CN117591688A - Inspection image filtering method and filtering device - Google Patents
Inspection image filtering method and filtering device Download PDFInfo
- Publication number
- CN117591688A CN117591688A CN202311298339.XA CN202311298339A CN117591688A CN 117591688 A CN117591688 A CN 117591688A CN 202311298339 A CN202311298339 A CN 202311298339A CN 117591688 A CN117591688 A CN 117591688A
- Authority
- CN
- China
- Prior art keywords
- image
- inspection
- data
- gallery
- patrol
- 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
- 238000007689 inspection Methods 0.000 title claims abstract description 230
- 238000001914 filtration Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000000605 extraction Methods 0.000 claims abstract description 32
- 238000003860 storage Methods 0.000 claims abstract description 31
- 238000013139 quantization Methods 0.000 claims abstract description 8
- 238000004364 calculation method Methods 0.000 claims description 21
- 238000012795 verification Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 9
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000012512 characterization method Methods 0.000 description 3
- 230000000750 progressive effect Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013506 data mapping Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/53—Querying
- G06F16/535—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/776—Validation; Performance evaluation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method and a device for filtering a patrol image, which solve the technical problems of low quality inspection efficiency and incomplete quality inspection coverage of the existing image. The method comprises the following steps: forming a feature extraction service, and performing feature extraction and quantization on the image to form image feature data; forming a gallery storage service, and establishing a storage structure and an index structure to respectively store images and image characteristic data to form patrol gallery updating; and forming an image check and repeat service, and checking and comparing the input image in a patrol drawing library according to the characteristic matching strategy. The index base of the image source data is established through the differences of the extracted image features, and a patrol gallery is formed, so that the searching process and the image features are effectively combined in the process of checking the increment images, the parallel searching performance and the patrol performance can be effectively maintained when the increment images are checked in batches, and the required patrol performance of the patrol gallery for continuously increasing the patrol images can be fully met.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a patrol image filtering method and a filtering device.
Background
In the prior art, in order to ensure the gas safety of gas users, related single specified staff need to regularly carry out safety inspection on user gas equipment (pipelines, flow meters, valves and the like) and gas consumption conditions, inspection staff need to use special gas mobile equipment (PDA) to photograph and save key positions in the security inspection process, and the number of terminal users and the scale of a pipe network are huge, so that new inspection images are increased more than millions of times per month. In order to ensure that the images uploaded by the inspection personnel meet the safety inspection regulations and are real and effective, the quality inspection team needs to perform image quality inspection work with 100% coverage rate so as to avoid the occurrence of fake behaviors such as repeated uploading of photos, uploading after cutting the same photos, and the like. For millions of incremental images, 100% full coverage of quality inspection cannot be achieved through manual comparison one by one under the condition that time and manpower are limited.
Disclosure of Invention
In view of the above problems, the embodiments of the present invention provide a method and a device for filtering inspection images, which solve the technical problems of low efficiency and insufficient coverage of quality inspection of existing images.
The inspection image filtering method provided by the embodiment of the invention comprises the following steps:
forming a feature extraction service, and performing feature extraction and quantization on the image to form image feature data;
forming a gallery storage service, and establishing a storage structure and an index structure to respectively store images and image characteristic data to form patrol gallery updating;
and forming an image check and repeat service, and checking and comparing the input image in a patrol drawing library according to the characteristic matching strategy.
In an embodiment of the present invention, the feature extraction service includes:
receiving a patrol image, and collecting time sequence information and geographic information in the patrol image to form field identification data of the collected patrol image;
extracting global feature descriptors and local feature descriptors of the inspection image through a DOLG model;
image feature data is formed using the feature descriptors.
In an embodiment of the present invention, the gallery storage service includes:
hashing the warehouse-in inspection image to form a unique identifier;
base64 encoding is carried out on the warehouse-in inspection image to form image character string data;
the image characteristic data is stored in an index structure to form retrieval data, and the image character string data is stored in a storage structure to form inspection data.
In an embodiment of the present invention, the image duplication checking service includes:
carrying out Hamming distance calculation one by one according to the image characteristic data of the input image and the retrieval data of the inspection gallery;
determining a first alternative inspection image set similar to the image characteristics of the input image according to the set first Hamming distance threshold interval;
and extracting the existing inspection image forming recommended comparison image set for manual verification according to the first alternative inspection image set.
