CN117058084A - Rail damage detection method and system - Google Patents

Rail damage detection method and system Download PDF

Info

Publication number
CN117058084A
CN117058084A CN202310999949.6A CN202310999949A CN117058084A CN 117058084 A CN117058084 A CN 117058084A CN 202310999949 A CN202310999949 A CN 202310999949A CN 117058084 A CN117058084 A CN 117058084A
Authority
CN
China
Prior art keywords
image
data
analysis
network
key information
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
Application number
CN202310999949.6A
Other languages
Chinese (zh)
Inventor
阮腾达
蔡国强
孙虎
苏奎松
邱慧敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiamen Jiaotest Intelligent Technology Co ltd
Original Assignee
Xiamen Jiaotest Intelligent Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xiamen Jiaotest Intelligent Technology Co ltd filed Critical Xiamen Jiaotest Intelligent Technology Co ltd
Priority to CN202310999949.6A priority Critical patent/CN117058084A/en
Publication of CN117058084A publication Critical patent/CN117058084A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • 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
    • G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
    • 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
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted 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
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20112—Image segmentation details

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a steel rail damage detection method and a steel rail damage detection system, and relates to the technical field of artificial intelligence. In the invention, the exemplary image data is subjected to key information mining operation to form an image key information description vector, and the user identity description data is subjected to feature space mapping operation to form a user identity mapping vector; combining the user identity mapping vector and the image key information description vector to form a multi-level combined description vector; based on the image validity analysis data, the image key information description vector and the multi-level merging description vector, performing network optimization operation to form a target image analysis network; outputting target image validity analysis data corresponding to the image data to be processed through a target image analysis network, and screening out the target image data to be processed based on the target image validity analysis data; and outputting a target damage detection result based on the target to-be-processed image data. Based on the above, the reliability of rail damage detection can be improved.

