CN115017347A - Hidden danger image processing method and system fusing Gaussian algorithm and Hash algorithm - Google Patents
Hidden danger image processing method and system fusing Gaussian algorithm and Hash algorithm Download PDFInfo
- Publication number
- CN115017347A CN115017347A CN202210452636.4A CN202210452636A CN115017347A CN 115017347 A CN115017347 A CN 115017347A CN 202210452636 A CN202210452636 A CN 202210452636A CN 115017347 A CN115017347 A CN 115017347A
- Authority
- CN
- China
- Prior art keywords
- image
- images
- algorithm
- similarity
- hash
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- 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
- G06F16/55—Clustering; Classification
-
- 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
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Library & Information Science (AREA)
- Image Analysis (AREA)
Abstract
The method comprises the steps of firstly identifying a picture with high similarity by adopting the Gaussian algorithm, then calculating the redundancy of the picture by the algorithm redundancy algorithm according to the similarity, rejecting the redundant picture, and storing an effective picture, thereby reducing the storage space and saving the storage space. The problem of traditional distributed storage data insecure, nonstandard is solved, simultaneously, realize the quick inquiry and the retrieval of image based on hash algorithm, can be used for transmission line image that visual monitoring shoot device, artifical shooting, unmanned aerial vehicle shot to save and retrieve, can reduce the human cost, improve the efficiency of data access.
Description
Technical Field
The disclosure relates to the technical field of power transmission line image management and analysis, in particular to a hidden danger image processing method and system fusing a Gaussian algorithm and a Hash algorithm.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The hidden troubles of the power transmission line channel are various in types, large in quantity and rapid in change, are main risk sources causing the fault tripping of the power transmission line, and bring great challenges to the safe and reliable operation of the power transmission line and a power grid. Simultaneously, along with unmanned aerial vehicle patrols and examines, visual monitoring is taken the device, the grid protects the normalized application of new technological means such as line personnel removal patrolling and examining platform, transmission line passageway hidden danger data source mode is abundant, has accumulated a large amount of multidimension degree image data simultaneously. However, the data representing the cause of the failure that is really effective in the massive monitoring data is very limited, and the shooting data is likely to be difficult to distinguish visually due to the influence of hardware equipment and shooting conditions. Because the effective data proportion is small and the effectiveness can not be ensured, the problem is also caused for manual monitoring, the time for fault processing is increased to a certain extent, and the efficiency of fault removal is reduced. With the acceleration of the process of power grid transformation, the coverage rate of monitoring equipment will be increased continuously in the future, and the monitoring data will also be increased rapidly.
The inventor finds that the influence of factors such as severe weather and the like on complex terrain, numerous roads and construction projects of the district where power transmission work area patrol personnel are located brings huge challenges to safe and reliable operation of the power transmission line. At present, transmission and inspection personnel still mainly carry out data processing and analysis through traditional methods such as manual import and table query, and the manual processing of mass data has the disadvantages of low working efficiency, high labor cost and difficulty in guaranteeing the integrity and reliability of hidden danger data.
Disclosure of Invention
In order to solve the problems, the hidden danger image processing method and system fusing the Gaussian algorithm and the Hash algorithm are provided, the intelligent storage and management method is provided for the hidden danger images of the power transmission lines from different sources, the problems that traditional distributed storage data are unsafe and irregular are solved, meanwhile, fast query and retrieval of the images are achieved based on the Hash algorithm, the method and system can be used for storing and retrieving the images of the power transmission lines shot by a visual monitoring device, a manual shooting device and an unmanned aerial vehicle, the labor cost can be reduced, and the data access efficiency is improved.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
one or more embodiments provide a hidden danger image processing method fusing a Gaussian algorithm and a Hash algorithm, and the hidden danger image processing method comprises an image data storage method, wherein the image data storage method comprises the following steps:
acquiring a power transmission line inspection image to be stored;
identifying the power transmission line inspection images to be stored by adopting a Gaussian algorithm to obtain matching data and similarity of characteristic points between the images;
calculating the redundancy between the images by combining trial matching and a minimum spanning tree algorithm according to the feature point matching data and the similarity between the images to obtain a redundant image of the image to be detected;
and (4) obtaining an image set T from the power transmission line inspection image with the redundant image deleted, and storing the image set T in a classified manner.
