Dynamic matching identification method and system for morphological characteristics of multi-modal traces

Abstract

The invention discloses a dynamic matching identification method and system for multi-modal trace morphological features, and relates to the technical field of feature matching, and the method comprises the steps: carrying out the color feature extraction of a trace image, and obtaining a color feature value; matching the reference color feature value set and comparing the reference color feature value set with the color feature values to obtain a color similarity set; judging whether the color coefficient meets a preset coefficient threshold value or not; if yes, obtaining a texture feature value set; and extracting a first texture feature value set corresponding to the first tool in the tool feature database, comparing the texture feature value set to obtain a first texture similarity, and when the similarity reaches a preset similarity limit value, taking a first type corresponding to the first tool as a tool type of the trace image. The technical problems that in the prior art, due to the fact that a trace image is interfered by factors such as materials, illumination, corrosion or thermochromism, the recognition accuracy is poor, and the tool type cannot be stably judged are solved, and the technical effect of improving the tool trace recognition precision and stability under the complex condition is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of feature matching, and particularly relates to a dynamic matching and recognition method and system for multi-modal trace morphological features. Background Art

[0002] After a tool is used on a trace-bearing object made of different materials such as metal, plastic, and fabric, various types of traces are formed, such as sole pattern imprints, tool sliding marks, tire patterns, etc. These traces show significant structural differences in images and are easily affected by factors such as the reflection characteristics of the material, changes in light, corrosion and oxidation, and high-temperature discoloration, resulting in color distortion, detail loss, or texture blur in the images, seriously interfering with the stable extraction of trace features. Especially in actual acquisition scenarios, there are often problems such as local occlusion, surface contamination, or color difference drift in the trace area, further reducing the separability and contrast of color and texture features. Traditional recognition methods relying on single visual features are difficult to maintain the accuracy and consistency of recognition in such complex environments, restricting the stable determination and automatic recognition ability of tool types. Summary of the Invention

[0003] The present application provides a dynamic matching and recognition method and system for multi-modal trace morphological features, which are used to solve the technical problems in the prior art that the recognition accuracy is poor due to the interference of factors such as material, light, corrosion, or thermal discoloration of trace images, and the tool type cannot be stably determined.

[0004] In view of the above problems, the present application provides a dynamic matching and recognition method and system for multi-modal trace morphological features.

[0005] In the first aspect of the present application, a dynamic matching and recognition method for multi-modal trace morphological features is provided. The method includes: Performing color feature extraction on the collected trace image to obtain color feature values, where the trace image refers to an image of the use trace of a tool on a trace-bearing object; matching the set of reference color feature values of the trace-bearing object, and comparing the color feature values with the set of reference color feature values to obtain a set of color similarity degrees; judging and analyzing whether the color coefficients obtained from the set of color similarity degrees meet a predetermined coefficient threshold; if they meet, issuing a texture feature collection instruction, and obtaining a set of texture feature values of the trace image according to the texture feature collection instruction; extracting a set of first texture feature values corresponding to a first tool in the tool feature database; comparing the set of texture feature values with the set of first texture feature values to obtain a first texture similarity; when the first texture similarity reaches a predetermined similarity limit value, taking the first type corresponding to the first tool as the tool type of the trace image.

[0006] In the second aspect of the present application, a dynamic matching and recognition system for multi-modal trace morphological features is provided. The system includes: A feature extraction module, configured to extract color features from the collected trace images to obtain color feature values, where the trace images refer to images of the usage traces of tools on the trace-bearing objects; a matching and comparison module, configured to match the set of reference color feature values of the trace-bearing objects, and compare the color feature values with the set of reference color feature values to obtain a set of color similarity degrees; a judgment module, configured to judge and analyze whether the color coefficients obtained from the set of color similarity degrees meet a predetermined coefficient threshold; a texture feature collection module, configured to, if it meets the requirement, issue a texture feature collection instruction, and obtain a set of texture feature values of the trace images according to the texture feature collection instruction; an extraction module, configured to extract a set of first texture feature values corresponding to a first tool in a tool feature database; a similarity determination module, configured to compare the set of texture feature values with the set of first texture feature values to obtain a first texture similarity degree; a tool type confirmation module, configured to, when the first texture similarity degree reaches a predetermined similarity limit value, use the first type corresponding to the first tool as the tool type of the trace images.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: In this application, color features are extracted from the collected trace images to obtain color feature values, where the trace images refer to images of the usage traces of tools on the trace-bearing objects; the set of reference color feature values of the trace-bearing objects is matched, and the color feature values are compared with the set of reference color feature values to obtain a set of color similarity degrees; it is judged and analyzed whether the color coefficients obtained from the set of color similarity degrees meet a predetermined coefficient threshold; if it meets the requirement, a texture feature collection instruction is issued, and a set of texture feature values of the trace images is obtained according to the texture feature collection instruction; a set of first texture feature values corresponding to a first tool in a tool feature database is extracted; the set of texture feature values is compared with the set of first texture feature values to obtain a first texture similarity degree; when the first texture similarity degree reaches a predetermined similarity limit value, the first type corresponding to the first tool is used as the tool type of the trace images. This invention solves the technical problems in the prior art that due to factors such as material, illumination, corrosion, or thermal discoloration of the trace images, the recognition accuracy is poor and the tool type cannot be stably judged. By introducing a dual feature fusion discrimination mechanism of color and texture, the trace images are dynamically matched with the tool template features, achieving the technical effect of improving the recognition accuracy and stability of tool traces under complex conditions. Description of the Drawings