In an embodiment of the present invention, the image duplication checking service includes:
extracting global feature data in image feature data of an input image and global feature data in search data of a patrol drawing library to perform hamming distance calculation one by one;
determining a second alternative inspection image set similar to the global image features of the input image according to the set second Hamming distance threshold interval;
extracting local feature data in the image feature data of the input image, and carrying out hamming distance calculation on the local feature data one by one in the search data of the second alternative inspection image set;
sorting the second alternative inspection image sets according to the Hamming distance, and determining a third alternative inspection image set similar to the local image features of the input image according to a set third Hamming distance threshold interval;
and extracting the existing inspection image forming recommended comparison image set for manual verification according to the third alternative inspection image set.
In an embodiment of the present invention, the image duplication checking service includes:
extracting global feature data in image feature data of an input image and global feature data in search data of a patrol drawing library to perform hamming distance calculation one by one;
determining a second alternative inspection image set similar to the global image features of the input image according to the set second Hamming distance threshold interval;
extracting local feature data in image feature data of an input image and local feature data in search data of a patrol drawing library to perform Hamming distance calculation one by one;
determining a fourth alternative inspection image set similar to the local image features of the input image according to the set fourth Hamming distance threshold interval;
sequencing the overlapped alternative inspection images in the second alternative inspection image set and the fourth alternative inspection image set to form a fifth alternative inspection image set;
and extracting the existing inspection image according to the selected prior-ordered alternative inspection images in the fifth alternative inspection image set to form a recommended comparison image set for manual verification.
The inspection image filtering device of the embodiment of the invention comprises:
the feature extraction module is used for forming feature extraction service, and performing feature extraction and quantization on the image to form image feature data;
the gallery updating module is used for forming a gallery storage service, establishing a storage structure and an index structure to respectively store images and image characteristic data to form patrol gallery updating;
and the characteristic check and repeat module is used for forming an image check and repeat service and checking and comparing the input image in the inspection gallery according to the characteristic matching strategy.
The inspection image filtering method provided by the embodiment of the invention comprises the following steps:
receiving a newly added inspection image, and forming corresponding newly added image feature data by utilizing feature extraction service;
performing traversal comparison on the newly added image characteristic data in a patrol gallery by using an image duplication checking service to determine an alternative patrol image set, and selecting and submitting manual verification judgment according to the number of recommended comparison images;
and orderly storing the image and the image characteristic data of the newly added inspection image judged to pass through by using the gallery storage service to form the increment updating of the inspection gallery.
The inspection image filtering device of the embodiment of the invention comprises:
the newly added image receiving module is used for receiving the newly added inspection image and forming corresponding newly added image feature data by utilizing the feature extraction service;
the newly added image checking and repeating module is used for performing traversal comparison on the newly added image characteristic data in the inspection gallery by utilizing an image checking and repeating service to determine an alternative inspection image set, and selecting and submitting manual check judgment according to the number of recommended comparison images;
and the newly added image warehousing module is used for orderly storing the image and the image characteristic data of the newly added inspection image judged by utilizing the gallery storage service to form the increment updating of the inspection gallery.
The inspection image filtering device of the embodiment of the invention comprises:
the memory is used for storing corresponding program codes in the processing process of the inspection image filtering method;
and a processor for executing the program code.
According to the inspection image filtering method and the filtering device, the index basis of the image source data is established through the differences of the extracted image features, the inspection gallery is formed, the searching process and the image features are effectively combined in the inspection process of the incremental images, when the incremental images are inspected in batches, the parallel searching performance and the inspection performance can be effectively maintained, and the inspection performance required by the continuous increase of the inspection images in the inspection gallery can be fully met.
Drawings
Fig. 1 is a flow chart of a method for filtering inspection images according to an embodiment of the invention.
Fig. 2 is a schematic diagram of an architecture of a patrol image filtering device according to an embodiment of the invention.
Fig. 3 is a flow chart of a method for filtering inspection images according to an embodiment of the invention.
Fig. 4 is a schematic diagram illustrating an application of the inspection image filtering method according to an embodiment of the invention.