Description

Rail damage detection method and system
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a steel rail damage detection method and system.
Background
There are various ways of detecting rail damage, for example, for external damage of a rail, an image acquisition and an image analysis can be performed to obtain a damage detection result. Specifically, the collected rail images can be analyzed by using a relatively mature artificial intelligence technology to output corresponding rail damage results. Wherein, artificial intelligence (Artificial Intelligence, AI for short) is a theory, method, technique and application system that uses digital computer or digital computer controlled calculation to simulate, extend and expand human intelligence, sense environment, acquire knowledge and use knowledge to obtain optimal results.
However, in the prior art, there is a problem that the reliability of rail damage detection is not high in the process of detecting rail damage based on rail images.
Disclosure of Invention
In view of the above, the present invention aims to provide a method and a system for detecting rail damage, so as to improve the reliability of rail damage detection to a certain extent.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme:
a rail damage detection method comprising:
extracting exemplary image data and user characteristic description data of an image analysis user, wherein the image analysis user belongs to a user for performing image validity analysis on the exemplary image data, the user characteristic description data comprises user identity description data of the image analysis user and image validity analysis data formed by the image analysis user for performing image validity analysis on the exemplary image data, and the exemplary image data is formed by performing damage detection operation on steel rails;
Performing key information mining operation on the exemplary image data through an initial image analysis network to form an image key information description vector corresponding to the exemplary image data, and performing feature space mapping operation on user identity description data of the image analysis user through the initial image analysis network to form a user identity mapping vector corresponding to the image analysis user;
combining the user identity mapping vectors to be combined into image key information description vectors corresponding to the exemplary image data to form corresponding multi-level combined description vectors, wherein the multi-level combined description vectors simultaneously carry image key information of the exemplary image data and user identity key information of the image analysis user;
based on the image validity analysis data, the image key information description vector and the multi-level combination description vector, performing network optimization operation on the initial image analysis network to form a target image analysis network corresponding to the initial image analysis network;
after a plurality of pieces of image data to be processed are obtained, respectively carrying out analysis operation on each piece of image data to be processed through the target image analysis network so as to output target image validity analysis data corresponding to each piece of image data to be processed, and screening out one piece of image data to be processed from the plurality of pieces of image data to be processed on the basis of the target image validity analysis data corresponding to each piece of image data to be processed, wherein the plurality of pieces of image data to be processed are all formed by carrying out damage detection operation on steel rails to be processed as target image data to be processed;
And performing damage detection operation on the steel rail to be processed based on the target image data to be processed so as to output a target damage detection result corresponding to the steel rail to be processed, wherein the target damage detection result is used for reflecting whether damage exists on the steel rail to be processed and the damage type of the steel rail to be processed.
In some preferred embodiments, in the above rail damage detection method, the step of performing, by using an initial image analysis network, a key information mining operation on the exemplary image data to form an image key information description vector corresponding to the exemplary image data includes:
performing image segmentation operation on the exemplary image data to form a plurality of image sub-data corresponding to the exemplary image data;
performing key information mining operation on each image sub-data in the plurality of image sub-data through an initial image analysis network to output a local key information description vector corresponding to each image sub-data, wherein the local key information description vector corresponding to the image sub-data is used for representing image key information of each exemplary image included in the image sub-data;
and determining an image key information description vector corresponding to the exemplary image data according to the local key information description vector corresponding to each image sub-data, wherein each local key information description vector in the image key information description vector is orderly arranged and combined based on the precedence relationship of the corresponding image sub-data in the exemplary image data, and the image key information description vector is used for representing the image key information of each exemplary image in the exemplary image data.
In some preferred embodiments, in the above rail damage detection method, the exemplary image data includes a plurality of exemplary images, and the step of performing an image segmentation operation on the exemplary image data to form a plurality of image sub-data corresponding to the exemplary image data includes:
determining a first image segmentation parameter and a second image segmentation parameter, wherein the first image segmentation parameter is used for representing the number of interval image frames between two adjacent image sub-data formed by segmentation, and the second image segmentation parameter is used for representing the number of image frames of an exemplary image included in the image sub-data formed by segmentation;
and performing image segmentation operation on the exemplary image data based on the first image segmentation parameter and the second image segmentation parameter to form a plurality of image sub-data corresponding to the exemplary image data.
In some preferred embodiments, in the above rail damage detection method, the image key information description vector includes a local key information description vector corresponding to each image sub-data; the step of merging the user identity mapping vectors into the image key information description vectors corresponding to the exemplary image data to form corresponding multi-level merged description vectors includes:
Combining the user identity mapping vector and the local key information description vector of each image sub-data to form a combined key information description vector corresponding to each image sub-data;
and based on the precedence relation of the image sub-data in the exemplary image data, orderly arranging and combining the merging key information description vectors corresponding to each image sub-data to form multi-level merging description vectors corresponding to the exemplary image data.
In some preferred embodiments, in the rail damage detection method, the local key information description vector corresponding to each image sub-data includes an image granularity level description vector of each exemplary image included in the corresponding image sub-data; the image sub-data to be processed belongs to any one of the image sub-data included in the exemplary image data;
the step of merging the user identity mapping vector and the local key information description vector of each image sub-data to form a merged key information description vector corresponding to each image sub-data includes:
respectively carrying out merging operation on the user identity mapping vector and the image sub-data to be processed, wherein the image granularity level description vector corresponds to each exemplary image, so as to form local multi-level merging description vectors corresponding to each exemplary image, and the local multi-level merging description vectors corresponding to the exemplary image simultaneously carry image key information corresponding to the exemplary image and user identity key information corresponding to the image analysis user;
And combining the partial multi-level combination description vectors corresponding to each exemplary image to form the combination key information description vector corresponding to the image sub-data to be processed.
In some preferred embodiments, in the above rail damage detection method, the step of performing a network optimization operation on the initial image analysis network based on the image validity analysis data, the image key information description vector and the multi-level merged description vector to form a target image analysis network corresponding to the initial image analysis network includes:
performing validity analysis operation on the exemplary image data according to the image key information description vector and the multi-level combination description vector through the initial image analysis network so as to output image validity characterization data corresponding to the exemplary image data;
according to the distinguishing information between the image validity characterization data and the image validity analysis data, determining a network learning cost index corresponding to the initial image analysis network, and performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network.
In some preferred embodiments, in the above rail damage detection method, the step of performing, by the initial image analysis network, validity analysis operation on the exemplary image data according to the image key information description vector and the multi-level merged description vector to output image validity characterization data corresponding to the exemplary image data includes:
performing validity analysis operation on the exemplary image data according to the image key information description vector through the initial image analysis network so as to output candidate validity characterization data corresponding to the exemplary image data;
performing validity analysis operation on the exemplary image data according to the multi-level combined description vector through the initial image analysis network so as to output undetermined validity representation data corresponding to the exemplary image data;
analyzing image validity characterization data corresponding to the exemplary image data based on the candidate validity characterization data and the pending validity characterization data;
and determining a network learning cost index corresponding to the initial image analysis network according to the distinguishing information between the image validity characterization data and the image validity analysis data, and performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network, wherein the method comprises the following steps:
Determining a network learning cost determining rule of the initial image analysis network;
marking the image validity characterization data and the image validity analysis data, and marking the data to be processed of the network learning cost determination rule to determine a network learning cost index corresponding to the initial image analysis network, wherein the network learning cost determination rule carries out analysis operation on a typical data set, and the typical data set refers to an image validity analysis data set of each exemplary image data of a plurality of image analysis users; the network learning cost determining rule comprises a first network learning cost determining rule and a second network learning cost determining rule which are mutually different, when the distribution dispersion of the image validity analysis data included in the typical data set is larger than a preset distribution dispersion, the network learning cost index is calculated through the first network learning cost determining rule, and when the distribution dispersion of the image validity analysis data included in the typical data set is smaller than or equal to the preset distribution dispersion, the network learning cost index is calculated through the second network learning cost determining rule;
And performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network.
In some preferred embodiments, in the above rail damage detection method, the image key information description vector includes a local key information description vector corresponding to each image sub-data, and the step of performing, by the initial image analysis network, validity analysis operation on the exemplary image data according to the image key information description vector to output candidate validity characterization data corresponding to the exemplary image data includes:
carrying out vector aggregation operation on the local key information description vector corresponding to each image sub-data through the initial image analysis network so as to form a sub-data aggregation description vector corresponding to each image sub-data;
according to the sub-data aggregation description vector corresponding to each image sub-data, carrying out validity analysis operation on the corresponding image sub-data so as to output local validity representation data corresponding to each image sub-data;
And carrying out fusion operation on the local validity characterization data corresponding to the plurality of image sub-data included in the exemplary image data, and outputting candidate validity characterization data corresponding to the exemplary image data.
In some preferred embodiments, in the above rail damage detection method, the step of performing a damage detection operation on the rail to be processed based on the target image data to be processed to output a target damage detection result corresponding to the rail to be processed includes:
performing damage detection operation on the steel rail to be processed based on the target image data to be processed through a target damage detection network formed by performing network optimization operation, so as to output a target damage detection result corresponding to the steel rail to be processed, wherein the target damage detection network is formed by performing network optimization operation on an initial damage detection network based on typical image data and actual damage state information of a typical steel rail corresponding to the typical image data;
the rail damage detection method further comprises the following steps:
after a plurality of image data to be analyzed are obtained, respectively carrying out analysis operation on each image data to be analyzed through the target image analysis network so as to output target image validity analysis data corresponding to each image data to be analyzed, wherein the plurality of image data to be analyzed are formed by carrying out damage detection operation on steel rails to be analyzed;
Analyzing each piece of image data to be analyzed through the target damage detection network respectively so as to output an initial damage detection result corresponding to each piece of image data to be analyzed;
and based on the target image effectiveness analysis data corresponding to each piece of analysis image data, carrying out fusion operation on the initial damage detection result corresponding to each piece of image data to be analyzed so as to output the target damage detection result corresponding to the steel rail to be analyzed.
The embodiment of the invention also provides a steel rail damage detection system, which comprises a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program so as to realize the steel rail damage detection method.
The rail damage detection method and system provided by the embodiment of the invention can carry out key information mining operation on the exemplary image data to form an image key information description vector, and carry out feature space mapping operation on the user identity description data to form a user identity mapping vector; combining the user identity mapping vector and the image key information description vector to form a multi-level combined description vector; based on the image validity analysis data, the image key information description vector and the multi-level merging description vector, performing network optimization operation to form a target image analysis network; outputting target image validity analysis data corresponding to the image data to be processed through a target image analysis network, and screening out the target image data to be processed based on the target image validity analysis data; and outputting a target damage detection result based on the target to-be-processed image data. Based on the foregoing, the basis of the rail damage detection is more reliable due to the determination of the validity of the image before the rail damage detection, and in addition, the user characteristic description data are fused in the process of performing the network optimization operation, so that the image validity analysis of the formed target image analysis network is easier to match with the actual application, and the reliability of the rail damage detection can be improved to a certain extent.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
Fig. 1 is a block diagram of a rail damage detection system according to an embodiment of the present invention.
Fig. 2 is a schematic flow chart of steps included in the rail damage detection method according to the embodiment of the present invention.
Fig. 3 is a schematic diagram of each module included in the rail damage detection device provided by the embodiment of the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected 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.
As shown in fig. 1, an embodiment of the present invention provides a rail damage detection system. Wherein, rail damage detection system can include memory and processor.
In detail, the memory and the processor are electrically connected directly or indirectly to realize transmission or interaction of data. For example, electrical connection may be made to each other via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute an executable computer program stored in the memory, so as to implement the rail damage detection method provided by the embodiment of the present invention.
It should be appreciated that in some specific embodiments, the Memory may be, but is not limited to, random access Memory (Random Access Memory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like.
It should be appreciated that in some specific embodiments, the processor may be a general purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a System on Chip (SoC), etc.; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
It should be appreciated that in some embodiments, the rail damage detection system may be a server with data processing capabilities.
With reference to fig. 2, the embodiment of the invention further provides a rail damage detection method, which can be applied to the rail damage detection system. The method steps defined by the flow related to the rail damage detection method can be realized by the rail damage detection system.
The specific flow shown in fig. 2 will be described in detail.
Step S110, exemplary image data and user feature description data of the image analysis user are extracted.
In the embodiment of the invention, the steel rail damage detection system can extract exemplary image data and user characteristic description data of an image analysis user. The image analysis user belongs to a user for performing image validity analysis on the exemplary image data, the user characteristic description data comprise user identity description data of the image analysis user and image validity analysis data formed by the image analysis user for analyzing the image validity of the exemplary image data, and the exemplary image data is formed by performing damage detection operation on steel rails. In addition, the user identity description data may be user attribute information of the image analysis user and user image related information, where the user attribute information may include account number, location, and the like, and the user image related information may refer to image related operations and behaviors performed by the image analysis user in history. And, the extracted exemplary image data may be plural, and the optimization operation may be sequentially performed on the initial image analysis network using the plural exemplary image data until a preset condition is satisfied.
Step S120, performing a key information mining operation on the exemplary image data through an initial image analysis network to form an image key information description vector corresponding to the exemplary image data, and performing a feature space mapping operation on the user identity description data of the image analysis user through the initial image analysis network to form a user identity mapping vector corresponding to the image analysis user.
In the embodiment of the invention, the steel rail damage detection system can perform key information mining operation on the exemplary image data through an initial image analysis network to form an image key information description vector corresponding to the exemplary image data, and perform feature space mapping operation on the user identity description data of the image analysis user through the initial image analysis network to form a user identity mapping vector corresponding to the image analysis user. That is, the exemplary image data is subjected to a mining operation of key information on the one hand, and the user identification description data is converted to be represented in the form of a vector on the other hand.
Step S130, performing a merging operation on the user identity mapping vector to merge into an image key information description vector corresponding to the exemplary image data, so as to form a corresponding multi-level merged description vector.
In the embodiment of the invention, the steel rail damage detection system can perform merging operation on the user identity mapping vector so as to merge the user identity mapping vector into the image key information description vector corresponding to the exemplary image data to form a corresponding multi-level merged description vector. The multi-level merging description vector carries the image key information of the exemplary image data and the user identity key information of the image analysis user at the same time, namely, the information at least comprises two dimensions, namely, the information of the exemplary image data, namely, the information of the user analyzing the exemplary image data.
And step S140, performing network optimization operation on the initial image analysis network based on the image validity analysis data, the image key information description vector and the multi-level combination description vector so as to form a target image analysis network corresponding to the initial image analysis network.
In the embodiment of the invention, the steel rail damage detection system can perform network optimization operation on the initial image analysis network based on the image validity analysis data, the image key information description vector and the multi-level combination description vector, for example, perform optimization adjustment on included network parameters to form a target image analysis network corresponding to the initial image analysis network.
Step S150, after obtaining a plurality of image data to be processed, performing an analysis operation on each image data to be processed through the target image analysis network, so as to output target image validity analysis data corresponding to each image data to be processed, and screening out one image data to be processed from the plurality of image data to be processed as target image data to be processed based on the target image validity analysis data corresponding to each image data to be processed.
In the embodiment of the present invention, after a plurality of pieces of image data to be processed are acquired, the rail damage detection system may perform an analysis operation on each piece of image data to be processed through the target image analysis network, so as to output target image validity analysis data corresponding to each piece of image data to be processed, and select one piece of image data to be processed from the plurality of pieces of image data to be processed as target image data to be processed, for example, the piece of image data to be processed with the highest validity represented by the target image validity analysis data may be selected. The plurality of image data to be processed are formed by performing damage detection operation on the steel rail to be processed, namely, are collected and formed by image collecting equipment.