One or more embodiments provide a hidden danger image processing system fusing a gaussian algorithm and a hash algorithm, comprising an image data storage module and an image retrieval module, wherein the data storage module is configured to perform the following steps:
acquiring a power transmission line inspection image to be stored;
identifying the power transmission line inspection images to be stored by adopting a Gaussian algorithm to obtain matching data and similarity of characteristic points between the images;
calculating the redundancy between the images by combining trial matching and a minimum spanning tree algorithm according to the feature point matching data and the similarity between the images to obtain a redundant image of the image to be detected;
and (4) carrying out classified storage on the power transmission line inspection image after the redundant image is deleted.
An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions, when executed by the processor, performing the steps of the above method.
Compared with the prior art, the beneficial effect of this disclosure is:
according to the method, the pictures with high similarity are firstly identified by adopting a Gaussian algorithm, then the redundancy of the pictures is calculated according to the similarity through the algorithm redundancy algorithm, the redundant pictures can be eliminated, and the effective pictures are stored, so that the storage capacity is reduced, the storage space is saved, meanwhile, through the similarity calculation, the redundant pictures of the pictures are searched based on the similarity, the effectiveness of image storage can be further improved, meanwhile, the mistaken deletion is avoided, and the image storage efficiency is improved.
Advantages of the present disclosure, as well as advantages of additional aspects, will be described in detail in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure and do not constitute a limitation thereof.
FIG. 1 is a flow chart of a processing method in embodiment 1 of the present disclosure;
Detailed Description
The present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise. It should be noted that, in the case of no conflict, the embodiments and features in the embodiments in the present disclosure may be combined with each other. The embodiments will be described in detail below with reference to the accompanying drawings.
Example 1
In one or more technical solutions disclosed in the embodiments, as shown in fig. 1, a hidden danger image processing method fusing a gaussian algorithm and a hash algorithm includes an image data storage method and an image retrieval method, where the data storage method includes the following steps:
a1, acquiring a power transmission line inspection image to be stored;
a2, identifying the power transmission line inspection images to be stored by adopting a Gaussian algorithm to obtain matching data and similarity of feature points between the images;
step A3, according to the feature point matching data and the similarity between the images, combining the trial matching and the minimum spanning tree algorithm to calculate the redundancy between the images, and obtaining the redundant image of the image to be detected;
and A4, obtaining an image set T from the power transmission line inspection image with the redundant image deleted, and storing the image set T in a classified manner.
In the embodiment, the pictures with high similarity are firstly identified by adopting a Gaussian algorithm, then the redundancy of the images is calculated by the algorithm redundancy algorithm according to the similarity, the redundant images can be eliminated, and the effective pictures are stored, so that the storage capacity is reduced, the storage space is saved, meanwhile, the redundant images of the images are searched based on the similarity through the similarity calculation, the effectiveness of image storage can be further improved, meanwhile, the mistaken deletion is avoided, and the image storage efficiency is improved.
In step a2, the inputs are: the hidden danger of the power transmission line patrols and examines the image and the similarity threshold value alpha; wherein, whether the picture has high similarity is judged by a threshold value. The output is: all similarity-processed image sets. The method mainly comprises the step of identifying the picture with high similarity through a Gaussian algorithm. The main process comprises the following steps: constructing a scale space, positioning key points, distributing directions, describing feature points, matching feature vectors and calculating similarity.
The gaussian algorithm steps are as follows:
(1) constructing a scale space, comprising: performing high-speed convolution on the image to obtain pixel points corresponding to each pixel position, detecting extreme points, and constructing a Gaussian pyramid;
firstly, a Gaussian pyramid is constructed, the scale space of an image is defined as the Gaussian convolution with variable scale, and the Gaussian convolution is carried out on each pixel position of the image as follows:
L(x,y,σ)＝G(x,y,σ)*I(x,y)
wherein, the first and the second end of the pipe are connected with each other,
in the formula, sigma is a scale space factor and represents the degree of Gaussian smoothing of an image; (x, y) represents the position of an image pixel; and m and n represent the dimension of the Gaussian template.
Optionally, for the power transmission line image, the gaussian smoothing is performed by using sigma with different sizes. Meanwhile, the sampling images are divided into different groups, each group is provided with a plurality of images, and the length and the width of the previous group of images can be 2 times of the width of the next group of images.