Fig. 5 is a schematic flow chart of an inspection image filtering device according to an embodiment of the invention.
Detailed Description
The present invention will be further described with reference to the drawings and the detailed description below, in order to make the objects, technical solutions and advantages of the present invention more apparent. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
An inspection image filtering method according to an embodiment of the present invention is shown in fig. 1. In fig. 1, an embodiment of the present invention includes:
step 100: a feature extraction service is formed, and feature extraction and quantization are performed on the image to form image feature data.
The main image features of the digitized image, such as color distribution, texture orientation, or shape summarization, can be extracted using computer image processing techniques. The image acquisition environment is subjected to targeted feature extraction optimization, so that good information expression of field content in the image can be obtained. The extracted features are orderly quantized to form feature data, so that a high-efficiency and accurate data basis can be provided for comparison between images.
Step 200: and forming a gallery storage service, and establishing a storage structure and an index structure to respectively store the image and the image characteristic data to form an inspection gallery update.
The images include historical existing images and incremental images of the patrol. Ordered storage of image source data is formed by a storage structure. The image feature data is converted into valid index data of the image source data by the index structure. The storage structure and index structure of the patrol gallery may be formed using general database technology or the storage and retrieval characteristics of a particular general data structure. The image source data may be in the form of different bearers for collecting image information on site.
Step 300: and forming an image check and repeat service, and checking and comparing the input image in a patrol drawing library according to the characteristic matching strategy.
Feature matching strategies include, but are not limited to, feature alignment type definitions, feature alignment threshold settings, feature alignment logic settings, and the like. And forming adjustment of check-up comparison logic and efficiency through a feature matching strategy.
According to the inspection image filtering method, the index base of the image source data is established through the differences of the extracted image features, the inspection gallery is formed, the searching process and the image features are effectively combined in the inspection process of the incremental images, when the incremental images are inspected in batches, the parallel searching performance and the inspection performance can be effectively maintained, and the inspection performance required by the continuous increase of the inspection images in the inspection gallery can be fully met.
As shown in fig. 1, in an embodiment of the present invention, step 100 includes:
step 110: and receiving the inspection image, and acquiring time sequence information and geographic information in the inspection image to form field identification data of the acquired inspection image.
The time sequence information and the geographic information are formed by time stamp information and geographic coordinate information which are generated when the gas special mobile device executes a photographing instruction during the formation of the inspection image. The field identification data may be used as field feature data for the inspection image.
Step 120: and extracting global feature descriptors and local feature descriptors of the inspection image through a DOLG model.
The DOLG model is a depth-orthogonal based local and global feature fusion framework. The algorithm first concentrates and extracts representative local information by using a multi-scale convolution and self-attention method. Components orthogonal to the global image representation are then extracted from the local information. Finally, the orthogonal components are complementarily connected with the global representation, and then aggregated to generate the final characterization.
Specifically, 256-dimensional global characterization descriptors and 14x14 256-dimensional local characterization descriptors containing field appearance information such as textures, shapes and the like are extracted through a DOLG model.
Step 130: image feature data is formed using the feature descriptors.
And performing binary quantization conversion on all feature descriptors to obtain binary image feature data of a set long sequence. The image feature data includes global feature data of a binary long-length-formulated sequence (a first part of the binary long-length-formulated sequence) and local feature data of the binary long-length-formulated sequence (a second part of the binary long-length-formulated sequence).
In one embodiment of the invention, the field identification data is formatted as necessary as an additional component of the image source data.
In one embodiment of the invention, the field identification data is converted into a binary fixed length sequence as an additional component of the image characteristic data.
The inspection image filtering method mainly utilizes the global feature descriptors and the local feature descriptors to form formatted binary fixed length sequence representation image features, so that the image features of the inspection image have a universal basis for flexible application on the basis of complete identification and extraction.
As shown in fig. 1, in an embodiment of the present invention, step 200 includes:
step 210: and carrying out hash processing on the warehouse-in inspection image to form a unique identifier.
The hashing process may be performed using a hashing algorithm. The hashing process may include the field identification data of the warehouse-in inspection image.
Step 220: and performing base64 coding on the warehouse-in inspection image to form image character string data.
The warehouse-in inspection image is serialized, so that the storage structure can be simplified, and the introduction of complex data types is avoided. The field identification data and the image string data may form a data connection.
Step 230: the image characteristic data is stored in an index structure to form retrieval data, and the image character string data is stored in a storage structure to form inspection data.
And forming a patrol gallery by writing the image characteristic data and the image character string data to obtain incremental update.
According to the inspection image filtering method, the image features are associated with the search information through constructing the search data, and the technical effect of image feature comparison, namely image filtering is synchronously achieved. Meanwhile, the character string data is used for storing the image source data, so that the design difficulty and the operation cost of the inspection gallery are reduced.
As shown in fig. 1, in an embodiment of the present invention, step 300 includes:
step 310: and carrying out hamming distance calculation one by one according to the image characteristic data of the input image and the retrieval data of the inspection gallery.
By utilizing the characteristic of binary long sequence of image characteristic data, the efficient calculation of hamming distance between the image characteristic data of the input image and the image characteristic data in the retrieval data can be formed.
According to preset data setting rules, global feature data formed by global feature descriptors, local feature data formed by local feature descriptors and field identification data contained in image feature data of an input image are corresponding to the data types of the image feature data of the image source data corresponding to the inspection gallery.
Step 320: and determining a first alternative inspection image set similar to the image characteristics of the input image according to the set first Hamming distance threshold interval.
In an embodiment of the present invention, the first hamming distance threshold interval may be the first 5 candidate inspection images whose hamming distances are closest (i.e., most similar) to each other after traversal comparison with the search data.
Step 330: and extracting the existing inspection image forming recommended comparison image set for manual verification according to the first alternative inspection image set.
According to the inspection image filtering method, the alternative range of the similar image is rapidly reduced by utilizing automatic data filtering of the machine, and the similar judgment is accurately formed by utilizing expert experience in combination with manual verification, so that the duplicate inspection efficiency can be effectively improved.
As shown in fig. 1, in an embodiment of the present invention, step 300 further includes:
step 340: extracting global feature data in image feature data of an input image and global feature data in search data of a patrol gallery to perform hamming distance calculation one by one.
Step 345: and determining a second alternative inspection image set similar to the global image features of the input image according to the set second hamming distance threshold interval.
In an embodiment of the present invention, the second hamming distance threshold interval may be the first 100 candidate inspection images whose hamming distances are closest to the global feature data in the search data after traversal comparison.
Step 350: extracting local feature data in the image feature data of the input image, and carrying out Hamming distance calculation on the local feature data one by one in the retrieval data of the second alternative inspection image set.
Step 355: and sorting the second alternative inspection image sets according to the Hamming distance, and determining a third alternative inspection image set similar to the local image features of the input image according to a set third Hamming (hamming) distance threshold interval.
In an embodiment of the present invention, the third hamming distance threshold interval may be the first 3 candidate inspection images whose hamming distances are closest to the local feature data traversal comparison in the second candidate inspection image set.
Step 360: and extracting the existing inspection image forming recommended comparison image set for manual verification according to the third alternative inspection image set.
According to the inspection image filtering method, the alternative range of the similar image is rapidly reduced by utilizing the automatic data filtering of the secondary progressive machine, and the similar judgment is accurately formed by utilizing expert experience in combination with manual verification, so that the duplicate inspection efficiency can be effectively improved.
As shown in fig. 1, in an embodiment of the present invention, step 300 further includes:
step 370: extracting global feature data in image feature data of an input image and global feature data in search data of a patrol gallery to perform hamming distance calculation one by one.
Step 375: and determining a second alternative inspection image set similar to the global image features of the input image according to the set second hamming distance threshold interval.
In an embodiment of the present invention, the second hamming distance threshold interval may be the first 10 candidate inspection images whose hamming distances are closest to the global feature data in the search data after traversal comparison.
Step 380: extracting local feature data in image feature data of an input image and local feature data in search data of a patrol drawing library to perform hamming distance calculation one by one.
Step 385: and determining a fourth alternative inspection image set similar to the local image features of the input image according to the set fourth hamming distance threshold interval.
In an embodiment of the present invention, the fourth hamming distance threshold interval may be the first 10 candidate inspection images whose hamming distances are closest to the local feature data in the search data after traversal comparison.
Step 390: and sequencing the overlapped alternative inspection images in the second alternative inspection image set and the fourth alternative inspection image set to form a fifth alternative inspection image set.
And sequencing the overlapped alternative patrol images according to the sequencing weights of the second alternative patrol image set and the fourth alternative patrol image set. For example, the ranking weights of the second candidate inspection image set and the fourth candidate inspection image set are 0.3 and 0.6, and the ranking of one coincident candidate inspection image in the two sets is 8 and 2, and the ranking in the fifth candidate inspection image set is 8×0.3+2×0.6=3.6≡4.
Step 395: and extracting the existing inspection image according to the selected prior-ordered alternative inspection images in the fifth alternative inspection image set to form a recommended comparison image set for manual verification.
According to the inspection image filtering method, the alternative range of similar images is rapidly reduced by utilizing parallel difference image characteristic traversal retrieval machine automatic data filtering, and similar judgment is accurately formed by combining fine sorting and manual verification by utilizing expert experience, so that the inspection efficiency can be effectively improved.
In an embodiment of the present invention, the comparison accuracy of the recommended comparison image formed by the existing inspection image is improved by the combination of the three check-up comparison processes.
In practical application, the newly added inspection images with consistent content can be accurately compared and confirmed by using the image duplication checking service, and when the recommended comparison images are unique, manual verification is not needed.
An exemplary inspection image filtering apparatus according to the present invention is shown in fig. 2. In fig. 2, an embodiment of the present invention includes:
a feature extraction module 10, configured to form a feature extraction service, and perform feature extraction and quantization on an image to form image feature data;
the gallery updating module 20 is configured to form a gallery storage service, and establish a storage structure and an index structure to store the image and the image feature data respectively to form a patrol gallery update;
the feature check module 30 is used for forming an image check service, and checking and comparing the input images in the inspection gallery according to a feature matching strategy.
As shown in fig. 2, in an embodiment of the present invention, the feature extraction module 10 includes:
the on-site feature extraction unit 11 is used for receiving the inspection image, and collecting time sequence information and geographic information in the inspection image to form on-site identification data of the collected inspection image;
an image feature extraction unit 12, configured to perform global feature descriptor extraction and local feature descriptor extraction of the inspection image through the DOLG model;
an image feature encoding unit 13 for forming image feature data using the feature descriptors.
As shown in fig. 2, in an embodiment of the present invention, the gallery updating module 20 includes:
an image identification unit 21, configured to hash the warehouse-in inspection image to form a unique identifier;
the image conversion unit 22 is used for performing base64 encoding on the warehouse-in inspection image to form image character string data;
the data mapping unit 23 is configured to store the image feature data into the index structure to form search data, and store the image character string data into the storage structure to form inspection data.
As shown in fig. 2, in an embodiment of the present invention, the feature check module 30 includes:
a first comparing unit 31, configured to perform hamming (hamming) distance calculation one by one according to the image feature data of the input image and the search data of the inspection gallery;
a first filtering unit 32, configured to determine a first candidate inspection image set that is similar to an image feature of the input image according to a set first hamming distance threshold interval;
the first recommending unit 33 is configured to extract an existing inspection image forming recommended comparison image set for manual verification according to the first alternative inspection image set.
As shown in fig. 2, in an embodiment of the present invention, the feature searching module 30 further includes:
a second comparing unit 34a, configured to extract global feature data in the image feature data of the input image and perform hamming (hamming) distance calculation on the global feature data in the search data of the inspection gallery one by one;
a second filtering unit 34b, configured to determine a second set of candidate inspection images that is similar to the global image feature of the input image according to a set second hamming distance threshold interval;
a progressive comparison unit 34c, configured to extract local feature data in the image feature data of the input image, and perform hamming (hamming) distance calculation on the local feature data one by one in the search data of the second candidate inspection image set;
a progressive filtering unit 34d, configured to sort the second set of candidate inspection images according to hamming distances, and determine a third set of candidate inspection images that is similar to the local image features of the input image according to a set third hamming (hamming) distance threshold interval;
and the second recommending unit 34e is configured to extract an existing inspection image forming recommended comparison image set for manual verification according to the third alternative inspection image set.