Step S160, performing a damage detection operation on the rail to be processed based on the target image data to be processed, so as to output a target damage detection result corresponding to the rail to be processed.
In the embodiment of the invention, the steel rail damage detection system can perform damage detection operation on the steel rail to be processed based on the target image data to be processed so as to output a target damage detection result corresponding to the steel rail to be processed. The target damage detection result is used for reflecting whether damage exists on the steel rail to be treated and the damage type of the steel rail, and can also comprise the damage degree and the like.
Based on the foregoing, the basis of the rail damage detection is more reliable due to the determination of the validity of the image before the rail damage detection, and in addition, the user characteristic description data is fused in the process of performing the network optimization operation, so that the image validity analysis of the formed target image analysis network is easier to match with the actual application, thereby improving the reliability of the rail damage detection to a certain extent and further improving the defects in the prior art.
It should be understood that, in some specific embodiments, step S120 in the foregoing implementation, that is, the step of performing, by the initial image analysis network, the key information mining operation on the exemplary image data to form the image key information description vector corresponding to the exemplary image data, may further include the implementation procedures described below:
Performing an image segmentation operation on the exemplary image data to form a plurality of image sub-data corresponding to the exemplary image data, wherein the exemplary image data may include a plurality of (frame) continuous exemplary images, and thus the plurality of exemplary images may be subjected to a segmentation operation (segmentation from a time dimension) to form a plurality of image sub-data, each of which may include one or more exemplary images;
performing key information mining operation on each image sub-data in the plurality of image sub-data through an initial image analysis network to output a local key information description vector corresponding to each image sub-data, wherein the local key information description vector corresponding to each image sub-data is used for representing image key information of each exemplary image included in the image sub-data, for example, convolution operation can be performed on each image sub-data through a convolution unit to form a local key information description vector corresponding to each image sub-data, and in addition, convolution operation can be performed on each exemplary image included in each image sub-data for each image sub-data, then, the results of the convolution operation, such as cascade combination operation, can be aggregated to obtain a local key information description vector;
And determining an image key information description vector corresponding to the exemplary image data according to the local key information description vector corresponding to each image sub-data, wherein each local key information description vector in the image key information description vector is orderly arranged and combined based on the sequence relation of the corresponding image sub-data in the exemplary image data, and the image key information description vector is used for representing the image key information of each exemplary image in the exemplary image data, for example, the image key information description vector can be { the local key information description vector corresponding to the image sub-data 1, the local key information description vector corresponding to the image sub-data 2, and the local key information description vector corresponding to the image sub-data 3 }.
It should be appreciated that, in some specific embodiments, the exemplary image data includes a plurality of exemplary images, based on which the step of performing an image segmentation operation on the exemplary image data to form a plurality of image sub-data corresponding to the exemplary image data may further include implementation procedures described below:
Determining a first image segmentation parameter and a second image segmentation parameter, wherein the first image segmentation parameter is used for representing the number of interval image frames between two adjacent image sub-data formed by segmentation, and the second image segmentation parameter is used for representing the number of image frames of an exemplary image included in the image sub-data formed by segmentation, so that the number of frames of the exemplary image included in each image sub-data is consistent;
and performing image segmentation operation on the exemplary image data based on the first image segmentation parameter and the second image segmentation parameter to form a plurality of image sub-data corresponding to the exemplary image data, wherein the first image segmentation parameter and the second image segmentation parameter can be configured according to actual requirements, and are not particularly limited and described herein.
It should be understood that, in some specific embodiments, the image key information description vector may include a local key information description vector corresponding to each image sub-data, based on this, step S130 in the foregoing implementation, that is, the step of performing a merging operation on the user identity mapping vector to merge into the image key information description vector corresponding to the exemplary image data to form a corresponding multi-level merged description vector, further may include an implementation process described below:
Respectively performing merging operation on the user identity mapping vector and the local key information description vector of each image sub-data to form a merged key information description vector corresponding to each image sub-data, for example, the user identity mapping vector and the local key information description vector of the image sub-data 1 may be subjected to merging operation to form a merged key information description vector corresponding to the image sub-data 1, the user identity mapping vector and the local key information description vector of the image sub-data 2 may be subjected to merging operation to form a merged key information description vector corresponding to the image sub-data 2, and the local key information description vector of the user identity mapping vector and the image sub-data 3 may be subjected to merging operation to form a merged key information description vector corresponding to the image sub-data 3;
based on the precedence relationship of the image sub-data in the exemplary image data, the merging key information description vectors corresponding to each image sub-data are orderly arranged and combined to form a multi-level merging description vector corresponding to the exemplary image data, for example, the multi-level merging description vector may be { the merging key information description vector corresponding to the image sub-data 1, the merging key information description vector corresponding to the image sub-data 2, and the merging key information description vector corresponding to the image sub-data 3 }.
It should be appreciated that in some specific embodiments, the local key information description vector corresponding to each of the image sub-data includes an image granularity level description vector of each of the exemplary images included in the corresponding image sub-data (the image granularity level description vector may be formed by performing a key information mining operation on the exemplary images). The step of merging the user identity mapping vector and the local key information description vector of each image sub-data to form a merged key information description vector corresponding to each image sub-data may further include the implementation procedure described below:
the user identity mapping vector and the image sub-data to be processed respectively comprise image granularity level description vectors corresponding to all the example images to form local multi-level combination description vectors corresponding to all the example images, wherein the local multi-level combination description vectors corresponding to the example images simultaneously carry image key information corresponding to the example images and user identity key information corresponding to the image analysis user, for example, the local multi-level combination description vectors corresponding to the user identity mapping vector and the example image 1 can be formed by combining the user identity mapping vector and the image granularity level description vectors corresponding to the example image 1, the local multi-level combination description vectors corresponding to the user identity mapping vector and the example image 2 can be formed by combining the local multi-level combination description vectors corresponding to the example image 2, and the local multi-level combination description vectors corresponding to the user identity mapping vector and the example image 3 can be formed by combining the local multi-level combination description vectors corresponding to the example image 3;
And combining to form a merging key information description vector corresponding to the image sub-data to be processed according to the local multi-level merging description vector corresponding to each exemplary image, wherein the merging key information description vector can be { a local multi-level merging description vector corresponding to an exemplary image 1, a local multi-level merging description vector corresponding to an exemplary image 2, and a local multi-level merging description vector corresponding to an exemplary image 3 }, for example.
Wherein, it should be understood that, in some specific embodiments, the step of respectively merging the user identity mapping vector and the image sub-data to be processed, including the image granularity level description vector corresponding to each exemplary image, to form the local multi-level merged description vector corresponding to each exemplary image may further include the following implementation procedures described below:
weighting the image granularity level description vector based on a first weighting vector to output a corresponding first weighting fusion vector, and respectively weighting the user identity mapping vector based on a second weighting vector and a third weighting vector to output a corresponding second weighting fusion vector and third weighting fusion vector, wherein the first weighting vector, the second weighting vector and the third weighting vector belong to network parameters of a corresponding neural network; and calculating the product between the first weighted fusion vector and the transpose vector of the second weighted vector, and performing multiplication fusion operation on the third weighted vector based on the product or the positive correlation value of the product to output a corresponding first-level similar fusion vector;
Filtering the image granularity level description vectors through a plurality of filtering units with different sizes respectively to form a plurality of first filtering description vectors, and filtering the user identity mapping vectors respectively to form a plurality of second filtering vectors;
for each filtering unit, performing weighting processing on a first filtering description vector corresponding to the filtering unit based on a first weighting vector to output a corresponding first weighting fusion vector, and performing weighting processing on a second filtering vector corresponding to the filtering unit based on a second weighting vector and a third weighting vector to output a corresponding second weighting fusion vector and a third weighting fusion vector, and calculating a product between the first weighting fusion vector and a transposed vector of the second weighting vector, and performing multiplication fusion operation on the third weighting vector based on the product or a positive correlation value of the product to output a corresponding second-level similarity fusion vector;
and carrying out mean value calculation on the first-level similar fusion vector and the second-level similar fusion vector corresponding to each filtering unit so as to output a corresponding local multi-level combined description vector.
It should be understood that, in some specific embodiments, the step of performing the mean calculation on the first-level similar fusion vector and the second-level similar fusion vector corresponding to each filtering unit to output the corresponding local multi-level merged description vector may further include the following implementation process described below:
performing average value calculation on the first-level similar fusion vector and the second-level similar fusion vector corresponding to each filtering unit to output a corresponding local multi-level combined description vector, so as to output a corresponding average value description vector (in other embodiments, the average value description vector may also be directly used as the corresponding local multi-level combined description vector);
acquiring steel rail detection audio data corresponding to the image sub-data to be processed, wherein the steel rail detection audio data belongs to audio data formed by audio acquisition after corresponding steel rails are hit, and the acquisition time of each exemplary image included in the image sub-data to be processed corresponding to the steel rail detection audio data is consistent with the acquisition time of the steel rail detection audio data;
performing key information mining operation on the steel rail detection audio data to output audio key information description vectors corresponding to the steel rail detection audio data;
And performing a merging operation on the audio key information description vector and the mean description vector to form a corresponding local multi-level merging description vector, wherein the merging operation can refer to the process of merging the user identity mapping vector into the image granularity level description vector to form the mean description vector, and is not repeated herein.
It should be appreciated that, in some specific embodiments, step S140 in the foregoing implementation, that is, the step of performing a network optimization operation on the initial image analysis network based on the image validity analysis data, the image key information description vector and the multi-level merged description vector to form a target image analysis network corresponding to the initial image analysis network, may further include the following implementation procedures described below:
performing validity analysis operation on the exemplary image data according to the image key information description vector and the multi-level combination description vector through the initial image analysis network so as to output image validity characterization data corresponding to the exemplary image data, wherein the image validity characterization data and the image validity analysis data are used for characterizing the validity of the exemplary image data, and the determined modes are inconsistent;
According to the distinguishing information between the image validity characterization data and the image validity analysis data, determining a network learning cost index corresponding to the initial image analysis network, for example, the network learning cost index can have a positive correlation with the distinguishing information, and performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network.
It should be appreciated that, in some specific embodiments, the step of performing, by the initial image analysis network, the validity analysis operation on the exemplary image data according to the image key information description vector and the multi-level merged description vector to output image validity characterization data corresponding to the exemplary image data may further include an implementation procedure described below:
performing validity analysis operation on the exemplary image data according to the image key information description vector through the initial image analysis network to output candidate validity characterization data corresponding to the exemplary image data, for example, processing the image key information description vector through a first output unit included in the initial image analysis network to output candidate validity characterization data, wherein the first output unit can perform full connection and activation processing;
Performing validity analysis operation on the exemplary image data according to the multi-level combined description vector through the initial image analysis network to output pending validity characterization data corresponding to the exemplary image data, and processing the multi-level combined description vector through the first output unit to obtain pending validity characterization data, or processing the multi-level combined description vector through a second output unit included in the initial image analysis network to output pending validity characterization data;
and analyzing the image validity characterization data corresponding to the exemplary image data based on the candidate validity characterization data and the undetermined validity characterization data, for example, average value or weighted summation calculation can be performed on the validity values of the candidate validity characterization data and the undetermined validity characterization data, and when weighted summation calculation is performed, the weighting coefficient corresponding to the undetermined validity characterization data can be larger than the weighting coefficient corresponding to the candidate validity characterization data.
It should be appreciated that, in some specific embodiments, the image key information description vector includes a local key information description vector corresponding to each image sub-data, based on which the step of performing, by the initial image analysis network, a validity analysis operation on the exemplary image data according to the image key information description vector to output candidate validity characterization data corresponding to the exemplary image data may further include the following implementation procedures described below:
Carrying out vector aggregation operation on the local key information description vector corresponding to each image sub-data through the initial image analysis network so as to form a sub-data aggregation description vector corresponding to each image sub-data, for example, carrying out superposition (such as mean calculation) or cascading combination operation on the local key information description vector corresponding to each image sub-data;
according to the sub-data aggregation description vector corresponding to each image sub-data, performing validity analysis operation on the corresponding image sub-data to output local validity characterization data corresponding to each image sub-data, for example, the sub-data aggregation description vector corresponding to each image sub-data can be processed through a first output unit included in the initial image analysis network;
and carrying out fusion operation on local validity characterization data corresponding to a plurality of image sub-data included in the exemplary image data, outputting candidate validity characterization data corresponding to the exemplary image data, and carrying out mean value calculation on the local validity characterization data corresponding to the plurality of image sub-data.
It should be understood that, in some specific embodiments, the step of determining, according to the distinguishing information between the image validity characterizing data and the image validity analyzing data, a network learning cost indicator corresponding to the initial image analyzing network, and performing a network optimization operation on the initial image analyzing network based on the network learning cost indicator to form a target image analyzing network corresponding to the initial image analyzing network may further include the following implementation procedures described below:
determining a network learning cost determining rule of the initial image analysis network;
marking the image validity characterization data and the image validity analysis data, and marking the data to be processed of the network learning cost determination rule to determine a network learning cost index corresponding to the initial image analysis network, wherein the network learning cost determination rule carries out analysis operation on a typical data set, and the typical data set refers to an image validity analysis data set of each exemplary image data of a plurality of image analysis users; the network learning cost determining rule comprises a first network learning cost determining rule and a second network learning cost determining rule which are mutually different, when the distribution dispersion of the image validity analysis data included in the typical data set is larger than a preset distribution dispersion, the network learning cost index is calculated through the first network learning cost determining rule, and when the distribution dispersion of the image validity analysis data included in the typical data set is smaller than or equal to the preset distribution dispersion, the network learning cost index is calculated through the second network learning cost determining rule, the specific value of the preset distribution dispersion is not limited, and the configuration can be carried out according to actual requirements;
And performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network, wherein on the basis, the determination rule of the network learning cost index is related to the distribution dispersion of the image validity analysis data, so that the focus of the target image analysis network formed based on the network learning cost index optimization is wider than concentrated in a few directions.
Wherein it should be understood that in some specific embodiments, the step of calculating the distribution dispersion may further include the implementation procedure described below:
for a plurality of image analysis users, performing de-duplication operation on the image validity analysis data included in each exemplary image data set to form an image validity analysis data subset;
for each image validity analysis data in the subset of image validity analysis data, determining the number of identical image validity analysis data of the image validity analysis data in the subset of image validity analysis data to obtain a corresponding target data number;
And carrying out average value calculation on the target data quantity corresponding to each image effectiveness analysis data in the image effectiveness analysis data subset to output corresponding average value data quantity, and carrying out average value calculation on the absolute difference value between the target data quantity corresponding to each image effectiveness analysis data and the average value data quantity to obtain corresponding distribution dispersion.
It should be understood, that in some specific embodiments, the step of marking the image validity characterization data and the image validity analysis data to mark the data to be processed of the network learning cost determination rule to determine the network learning cost indicator corresponding to the initial image analysis network may further include the following implementation procedure described below:
and when the distribution dispersion of the image effectiveness analysis data included in the typical data set is smaller than or equal to the preset distribution dispersion, performing square sum calculation of differences on the image effectiveness characterization data and the corresponding image effectiveness analysis data included in the typical data set to output a corresponding network learning cost index.
It should be understood, that in some specific embodiments, the step of marking the image validity characterization data and the image validity analysis data to mark the data to be processed of the network learning cost determination rule to determine the network learning cost indicator corresponding to the initial image analysis network may further include the following implementation procedure described below:
when the distribution dispersion of the image validity analysis data included in the typical data set is larger than the preset distribution dispersion, calculating the average value of the image validity analysis data included in the typical data set to obtain a first average value, calculating the average value of the corresponding image validity characterization data to obtain a second average value, calculating the square value of the difference between the first average value and the second average value to obtain a first square value, calculating the ratio between the first square value and a preset adjustment parameter, wherein the adjustment parameter can be configured according to actual requirements, and performing an exponential operation on the negative correlation value of the ratio to output a first exponential operation value, wherein the sum value between the ratio and the negative correlation value is equal to a preset value, such as 0, 1, 2, 3 and the like;
Respectively calculating square values of differences between each image validity analysis data and each corresponding image validity characterization data included in the typical data set to obtain each corresponding second square value, respectively calculating a ratio between each second square value and the adjustment parameter, performing exponential operation on a negative correlation value of the ratio to output each corresponding second exponent operation value, and calculating a sum value of each second exponent operation value to obtain a target exponent operation value;
calculating the ratio between the first index operation value and the target index operation value, carrying out logarithmic operation on the ratio to obtain a corresponding logarithmic result value, and finally, determining a network learning cost index corresponding to the initial image analysis network based on the logarithmic result value, wherein the network learning cost index can have a corresponding relationship with the logarithmic result value in a negative correlation way, and if the sum value of the network learning cost index and the logarithmic result value is equal to a preset value.
It should be understood that, in some specific embodiments, step S160 in the foregoing implementation, that is, the step of performing, based on the target to-be-processed image data, a damage detection operation on the to-be-processed rail to output a target damage detection result corresponding to the to-be-processed rail, may further include the implementation process described below:
The target damage detection network is formed by performing network optimization operation, and based on the target to-be-processed image data, the damage detection operation is performed on the to-be-processed steel rail to output a target damage detection result corresponding to the to-be-processed steel rail, the target damage detection network is formed by performing network optimization operation on an initial damage detection network based on typical image data and actual damage state information of a typical steel rail corresponding to the typical image data, for example, steel rail damage analysis can be performed on the typical image data based on the initial damage detection network to output corresponding predicted damage state information, and then, network parameters included in the initial damage detection network can be optimally updated based on the difference between the predicted damage state information and the actual damage state information.
It should be appreciated that in some specific embodiments, the rail damage detection method may further include the following implementation procedures:
after a plurality of image data to be analyzed are obtained, respectively carrying out analysis operation on each image data to be analyzed through the target image analysis network so as to output target image validity analysis data corresponding to each image data to be analyzed, wherein the plurality of image data to be analyzed are formed by carrying out damage detection operation on steel rails to be analyzed;
Analyzing each piece of image data to be analyzed through the target damage detection network respectively so as to output an initial damage detection result corresponding to each piece of image data to be analyzed;
and based on the target image validity analysis data corresponding to each piece of analysis image data, carrying out fusion operation on the initial damage detection result corresponding to each piece of image data to be analyzed so as to output the target damage detection result corresponding to the steel rail to be analyzed, for example, the target image validity analysis data can be used as the importance degree of the corresponding initial damage detection result, so that each initial damage detection result can be fused based on the importance degree to obtain the target damage detection result.
With reference to fig. 3, the embodiment of the invention further provides a rail damage detection device, which can be applied to the rail damage detection system. Wherein, rail damage detection device can include:
an exemplary data extraction module, configured to extract exemplary image data and user feature description data of an image analysis user, where the image analysis user belongs to a user who performs image validity analysis on the exemplary image data, and the user feature description data includes user identity description data of the image analysis user and image validity analysis data formed by the image analysis user performing analysis on the image validity of the exemplary image data, and the exemplary image data is formed by performing a damage detection operation on a rail;
The key information mining module is used for performing key information mining operation on the exemplary image data through an initial image analysis network to form an image key information description vector corresponding to the exemplary image data, and performing feature space mapping operation on the user identity description data of the image analysis user through the initial image analysis network to form a user identity mapping vector corresponding to the image analysis user;
the vector merging module is used for carrying out merging operation on the user identity mapping vector so as to merge the user identity mapping vector into an image key information description vector corresponding to the exemplary image data to form a corresponding multi-level merging description vector, wherein the multi-level merging description vector simultaneously carries image key information of the exemplary image data and user identity key information of the image analysis user;
the network optimization module is used for performing network optimization operation on the initial image analysis network based on the image validity analysis data, the image key information description vector and the multi-level combination description vector so as to form a target image analysis network corresponding to the initial image analysis network;
The image data screening module is used for respectively carrying out analysis operation on each piece of image data to be processed through the target image analysis network after a plurality of pieces of image data to be processed are acquired so as to output target image validity analysis data corresponding to each piece of image data to be processed, and screening one piece of image data to be processed from the plurality of pieces of image data to be processed on the basis of the target image validity analysis data corresponding to each piece of image data to be processed, wherein the plurality of pieces of image data to be processed belong to steel rails to be processed and are formed by carrying out damage detection operation;
the steel rail damage detection module is used for carrying out damage detection operation on the steel rail to be processed based on the target image data to be processed so as to output a target damage detection result corresponding to the steel rail to be processed, wherein the target damage detection result is used for reflecting whether damage exists on the steel rail to be processed and the damage type of the steel rail to be processed.
In summary, according to the method and the system for detecting rail damage provided by the invention, the exemplary image data can be subjected to key information mining operation to form the image key information description vector, and the user identity description data is subjected to feature space mapping operation to form the user identity mapping vector; combining the user identity mapping vector and the image key information description vector to form a multi-level combined description vector; based on the image validity analysis data, the image key information description vector and the multi-level merging description vector, performing network optimization operation to form a target image analysis network; outputting target image validity analysis data corresponding to the image data to be processed through a target image analysis network, and screening out the target image data to be processed based on the target image validity analysis data; and outputting a target damage detection result based on the target to-be-processed image data. Based on the foregoing, the basis of the rail damage detection is more reliable due to the determination of the validity of the image before the rail damage detection, and in addition, the user characteristic description data are fused in the process of performing the network optimization operation, so that the image validity analysis of the formed target image analysis network is easier to match with the actual application, and the reliability of the rail damage detection can be improved to a certain extent.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A method of detecting rail damage comprising:
extracting exemplary image data and user characteristic description data of an image analysis user, wherein the image analysis user belongs to a user for performing image validity analysis on the exemplary image data, the user characteristic description data comprises user identity description data of the image analysis user and image validity analysis data formed by the image analysis user for performing image validity analysis on the exemplary image data, and the exemplary image data is formed by performing damage detection operation on steel rails;
performing key information mining operation on the exemplary image data through an initial image analysis network to form an image key information description vector corresponding to the exemplary image data, and performing feature space mapping operation on user identity description data of the image analysis user through the initial image analysis network to form a user identity mapping vector corresponding to the image analysis user;
Combining the user identity mapping vectors to be combined into image key information description vectors corresponding to the exemplary image data to form corresponding multi-level combined description vectors, wherein the multi-level combined description vectors simultaneously carry image key information of the exemplary image data and user identity key information of the image analysis user;
based on the image validity analysis data, the image key information description vector and the multi-level combination description vector, performing network optimization operation on the initial image analysis network to form a target image analysis network corresponding to the initial image analysis network;
after a plurality of pieces of image data to be processed are obtained, respectively carrying out analysis operation on each piece of image data to be processed through the target image analysis network so as to output target image validity analysis data corresponding to each piece of image data to be processed, and screening out one piece of image data to be processed from the plurality of pieces of image data to be processed on the basis of the target image validity analysis data corresponding to each piece of image data to be processed, wherein the plurality of pieces of image data to be processed are all formed by carrying out damage detection operation on steel rails to be processed as target image data to be processed;
And performing damage detection operation on the steel rail to be processed based on the target image data to be processed so as to output a target damage detection result corresponding to the steel rail to be processed, wherein the target damage detection result is used for reflecting whether damage exists on the steel rail to be processed and the damage type of the steel rail to be processed.
2. The method for detecting rail damage according to claim 1, wherein the step of performing a key information mining operation on the exemplary image data through an initial image analysis network to form an image key information description vector corresponding to the exemplary image data comprises:
performing image segmentation operation on the exemplary image data to form a plurality of image sub-data corresponding to the exemplary image data;
performing key information mining operation on each image sub-data in the plurality of image sub-data through an initial image analysis network to output a local key information description vector corresponding to each image sub-data, wherein the local key information description vector corresponding to the image sub-data is used for representing image key information of each exemplary image included in the image sub-data;
and determining an image key information description vector corresponding to the exemplary image data according to the local key information description vector corresponding to each image sub-data, wherein each local key information description vector in the image key information description vector is orderly arranged and combined based on the precedence relationship of the corresponding image sub-data in the exemplary image data, and the image key information description vector is used for representing the image key information of each exemplary image in the exemplary image data.
3. A rail damage detection method as recited in claim 2, wherein the exemplary image data includes a plurality of exemplary images, and the step of performing an image segmentation operation on the exemplary image data to form a plurality of image sub-data corresponding to the exemplary image data includes:
determining a first image segmentation parameter and a second image segmentation parameter, wherein the first image segmentation parameter is used for representing the number of interval image frames between two adjacent image sub-data formed by segmentation, and the second image segmentation parameter is used for representing the number of image frames of an exemplary image included in the image sub-data formed by segmentation;
and performing image segmentation operation on the exemplary image data based on the first image segmentation parameter and the second image segmentation parameter to form a plurality of image sub-data corresponding to the exemplary image data.
4. The method for detecting rail damage according to claim 1, wherein the image key information description vector comprises a local key information description vector corresponding to each image sub-data; the step of merging the user identity mapping vectors into the image key information description vectors corresponding to the exemplary image data to form corresponding multi-level merged description vectors includes:
Combining the user identity mapping vector and the local key information description vector of each image sub-data to form a combined key information description vector corresponding to each image sub-data;
and based on the precedence relation of the image sub-data in the exemplary image data, orderly arranging and combining the merging key information description vectors corresponding to each image sub-data to form multi-level merging description vectors corresponding to the exemplary image data.
5. The method for detecting rail damage according to claim 4, wherein the local key information description vector corresponding to each image sub-data comprises an image granularity level description vector of each exemplary image included in the corresponding image sub-data; the image sub-data to be processed belongs to any one of the image sub-data included in the exemplary image data;
the step of merging the user identity mapping vector and the local key information description vector of each image sub-data to form a merged key information description vector corresponding to each image sub-data includes:
respectively carrying out merging operation on the user identity mapping vector and the image sub-data to be processed, wherein the image granularity level description vector corresponds to each exemplary image, so as to form local multi-level merging description vectors corresponding to each exemplary image, and the local multi-level merging description vectors corresponding to the exemplary image simultaneously carry image key information corresponding to the exemplary image and user identity key information corresponding to the image analysis user;
And combining the partial multi-level combination description vectors corresponding to each exemplary image to form the combination key information description vector corresponding to the image sub-data to be processed.
6. The method for detecting rail damage according to claim 1, wherein the step of performing a network optimization operation on the initial image analysis network based on the image validity analysis data, the image key information description vector and the multi-level merged description vector to form a target image analysis network corresponding to the initial image analysis network comprises:
performing validity analysis operation on the exemplary image data according to the image key information description vector and the multi-level combination description vector through the initial image analysis network so as to output image validity characterization data corresponding to the exemplary image data;
according to the distinguishing information between the image validity characterization data and the image validity analysis data, determining a network learning cost index corresponding to the initial image analysis network, and performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network.
7. The method for detecting rail damage of claim 6, wherein the step of performing validity analysis on the exemplary image data by the initial image analysis network according to the image key information description vector and the multi-level merged description vector to output image validity characterization data corresponding to the exemplary image data comprises:
performing validity analysis operation on the exemplary image data according to the image key information description vector through the initial image analysis network so as to output candidate validity characterization data corresponding to the exemplary image data;
performing validity analysis operation on the exemplary image data according to the multi-level combined description vector through the initial image analysis network so as to output undetermined validity representation data corresponding to the exemplary image data;
analyzing image validity characterization data corresponding to the exemplary image data based on the candidate validity characterization data and the pending validity characterization data;
and determining a network learning cost index corresponding to the initial image analysis network according to the distinguishing information between the image validity characterization data and the image validity analysis data, and performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network, wherein the method comprises the following steps:
Determining a network learning cost determining rule of the initial image analysis network;
marking the image validity characterization data and the image validity analysis data, and marking the data to be processed of the network learning cost determination rule to determine a network learning cost index corresponding to the initial image analysis network, wherein the network learning cost determination rule carries out analysis operation on a typical data set, and the typical data set refers to an image validity analysis data set of each exemplary image data of a plurality of image analysis users; the network learning cost determining rule comprises a first network learning cost determining rule and a second network learning cost determining rule which are mutually different, when the distribution dispersion of the image validity analysis data included in the typical data set is larger than a preset distribution dispersion, the network learning cost index is calculated through the first network learning cost determining rule, and when the distribution dispersion of the image validity analysis data included in the typical data set is smaller than or equal to the preset distribution dispersion, the network learning cost index is calculated through the second network learning cost determining rule;
And performing network optimization operation on the initial image analysis network based on the network learning cost index to form a target image analysis network corresponding to the initial image analysis network.
8. The method for detecting rail damage of claim 7, wherein the image key information description vector includes a local key information description vector corresponding to each image sub-data, and the step of performing validity analysis operation on the exemplary image data according to the image key information description vector by the initial image analysis network to output candidate validity characterization data corresponding to the exemplary image data includes:
carrying out vector aggregation operation on the local key information description vector corresponding to each image sub-data through the initial image analysis network so as to form a sub-data aggregation description vector corresponding to each image sub-data;
according to the sub-data aggregation description vector corresponding to each image sub-data, carrying out validity analysis operation on the corresponding image sub-data so as to output local validity representation data corresponding to each image sub-data;
And carrying out fusion operation on the local validity characterization data corresponding to the plurality of image sub-data included in the exemplary image data, and outputting candidate validity characterization data corresponding to the exemplary image data.
9. The method for detecting rail damage according to any one of claims 1 to 8, wherein the step of performing damage detection operation on the rail to be processed based on the target image data to be processed to output a target damage detection result corresponding to the rail to be processed includes:
performing damage detection operation on the steel rail to be processed based on the target image data to be processed through a target damage detection network formed by performing network optimization operation, so as to output a target damage detection result corresponding to the steel rail to be processed, wherein the target damage detection network is formed by performing network optimization operation on an initial damage detection network based on typical image data and actual damage state information of a typical steel rail corresponding to the typical image data;
the rail damage detection method further comprises the following steps:
after a plurality of image data to be analyzed are obtained, respectively carrying out analysis operation on each image data to be analyzed through the target image analysis network so as to output target image validity analysis data corresponding to each image data to be analyzed, wherein the plurality of image data to be analyzed are formed by carrying out damage detection operation on steel rails to be analyzed;
Analyzing each piece of image data to be analyzed through the target damage detection network respectively so as to output an initial damage detection result corresponding to each piece of image data to be analyzed;
and based on the target image effectiveness analysis data corresponding to each piece of analysis image data, carrying out fusion operation on the initial damage detection result corresponding to each piece of image data to be analyzed so as to output the target damage detection result corresponding to the steel rail to be analyzed.
10. A rail damage detection system comprising a processor and a memory, the memory being for storing a computer program, the processor being for executing the computer program to implement the rail damage detection method of any one of claims 1 to 9.
CN202310999949.6A 2023-08-09 2023-08-09 Rail damage detection method and system Pending CN117058084A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310999949.6A CN117058084A (en) 2023-08-09 2023-08-09 Rail damage detection method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310999949.6A CN117058084A (en) 2023-08-09 2023-08-09 Rail damage detection method and system