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0009] Figure 1 Schematic flow chart of the dynamic matching and recognition method for multi-modal trace morphological features provided by the embodiments of the present application; Figure 2 Schematic structural diagram of the dynamic matching and recognition system for multi-modal trace morphological features provided by the embodiments of the present application.

[0010] Explanation of reference numerals: Feature extraction module 11, matching and comparison module 12, judgment module 13, texture feature collection module 14, extraction module 15, similarity determination module 16, tool type confirmation module 17. Specific implementation manners

[0011] The present application provides a dynamic matching and recognition method and system for multi-modal trace morphological features. To solve the technical problems in the prior art that the recognition accuracy is poor and the tool type cannot be stably judged due to the interference of factors such as material, light, corrosion, or thermal discoloration on the trace image, by introducing a dual feature fusion discrimination mechanism of color and texture, the trace image is dynamically matched with the tool template features, achieving the technical effect of improving the recognition accuracy and stability of tool traces under complex conditions.

[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0013] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0014] Embodiment 1, as Figure 1 shown, the present application provides a dynamic matching and recognition method for multi-modal trace morphological features, and the method includes: Step S100: Extract color features from the collected trace image to obtain color feature values, where the trace image refers to an image of the usage trace of a tool on a trace-bearing object.

[0015] In the embodiments of the present application, a trace image refers to image information with indentations, patterns, scratches, or structural textures formed after a tool is used on a trace-bearing object (such as a metal surface, rubber product, fabric material, etc.), including morphological forms such as sole pattern imprints, strip-shaped friction traces, or surface imprints generated by the edge of a tool structure. When collecting such trace images, a high-resolution industrial camera or a microscopic image acquisition device is used to capture the target trace area under uniform lighting conditions and background environment to avoid image color deviation caused by surface reflection, material light absorption differences, etc., and ensure that the collected images have good color consistency and detail retention.

[0016] Subsequently, color features are extracted from the collected trace image. By calculating the discrete cosine transform (DCT) coefficients of the image and extracting the direct current coefficient therefrom as the color feature value.

[0017] Furthermore, in the method provided by the embodiments of the application, when extracting color features from the collected trace image to obtain color feature values, it further includes: Obtain the discrete cosine transform coefficients of the trace image; characterize the color feature value with the direct current coefficient in the discrete cosine transform coefficients.

[0018] In the embodiments of the present application, first, preprocessing is performed on the collected trace image, including grayscale operation of the image to reduce channel complexity and retain overall brightness information. Subsequently, a two-dimensional discrete cosine transform (2D-DCT) is applied to the entire grayscale image to transform the image from the spatial domain to the frequency domain, thereby obtaining a set of discrete cosine transform coefficients. These coefficients form a coefficient matrix with the same size as the original image, where each coefficient corresponds to the energy distribution of different spatial frequency components.

[0019] After the transformation is completed, the direct current coefficient (DC coefficient) at the upper left corner position is extracted from the obtained DCT coefficient matrix. This coefficient is a concentrated expression of the overall color or brightness level of the image. Finally, this direct current coefficient is used as the color feature value of the trace image.

[0020] Step S200: Match the set of reference color feature values of the trace-bearing object, and compare the color feature value with the set of reference color feature values to obtain a set of color similarity degrees.

[0021] In the embodiments of the present application, during the process of matching the reference color feature value set of the trace-bearing object, the color reference data of the trace-bearing object in multiple states is first constructed. Specifically, it includes obtaining the original image of the trace-bearing object, the corrosion image set, and the high-temperature image set, and respectively extracting the original color feature values, the corrosion color feature value set, and the high-temperature color feature value set from them. The three together constitute the reference color feature value set.

[0022] Subsequently, the color feature values extracted from the collected trace images are sequentially compared with various color feature values in the reference set, and methods such as Euclidean distance are used to calculate the similarity index, so as to obtain a set of results reflecting the degree of color similarity, that is, the color similarity set.

[0023] Furthermore, in the method provided by the embodiments of the application, matching the reference color feature value set of the trace-bearing object further includes: Obtaining the original image of the trace-bearing object and the original color feature values of the original image; constructing the corrosion image set of the trace-bearing object and sequentially obtaining the corrosion color feature value sets of each corrosion image in the corrosion image set; constructing the high-temperature image set of the trace-bearing object and sequentially obtaining the high-temperature color feature value sets of each high-temperature image in the high-temperature image set; based on the original color feature values, the corrosion color feature value sets, and the high-temperature color feature value sets, constructing the reference color feature value set.