As shown in fig. 2, in an embodiment of the present invention, the feature searching module 30 further includes:
a third comparing unit 37a, configured to extract global feature data in image feature data of the input image and perform hamming distance calculation on global feature data in search data of the inspection gallery one by one;
a third filtering unit 37b, configured to determine a second set of candidate inspection images that is similar to the global image feature of the input image according to a set second hamming distance threshold interval;
a fourth comparing unit 37c for extracting local feature data in the image feature data of the input image and local feature data in the search data of the inspection gallery, and performing hamming distance calculation one by one;
a fourth filtering unit 37d, configured to determine a fourth candidate patrol image set that is similar to the local image feature of the input image according to a set fourth hamming (hamming) distance threshold interval;
a parallel filtering unit 37e, configured to sort the candidate inspection images overlapped in the second candidate inspection image set and the fourth candidate inspection image set to form a fifth candidate inspection image set;
and the third recommending unit 37f is configured to extract an existing inspection image forming recommended comparison image set for manual verification according to the selected and ordered previous alternative inspection images in the fifth alternative inspection image set.
An inspection image filtering method according to an embodiment of the present invention is shown in fig. 3. In fig. 3, the service provided by the method for filtering the inspection image according to the embodiment of the present invention includes:
step 400: and receiving the newly added inspection image, and forming corresponding newly added image feature data by utilizing a feature extraction service.
Step 500: and traversing and comparing the newly added image characteristic data in the inspection gallery by using an image checking and repeating service to determine an alternative inspection image set, and selecting and submitting manual check judgment according to the number of recommended comparison images.
When the number of the alternative inspection images is more than 1, submitting a manual check to judge that the newly added inspection images pass according to the manual approval.
Step 600: and orderly storing the image and the image characteristic data of the newly added inspection image judged to pass through by using the gallery storage service to form the increment updating of the inspection gallery.
The inspection image filtering method of the embodiment of the invention forms a flexible processing process of inspection image filtering and inspection gallery updating. The method not only can be used for the check and repeat processing of the newly added inspection image, but also can be applied to the initial construction of the inspection gallery and the gallery combination. The formed gallery resources can be further utilized and developed as on-site dynamic resources.
An application of the inspection image filtering method according to an embodiment of the present invention is shown in fig. 4. In fig. 4, the current inspection picture is compared with the content of the historical inspection picture library, the highly similar historical picture is automatically found, and the inspection result is formed by the manual inspection of quality control, so that whether the current inspection picture is qualified or not is judged.
An exemplary inspection image filtering apparatus according to the present invention is shown in fig. 5. In fig. 5, an embodiment of the present invention includes:
the newly added image receiving module 40 is configured to receive the newly added inspection image, and form corresponding newly added image feature data by using a feature extraction service;
the newly added image checking and repeating module 50 is used for performing traversal comparison on the newly added image characteristic data in the inspection gallery by utilizing an image checking and repeating service to determine an alternative inspection image set, and selecting and submitting manual check judgment according to the number of recommended comparison images;
and the newly added image warehousing module 60 is used for orderly storing the image and the image characteristic data of the newly added inspection image judged to pass through by using the gallery storage service to form the increment update of the inspection gallery.
An embodiment of the invention provides a patrol image filtering device, comprising:
the memory is used for storing program codes corresponding to the processing process of the inspection image filtering method in the embodiment;
and the processor is used for executing the program codes corresponding to the processing procedure of the inspection image filtering method in the embodiment.
The processor may employ a DSP (Digital Signal Processor) digital signal processor, an FPGA (Field-Programmable Gate Array) Field programmable gate array, a MCU (Microcontroller Unit) system board, a SoC (system on a chip) system board, or an PLC (Programmable Logic Controller) minimum system including I/O.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
Claims (10)
1. The inspection image filtering method is characterized by comprising the following steps of:
forming a feature extraction service, and performing feature extraction and quantization on the image to form image feature data;
forming a gallery storage service, and establishing a storage structure and an index structure to respectively store images and image characteristic data to form patrol gallery updating;
and forming an image check and repeat service, and checking and comparing the input image in a patrol drawing library according to the characteristic matching strategy.
2. The inspection image filtering method of claim 1, wherein the feature extraction service comprises:
receiving a patrol image, and collecting time sequence information and geographic information in the patrol image to form field identification data of the collected patrol image;
extracting global feature descriptors and local feature descriptors of the inspection image through a DOLG model;
image feature data is formed using the feature descriptors.
3. The inspection image filtering method as claimed in claim 1, wherein the gallery storage service comprises:
hashing the warehouse-in inspection image to form a unique identifier;
base64 encoding is carried out on the warehouse-in inspection image to form image character string data;
the image characteristic data is stored in an index structure to form retrieval data, and the image character string data is stored in a storage structure to form inspection data.
4. The inspection image filtering method as claimed in claim 1, wherein the image inspection service comprises:
carrying out Hamming distance calculation one by one according to the image characteristic data of the input image and the retrieval data of the inspection gallery;
determining a first alternative inspection image set similar to the image characteristics of the input image according to the set first Hamming distance threshold interval;
and extracting the existing inspection image forming recommended comparison image set for manual verification according to the first alternative inspection image set.
5. The inspection image filtering method as claimed in claim 1, wherein the image inspection service comprises:
extracting global feature data in image feature data of an input image and global feature data in search data of a patrol drawing library to perform hamming distance calculation one by one;
determining a second alternative inspection image set similar to the global image features of the input image according to the set second Hamming distance threshold interval;
extracting local feature data in the image feature data of the input image, and carrying out hamming distance calculation on the local feature data one by one in the search data of the second alternative inspection image set;
sorting the second alternative inspection image sets according to the Hamming distance, and determining a third alternative inspection image set similar to the local image features of the input image according to a set third Hamming distance threshold interval;
and extracting the existing inspection image forming recommended comparison image set for manual verification according to the third alternative inspection image set.
6. The inspection image filtering method as claimed in claim 1, wherein the image inspection service comprises:
extracting global feature data in image feature data of an input image and global feature data in search data of a patrol drawing library to perform hamming distance calculation one by one;
determining a second alternative inspection image set similar to the global image features of the input image according to the set second Hamming distance threshold interval;
extracting local feature data in image feature data of an input image and local feature data in search data of a patrol drawing library to perform Hamming distance calculation one by one;
determining a fourth alternative inspection image set similar to the local image features of the input image according to the set fourth Hamming distance threshold interval;
sequencing the overlapped alternative inspection images in the second alternative inspection image set and the fourth alternative inspection image set to form a fifth alternative inspection image set;
and extracting the existing inspection image according to the selected prior-ordered alternative inspection images in the fifth alternative inspection image set to form a recommended comparison image set for manual verification.
7. An inspection image filtering device, comprising:
the feature extraction module is used for forming feature extraction service, and performing feature extraction and quantization on the image to form image feature data;
the gallery updating module is used for forming a gallery storage service, establishing a storage structure and an index structure to respectively store images and image characteristic data to form patrol gallery updating;
and the characteristic check and repeat module is used for forming an image check and repeat service and checking and comparing the input image in the inspection gallery according to the characteristic matching strategy.
8. The inspection image filtering method is characterized by comprising the following steps of:
receiving a newly added inspection image, and forming corresponding newly added image feature data by utilizing feature extraction service;
performing traversal comparison on the newly added image characteristic data in a patrol gallery by using an image duplication checking service to determine an alternative patrol image set, and selecting and submitting manual verification judgment according to the number of recommended comparison images;
and orderly storing the image and the image characteristic data of the newly added inspection image judged to pass through by using the gallery storage service to form the increment updating of the inspection gallery.
9. An inspection image filtering device, comprising:
the newly added image receiving module is used for receiving the newly added inspection image and forming corresponding newly added image feature data by utilizing the feature extraction service;
the newly added image checking and repeating module is used for performing traversal comparison on the newly added image characteristic data in the inspection gallery by utilizing an image checking and repeating service to determine an alternative inspection image set, and selecting and submitting manual check judgment according to the number of recommended comparison images;
and the newly added image warehousing module is used for orderly storing the image and the image characteristic data of the newly added inspection image judged by utilizing the gallery storage service to form the increment updating of the inspection gallery.