Publications (1)

Publication Number Publication Date
CN117058084A true CN117058084A (en) 2023-11-14

Family

ID=88654563

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310999949.6A Pending CN117058084A (en) 2023-08-09 2023-08-09 Rail damage detection method and system

Country Status (1)

Country Link
CN (1) CN117058084A (en)

Similar Documents

Publication Publication Date Title
CN111738351B (en) Model training method and device, storage medium and electronic equipment
CN109375609B (en) Abnormal attack detection method and device
CN111460250A (en) Image data cleaning method, image data cleaning device, image data cleaning medium, and electronic apparatus
CN112329713A (en) Network flow abnormity online detection method, system, computer equipment and storage medium
CN116091796B (en) Unmanned aerial vehicle acquisition data processing method and system based on artificial intelligence
CN114553591B (en) Training method of random forest model, abnormal flow detection method and device
CN116126947B (en) Big data analysis method and system applied to enterprise management system
CN116664335B (en) Intelligent monitoring-based operation analysis method and system for semiconductor production system
CN116821777A (en) Novel basic mapping data integration method and system
CN117095198A (en) Remote sensing image retrieval network training method, application method, electronic equipment and medium
CN113297061A (en) Test case generation method, device and system, offline server and storage medium
CN116757870A (en) Intelligent energy monitoring data processing method and system for energy Internet of things
CN111368894A (en) FCBF feature selection method and application thereof in network intrusion detection
CN116070149A (en) Data analysis method and system based on artificial intelligence and cloud platform
CN117058084A (en) Rail damage detection method and system
CN115375886A (en) Data acquisition method and system based on cloud computing service
CN115484044A (en) Data state monitoring method and system
CN115757900A (en) User demand analysis method and system applying artificial intelligence model
CN111291370B (en) Network data intrusion detection method, system, terminal and storage medium
CN114785616A (en) Data risk detection method and device, computer equipment and storage medium
CN116958838B (en) Forest resource monitoring method and system based on unmanned aerial vehicle aerial survey technology
CN115761598B (en) Big data analysis method and system based on cloud service platform
CN116501285B (en) AI dialogue processing method based on virtual digital image interaction and digitizing system
CN113343963B (en) Intelligent trunk fingerprint identification method and system based on high-quality template
CN117150065B (en) Image information acquisition method 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