
| CN115658940A (en) * | 2022-10-24 | 2023-01-31 | 国网上海市电力公司 | Method, device and storage medium for constructing and updating heterogeneous detection feature database |
| CN115641538A (en) * | 2022-10-31 | 2023-01-24 | 杭州励飞软件技术有限公司 | Equipment abnormity detection method and device, computer equipment and storage medium |
| CN116291724A (en) * | 2022-12-29 | 2023-06-23 | 北京航空航天大学云南创新研究院 | Real-time early warning detection method and system for highway tunnel construction |
| CN116612243A (en) * | 2023-07-21 | 2023-08-18 | 武汉国遥新天地信息技术有限公司 | Method for inhibiting and processing abnormal points of three-dimensional track of optical motion capture system |
Non-Patent Citations (3)
| Title |
|---|
| JORGE MEIRA；CARLOS EIRAS-FRANCO；VERÓNICA BOLÓN-CANEDO；GORETI MARREIROS；AMPARO ALONSO-BETANZOS: "Fast anomaly detection with locality-sensitive hashing and hyperparameter autotuning", INFORMATION SCIENCES, vol. 607, pages 1245 - 1264, XP087142825, DOI: 10.1016/j.ins.2022.06.035 * |
| 丁汀;颜登程;张以文;周珊: "融合煤矿多维时序数据的瓦斯异常检测算法", 计算机集成制造系统, no. 6, pages 211 - 219 * |
| 侯博文;葛唯益;纪建;苗启广: "利用局部敏感哈希提高隔离森林算法效率", 指挥信息系统与技术, no. 3, pages 26 - 32 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117195607A (en) * | 2023-11-08 | 2023-12-08 | 苏州长光华芯光电技术股份有限公司 | Reconstruction method of micro-nano air hole three-dimensional model |
| CN117195607B (en) * | 2023-11-08 | 2024-02-27 | 苏州长光华芯光电技术股份有限公司 | Reconstruction method of micro-nano air hole three-dimensional model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116823816B (en) | 2023-11-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wan et al. | Industrial image anomaly localization based on Gaussian clustering of pretrained feature | |
| Li et al. | Real-time pose estimation of deformable objects using a volumetric approach | |
| CA3121440A1 (en) | Assembly body change detection method, device and medium based on attention mechanism | |
| Tau et al. | Dense correspondences across scenes and scales | |
| CN107481279A (en) | A kind of monocular video depth map computational methods | |
| Zhao et al. | Learning deep network for detecting 3d object keypoints and 6d poses | |
| CN116823816B (en) | Detection equipment and detection method based on security monitoring static memory | |
| Fu et al. | Learning to reduce scale differences for large-scale invariant image matching | |
| CN111368733B (en) | Three-dimensional hand posture estimation method based on label distribution learning, storage medium and terminal | |
| CN115775220A (en) | Method and system for detecting anomalies in images using multiple machine learning programs | |
| Ekekrantz et al. | Adaptive iterative closest keypoint | |
| Yalic et al. | Automatic Object Segmentation on RGB-D Data using Surface Normals and Region Similarity. | |
| CN113011359B (en) | Method for simultaneously detecting plane structure and generating plane description based on image and application | |
| CN108564043B (en) | Human body behavior recognition method based on space-time distribution diagram | |
| Cao et al. | Dgecn++: A depth-guided edge convolutional network for end-to-end 6d pose estimation via attention mechanism | |
| Srivastava et al. | Drought stress classification using 3D plant models | |
| Li et al. | Rethinking scene representation: A saliency-driven hierarchical multi-scale resampling for RGB-D scene point cloud in robotic applications | |
| CN117351078A (en) | Target size and 6D gesture estimation method based on shape priori | |
| Cao et al. | Stable image matching for 3D reconstruction in outdoor | |
| CN115965749B (en) | Three-dimensional reconstruction equipment based on thunder vision fuses | |
| Liu et al. | Deep learning of directional truncated signed distance function for robust 3D object recognition | |
| Jisen | A study on target recognition algorithm based on 3D point cloud and feature fusion | |
| Liu et al. | Energy-based global ternary image for action recognition using sole depth sequences | |
| Chen et al. | PointMM: Point Cloud Semantic Segmentation CNN under Multi-Spatial Feature Encoding and Multi-Head Attention Pooling | |
| Li et al. | Feature-based SLAM for dense mapping |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |