CN115760697A - Titanium-containing TRIP steel image detection method and system and electronic equipment - Google Patents
Titanium-containing TRIP steel image detection method and system and electronic equipment Download PDFInfo
- Publication number
- CN115760697A CN115760697A CN202211276811.5A CN202211276811A CN115760697A CN 115760697 A CN115760697 A CN 115760697A CN 202211276811 A CN202211276811 A CN 202211276811A CN 115760697 A CN115760697 A CN 115760697A
- Authority
- CN
- China
- Prior art keywords
- image
- frame
- steel
- sample
- detected
- 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
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P10/00—Technologies related to metal processing
- Y02P10/20—Recycling
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a titanium-containing TRIP steel image detection method, which is characterized in that a test sample is subjected to corrosion test by using corrosive liquid, characteristic points of a metal material image motion area in unit time are extracted based on TRIP steel atomic pixel statistics, a motion change speed value is used as a basic difference characteristic of titanium inclusion, instantaneous difference characteristics of motion states of test sample segments are extracted based on the change speed of a difference block in two continuous contexts, and therefore the abnormality of the test sample is detected; the response speed when detecting titanium inclusion occurrence has better identification and detection effects.
Description
Technical Field
The application relates to the field of image processing, in particular to a titanium-containing TRIP steel image detection method and system and electronic equipment.
Background
In recent years, the application of TRIP steel having high strength and high toughness to the automobile industry has increased year by year, and the market prospect is promising. The content, volume, morphology, size, existing state and other factors of retained austenite (sometimes martensite-austenite, also called MA island because of island shape) in TRIP steel play a key role in the influence of the structure and performance of the TRIP steel. The difficulty of analyzing the TRIP steel structure by adopting a quantitative metallographic technique is identification and extraction of a characteristic substance structure. By using the traditional corrosion method, only a mass spectrum image of a tissue can be obtained, and all phases are difficult to distinguish due to insufficient obvious contrast, even partial details are covered, so that the quantitative analysis of the characteristic substances is not facilitated.
With the development of computer science, the image analysis technology is gradually developed from the original traditional method to the deep learning method, and research shows that the deep learning method is far superior to human in the aspect of image recognition technology. However, deep learning requires a large amount of data for training a learning model, but the acquisition of material data is difficult, and the data is generally small sample data, so how to utilize the existing data at the present stage to realize the identification of the steel microstructure is a key in the field of materials science. Specifically, the existing TRIP steel image detection method has a need for increasing the response speed and the identification detection effect when detecting titanium inclusion impurities.
Disclosure of Invention
In order to solve the technical problems in the background art, the invention provides a titanium-containing TRIP steel image detection method, a titanium-containing TRIP steel image detection system and electronic equipment.
In order to achieve the purpose, the invention adopts the following technical scheme:
according to a preliminary aspect of the present invention, the present invention claims a method for detecting an image of a titanium-containing TRIP steel, which is characterized by comprising:
obtaining a steel sample to be detected, and carrying out corrosion treatment on the steel sample to be detected;
adopting a high-speed image sensor to capture continuous image frame images of the corroded steel sample to be detected;
image preprocessing is carried out on the captured continuous image frame images, and important frame images are determined;
a plurality of characteristic points of a metal material image motion region in the important frame image in unit time are obtained through statistical extraction based on TRIP steel atomic pixels;
obtaining basic difference characteristics of the titanium inclusion of the plurality of characteristic points by utilizing the motion change speed value;
extracting instantaneous difference characteristics of the motion state of the test sample segment based on the change speed of the difference characteristics in two continuous contexts;
and obtaining the quality of the titanium inclusions of the steel sample to be detected based on the instantaneous difference characteristics.
Further, the obtaining of the steel sample to be detected and the corrosion treatment of the steel sample to be detected specifically include:
carrying out corrosion marking treatment on a steel sample to be treated;
the method comprises the steps of enabling the melting concentration of a steel sample to be processed to be at a reference concentration, and conducting corrosion treatment on the steel sample at a preliminarily specified time under the reference concentration by using a catalyst, wherein the reference concentration is higher than a conventional substance concentration and lower than the reference concentration;
standing the steel sample after corrosion treatment is obtained, and carrying out corrosion treatment on the steel sample at a high concentration within a detection specified time which is not less than the preliminary specified time by using a corrosion liquid, wherein the high concentration is higher than the reference concentration;
and after obtaining the steel sample subjected to corrosion treatment, carrying out image shooting treatment on the steel sample.
Further, the method for capturing continuous image frame images of the steel sample to be detected after corrosion treatment by using the high-speed image sensor specifically comprises the following steps:
acquiring discrete frame characteristics of a corrosion process image of a steel sample to be processed and continuous frame characteristics based on Gaussian distribution;
integrating the discrete frame characteristics of the corrosion process image of the steel sample to be processed with the continuous frame characteristics based on Gaussian distribution to obtain an image frame sequence of the corrosion process image of the steel sample to be processed;
inputting the image frame sequence of the corrosion process image of the steel sample to be processed into a preset neural network to obtain a motion sequence;
and obtaining the description information of the corrosion process image of the steel sample to be processed according to the motion sequence.
Further, the image preprocessing is performed on the captured continuous image frame images to determine important frame images, and specifically includes:
counting the mass distribution value distribution data of different mass distribution values of each marked isotope in each frame of image frame aiming at any image frame of the image in the corrosion process and the image frame before the image frame;
calculating the similarity of any image frame of the corrosion process image and the image frame before the image frame according to different mass distribution value distribution data of the same marked isotope;
when the similarity is smaller than a preliminary threshold value, determining any image frame of the corrosion process image as a candidate frame;
calculating the quality unbalance degree of the candidate frame, wherein the quality unbalance degree represents the quality distribution condition of the candidate frame;
and acquiring the candidate frame with the quality unbalance degree larger than the detection threshold as the important frame.
Further, the extracting and obtaining of the plurality of feature points of the metal material image motion area in the important frame image in the unit time based on the TRIP steel atomic pixel statistics specifically includes:
obtaining a TRIP steel atomic pixel image according to the important frame image;
integrating the important frame image and the TRIP steel atomic pixel image to obtain an integrated image;
and inputting the integrated image into a pre-trained CNN characteristic extraction model for characteristic extraction to obtain image characteristics.
Further, the obtaining of the basic difference characteristics of the titanium inclusion of the plurality of feature points by using the motion variation speed value specifically includes:
acquiring original data of image characteristics;
extracting important frame image characteristics and important frame image characteristic difference values based on the acquired original data;
calculating the difference and Euclidean distance between the extracted important frame image characteristics and the important frame image characteristic difference value;
and completing the detection of the image characteristic difference according to the obtained difference and the Euclidean distance.
Specifically, the extracting the transient difference feature of the motion state of the test sample segment based on the change speed of the difference feature in two continuous contexts specifically includes:
acquiring a change speed value between any two continuous image frames in the important frames;
determining the next frame of image frame of the continuous two frames of data with the maximum change speed value as an instant image frame;
and comparing the instantaneous image frame with the preliminary image frame in the important frame to obtain the instantaneous difference characteristic.
Specifically, the obtaining of the titanium inclusion quality of the steel sample to be detected based on the transient difference characteristics specifically includes:
acquiring basic attribute parameters of the corrosion image of the sample to be detected and basic attribute parameters of the sample to be detected;
and acquiring the impurity quality of the steel sample to be detected based on the instantaneous difference characteristics, the basic attribute parameters of the corrosion image of the sample to be detected and the basic attribute parameters of the sample to be detected.
According to the detection aspect of the invention, the invention claims a titanium-containing TRIP steel image detection system, which comprises:
the corrosion module is used for acquiring a steel sample to be detected and carrying out corrosion treatment on the steel sample to be detected;
the image capturing module is used for capturing continuous image frames of the steel sample to be detected after corrosion treatment by adopting a high-speed image sensor;
the important frame determining module is used for carrying out image preprocessing on the captured continuous image frame images to determine important frame images;
the characteristic extraction module is used for obtaining a plurality of characteristic points of a metal material image motion area in the important frame image in unit time based on TRIP steel atomic pixel statistics extraction;
the difference comparison module is used for obtaining the basic difference characteristics of the titanium inclusion of the plurality of characteristic points by utilizing the motion change speed value;
the instantaneous comparison module is used for extracting instantaneous difference characteristics of the motion state of the test sample segment based on the change speed of the difference characteristics in two continuous contexts;
and the impurity identification module is used for obtaining the quality of the titanium inclusion of the steel sample to be detected based on the instantaneous difference characteristics.
According to a third aspect of the invention, the invention claims an electronic device comprising: the image detection method comprises a memory and a processor, wherein the memory is used for storing a computer executable program, the processor reads part or all of the computer executable program from the memory and executes the computer executable program, and the processor can realize the titanium-containing TRIP steel image detection method when executing part or all of the computer executable program.
The invention relates to a titanium-containing TRIP steel image detection method and system, corrosion testing is carried out on a test sample by using corrosive liquid, feature points of a metal material image motion area in unit time are obtained by statistics and extraction based on TRIP steel atomic pixels, a motion change speed value is used as a basic difference feature of titanium inclusion, instantaneous difference features of motion states of test sample segments are extracted based on the change speeds of difference blocks in two continuous contexts, and therefore the abnormality of the test sample is detected; the response speed when detecting titanium inclusion occurrence has better identification and detection effects.
Drawings
FIG. 1 is a flow chart of the image detection method of TRIP steel containing titanium;
FIG. 2 is a flowchart illustrating an exemplary embodiment of a method for detecting a titanium-containing TRIP steel image according to an exemplary embodiment of the present disclosure;
FIG. 3 is a flowchart illustrating a third exemplary embodiment of an image inspection method for a TRIP steel containing titanium according to the present invention;
FIG. 4 is a flowchart illustrating a fourth exemplary embodiment of an image inspection method for a TRIP steel containing titanium according to the present invention;
FIG. 5 is a schematic diagram illustrating a fourth embodiment of the method for inspecting an image of a TRIP steel containing titanium according to the present invention;
FIG. 6 is a flowchart illustrating a fifth exemplary embodiment of an image inspection method for a TRIP steel containing titanium according to the present invention;
FIG. 7 is a flowchart illustrating a sixth exemplary embodiment of an image inspection method for a TRIP steel containing titanium according to the present invention;
FIG. 8 is a flowchart illustrating a method for inspecting images of a TRIP steel containing titanium according to a seventh embodiment of the present invention;
FIG. 9 is a block diagram of the image inspection system for TRIP steel containing titanium;
fig. 10 is a system configuration diagram of an electronic apparatus of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. It will be understood that, as used herein, the terms "preliminary," "detecting," and the like may be used herein to describe various elements, but these elements are not limited by these terms unless otherwise specified. These terms are only used to distinguish one element from another. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Referring to fig. 1, the invention provides a method for detecting an image of a titanium-containing TRIP steel, which is characterized by comprising the following steps:
obtaining a steel sample to be detected, and carrying out corrosion treatment on the steel sample to be detected;
adopting a high-speed image sensor to capture continuous image frame images of the corroded steel sample to be detected;
image preprocessing is carried out on the captured continuous image frame images, and important frame images are determined;
a plurality of characteristic points of a metal material image motion area in the important frame image in unit time are obtained through statistic extraction based on TRIP steel atomic pixels;
obtaining the basic difference characteristics of the titanium inclusion of the plurality of characteristic points by utilizing the motion change speed value;
extracting instantaneous difference characteristics of the motion state of the test sample segment based on the change speed of the difference characteristics in two continuous contexts;
and obtaining the quality of the titanium inclusions of the steel sample to be detected based on the instantaneous difference characteristics.
TRIP steel, i.e., transformation induced plasticity steel, whose microstructure is mainly a multi-phase structure composed of ferrite, bainite, retained austenite and a small amount of martensite; ferrite is approximately 50% to 60%, bainite and a small amount of martensite are approximately 25% to 40%, and retained austenite is approximately 5% to 15%.
In addition to the necessary alloying of the TRIP steel, the inclusion in the TRIP steel should be removed as much as possible; the contents of O, N, P, S and Al in the steel can be controlled according to the low alloy steel standard, and particularly the sulfide form should be noted to prevent adverse effects on the steel properties.
The titanium content in TRIP steel is generally 1% -2%, the lower content of TRIP steel is not beneficial to the formation of residual austenite, the TRIP effect cannot be obtained, but the titanium cannot be added too much, otherwise, the volume of the residual austenite is increased, the strength is reduced, and the plasticity is improved.
The element detection is carried out according to the atomic motion condition of each element, and the atomic motion in the corrosion process is amplified so as to more conveniently detect each element substance and carry out content evaluation.
Further, referring to fig. 2, the obtaining a steel sample to be detected and performing corrosion treatment on the steel sample to be detected specifically include:
carrying out corrosion marking treatment on a steel sample to be treated;
the method comprises the steps of enabling the melting concentration of a steel sample to be processed to be at a reference concentration, and conducting corrosion treatment on the steel sample at a preliminarily specified time under the reference concentration by using a catalyst, wherein the reference concentration is higher than a conventional substance concentration and lower than the reference concentration;
standing the steel sample after corrosion treatment is obtained, and carrying out corrosion treatment on the steel sample at a high concentration within a detection specified time which is not less than the preliminary specified time by using a corrosion liquid, wherein the high concentration is higher than the reference concentration;
and after obtaining the steel sample subjected to corrosion treatment, carrying out image shooting treatment on the steel sample.
Specifically, the steel sample to be processed is subjected to corrosion marking treatment, specifically, the steel sample to be processed in a molten state is subjected to marking treatment by using an isotope, and atomic motion characteristics of an area to be observed are marked;
the prepared steel sample is melted to a reference concentration, which is exemplified by a melting concentration between 1500 c and 1600 c, which can be achieved in various ways. The melt concentration of the steel sample is increased according to a time schedule to a higher reference concentration according to a preset melting rate (e.g., 5-15 ℃/min or other time), which may be accomplished, for example, by operating the melt delivery apparatus in the melting furnace, and may be maintained at the higher melt concentration for a specified period of time (e.g., 20-40 minutes or other time). Then, it can be exemplified that the steel sample is sent to the catalyst to be used below for preliminary corrosion treatment using a clean tool.
The catalyst is provided to carry out corrosion catalysis of the steel sample and may be any suitable corrosion catalytic material exemplified by alkaline, acidic, and the like. As an example, an iron trichloride etching solution may be used in this embodiment. Specifically, ferric trichloride powder with a mass purity of not less than 99% can be added into a suitable container such as an anti-corrosion reaction container, and can be added according to a half of the volume of the container or other ratios according to specific requirements to be mixed with water to form the ferric trichloride corrosive liquid with corrosion catalysis. The vessel is then placed into an etching apparatus for etching and may be allowed to increase from the normal material concentration to a reference concentration (e.g., 300-400 c, etc., which is not greater than the reference concentration to be used hereinafter) according to a preset melting rate (e.g., 5-15 c/min or other time), and after reaching that concentration for a specified time (e.g., 1-3 hours or other time), the iron trichloride etching solution may be in the material state at that time.
After the ferric trichloride etching solution in the above state is contacted with the steel sample, it will perform rapid etching treatment on the surface of the steel sample, and the working melting concentration at this time can be controlled at a reference concentration (for example, 350-500 ℃ C. And the like) and can be kept for a specified time (for example, 10-30 minutes or other time) so as to promote the effect of the surface etching treatment at this time to be more sufficient and effective.
Taking out the steel sample subjected to the corrosion treatment from the ferric trichloride corrosive liquid and standing, wherein natural standing, air cooling and other feasible modes can be adopted as an example. Then, the steel sample may be continued to be subjected to etching treatment using an etching solution different from the above-described ferric trichloride etching solution. For example, the etching solution can be a suitable acidic material such as aqua regia, and the like, and the etching solution can be concentrated aqua regia with a concentration ranging from 25% to 35% to etch the steel sample, so that the cross-etching effect can be formed on the surface of the steel sample together with the above-mentioned ferric trichloride etching solution, because the molten ferric trichloride etching solution can dissolve the TRIP steel, and the corrosion defect can be formed on the surface of the steel sample when the molten ferric trichloride etching solution is used alone, and the aqua regia has the double effects of neutralizing the ferric trichloride etching solution and etching the TRIP steel, so that the molten ferric trichloride etching solution can be slowly neutralized, and complete and clear grain boundaries can be formed at the grain boundaries of the steel sample by slow etching, and the outstanding effect is not achieved at all by the prior art method. Therefore, the microstructure structure of the crystal phase on the surface of the TRIP steel can be safely and effectively corroded and exposed, so that the microstructure structure with the compact TRIP steel surface can be clearly observed and analyzed, a worker can be prompted to accurately know the size and distribution of crystal grains, the distribution of air holes and impurity phases and the like, and a reliable basis is provided for implementation process improvement and other aspects.
The above-described etching liquid is subjected to a detection etching treatment of the steel sample at a high concentration (exemplified by a normal substance concentration or other molten concentration value higher), and the present etching treatment time may be set to be maintained for a specified time, exemplified by 1 to 4 hours or other time, and the time period may be not less than the time taken when the steel sample is subjected to the etching treatment using the catalyst.
And after obtaining the steel sample subjected to corrosion treatment, carrying out image shooting treatment on the steel sample by adopting a high-speed image sensor.
Further, referring to fig. 3, the capturing of the continuous image frame image of the steel sample to be detected after the corrosion treatment by using the high-speed image sensor specifically includes:
acquiring discrete frame characteristics of a corrosion process image of a steel sample to be processed and continuous frame characteristics based on Gaussian distribution;
integrating the discrete frame characteristics of the corrosion process image of the steel sample to be processed with the continuous frame characteristics based on Gaussian distribution to obtain an image frame sequence of the corrosion process image of the steel sample to be processed;
inputting the image frame sequence of the corrosion process image of the steel sample to be processed into a preset neural network to obtain a motion sequence;
and obtaining the description information of the corrosion process image of the steel sample to be processed according to the motion sequence.
Representing the characteristics of discrete frames B t And a feature representation S of successive frames based on a Gaussian distribution c Are integrated together.
The neural network is preset to be LSTM. The input being a sequence of image frames R t ＝(r 1 ,r 2 ,...r n ) The output being a sequence of movements (y) 1 ,y 2 ,...y m ). It should be noted that both the input and output are variable and may be of different lengths.
The input sequence of frames is encoded, one at a time, with a potential vector representing the feature information of the image frame, and then the input feature information is decoded, each time resulting in a motion, and finally a complete sentence. This embodiment uses an LSTM cell to input R at time t t LSTM calculates a hidden state h t And a memory state c t 。
The modeling is performed by means of an LDA Topic Model (Topic Model). LDA is a statistical model that forms topic information based on the conditional probability that the same and different phrases occur in different texts. The general task of the method includes that twenty-five frames are randomly grabbed to carry out artificial TRIP steel atom marking, subject information is extracted by using descriptions of known artificial TRIP steel atom marking to serve as a priori, and then other frame descriptions are generated in a guiding mode by combining an attention mechanism. The rest frames can be guided to generate the rest frame description fitting the theme by marking the extracted theme of the twenty-five frames by the TRIP steel atom. Therefore, twenty-five frames are randomly grabbed to carry out TRIP steel atom marking, the description diversity can be increased, and the theme can be richer.
The modeling process comprises the following steps:
random frame grabbing: l = (L) 1 ,l 2 ,l 3 ....l n-1 ,l n ) Wherein n represents the total number of frames, l i Representing random grabbing of frames, the present embodiment grabs twenty-five frames.
TRIP Steel atomic signature description: frame description D = (D) 1 ,d 2 ,d 3 ...d n-1 ,d n ) Wherein n represents the total number of frames, d i A descriptive statement representing a random grab frame, the TRIP steel atom of this embodiment marks twenty-five frames.
Extracting topics based on TRIP steel atomic marker description: after random frame grabbing and TRIP steel atom mark description are completed, theme model modeling is carried out on the grabbed frame description of twenty-five frames, and theme distribution is expressed as follows: topic h ＝{z 1 ,z 2 ,...z m }。
And performing TRIP steel atom marking on twenty-five frames in all the captured frames, and regarding the content of the TRIP steel atom marking as a document. And extracting theme information by using the LDA theme model. The traditional method is to input the TRIP steel melting characteristics into an encoder to encode the TRIP steel melting characteristics for network initialization, at the next moment, the decoded output at the previous moment is taken as the decoded output at the later moment and input into an LSTM unit, and finally the output of the LSTM decoder is transmitted to a softmax layer to be guided to the current highest-scoring moving atom.
Modeling the extracted subject information to form a subject feature Topic h And the image frame description is integrated with the TRIP steel melting characteristics and input into LSTM decoding to guide the generation of the image frame description except for the artificial TRIP steel atom mark. The general task of this embodiment is to obtain descriptions of the rest frames after randomly grabbing twenty-five frames of TRIP steel atoms for an image, so that the descriptions not only contain rich TRIP steel melting characteristics, but also have theme information blended in to make the descriptions more vivid and smooth.
Further, referring to fig. 4, the performing image preprocessing on the captured continuous image frame images to determine an important frame image specifically includes:
counting the mass distribution value distribution data of different mass distribution values of each marked isotope in each frame of image frame aiming at any image frame of the image in the corrosion process and the image frame before the image frame;
calculating the similarity of any image frame of the corrosion process image and the image frame before the image frame according to different mass distribution value distribution data of the same marked isotope;
when the similarity is smaller than a preliminary threshold value, determining any image frame of the corrosion process image as a candidate frame;
calculating the quality unbalance degree of the candidate frame, wherein the quality unbalance degree represents the quality distribution condition of the candidate frame;
and acquiring the candidate frames with the quality unbalance degrees larger than the detection threshold value as important frames.
Specifically, the mass distribution value distribution data may include the number of occurrences of different mass distribution values of the same isotope label or the frequency of occurrences of different mass distribution values, which are used to represent the quality characteristics of the image frame. The occurrence frequency of any mass distribution value is specifically a ratio of the occurrence frequency of any mass distribution value to the number of the total TRIP steel atom pixels of the image frame, and when the mass distribution value distribution data is the occurrence frequency of different mass distribution values, the mass distribution value distribution data may be specifically a mass histogram.
For two adjacent frames of image frames, the similarity of the two adjacent frames of image frames can be obtained according to the distribution data of different quality distribution values of the same marked isotope in the two adjacent frames of image frames. There are many possible implementations of the similarity calculation, which will be described in detail in the following embodiments.
If the similarity is smaller than the first threshold, it indicates that the similarity of the image frame of the previous frame of any image frame of the image frame is smaller, the image frame may be an important frame, and thus it may be determined that the image frame is a candidate frame of the important frame. And if the similarity is larger than the first threshold value, the image frame is similar to the image frame of the previous frame of any image frame of the image frame, and the image frame is not an important frame, namely the image frame is discarded.
And selecting the candidate frames with rich quality or rich brightness as the important frames by calculating the quality imbalance of the obtained candidate frames. The quality unbalance degree can be used for representing the quality distribution condition of the candidate frame, if the quality unbalance degree is larger than the second threshold value, the candidate frame is rich in quality, larger single quality cannot occur, the contained information amount is large, and therefore the candidate frame is an important frame. After each candidate frame is obtained, the mass imbalance can be calculated, and if the mass imbalance is greater than the second threshold, the candidate frame is an important frame.
For any image frame of the image in the corrosion process, comparing the image frame of the previous frame of the image in the corrosion process with the distribution data of the mass distribution value to determine whether any image frame of the image in the corrosion process can be used as a candidate frame of the important frame, and then judging the mass unbalance degree of the candidate frame, namely determining whether the candidate frame can be used as the important frame, so that the method is not only suitable for obtaining the important frame of the offline image data, but also can process any image data in real time, and improves the effectiveness and the universality of obtaining the important frame.
And the quality characteristics of the quality distribution value distribution data are utilized to judge the similarity of two adjacent frames of image frames, only the quality distribution value is counted, the complexity is low, the extraction speed is high, the quality distribution value distribution data are global characteristics of the image frames, and the small-amplitude change of a main body in the image does not cause the change of the global characteristics, so the robustness is higher.
When the image frame includes the mass distribution value distribution data of a plurality of masses, as another embodiment, the calculating the similarity between any image frame of the erosion process image and the image frame of the previous image frame of the any image frame according to the mass distribution value distribution data of the same labeled isotope is specifically: calculating the similarity coefficient of any image frame of the corrosion process image and the image frame mass distribution value distribution data of the previous frame of the image frame corresponding to the same marked isotope according to the mass distribution value distribution data of the same marked isotope; and taking the average value, the maximum value or the minimum value of the quality distribution value distribution data similarity coefficients corresponding to different qualities of any image frame of the corrosion process image and the previous image frame of any image frame as the similarity of any image frame of the corrosion process image and the previous image frame of any image frame. The sum of the similarity coefficients of the different mass distribution value distribution data may also be selected as the similarity, which is not limited in the present application, and the similarity obtained according to the intersection coefficients of the mass distribution value distribution data of different masses is within the protection scope of the present application.
When the image frame is a mass spectrum image, the image frame only has mass distribution value distribution data corresponding to one mass-to-charge ratio, namely mass-to-charge ratio distribution data, so that the similarity coefficient of the corresponding mass-to-charge ratio of the image frames of two adjacent frames is taken as the similarity of the image frames of the two adjacent frames. When the statistics of the mass distribution value distribution data is the occurrence frequency of each mass distribution value, the mass distribution value distribution data can be represented by a mass spectrum data map.
Further, referring to fig. 5, the extracting and obtaining a plurality of feature points of the metal material image motion region in the important frame image in unit time based on the TRIP steel atomic pixel statistics specifically includes:
obtaining a TRIP steel atomic pixel image according to the important frame image;
integrating the important frame image and the TRIP steel atomic pixel image to obtain an integrated image;
and inputting the integrated image into a pre-trained CNN characteristic extraction model for characteristic extraction to obtain image characteristics.
And periodically extracting one frame image from the image frame images as an important frame image according to a preset frequency. The image data has a large amount of repeated and invalid data, and the convolutional neural network CNN has a large calculation amount and is independent in time dimension because the processing of different frames is time-invariant; therefore, convolution operations of different frames can be processed in parallel, repeated data is removed, image frames are extracted, calculation is accelerated, deployment is convenient, and end-to-end learning is completed.
The video important frame is a frame used for video compression and video encoding and decoding, and contains complete information, and other non-important frames are compressed by using a difference value with the important frame. The image frame can be specifically divided into three types of IPB frames: the I frame represents an important frame and is the most complete frame picture; the P frame represents a single guide frame, and the previous I frame or P frame is utilized to carry out context guide coding in a motion guide mode; b frame represents bidirectional guide frame, using bidirectional frame to make guide coding; in general, the important frame I frame is the most information frame and the most usage frame. The embodiment of the invention selects to extract I frames, and the preset frequency is to extract one frame every second.
In 10 continuous important frame images, respectively randomly extracting 1 frame image from the first 5 frame images, randomly extracting 1 frame image from the second 5 frame images, obtaining a horizontal vector and a vertical vector of a TRIP steel atomic pixel according to the extracted 2 frame image important frame images, taking the horizontal vector and the vertical vector as first dimension data and second dimension data of the TRIP steel atomic pixel image, adding third dimension data, setting the value of the third dimension data of the TRIP steel atomic pixel image to be 0, and obtaining the TRIP steel atomic pixel image corresponding to the 10 frame image.
The TRIP steel atomic pixel represents the motion speed and the motion direction of each TRIP steel atomic pixel in two adjacent frame images, is used for describing the transient motion transition (motion direction and motion offset) of a motion atom, is essentially a two-dimensional vector field, each vector represents the displacement of the point in the scene from the previous frame to the next frame, can be decomposed into a horizontal vector and a vertical vector, and the parts of the horizontal vector (namely, the first dimension) and the vertical vector (the second dimension) of the TRIP steel atomic pixel are rescaled to the range of [0, 255] and are regarded as the channel of the TRIP steel atomic pixel image. The important frame image comprises RGB three channels, namely the numerical value of the TRIP steel atomic pixel point of the image is represented by three-dimensional data.
Because the distribution of TRIP steel atomic pixels is different from that of RGB images, the activation values of the first convolution layer have different distributions, and after initialization is carried out by using a pre-training model, in order to prevent overfitting and enable optimization solution to be stable and quick, the embodiment of the invention adopts regularization on image data of important frame images, wherein the main idea of regularization is to calculate the 2-norm of each image sample, and then divide each element in the sample by the norm, so that the 2-norm (L2-norm) of each processed sample is equal to 1;
the image has a strong time sequence relation, the context frames are all connected, the attention mechanism module is added to enhance the reading capability of the model for time sequence characteristics, and the characteristics of each frame extracted by the convolutional neural network are also connected in time sequence among channels. Embodiments of the present invention model these one by one and combine them to extract more image features as input to subsequent networks.
Further, referring to fig. 6, the obtaining of the basic difference characteristics of the titanium inclusion at the plurality of feature points by using the motion variation speed value specifically includes:
acquiring original data of image characteristics;
extracting important frame image characteristics and important frame image characteristic difference values based on the acquired original data;
calculating the difference and Euclidean distance between the extracted important frame image characteristics and the important frame image characteristic difference value;
and completing the detection of the image characteristic difference according to the obtained difference and the Euclidean distance.
Searching and collecting a data set in the aspect of relevant depth motion recognition, carrying out format unification on the collected data set, generating a label, positioning a picture TRIP steel, cutting the picture and other preprocessing. The data set is then partitioned. The adopted comparison data sets are public data sets, most of the data sets are TRIP steel videos collected from other public video websites, and then the motion recognition videos are created on the basis through different motion recognition means. Because the collected data in the data sets are all streamed on the internet in the form of videos and have no uniform format and label with different division before each other, before the formal experiment starts, each related data set needs to be subjected to unified data preprocessing, namely, the data format is unified; for videos of different formats and different sizes in a data set, all videos are extracted into pictures in units of frames.
In order to improve the effect of network training and suppress the interference of irrelevant video backgrounds except for the TRIP steel part in the training, for all picture frames extracted from the data set video, the embodiment uses the face recognition library to position and cut TRIP steel areas in the picture frames, and uniformly adjusts the size of the picture frames into 128 × 128 for storage. After the data format is unified, label processing is performed on each picture according to the difference of the videos in the data set and the division of the training data.
The overall network architecture is a dual-stream network composed of two Convolutional Neural Networks (CNN) subnetworks, which are respectively an intra-frame subnetwork for extracting intra-frame spatial features and a context subnetwork for extracting context temporal features. The intra-frame sub-network extracts important frame image features from RGB images of the video, and the context sub-network extracts important frame image feature difference values from a logistic regression graph of the video; for context flows, logical regression is used to represent the context flow. Since the basic principle of depth motion recognition in video processing is to process motion recognition images of each frame and then concatenate the motion recognition images generated by them. This will result in regions of the TRIP steel with more varied atoms that may be deformed more by context inconsistencies. Compared with other traditional technologies based on intra-frame image characteristics, the method has the advantages that the logistic regression characteristics can be better focused on the area with larger TRIP steel atom change, and the detection accuracy is improved.
Judging whether the video is changed or not by judging the difference between the image characteristics of the important frame and the difference values of the image characteristics of the important frame; the difference between the difference value of the image feature of the important frame and the difference value of the image feature of the important frame is captured by a contrast loss function, so that the Euclidean distance between the difference values of the image feature of the intra frame and the image feature of the important frame is smaller for a static video and larger for a motion recognition video. The intra-frame sub-network extracts important frame image features from an RGB (red, green and blue) image of a video, and the context network extracts important frame image feature difference values from a video logistic regression image which is preprocessed in advance; the network architecture of the intra-frame context subnetwork is based on a ResNet network using separable convolution, and the adopted residual learning mechanism solves the problems of slow network convergence, degraded training effect and the like.
In the process of extracting the feature difference value of the important frame image, a logistic regression algorithm is adopted, and specifically: determining the weight of each pixel point according to the pixel values and coordinates of other pixel points in the neighborhood of each pixel point, and then expanding the coordinates of the point by using a polynomial. The dependent variable is shown in equation (1) by regarding the image as a function of a two-dimensional signal (the output image is a grayscale image). The image is then approximately modeled using a quadratic polynomial as shown in equation (2), where A represents a 2x2 symmetric matrix, b is a 2x 1 vector matrix, and c is a scalar:
the two-dimensional signal space (Cartesian coordinate system) of the original image requires a six-dimensional vector as a coefficient transformation to (1, x, y, x) 2 ，y 2 Xy) is the space of the basis function, and the positions x and y of different pixel points are substituted to calculate the gray values of the different pixel points. In order to obtain the six-bit coefficient of each pixel point in each frame of image, the logistic regression algorithm sets a neighborhood of (2n + 1) × (2n + 1) around each pixel point, and then (2n + 1) in the pixel point domain is used 2 And fitting by taking the pixel points as sample points of a least square method.
In a gray value matrix with the size of (2n + 1) x (2n + 1) in the neighborhood of a pixel point, the matrixes are split and combined into (2n + 1) 2 The vector f for x 1 is given in column-first order (1,x, y, x) 2 ，y 2 Xy) as the basis function and the dimension of the transformation matrix B is (2n + 1) 2 X 6 (i.e. 6 column vectors b) i A matrix composed of all), the dimension of the coefficient vector r shared in the neighborhood is 6 × 1, then there is formula (4):
f＝B×r＝(b 1 b 2 b 3 b 4 b 5 b 6 )×r (4)
when the least square method is used for solving, the logistic regression algorithm gives weight to the sample error of each pixel point in the neighborhood by using two-dimensional Gaussian distribution. In the (2n + 1) x (2n + 1) matrix with two-dimensional Gaussian distribution in each pixel point neighborhood, the matrix is split into (2n + 1) in the order of column priority 2 Vector a of x 1. As shown in equation (5), the original transformation matrix B of the basis functions will be transformed into:
B＝(a·b 1 a·b 2 a·b 3 a·b 4 a·b 5 a·b 6 ) (5)
and converting the basis function matrix B again in a dual mode, so that the coefficient vector of each pixel point in a single image can be obtained. And then, CRF can be obtained through parameter vector calculation and local fuzzification processing, and after CRF is obtained, in order to enable the input data structure of the context flow to correspond to the RGB three-layer data structure of the frame flow, the two layers of logistic regression data matrixes are supplemented into a three-layer matrix.
The network architecture of the intra-frame subnetwork is composed of 36 convolutional layers, the network is based on an Xception network model, and the network shows strong learning capability in the aspect of image vision. The context sub-network is used for exploring the space-time long-distance context correlation of TRIP steel key areas in the video frames so as to enhance the capability of learning the representation. The method uses logistic regression to deduce the displacement process and direction of the pixel points in the image through the context of two frames of images, and then captures the features from the logistic regression through the context sub-network.
Specifically, referring to fig. 7, the extracting the instantaneous difference feature of the motion state of the test sample segment based on the change speed of the difference feature in two consecutive contexts specifically includes:
acquiring a change speed value between any two continuous image frames in the important frames;
determining the next frame of image frame of the continuous two frames of data with the maximum change speed value as an instant image frame;
and comparing the instantaneous image frame with the preliminary image frame in the important frame to obtain the instantaneous difference characteristic.
Specifically, the change speed value is a change percentage of the image characteristic difference value of any two consecutive image frames;
the temporal difference feature is a percentage change in the image feature difference value between the temporal image frame and a preliminary image frame of the important frames.
Specifically, referring to fig. 8, the obtaining of the titanium inclusion quality of the steel sample to be detected based on the transient difference characteristic specifically includes:
acquiring basic attribute parameters of the corrosion image of the sample to be detected and basic attribute parameters of the sample to be detected;
and acquiring the impurity quality of the steel sample to be detected based on the instantaneous difference characteristics, the basic attribute parameters of the corrosion image of the sample to be detected and the basic attribute parameters of the sample to be detected.
The basic attribute parameters of the sample to be detected comprise weight and density;
the basic attribute parameters of the corrosion image of the sample to be detected comprise corrosion ending duration and corrosion liquid amount.
And integrating the motion change characteristic rate of TRIP steel atoms in the transient difference characteristic, the weight and the density of the sample to be detected, the corrosion end time of the corrosion image of the sample to be detected and the corrosion liquid amount to obtain the quality of the titanium impurities of the steel sample to be detected.
Referring to fig. 9, the invention provides a titanium-containing TRIP steel image detection system, comprising:
the corrosion module is used for acquiring a steel sample to be detected and carrying out corrosion treatment on the steel sample to be detected;
the image capturing module is used for capturing continuous image frames of the steel sample to be detected after corrosion treatment by adopting a high-speed image sensor;
the important frame determining module is used for carrying out image preprocessing on the captured continuous image frame images to determine important frame images;
the characteristic extraction module is used for obtaining a plurality of characteristic points of a metal material image motion area in the important frame image in unit time based on TRIP steel atomic pixel statistics extraction;
the difference comparison module is used for obtaining the basic difference characteristics of the titanium inclusion of the plurality of characteristic points by utilizing the motion change speed value;
the instantaneous comparison module is used for extracting instantaneous difference characteristics of the motion state of the test sample segment based on the change speed of the difference characteristics in two continuous contexts;
and the impurity identification module is used for obtaining the quality of the titanium inclusion of the steel sample to be detected based on the instantaneous difference characteristics.
Referring to fig. 10, the present invention claims an electronic device 901, including: the memory 902 and the processor 903 are used for storing the computer executable program, the processor reads part or all of the computer executable program from the memory and executes the computer executable program, and the processor can realize the titanium-containing TRIP steel image detection method when executing part or all of the computer executable program.
Those skilled in the art will appreciate that the disclosure of the present disclosure is susceptible to numerous variations and modifications. For example, the various devices or components described above may be implemented in hardware, software, firmware, or a combination of some or all of the three.
Flow charts are used in this disclosure to illustrate steps of methods according to embodiments of the disclosure. It should be understood that the preceding or following steps are not necessarily performed in exact order. Rather, various steps may be processed in reverse order or concurrently. Also, other operations may be added to these processes.
It will be understood by those skilled in the art that all or part of the steps of the above methods may be performed by sequential hardware instructions of a computer program, which may be stored in a computer readable storage medium, such as a read-only memory, a magnetic or optical disk, and the like. Alternatively, all or part of the steps of the above embodiments may be implemented using one or more integrated circuits. Accordingly, each module/unit in the above embodiments may be implemented in the form of hardware, and may also be implemented in the form of a software functional module. The present disclosure is not limited to any specific form of combination of hardware and software.
Unless otherwise defined, all terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of sequential techniques and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The foregoing is illustrative of the present disclosure and is not to be construed as limiting thereof. Although a few exemplary embodiments of this disclosure have been described, those skilled in the art will readily appreciate that many modifications are possible in the exemplary embodiments without materially departing from the novel teachings and advantages of this disclosure. Accordingly, all such modifications are intended to be included within the scope of this disclosure as defined in the claims. It is to be understood that the foregoing is illustrative of the present disclosure and is not to be construed as limited to the specific embodiments disclosed, and that modifications to the disclosed embodiments, as well as other embodiments, are intended to be included within the scope of the appended claims. The disclosure is defined by the claims and their equivalents.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an illustrative embodiment," "an example," "a specific example," or "some examples" or the like mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
Claims (10)
1. The image detection method of the titanium-containing TRIP steel comprises the following steps:
obtaining a steel sample to be detected, and carrying out corrosion treatment on the steel sample to be detected;
adopting a high-speed image sensor to capture continuous image frame images of the corroded steel sample to be detected;
performing image preprocessing on the captured continuous image frame images to determine important frame images;
a plurality of characteristic points of a metal material image motion area in the important frame image in unit time are obtained through statistic extraction based on TRIP steel atomic pixels;
obtaining basic difference characteristics of the titanium inclusion of the plurality of characteristic points by utilizing the motion change speed value;
extracting instantaneous difference characteristics of the motion state of the test sample segment based on the change speed of the difference characteristics in two continuous contexts;
and obtaining the quality of the titanium inclusions of the steel sample to be detected based on the instantaneous difference characteristics.
2. The method for detecting the titanium-containing TRIP steel image as claimed in claim 1, wherein the step of obtaining the steel sample to be detected and performing corrosion treatment on the steel sample to be detected specifically comprises:
carrying out corrosion marking treatment on a steel sample to be treated;
the method comprises the steps of enabling the melting concentration of a steel sample to be processed to be at a reference concentration, and conducting corrosion treatment on the steel sample at a preliminarily specified time under the reference concentration by using a catalyst, wherein the reference concentration is higher than a conventional substance concentration and lower than the reference concentration;
standing the obtained steel sample after corrosion treatment, and carrying out corrosion treatment on the steel sample at a high concentration for a detection specified time which is not less than the preliminary specified time by using a corrosion liquid, wherein the high concentration is higher than the reference concentration;
and after obtaining the steel sample subjected to corrosion treatment, carrying out image shooting treatment on the steel sample.
3. The method for detecting the TRIP steel image containing titanium according to claim 1, wherein the step of capturing the continuous image frame image of the steel sample to be detected after the corrosion treatment by using a high-speed image sensor comprises the following steps:
acquiring discrete frame characteristics of a corrosion process image of a steel sample to be processed and continuous frame characteristics based on Gaussian distribution;
integrating the discrete frame characteristics of the corrosion process image of the steel sample to be processed with the continuous frame characteristics based on Gaussian distribution to obtain an image frame sequence of the corrosion process image of the steel sample to be processed;
inputting the image frame sequence of the corrosion process image of the steel sample to be processed into a preset neural network to obtain a motion sequence; and obtaining the description information of the corrosion process image of the steel sample to be processed according to the motion sequence.
4. The method for detecting the TRIP steel image containing titanium according to claim 1, wherein the image preprocessing is performed on the captured continuous image frame images to determine important frame images, and the method specifically comprises the following steps:
counting the mass distribution value distribution data of different mass distribution values of each marked isotope in each frame of image frame aiming at any image frame of the image in the corrosion process and the image frame before the image frame;
calculating the similarity of any image frame of the corrosion process image and the image frame before the image frame according to different mass distribution value distribution data of the same marked isotope;
when the similarity is smaller than a preliminary threshold value, determining any image frame of the corrosion process image as a candidate frame;
calculating the quality unbalance degree of the candidate frame, wherein the quality unbalance degree represents the quality distribution condition of the candidate frame;
and acquiring the candidate frame with the quality unbalance degree larger than the detection threshold as the important frame.
5. The method for detecting the TRIP steel image containing titanium according to claim 1, wherein the method for extracting and obtaining the plurality of feature points of the metal material image motion area in the important frame image in unit time based on the TRIP steel atomic pixel statistics specifically comprises the following steps:
obtaining a TRIP steel atomic pixel image according to the important frame image;
integrating the important frame image and the TRIP steel atomic pixel image to obtain an integrated image;
and inputting the integrated image into a pre-trained CNN characteristic extraction model for characteristic extraction to obtain image characteristics.
6. The method for detecting the TRIP steel image containing titanium according to claim 1, wherein the step of obtaining the basic difference characteristics of the titanium inclusion of the plurality of feature points by using the motion change velocity values specifically comprises the following steps:
acquiring original data of image characteristics;
extracting important frame image characteristics and important frame image characteristic difference values based on the acquired original data;
calculating the difference and Euclidean distance between the extracted important frame image characteristics and the important frame image characteristic difference value;
and completing the detection of the image characteristic difference according to the obtained difference and the Euclidean distance.
7. The method for detecting the TRIP steel image containing titanium according to claim 1, wherein the step of extracting the transient difference characteristic of the motion state of the test sample segment based on the change speed of the difference characteristic in two continuous contexts specifically comprises the following steps:
acquiring the change speed value between any two continuous image frames in the important frames;
determining the next frame of image frame of the continuous two frames of data with the maximum change speed value as an instant image frame;
and comparing the instantaneous image frame with the preliminary image frame in the important frame to obtain the instantaneous difference characteristic.
8. The method for detecting the TRIP steel image containing titanium according to claim 1, wherein the step of obtaining the titanium inclusion quality of the steel sample to be detected based on the transient difference characteristics comprises the following steps:
acquiring basic attribute parameters of the corrosion image of the sample to be detected and basic attribute parameters of the sample to be detected;
and acquiring the impurity quality of the steel sample to be detected based on the instantaneous difference characteristics, the basic attribute parameters of the corrosion image of the sample to be detected and the basic attribute parameters of the sample to be detected.
9. A titanium-containing TRIP steel image detection system comprises:
the corrosion module is used for acquiring a steel sample to be detected and carrying out corrosion treatment on the steel sample to be detected;
the image capture module is used for capturing continuous image frames of the corroded steel sample to be detected by adopting a high-speed image sensor;
the important frame determining module is used for carrying out image preprocessing on the captured continuous image frame images to determine important frame images;
the characteristic extraction module is used for obtaining a plurality of characteristic points of a metal material image motion area in the important frame image in unit time based on TRIP steel atomic pixel statistics extraction;
the difference comparison module is used for obtaining the basic difference characteristics of the titanium inclusion of the plurality of characteristic points by utilizing the motion change speed value;
the instantaneous comparison module is used for extracting instantaneous difference characteristics of the motion state of the test sample segment based on the change speed of the difference characteristics in two continuous contexts;
and the impurity identification module is used for obtaining the quality of the titanium inclusion of the steel sample to be detected based on the instantaneous difference characteristics.
10. An electronic device, comprising: the image detection method for the titanium-containing TRIP steel comprises a memory and a processor, wherein the memory is used for storing a computer executable program, the processor reads part or all of the computer executable program from the memory and executes the computer executable program, and when the processor executes part or all of the computer executable program, the image detection method for the titanium-containing TRIP steel can be realized according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211276811.5A CN115760697A (en) | 2022-10-18 | 2022-10-18 | Titanium-containing TRIP steel image detection method and system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211276811.5A CN115760697A (en) | 2022-10-18 | 2022-10-18 | Titanium-containing TRIP steel image detection method and system and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115760697A true CN115760697A (en) | 2023-03-07 |
Family
ID=85353762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211276811.5A Pending CN115760697A (en) | 2022-10-18 | 2022-10-18 | Titanium-containing TRIP steel image detection method and system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115760697A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117876366A (en) * | 2024-03-11 | 2024-04-12 | 宝鸡子扬双金属材料有限公司 | Titanium tube quality detection method and system based on image processing |
-
2022
- 2022-10-18 CN CN202211276811.5A patent/CN115760697A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117876366A (en) * | 2024-03-11 | 2024-04-12 | 宝鸡子扬双金属材料有限公司 | Titanium tube quality detection method and system based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111126453B (en) | Fine-grained image classification method and system based on attention mechanism and cut filling | |
| WO2020108362A1 (en) | Body posture detection method, apparatus and device, and storage medium | |
| US8027542B2 (en) | High speed video action recognition and localization | |
| CN112016500A (en) | Group abnormal behavior identification method and system based on multi-scale time information fusion | |
| CN111353395A (en) | Face changing video detection method based on long-term and short-term memory network | |
| CN113065474B (en) | Behavior recognition method and device and computer equipment | |
| CN110120064A (en) | A kind of depth related objective track algorithm based on mutual reinforcing with the study of more attention mechanisms | |
| CN108200432A (en) | A kind of target following technology based on video compress domain | |
| CN110555868A (en) | method for detecting small moving target under complex ground background | |
| CN115393396B (en) | Unmanned aerial vehicle target tracking method based on mask pre-training | |
| CN113610087B (en) | Priori super-resolution-based image small target detection method and storage medium | |
| CN111986180A (en) | Face forged video detection method based on multi-correlation frame attention mechanism | |
| CN115760697A (en) | Titanium-containing TRIP steel image detection method and system and electronic equipment | |
| CN112419174A (en) | Image character removing method, system and device based on gate cycle unit | |
| CN114359333A (en) | Moving object extraction method and device, computer equipment and storage medium | |
| CN111462090A (en) | Multi-scale image target detection method | |
| CN111091583B (en) | Long-term target tracking method | |
| CN116630850A (en) | Twin target tracking method based on multi-attention task fusion and bounding box coding | |
| CN116934796A (en) | Visual target tracking method based on twinning residual error attention aggregation network | |
| CN111127355A (en) | Method for finely complementing defective light flow graph and application thereof | |
| CN114612456B (en) | Billet automatic semantic segmentation recognition method based on deep learning | |
| CN113469913B (en) | Hot-rolled strip steel surface water drop removing method based on gradual cycle generation countermeasure network | |
| CN112164097B (en) | Ship video detection sample collection method | |
| Zhou et al. | ACR-Net: attention integrated and cross-spatial feature fused rotation network for tubular solder joint detection | |
| CN113743188A (en) | Internet video low-custom behavior detection method based on feature fusion |
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 |