Secondly, a gaussian difference pyramid is constructed. Wherein, the difference operator in the Gaussian difference scale is as follows:
and finally, detecting an extreme point: and comparing the pixel point of the image to be detected with the pixel point of the adjacent image, and detecting to obtain an extreme point. In the step of extreme point detection, the extreme point in the discrete space is detected, and is not necessarily the true extreme point.
Optionally, the gray values of 26 neighborhood pixels of the image to be detected and the two previous and next images are compared one by one, and an extreme value is detected.
(2) Positioning key points: and calculating extreme value offset aiming at extreme value points in the Gaussian difference pyramid, and identifying stable extreme value points as key points according to a set offset threshold. The true extreme points can be determined by locating the keypoints.
Specifically, the extremum offset can be obtained by using a Taylor expansion.
Optionally, the offset threshold may be set to 0.7, and when the offset of any dimension is greater than 0.7, the position of the current keypoint is changed, and fitting is repeated at the new position until convergence. If the set iteration number is exceeded or the absolute value of the offset is too small, an unstable point exists, and the point can be regarded as a non-extreme point.
In this embodiment, the extreme point in the discrete space is not a true extreme point, and in order to improve the stability of the key point, repeated fitting is performed through the fitting scale space function, and the extreme value offset is obtained by using the Taylor expansion equation, so that the accuracy of key point positioning is improved.
(3) The distribution direction: distributing directions for the key points of the high-difference pyramid, establishing histograms in different directions according to radians of the gradient directions of the key points, and determining the feature points.
Specifically, a direction can be assigned to each point according to a local characteristic calculation result of the key points in the gaussian difference pyramid, so that the point has rotational invariance. The gradient model and direction of the local property calculation are as follows:
in the formula: the positive directions of x and y are respectively right and upper; l is a gray value of the key point mapped in the scale space; m (x, y) is the gradient magnitude; theta (x, y) is the radian of the gradient direction in which the key point is located.
360 degrees can be divided into 36 areas in turn according to the counterclockwise direction, histograms in different directions are obtained, and feature points are obtained. Specifically, m (x, y) may be added according to a gaussian distribution with σ equal to 1.5_ octv and a rule of 3 σ, a neighborhood window radius is 3 × 1.5_ octv, and only a direction with a peak value greater than 80% of a main direction peak value is reserved as a secondary direction of the key point. And finishing the process, namely obtaining SIFT feature points.
(4) Describing feature points: and adding each piece of directional gradient information to each feature point as a feature point description.
Optionally, gradient information in 8 directions can be calculated in a4 × 4 window in a keypoint scale space, and 128-dimensional vector characterization is performed, that is, the descriptor of the keypoint is obtained.
(5) Matching the feature vectors: and matching the feature vectors by adopting a nearest neighbor distance method, and matching the feature vectors with the feature points.
Calculating Euclidean distances between the feature vectors of the two neighborhood points aiming at each feature point to be detected, wherein if the Euclidean distances are smaller than a set threshold value, the feature vectors are successfully matched;
specifically, the set threshold of the euclidean distance may be 0.6, and the ratio of the euclidean distance between the sampling point and the feature vectors of the two neighboring points is calculated and compared with the set threshold of 0.6. If the ratio is smaller than the threshold, the feature vector matching is determined to be successful.
(6) Calculating the similarity: and calculating the similarity between the images according to the matching number of the image feature points.
The number of point feature matches represents the similarity between images, and the greater the number of feature point matches, the higher the similarity between representative images, and the greater the overlap area. The feature matching relationship is that the more the number of the relationship feature matches between the feature matches in step (5) is, the higher the accuracy is, the more accurate the obtained transformation relationship is, but the higher the performance requirement on the computer is, and the similarity between the images is defined as follows:
s (i, j) represents the similarity between images i and j, n i Representing the number of feature points in image i, n j Representing the number of the characteristic points in the image j, if the characteristic point x is successfully matched with the characteristic point y, M xy 1, otherwise M xy ＝0。
The embodiment provides an algorithm for quickly calculating the image redundancy, the algorithm adopts the previously calculated feature points to perform feature matching, combines trial matching and a minimum spanning tree algorithm, and iteratively detects a potential similarity relation, thereby improving the calculation efficiency to a certain extent.