[0024] In the embodiments of the present application, to obtain the original image of the trace-bearing object, image acquisition is first performed under standardized shooting conditions. Specifically, an image acquisition device with high-resolution imaging capabilities (such as an industrial camera or a microscopic image system) is used to image the surface of the trace-bearing object that has not been affected by external physical and chemical influences in a controlled environment with constant illumination, no stray reflections, and a uniform background, so as to obtain the original image in its true color state. After the image acquisition is completed, it enters the color feature extraction stage. During the color feature extraction process, the entire image is input as a whole. First, the image is grayscale processed to retain the overall brightness information of the image and reduce the channel complexity. Subsequently, a two-dimensional discrete cosine transform (DCT) is performed on the entire grayscale image to transform the image from the spatial domain to the frequency domain, so as to obtain the energy distribution of the image at different frequency components. In the DCT coefficient matrix generated after the transformation, the direct current coefficient in the upper left corner reflects the average brightness level or color trend of the entire image. This direct current coefficient is directly extracted and used as the original color feature value of the original image.

[0025] Next, a controlled corrosion operation is applied to the surface of the trace-bearing object to simulate the change processes such as oxidation, acid etching, or wear commonly seen in the natural environment. The corrosion methods can include chemical corrosion (such as treatment with diluted acid solution) or physical corrosion (such as surface micro-abrasion). At different corrosion stages or time nodes, image acquisition is performed on the surface of the trace-bearing object to form a corrosion image set reflecting different corrosion states. For each image in the corrosion image set, grayscale conversion and overall DCT processing are performed in the same steps as the original image, and the DC coefficient of the image is extracted as the corrosion color feature value of the image. Finally, the color feature values in multiple corrosion states are combined to form a corrosion color feature value set.

[0026] Subsequently, a controlled high-temperature treatment is performed on the trace-bearing object. By means of an infrared heater, a hot plate, etc., it is heated to multiple set temperature points (such as 100°C, 150°C, 200°C), and surface images are taken under a stable thermal field state to obtain a high-temperature image set. The above-mentioned processing flow is repeated for each high-temperature image, the overall DCT DC coefficient of it is extracted as the high-temperature color feature value, and the color feature results of all high-temperature images are summarized to form a high-temperature color feature value set.

[0027] Finally, the primary color feature values of the original image, the corrosion color feature value set of the corrosion image set, and the high-temperature color feature value set of the high-temperature image set are uniformly sorted out to obtain a reference color feature value set.

[0028] Furthermore, in the method provided by the application embodiment, before determining whether the color coefficient obtained by analyzing the color similarity set meets a predetermined coefficient threshold, it further includes: Extract the first color similarity in the color similarity set, where the first color similarity refers to the similarity between the color feature value and the primary color feature value; extract the second color similarity in the color similarity set, where the second color similarity refers to the color similarity between the trace image and the target corrosion image in the corrosion image set, and the target corrosion image refers to the corrosion image with the highest similarity to the color feature value; extract the third color similarity in the color similarity set, where the third color similarity refers to the color similarity between the trace image and the target high-temperature image in the high-temperature image set, and the target high-temperature image refers to the high-temperature image with the highest similarity to the color feature value; and obtain the color coefficient according to the first color similarity, the second color similarity, and the third color similarity.

[0029] In the embodiment of the present application, the first color similarity is extracted first. The first color similarity refers to the similarity between the color feature value and the primary color feature value. By calculating the Euclidean distance between these two feature values, a numerical value representing their color deviation is obtained as the original evaluation value of the first color similarity.

[0030] Next, extract the second color similarity. The second color similarity refers to the color similarity between the trace image and the target etched image in the etched image set. First, calculate the Euclidean distance between the color feature values of the trace image and the corresponding etched color feature values of each image in the etched image set in sequence, and find the image with the smallest distance from them. This image is the target etched image, and the corresponding minimum distance value reflects the degree of fit between the color features in the etched state and the trace image, which is the second color similarity.

[0031] Subsequently, extract the third color similarity. The third color similarity refers to the color similarity between the trace image and the target high-temperature image in the high-temperature image set. The operation process is the same as that of the second color similarity. Calculate the Euclidean distance between the color feature values of the trace image and the high-temperature color feature values of each high-temperature image, and find the image corresponding to the minimum distance as the target high-temperature image. Its Euclidean distance is the third color similarity, which reflects whether the trace image presents the color characteristics under high-temperature action.

[0032] Finally, calculate the color coefficient according to the extracted first color similarity, second color similarity, and third color similarity. Specifically, first calculate the mean value of the second color similarity and the third color similarity to obtain the first mean value; then perform an arithmetic average of the first mean value and the first color similarity to form a fusion evaluation value; finally, perform a normalization process on the fusion value to output the final color coefficient.

[0033] Furthermore, in the method provided by the application embodiment, obtaining the color coefficient according to the first color similarity, the second color similarity, and the third color similarity further includes: Obtain the mean value of the second color similarity and the third color similarity, denoted as the first mean value; take the mean value of the first mean value and the first color similarity, and perform a normalization process to obtain the color coefficient.

[0034] In the embodiment of the present application, first perform an arithmetic average operation on the second color similarity and the third color similarity to obtain the first mean value.

[0035] Next, calculate the mean value of the first mean value and the first color similarity to form a fusion evaluation value. Finally, perform a minimum-maximum normalization process on the fusion evaluation value, and map it to a unified numerical range according to the set theoretical maximum and minimum value intervals to generate the final color coefficient.

[0036] Furthermore, in the method provided by the application embodiment, before judging whether the color coefficient obtained by analyzing the color similarity set meets a predetermined coefficient threshold, it further includes: Read the predetermined object features, collect multi-dimensional features of the trace-bearing object based on the predetermined object features to obtain an object feature set; perform normalized weighted analysis on the object feature set to obtain an object performance feedback coefficient; adjust the color coefficient with the object performance feedback coefficient as the weight.

[0037] In the embodiment of the present application, first, read the predetermined object features, which are pre-stored in the device database and mainly include the structures and physical properties that have a significant impact on trace formation, such as material types (such as metals, plastics), hardness grades (such as Brinell HB values), surface states (such as roughness Ra values, whether coated), geometric structures (such as flat surfaces, cylindrical surfaces), and thicknesses, etc. These features are loaded through identification numbers, user inputs, or tag recognition methods to form a predetermined feature template as a reference standard.

[0038] Subsequently, collect multi-dimensional features of the trace-bearing object based on the predetermined object features. This process uses the direct measurement method and uses integrated sensors or detection devices to measure the object on-site. For example, use a surface roughness meter to obtain the Ra value, use an ultrasonic thickness gauge to obtain thickness data, and analyze the geometric structure of the object through image recognition. Fill in various measurement results in sequence according to the predetermined feature items to form an object feature set with the same structure as the template.

[0039] Then, perform normalized weighted analysis on the above object feature set to obtain an object performance feedback coefficient. The normalization uses the min-max normalization method, that is, according to the physical feasible range where each feature is located (for example, the metal hardness is usually between 50 and 300 HB), linearly scale the measured value to the interval [0, 1]. Subsequently, use the fixed weight weighting method to assign preset weights to each normalized feature. For example, the hardness weight is 0.35, the surface state is 0.25, the shape structure is 0.2, and the thickness is 0.2. By multiplying the normalized feature values by the corresponding weights and summing, a value between 0 and 1 is obtained, which is the object performance feedback coefficient.

[0040] Finally, adjust the color coefficient with the object performance feedback coefficient as the weight. This step uses the linear multiplication method, that is, multiply the color coefficient obtained in the previous step by the object performance feedback coefficient to generate the adjusted color coefficient.

[0041] Step S300: Determine whether the color coefficient obtained by analyzing the color similarity set meets a predetermined coefficient threshold.

[0042] In the embodiment of the present application, after adjusting the color coefficient obtained from the color similarity set, compare the adjusted color coefficient with the predetermined coefficient threshold preset by technical experts to determine whether it meets the predetermined coefficient threshold. If the color coefficient is greater than or equal to the predetermined coefficient threshold, it is considered to meet the requirement. If it is less than the threshold, it is considered not to meet the requirement.

[0043] Further, in the method provided by the application embodiment, after determining whether the color coefficient obtained by analyzing the color similarity set meets the predetermined coefficient threshold, the following steps are further included: If it does not meet the requirement, a repair instruction is issued; based on the repair instruction, an image repair plan is retrieved to perform repair processing on the trace image; wherein, the image repair plan includes a physical repair plan and / or a chemical repair plan, and the physical repair plan includes plasma cleaning and laser cleaning, and the chemical repair plan includes composite pickling and redox.

[0044] In the embodiment of the present application, when it is determined that the color coefficient does not reach the predetermined coefficient threshold, a repair instruction is issued at this time. This repair instruction serves as a process control signal to retrieve and execute a preset image repair plan to improve the clarity and color accuracy of the trace area in the image, so as to meet the subsequent recognition requirements.

[0045] According to the type of image pollution or interference, the image repair plan is divided into two categories: a physical repair plan and / or a chemical repair plan, and the execution path is automatically matched according to the trace formation material, pollution form and cause. In the physical repair plan, if the image quality degradation is due to physical interferences such as surface dust, particle attachment or weakly adherent coatings, the plasma cleaning plan is preferentially invoked. This method uses high-energy active particles (such as oxygen ions or argon ions) in the plasma to perform low-temperature and non-destructive surface treatment on the trace area, removing micron-scale pollutants while retaining the trace texture features.

[0046] If the pollution is more stable coating residues, oxidation patches or aged film layers, laser cleaning is used for treatment. This method adjusts the power, wavelength and focusing area of the pulsed laser beam, and uses photothermal action or photolysis effect to achieve precise peeling of surface pollutants.

[0047] In the chemical repair plan, if there are oxide films, chemical reaction deposits or color migration interferences on the surface of the trace image, the composite pickling plan is invoked for treatment. Composite pickling uses a variety of acid solutions (such as a nitric acid-phosphoric acid mixture) to react with pollutants to achieve the removal of metal oxide layers and deposits.

[0048] If the image is affected by heat or contaminated by chemical reducing substances, redox treatment is adopted. This method applies an oxidant (such as hydrogen peroxide) or a reducing agent (such as sulfite) to react with the pollution layer through redox reaction, causing its structural change and then being removed, which is commonly used in scenarios such as stain removal or color layer reduction.

[0049] After completing the physical or chemical repair, the image is re-acquired and the color feature extraction and color coefficient calculation processes are re-executed to ensure that the image quality has reached the recognition requirements.

[0050] Step S400: If it meets the requirements, issue a texture feature collection instruction, and obtain the texture feature value set of the trace image according to the texture feature collection instruction.

[0051] In the embodiment of the present application, when it is determined that the color coefficient meets the predetermined coefficient threshold, a texture feature collection instruction is issued, and the texture information extraction process of the trace image is entered. Specifically, first, rasterize the trace image, dividing the image into multiple raster blocks; then, based on a predetermined sampling strategy, uniformly sample these raster blocks to obtain a set of representative target sample blocks. During the extraction process, select the first sample block among them and perform a discrete cosine transform (DCT) on it, and extract the first alternating current coefficient (AC coefficient) as the first texture feature value of the sample block. Further, construct a complete texture feature value set according to this feature value.

[0052] Furthermore, in the method provided by the embodiment of the application, obtaining the texture feature value set of the trace image according to the texture feature collection instruction further includes: Rasterize the trace image to obtain a set of raster blocks; uniformly sample the set of raster blocks based on a predetermined sampling strategy to obtain a set of target sample blocks; extract the first sample block from the set of target sample blocks, and use the first alternating current coefficient in the first discrete cosine transform coefficient of the first sample block as the first texture feature value; based on the first texture feature value according to the texture feature collection instruction, form the texture feature value set.

[0053] In the embodiment of the present application, first rasterize the trace image, that is, divide the entire image into multiple adjacent non-overlapping image blocks according to a fixed size (such as 8×8 pixels), which is called a set of raster blocks.

[0054] Next, according to the built-in uniform sampling strategy, subset selection is performed on the set of raster blocks. This strategy uses a row-column interval equal-distance sampling method, that is, extract spatially representative image blocks from the entire set of raster blocks according to a set step size (such as selecting one every other block) to form a set of target sample blocks.

[0055] Then select the first image block from the set of target sample blocks as the first sample block, and perform a two-dimensional discrete cosine transform (2D-DCT) on this block. Perform DCT calculation in this image block to generate a corresponding frequency coefficient matrix. The upper left corner coefficient of this matrix is the direct current coefficient (DC), representing the average brightness of the image block; and the first alternating current coefficient (AC) adjacent to the DC coefficient, that is, the first frequency component in the lowest non-zero frequency direction in the matrix, reflects the basic texture change of the image block in the horizontal or vertical direction.

[0056] Extract the first AC coefficient as the first texture feature value to characterize the basic texture structure of the image patch. Repeat this operation in the entire set of target sample patches. Extract the first AC coefficient for each sample patch and aggregate the texture feature values of all sample patches to finally form a complete set of texture feature values.

[0057] Step S500: Extract the set of first texture feature values corresponding to the first tool in the tool feature database.

[0058] In the embodiment of the present application, extract the set of first texture feature values corresponding to the first tool from the tool feature database. Herein, the first tool refers to the tool object selected in the preset order in the current recognition process and is one of the matching candidate tools; the set of first texture feature values refers to the set of texture feature data obtained by a specific processing process from the trace image formed after the tool acts on the trace-bearing object under standard conditions.

[0059] In the tool feature database, typical trace image samples generated by multiple types of tools on the standard contact surface are pre-stored. These traces include sole pattern imprints, strip traces left by the sliding of the tool edge, or local image structures formed by specific pattern contours. In the database construction stage, these images undergo a unified processing process, including image rasterization, extraction of sample image patches based on a uniform sampling strategy, performing two-dimensional discrete cosine transform (2D-DCT) on each image patch, and extracting the representative first AC coefficient (AC coefficient) therefrom to capture the local frequency characteristics. When performing this step, retrieve the standard trace image entry corresponding to the first tool in the database and directly read the stored set of AC coefficients. These reflect the spatial frequency changes of the trace image in multiple regions and form the set of first texture feature values.

[0060] Step S600: Compare the set of texture feature values with the set of first texture feature values to obtain the first texture similarity.

[0061] In the embodiment of the present application, perform curve fitting on the set of texture feature values of the trace image and the set of first texture feature values of the first tool in sequence to respectively construct the corresponding texture curve and the first texture curve; subsequently, use the first Fréchet distance between the two texture curves as a measurement index to calculate the similarity degree of their morphological trends, thereby obtaining the final first texture similarity.

[0062] Furthermore, in the method provided by the embodiment of the application, comparing the set of texture feature values with the set of first texture feature values to obtain the first texture similarity further includes: The texture feature value set and the first texture feature value set are successively subjected to curve fitting to obtain a texture curve and a first texture curve respectively. The first Fréchet distance between the texture curve and the first texture curve is used to characterize the first texture similarity.

[0063] In an embodiment of the present application, first, the texture feature value set extracted from the trace image is processed. The feature values in the set are arranged in sequence according to the extraction order of the image blocks to form a one-dimensional sequence, and a linear interpolation method is used to perform continuous processing on the sequence, thereby generating a texture curve.

[0064] The same linear interpolation operation is also performed on the first texture feature value set corresponding to the first tool. The feature values in the set are arranged in a fixed order, and a continuous curve is constructed by the same interpolation method to form a standard comparison curve, that is, the first texture curve.

[0065] After the construction of the two curves is completed, the Fréchet distance algorithm is used to perform structure matching calculation on these two curves. Taking the trace image texture curve and the first tool texture curve as inputs, calculate the maximum offset distance during their traversal. This distance value is used as a quantitative expression of the structural difference and can accurately reflect the overall matching degree of the texture trends of the two. Subsequently, normalization processing is performed on this distance value, and it is converted to the [0, 1] standard range by the maximum distance normalization method to obtain the final first texture similarity, which is used to represent the structural matching degree of the texture feature value set and the first texture feature value set in the texture dimension.

[0066] Step S700: When the first texture similarity reaches a predetermined similarity limit value, the first type corresponding to the first tool is used as the tool type of the trace image.

[0067] In an embodiment of the present application, according to the first texture similarity calculated in the previous steps, it is judged whether it reaches the set predetermined similarity limit value to determine whether the current trace image can be determined to be formed by a specific tool. Among them, the first tool is a candidate tool sample randomly selected from the tool feature database or automatically selected in the order of the recognition process, and does not represent the only tool; the first type is the tool classification label corresponding to the tool in the database, which is used to describe its structural category or usage attribute.

[0068] In the recognition process, first, the texture similarity between the current trace image and the first tool is calculated, and the result is compared with a pre-set similarity threshold. The predetermined similarity limit value is pre-set by technical experts and is used as a judgment boundary for distinguishing whether the texture structures match or not. For example, it is set at 0.8 to ensure that the recognition has sufficient discriminant credibility.

[0069] When the similarity value is greater than or equal to this threshold, it is determined that the texture features of the current trace image have a high structural consistency with the standard template of the first tool, that is, it is considered that this tool is very likely to be the tool source that formed this trace. At this time, the tool type corresponding to the first tool, that is, the classification label in its database, is used as the recognition result of this trace image, that is, the tool type corresponding to this image is determined. This process realizes the automatic matching of the trace image and the tool type through the threshold control logic, and automatically completes the intelligent classification and attribution judgment of the tool trace.

[0070] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects: The present application extracts color features from the collected trace images to obtain color feature values, where the trace images refer to images of the use traces of tools on the trace-bearing objects; matches the set of reference color feature values of the trace-bearing objects, and compares the color feature values with the set of reference color feature values to obtain a set of color similarities; determines whether the color coefficients obtained by analyzing the set of color similarities meet a predetermined coefficient threshold; if they meet, issues a texture feature collection instruction, and obtains a set of texture feature values of the trace image according to the texture feature collection instruction; extracts the set of first texture feature values corresponding to the first tool in the tool feature database; compares the set of texture feature values with the set of first texture feature values to obtain a first texture similarity; when the first texture similarity reaches a predetermined similarity limit value, the first type corresponding to the first tool is used as the tool type of the trace image. The present invention solves the technical problems in the prior art that the recognition accuracy is poor and the tool type cannot be stably judged due to the interference of factors such as material, light, corrosion, or thermal discoloration of the trace image. By introducing a dual feature fusion discrimination mechanism of color and texture, the trace image and the tool template features are dynamically matched, achieving the technical effect of improving the recognition accuracy and stability of tool traces under complex conditions.

[0071] Embodiment 2, based on the same inventive concept as the dynamic matching and recognition method for multi-modal trace morphological features in the foregoing embodiment, as Figure 2 shown, the present application provides a dynamic matching and recognition system for multi-modal trace morphological features. The system in the embodiments of the present application and the method embodiments are based on the same inventive concept. Among them, the system includes: A feature extraction module 11 is used to extract color features from the collected trace images to obtain color feature values. Herein, the trace images refer to the images of the usage traces of tools on the trace-bearing objects. A matching and comparison module 12 is used to match the set of reference color feature values of the trace-bearing objects and compare the color feature values with the set of reference color feature values to obtain a set of color similarity values. A judgment module 13 is used to judge and analyze whether the color coefficient obtained from the set of color similarity values meets a predetermined coefficient threshold. A texture feature collection module 14 is used to, if it meets the requirement, issue a texture feature collection instruction and obtain a set of texture feature values of the trace images according to the texture feature collection instruction. An extraction module 15 is used to extract a set of first texture feature values corresponding to the first tool in the tool feature database. A similarity determination module 16 is used to compare the set of texture feature values with the set of first texture feature values to obtain a first texture similarity. A tool type confirmation module 17 is used to, when the first texture similarity reaches a predetermined similarity limit value, use the first type corresponding to the first tool as the tool type of the trace images.

[0072] Furthermore, the system is also used to implement the following functions: Obtain the discrete cosine transform coefficients of the trace images; use the direct current coefficient in the discrete cosine transform coefficients to represent the color feature values.

[0073] Furthermore, the system is also used to implement the following functions: Obtain the original images of the trace-bearing objects and obtain the original color feature values of the original images; form a set of corrosion images of the trace-bearing objects and sequentially obtain the sets of corrosion color feature values of the corrosion images in the set of corrosion images; form a set of high-temperature images of the trace-bearing objects and sequentially obtain the sets of high-temperature color feature values of the high-temperature images in the set of high-temperature images; based on the original color feature values, the sets of corrosion color feature values and the sets of high-temperature color feature values, form the set of reference color feature values.

[0074] Furthermore, the system is also used to implement the following functions: Extract the first color similarity in the color similarity set, where the first color similarity refers to the similarity between the color feature value and the primary color feature value; extract the second color similarity in the color similarity set, where the second color similarity refers to the color similarity between the trace image and the target corrosion image in the corrosion image set, and the target corrosion image refers to the corrosion image with the highest similarity to the color feature value; extract the third color similarity in the color similarity set, where the third color similarity refers to the color similarity between the trace image and the target high-temperature image in the high-temperature image set, and the target high-temperature image refers to the high-temperature image with the highest similarity to the color feature value; obtain the color coefficient according to the first color similarity, the second color similarity, and the third color similarity.

[0075] Further, the system is also used to implement the following functions: Obtain the mean value of the second color similarity and the third color similarity, denoted as the first mean value; take the mean value of the first mean value and the first color similarity, and perform normalization processing to obtain the color coefficient.

[0076] Further, the system is also used to implement the following functions: Perform rasterization processing on the trace image to obtain a raster block set; perform uniform sampling on the raster block set based on a predetermined sampling strategy to obtain a target sample block set; extract the first sample block in the target sample block set, and use the first alternating current coefficient in the first discrete cosine transform coefficient of the first sample block as the first texture feature value; form the texture feature value set based on the first texture feature value according to the texture feature collection instruction.

[0077] Further, the system is also used to implement the following functions: Perform curve fitting processing on the texture feature value set and the first texture feature value set in sequence to obtain a texture curve and a first texture curve respectively; use the first Fréchet distance between the texture curve and the first texture curve to represent the first texture similarity.

[0078] Further, the system is also used to implement the following functions: If it does not meet the requirements, issue a repair instruction; based on the repair instruction, retrieve an image repair plan to perform repair processing on the trace image; where the image repair plan includes a physical repair plan and / or a chemical repair plan, and the physical repair plan includes plasma cleaning and laser cleaning, and the chemical repair plan includes composite pickling and redox.

[0079] Further, the system is also used to implement the following functions: Read the predetermined object features, collect multi-dimensional features of the trace-bearing object based on the predetermined object features to obtain an object feature set; perform normalized weighted analysis on the object feature set to obtain an object performance feedback coefficient; adjust the color coefficient with the object performance feedback coefficient as the weight.

[0080] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification has been made. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0081] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included within the protection scope of the present application.

[0082] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications therein.

Claims

1. A dynamic matching and recognition method for multi-modal trace morphological features, characterized in that Including: Performing color feature extraction on the collected trace image to obtain color feature values, where the trace image refers to an image of the usage trace of a tool on a trace-bearing object; Matching the set of reference color feature values of the trace-bearing object, and comparing the color feature values with the set of reference color feature values to obtain a set of color similarity degrees; Judging and analyzing whether the color coefficient obtained from the set of color similarity degrees meets a predetermined coefficient threshold; If it meets the requirement, issuing a texture feature collection instruction, and obtaining a set of texture feature values of the trace image according to the texture feature collection instruction; Extracting a set of first texture feature values corresponding to the first tool in the tool feature database; Comparing the set of texture feature values with the set of first texture feature values to obtain a first texture similarity degree; When the first texture similarity degree reaches a predetermined similarity limit value, taking the first type corresponding to the first tool as the tool type of the trace image.