10. An inspection image filtering device, comprising:
a memory for storing corresponding program codes in the processing procedure of the inspection image filtering method according to any one of claims 1 to 6 and 8;
and a processor for executing the program code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311298339.XA CN117591688A (en) | 2023-10-09 | 2023-10-09 | Inspection image filtering method and filtering device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311298339.XA CN117591688A (en) | 2023-10-09 | 2023-10-09 | Inspection image filtering method and filtering device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117591688A true CN117591688A (en) | 2024-02-23 |
Family
ID=89918944
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311298339.XA Pending CN117591688A (en) | 2023-10-09 | 2023-10-09 | Inspection image filtering method and filtering device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117591688A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113204665A (en) * | 2021-04-28 | 2021-08-03 | 北京百度网讯科技有限公司 | Image retrieval method, image retrieval device, electronic equipment and computer-readable storage medium |
| CN115052160A (en) * | 2022-04-22 | 2022-09-13 | 江西中烟工业有限责任公司 | Image coding method and device based on cloud data automatic downloading and electronic equipment |
| CN115129915A (en) * | 2021-03-26 | 2022-09-30 | 北京新氧科技有限公司 | Repeated image retrieval method, device, equipment and storage medium |
| CN116304150A (en) * | 2023-03-22 | 2023-06-23 | 马上消费金融股份有限公司 | Retrieval method and device, electronic equipment and computer readable storage medium |
-
2023
- 2023-10-09 CN CN202311298339.XA patent/CN117591688A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115129915A (en) * | 2021-03-26 | 2022-09-30 | 北京新氧科技有限公司 | Repeated image retrieval method, device, equipment and storage medium |
| CN113204665A (en) * | 2021-04-28 | 2021-08-03 | 北京百度网讯科技有限公司 | Image retrieval method, image retrieval device, electronic equipment and computer-readable storage medium |
| CN115052160A (en) * | 2022-04-22 | 2022-09-13 | 江西中烟工业有限责任公司 | Image coding method and device based on cloud data automatic downloading and electronic equipment |
| CN116304150A (en) * | 2023-03-22 | 2023-06-23 | 马上消费金融股份有限公司 | Retrieval method and device, electronic equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109493320B (en) | Remote sensing image road extraction method and system based on deep learning, storage medium and electronic equipment | |
| CN111062361B (en) | River and lake shoreline sand production monitoring and analyzing method and device | |
| CN112949413A (en) | City landscape element classification and locality measurement method based on street view picture | |
| CN101350012A (en) | Method and system for matching address | |
| CN114862169A (en) | BIM-based tunnel engineering safety risk monitoring system and method | |
| CN111797188B (en) | Urban functional area quantitative identification method based on open source geospatial vector data | |
| CN109857782A (en) | A kind of Monitor of Logging Data Processing System | |
| CN116522272A (en) | Multi-source space-time data transparent fusion method based on urban information unit | |
| CN115809833A (en) | Intelligent monitoring method and device for capital construction project based on portrait technology | |
| CN113409001A (en) | Method for controlling automatic pricing of construction engineering quantity list | |
| CN114463499B (en) | Building three-dimensional modeling method and device | |
| Wang et al. | Automatic identification and location of tunnel lining cracks | |
| CN115937708A (en) | High-definition satellite image-based roof information automatic identification method and device | |
| CN109102303B (en) | Risk detection method and related device | |
| CN112199376B (en) | Standard knowledge base management method and system based on cluster analysis | |
| CN113487638A (en) | Ground feature edge detection method of high-precision semantic segmentation algorithm U2-net | |
| CN117591688A (en) | Inspection image filtering method and filtering device | |
| Li et al. | Automated bridge crack detection based on improving encoder–decoder network and strip pooling | |
| CN115829209A (en) | Environment-friendly intelligent warehouse environment-friendly quality analysis method and device based on carbon path | |
| CN111078683A (en) | Interpolation search-based power grid ledger data filling and counting method and device | |
| CN113486105B (en) | CIM-oriented natural resource history business data integration method and device | |
| CN114862053A (en) | Material supplier pushing method, system and equipment | |
| CN105824871A (en) | Picture detecting method and equipment | |
| CN118378103B (en) | Geographic information system data matching management method based on artificial intelligence | |
| CN115713664B (en) | Intelligent marking method and device for fire inspection and acceptance |
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 |