The steps of the similarity calculation and the redundancy calculation in the embodiment can be executed in parallel, and the splicing efficiency is further improved. And the similarity matrix of the image set is the basis of a subsequent redundant image screening algorithm.
Optionally, a similarity matrix of the image set is constructed according to the calculated similarity
In step a3, the method for obtaining a redundant image of an image to be detected by calculating redundancy between images according to feature point matching data and similarity between images in combination with trial matching and a minimum spanning tree algorithm specifically comprises:
step A31, judging whether the image set grouped according to the similarity has an area with over-high overlapping degree according to the set similarity threshold; wherein the image set is: a set consisting of corresponding images in a similarity matrix constructed based on the similarity;
step A32, if there is an area with too high overlap, determining two images with the highest similarity in the area, calculating the similarity between the rest images in the image set and the two images, and deleting the images with the similarity larger than the set similarity threshold.
The specific implementation steps of step a3 are as follows:
1) setting a similarity threshold;
the similarity threshold is used for judging whether an overlapping area between two images in the image set is too high or not;
2) obtaining two images (including an image a and an image b) with the maximum similarity in the current image set, judging whether the similarity values of the two images are larger than a similarity threshold value, if so, representing that a local area with an excessively high overlapping area exists in the image set, and turning to the step 3), otherwise, representing that no excessively high overlapping area exists between all adjacent images in the image set, and turning to the step 5);
3) according to the formula
Analyzing the similarity relation between the image a and the image b and the comparison image, and according to the analysis result, obtaining the similarity relation between the imagesJudging one image into a redundant image, removing a current image set, and turning to the step 4); wherein, a and b represent two images with the maximum similarity in the current image set, and n represents the number of the images.
4) Updating the similarity matrix of the current image set, and turning to the step 2);
5) and stopping the screening algorithm and outputting the current image set T.
The further technical scheme includes that a distributed storage platform is constructed, and optionally, the distributed storage platform comprises an online storage area and a near-line storage area.
The storage device for online storage and the stored data are kept in an online state all the time, so that the data can be read by a user at will, and the requirement of a computing module on high speed of data access is met. The online storage area is used for storing data which is being analyzed, data which provides support for online calculation, data accessed at high frequency and the like;
near line storage is used to store data that is not used often, or that has not been accessed in large quantities. Among them, incoming storage requires fast addressing and high transfer rate. The near-line memory area is used for storing the analyzed data of the low-frequency access.
The online storage and the near-line storage can be divided according to the storage configuration instruction or the data acquisition time.
Step A4, the power transmission line inspection images with the redundant images deleted are classified and stored; the classified storage can be implemented by constructing a power transmission line storage model for storage, namely, the classified storage is implemented according to classification standards contained in the power transmission line storage model, and the classification standards of the power transmission line storage model comprise: the method comprises the steps of power transmission line account information, power transmission line channel types, terrains, power transmission line channel attributions, information image shooting types and defect types.
Aiming at the image shooting type, three types of images including an unmanned aerial vehicle inspection image, a manual inspection shooting image and a monitoring device automatic monitoring shooting image are established, and the naming can adopt a unified rule during storage: "line name- # bar number-shooting time".
Optionally, the power transmission line ledger information may include a line name, a voltage class, and a tower coordinate.
Optionally, the type of the power transmission line channel condition monitoring module includes: the fast growing region of trees and bamboos, the region where floating objects are easy to generate and break outside, the region where mountain (smoke) fire is high, the three-span section, the fishing region and the farmland.
Optionally, the defect types include: accessory facilities, stay wires, foundations, towers, hardware fittings, accessory equipment and grounding.
The above steps of this embodiment are steps of outputting and storing, and this embodiment further includes a step of quickly reading stored data, that is, an image retrieval method, including the following steps:
step B1, feature extraction: performing feature extraction on the images of the acquired and constructed image database to construct an image feature database;
the features of all the images in the image database are extracted one by one, each image is represented by a vector, for example, 512 features are extracted, and each image is represented by a 512-dimensional vector and added to the feature library.
Step B2, Hash encoding: training and learning a Hash function according to the images of the image feature library; coding the images in the stored image set T based on the trained hash function to obtain the hash code of each stored image;
the image feature library is used for constructing a training set and a testing set for training and testing.