2. The method according to claim 1, wherein Performing color feature extraction on the collected trace image to obtain color feature values, including: Obtaining the discrete cosine transform coefficients of the trace image; Characterizing the color feature values with the direct current coefficients in the discrete cosine transform coefficients.

3. The method according to claim 1, wherein Matching the set of reference color feature values of the trace-bearing object, including: Obtaining the original image of the trace-bearing object, and obtaining the original color feature values of the original image; Constructing a set of corrosion images of the trace-bearing object, and sequentially obtaining sets of corrosion color feature values of each corrosion image in the set of corrosion images; Constructing a set of high-temperature images of the trace-bearing object, and sequentially obtaining sets of high-temperature color feature values of each high-temperature image in the set of high-temperature images; Based on the original color feature values, the sets of corrosion color feature values, and the sets of high-temperature color feature values, constructing the set of reference color feature values.

4. The method according to claim 3, wherein Before judging and analyzing whether the color coefficient obtained from the set of color similarity degrees meets a predetermined coefficient threshold, including: Extracting a first color similarity degree from the set of color similarity degrees, where the first color similarity degree refers to the similarity degree between the color feature values and the original color feature values; Extracting a second color similarity degree from the set of color similarity degrees, where the second color similarity degree refers to the color similarity degree between the trace image and the target corrosion image in the set of corrosion images, and the target corrosion image refers to the corrosion image with the highest similarity degree to the color feature values; Extracting a third color similarity degree from the set of color similarity degrees, where the third color similarity degree refers to the color similarity degree between the trace image and the target high-temperature image in the set of high-temperature images, and the target high-temperature image refers to the high-temperature image with the highest similarity degree to the color feature values; Obtaining the color coefficient according to the first color similarity degree, the second color similarity degree, and the third color similarity degree.

5. The method according to claim 4, wherein Obtaining the color coefficient according to the first color similarity degree, the second color similarity degree, and the third color similarity degree, including: Obtaining the mean value of the second color similarity degree and the third color similarity degree, denoted as the first mean value; Obtaining the mean value of the first mean value and the first color similarity degree, and performing normalization processing to obtain the color coefficient.

6. The method according to claim 1, wherein Obtain the texture feature value set of the trace image according to the texture feature collection instruction, including: Perform rasterization processing on the trace image to obtain a raster block set; Perform uniform sampling on the raster block set based on a predetermined sampling strategy to obtain a target sample block set; Extract the first sample block from the target sample block set, and use the first alternating current coefficient in the first discrete cosine transform coefficient of the first sample block as the first texture feature value; Construct the texture feature value set based on the first texture feature value according to the texture feature collection instruction.

7. The method according to claim 1, characterized in that, Compare the texture feature value set with the first texture feature value set to obtain the first texture similarity, including: Perform curve processing on the texture feature value set and the first texture feature value set in sequence to obtain a texture curve and a first texture curve respectively; Characterize the first texture similarity with the first Fréchet distance between the texture curve and the first texture curve.

8. The method according to claim 1, wherein After judging and analyzing whether the color coefficient obtained from the color similarity set meets a predetermined coefficient threshold, it further includes: If it does not meet the requirement, issue a repair instruction; Based on the repair instruction, retrieve an image repair plan to perform repair processing on the trace image; Wherein, the image repair plan includes a physical repair plan and / or a chemical repair plan, and the physical repair plan includes plasma cleaning and laser cleaning, and the chemical repair plan includes composite pickling and redox.

9. The method according to claim 1, characterized in that Before judging and analyzing whether the color coefficient obtained from the color similarity set meets a predetermined coefficient threshold, it further includes: Read the predetermined object characteristics, and perform multi-dimensional feature collection on the trace-bearing object based on the predetermined object characteristics to obtain an object feature set; Perform normalized weighted analysis on the object feature set to obtain an object performance feedback coefficient; Adjust the color coefficient with the object performance feedback coefficient as the weight.

10. A dynamic matching and recognition system for multi-modal trace morphological features, characterized in that, The system includes: A feature extraction module, configured to extract color features from the collected trace image to obtain color feature values, where the trace image refers to an image of the use trace of a tool on a trace-bearing object; A matching and comparison module, configured to match the reference color feature value set of the trace-bearing object, and compare the color feature values with the reference color feature value set to obtain a color similarity set; A judgment module, configured to judge and analyze whether the color coefficient obtained from the color similarity set meets a predetermined coefficient threshold; A texture feature collection module, configured to, if it meets the requirement, issue a texture feature collection instruction, and obtain the texture feature value set of the trace image according to the texture feature collection instruction; An extraction module, configured to extract the first texture feature value set corresponding to the first tool in the tool feature database; A similarity determination module, configured to compare the texture feature value set with the first texture feature value set to obtain the first texture similarity; A tool type confirmation module, configured to, when the first texture similarity reaches a predetermined similarity limit value, use the first type corresponding to the first tool as the tool type of the trace image.

Images