The training set is used for training the learning hash function to construct a hash function set H (x) h 1 (x),h 2 (x),…,h x (x) (ii) a Substituting the hash function set obtained through learning into each feature in the test set to obtain a corresponding hash code for testing precision and recall;
step B3, Hamming distance identification: and carrying out Hash coding on the image to be inquired according to a Hash function to obtain a Hash code of the image to be inquired, calculating the Hash code of the image to be inquired and the Hash codes corresponding to the images in the image set T, and carrying out Hamming distance calculation one by one, wherein the image corresponding to the minimum Hamming distance is used as a retrieval result.
Further, the method also comprises the following step of rearranging according to the Hamming distance: reordering the images with the Hamming distance not greater than a certain set value obtained in the step B3 to output the set number, and outputting the corresponding images from small to large according to the Hamming distance as the retrieval result; this step is an optional step, which is typical in large-scale image retrieval applications. This step may use euclidean distance as a similarity measure, with the top M results obtained in step (3); or selecting results with Hamming distance not more than a certain set value, reordering and outputting the first M retrieval results.
In the data retrieval method of the embodiment, the image retrieval based on the hash utilizes a mapping method, the original image is represented by the low-dimensional binary hash code, the direct storage of the high-dimensional image feature vector is avoided, and the generated hash code is utilized to construct the index, so that the feature storage overhead is reduced, and the efficiency of feature matching is improved.
Example 2
Based on embodiment 1, the present embodiment provides a hidden danger image processing system fusing a gaussian algorithm and a hash algorithm, including an image data storage module and an image retrieval module, where the data storage module is configured to execute the following steps:
a1, acquiring a power transmission line inspection image to be stored;
step A2, identifying the power transmission line inspection images to be stored by adopting a Gaussian algorithm to obtain matching data and similarity of characteristic points between the images;
step A3, according to the feature point matching data and the similarity between the images, combining the trial matching and the minimum spanning tree algorithm to calculate the redundancy between the images, and obtaining the redundant image of the image to be detected;
and A4, obtaining an image set T from the power transmission line inspection image with the redundant image deleted, and storing the image set T in a classified manner.
Further, the image retrieval module is configured to perform the steps of:
step B1, feature extraction: extracting the characteristics of the images obtained from the image database to construct an image characteristic library;
step B2, Hash encoding: training and learning a Hash function according to the images of the image feature library; coding the images in the stored image set T based on the trained hash function to obtain the hash code of each stored image;
step B3, Hamming distance identification: and carrying out hash coding on the image to be queried according to a hash function to obtain a hash code of the image to be queried, calculating the hash code of the image to be queried and hash codes corresponding to the images in the image database, and carrying out Hamming distance calculation one by one, wherein the image corresponding to the minimum Hamming distance is used as a retrieval result.
Further, the method also comprises the following step of rearranging according to the Hamming distance: reordering the images with the Hamming distance not more than a certain set value obtained in the step B3, and outputting the corresponding images from small to large according to the Hamming distance as a retrieval result; this step enables a plurality of search images to be output for selection by the searcher.
Example 3
Based on embodiment 1, this embodiment provides an electronic device, which includes a memory, a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps of the method of embodiment 1 are completed.
The memory is a distributed storage platform, and optionally, the distributed storage platform includes an online storage area and a near-line storage area.
The storage device for online storage and the stored data are kept in an online state all the time, so that the data can be read by a user at will, and the requirement of a computing module on high speed of data access is met. The online storage area is used for storing data which is being analyzed, data which provides support for online calculation, data accessed at high frequency and the like;
near line storage is used to store data that is not used often, or that has not been accessed in large quantities. Among them, incoming storage requires fast addressing and high transfer rate. The near-line memory area is used for storing the analyzed data of the low-frequency access.
The online storage and the near-line storage can be divided according to the storage configuration instruction or the data acquisition time.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Although the embodiments of the present disclosure have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present disclosure, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive changes in the technical solutions of the present disclosure.
Claims (10)
1. The hidden danger image processing method fusing the Gaussian algorithm and the Hash algorithm is characterized by comprising an image data storage method, wherein the image data storage method comprises the following steps:
acquiring a power transmission line inspection image to be stored;
identifying the power transmission line inspection images to be stored by adopting a Gaussian algorithm to obtain matching data and similarity of characteristic points between the images;
calculating the redundancy between the images by combining trial matching and a minimum spanning tree algorithm according to the feature point matching data and the similarity between the images to obtain a redundant image of the image to be detected;
and (4) obtaining an image set T from the power transmission line inspection image with the redundant image deleted, and storing the image set T in a classified manner.
2. The hidden danger image processing method combining the gaussian algorithm and the hash algorithm as claimed in claim 1, further comprising an image retrieval method, wherein the image retrieval method comprises the steps of:
extracting the characteristics of the images obtained from the image database to construct an image characteristic library;
training and learning a hash function according to the images of the image feature library; coding the images in the stored image set T based on the trained hash function to obtain the hash code of each stored image;
and carrying out Hash coding on the image to be inquired according to a Hash function to obtain a Hash code of the image to be inquired, calculating the Hash code of the image to be inquired and the Hash codes corresponding to the images in the image set T, and carrying out Hamming distance calculation one by one, wherein the image corresponding to the minimum Hamming distance is used as a retrieval result.
3. The hidden danger image processing method combining gaussian algorithm and hash algorithm as claimed in claim 2, wherein images with hamming distances not greater than a certain set value obtained by calculation are reordered, and a set number of images are outputted as a search result from small to large according to the hamming distances.
4. The hidden danger image processing method fusing the gaussian algorithm and the hash algorithm according to claim 1, wherein the method for identifying the power transmission line inspection images to be stored by adopting the gaussian algorithm to obtain matching data of feature points between the images and similarity comprises the following steps:
constructing a scale space, comprising: performing high-speed convolution on the image to obtain pixel points corresponding to each pixel position, detecting extreme points, and constructing a Gaussian pyramid;
calculating extreme value offset aiming at extreme value points in the Gaussian difference pyramid, and identifying stable extreme value points as key points according to a set offset threshold;
distributing directions for key points of the high difference pyramid, establishing histograms in different directions according to radians of gradient directions of the key points, and determining feature points;
adding each direction gradient information for each characteristic point as characteristic point description;
matching the feature vectors by adopting a nearest neighbor distance method, and matching the feature vectors of the feature points;
and calculating the similarity between the images according to the matching number of the image feature points.
5. The hidden danger image processing method combining the gaussian algorithm and the hash algorithm according to claim 4, wherein: and solving the extreme value offset by using a Taylor expansion equation.
6. The hidden danger image processing method fusing the gaussian algorithm and the hash algorithm as claimed in claim 1, wherein constructing the scale space comprises: the method for calculating the redundancy between the images by combining trial matching and a minimum spanning tree algorithm according to the feature point matching data and the similarity between the images to obtain the redundant image of the image to be detected specifically comprises the following steps:
judging whether the image sets grouped according to the similarity have regions with high overlapping degree or not according to a set similarity threshold; wherein the image set is: a set consisting of corresponding images in a similarity matrix constructed based on the similarity;
if the area with the high overlapping degree exists, two images with the highest similarity in the area are determined, the similarity between the rest images in the image set and the two images is calculated, and the images with the similarity larger than the set similarity threshold value are deleted.
7. The hidden danger image processing method combining the gaussian algorithm and the hash algorithm according to claim 1, wherein the transmission line inspection image from which the redundant image is deleted is classified and stored; the classified storage can be used for constructing a power transmission line storage model for storage and performing classified storage according to classification standards contained in the power transmission line storage model;
or, the classification standard of the power transmission line storage model comprises: the type and the terrain of the power transmission line channel, the attribution of the power transmission line channel, the shooting type of the information image and the defect type.
8. The hidden danger image processing system fusing the Gaussian algorithm and the Hash algorithm is characterized by comprising an image data storage module and an image retrieval module, wherein the data storage module is configured to execute the following steps:
acquiring a power transmission line inspection image to be stored;
identifying the power transmission line inspection images to be stored by adopting a Gaussian algorithm to obtain matching data and similarity of characteristic points between the images;
calculating the redundancy between the images by combining trial matching and a minimum spanning tree algorithm according to the feature point matching data and the similarity between the images to obtain a redundant image of the image to be detected;
and (4) carrying out classified storage on the power transmission line inspection image after the redundant image is deleted.
9. The fused gaussian and hash hidden danger image processing system of claim 8 wherein the image retrieval module is configured to perform the steps of:
extracting the characteristics of the images obtained from the image database to construct an image characteristic library;
training and learning a hash function according to the images of the image feature library; coding the images in the stored image set T based on the trained hash function to obtain the hash code of each stored image;
carrying out Hash coding on an image to be queried according to a Hash function to obtain a Hash code of the image to be queried, calculating the Hash code of the image to be queried and the Hash codes corresponding to the images in the image set T one by one to carry out Hamming distance calculation, and taking the image corresponding to the minimum Hamming distance as a retrieval result;
or, the method also comprises the following steps of: and rearranging and outputting the set number of the images of which the calculated Hamming distance is not more than a set value, and outputting the corresponding images from small to large according to the Hamming distance as a retrieval result.
10. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the steps of any of the methods of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210452636.4A CN115017347A (en) | 2022-04-27 | 2022-04-27 | Hidden danger image processing method and system fusing Gaussian algorithm and Hash algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210452636.4A CN115017347A (en) | 2022-04-27 | 2022-04-27 | Hidden danger image processing method and system fusing Gaussian algorithm and Hash algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115017347A true CN115017347A (en) | 2022-09-06 |
Family
ID=83067786
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210452636.4A Pending CN115017347A (en) | 2022-04-27 | 2022-04-27 | Hidden danger image processing method and system fusing Gaussian algorithm and Hash algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115017347A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117906615A (en) * | 2024-03-15 | 2024-04-19 | 苏州艾吉威机器人有限公司 | Fusion positioning method and system of intelligent carrying equipment based on environment identification code |
-
2022
- 2022-04-27 CN CN202210452636.4A patent/CN115017347A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117906615A (en) * | 2024-03-15 | 2024-04-19 | 苏州艾吉威机器人有限公司 | Fusion positioning method and system of intelligent carrying equipment based on environment identification code |
| CN117906615B (en) * | 2024-03-15 | 2024-06-04 | 苏州艾吉威机器人有限公司 | Fusion positioning method and system of intelligent carrying equipment based on environment identification code |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107885764B (en) | Rapid Hash vehicle retrieval method based on multitask deep learning | |
| CN111914813B (en) | Power transmission line inspection image naming method and system based on image classification | |
| Schlegel et al. | HBST: A hamming distance embedding binary search tree for feature-based visual place recognition | |
| Du et al. | Computer-aided plant species identification (CAPSI) based on leaf shape matching technique | |
| US20200004781A1 (en) | Method and Apparatus for Multi-Dimensional Content Search and Video Identification | |
| US20120314961A1 (en) | Scalable near duplicate image search with geometric constraints | |
| CN104794219A (en) | Scene retrieval method based on geographical position information | |
| CN112347284B (en) | Combined trademark image retrieval method | |
| KR101191223B1 (en) | Method, apparatus and computer-readable recording medium by for retrieving image | |
| CN111046950A (en) | Image processing method and device, storage medium and electronic device | |
| CN115527036A (en) | Power grid scene point cloud semantic segmentation method and device, computer equipment and medium | |
| CN115658940A (en) | Method, device and storage medium for constructing and updating heterogeneous detection feature database | |
| CN115017347A (en) | Hidden danger image processing method and system fusing Gaussian algorithm and Hash algorithm | |
| US9412020B2 (en) | Geometric coding for billion-scale partial-duplicate image search | |
| CN114491200A (en) | Method and device for matching heterogeneous interest points based on graph neural network | |
| Han et al. | A novel loop closure detection method with the combination of points and lines based on information entropy | |
| Hou et al. | Use of roadway scene semantic information and geometry-preserving landmark pairs to improve visual place recognition in changing environments | |
| Li et al. | Hot-NetVLAD: Learning discriminatory key points for visual place recognition | |
| CN116524725B (en) | Intelligent driving traffic sign image data identification system | |
| CN112084353A (en) | Bag-of-words model method for rapid landmark-convolution feature matching | |
| Heo et al. | Shortlist selection with residual-aware distance estimator for k-nearest neighbor search | |
| CN115186138A (en) | Comparison method and terminal for power distribution network data | |
| CN111783891B (en) | Customized object detection method | |
| Abdollahyan et al. | Visual localization in the presence of appearance changes using the partial order kernel | |
| CN114187501A (en) | Package detection method, device and system |
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 |