Method, system and equipment for evaluating effective information amount of industrial image

Through edge feature extraction and gradient amplitude calculation, the information loss of industrial images is evaluated, which solves the problem of information loss during image conversion in industrial non-destructive testing, and improves detection accuracy and technological development.

CN118941492APending Publication Date: 2024-11-12BEIJING INST OF TECH
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410603126.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-05-15
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In the field of industrial non-destructive testing, serious information loss during high-precision image conversion process, resulting in the impact of the accuracy and reliability of weld detection. The existing evaluation methods are unstable and have poor correlation with human perception.

Method used

By acquiring the original image and processing images, extracting edge features, calculating gradient amplitude map and information loss values, and evaluating the effective information amount of industrial images, it provides an evaluation method and system for the effective information amount of industrial images, which is used to optimize the pre-processing process , improve detection accuracy.

Benefits of technology

This method has a strong correlation with human perception, and can focus on evaluating information losses in specific areas, apply it to industrial non-destructive testing, improve detection accuracy, promote the development of non-destructive testing technology, and provide more accurate benchmarks and guidance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941492A_ABST
    Figure CN118941492A_ABST
Patent Text Reader

Abstract

The invention provides an industrial image effective information amount evaluation method, system and equipment, and the method comprises the steps: obtaining an original image and a processing image, and obtaining an original thumbnail and a processing thumbnail according to the original image and the processing image; performing edge feature extraction on the original thumbnail and the processed thumbnail to obtain an original edge feature graph and a processed edge feature graph; according to the original edge feature map and the processed edge feature map, obtaining a gradient magnitude map; obtaining a processing information loss value according to the gradient magnitude image; and evaluating the effective information amount of the industrial image according to the processing information loss value. The method can be applied to the field of industrial nondestructive testing, and is used for evaluating whether enough information and edge features of defects, namely effective information, are reserved in an industrial image processing process, so that subsequent application such as image recognition and target detection is facilitated.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of industrial nondestructive testing, and in particular to a method, system and device for evaluating the effective information volume of industrial images. Background Art

[0002] Industrial manufacturing is the foundation of economic development, and product quality is crucial to the competitiveness of enterprises. Nondestructive testing, as a product quality testing method, is widely used in the field of industrial manufacturing. Nondestructive testing technology is a non-destructive testing method that can evaluate and test various performance indicators without affecting the integrity of the object being tested.

[0003] In the current field of industrial nondestructive testing, on the one hand, most of the current nondestructive testing equipment (such as X-ray) outputs high-precision (16-bit single-channel images) original images, while the current mainstream detection models are more suitable for low-precision images (8-bit 3-channel images); on the other hand, usually, images need to be enhanced by multiple data methods before inputting the model to highlight the details of the image. However, the above two differences may lead to information loss in the process of image conversion, which in turn affects the accuracy and reliability of weld detection.

[0004] At present, there are several common methods for evaluating the information loss of public image datasets: Structural Similarity Index (SSIM): SSIM is a widely used indicator for image quality evaluation, which takes into account information such as brightness, contrast and structure of the entire image. By comparing the SSIM values ​​of the original image and the transformed image, the degree of information loss of the entire image can be quantified. Peak Signal-to-Noise Ratio (PSNR): PSNR is another common image quality evaluation indicator, which mainly focuses on the mean square error between the original image and the reconstructed image. However, the results of this method are unstable and have poor correlation with human perception. Histogram analysis: By comparing the histograms of the original and transformed images, the changes in the grayscale distribution of the image can be observed. Information loss will cause changes in the histogram, which can be quantified by histogram comparison or related indicators. However, in the evaluation of information loss of weld images, more attention is paid to the information loss area in the weld area, and the information loss of the background and other non-critical areas has little effect on the accuracy of detection. Summary of the invention

[0005] Based on this, it is necessary to provide a method, system and device for evaluating the effective information content of industrial images in response to the above technical problems.

[0006] A method for evaluating the effective information content of an industrial image comprises the following steps:

[0007] Acquire an original image and a processed image, and obtain an original thumbnail and a processed thumbnail according to the original image and the processed image;

[0008] Performing edge feature extraction on the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map;

[0009] Obtaining a gradient amplitude map according to the original edge feature map and the processed edge feature map;

[0010] According to the gradient amplitude map, a processing information loss value is obtained;

[0011] The effective information amount of the industrial image is evaluated according to the processing information loss value.

[0012] In one embodiment, acquiring the original image and processing the image comprises:

[0013] An original image data set is acquired, and data enhancement and digital conversion are performed on the original image data set to obtain a processed image data set; wherein the original image data set includes a plurality of original images, and the processed image data set includes a plurality of processed images.

[0014] In one embodiment, obtaining an original thumbnail and a processed thumbnail according to the original image and the processed image includes:

[0015] Acquire a Gaussian smoothing kernel according to the original image and the processed image, and perform a convolution operation on the original image and the processed image according to the Gaussian smoothing kernel to obtain an original denoised image and a processed denoised image;

[0016] Downsampling operations are performed on the original denoised image and the processed denoised image to obtain the original thumbnail image and the processed thumbnail image.

[0017] In one embodiment, edge feature extraction is performed on the original thumbnail and the processed thumbnail to obtain the original edge feature map and the processed edge feature map, including:

[0018] Obtain a convolution kernel containing a Sobel operator, and perform convolution operations in the horizontal direction and the vertical direction according to the convolution kernel to obtain a horizontal edge feature information map and a vertical edge feature information map of the original thumbnail and the processed thumbnail in the horizontal direction and the vertical direction;

[0019] The horizontal edge feature information map and the vertical edge feature information map are superimposed to obtain an original edge feature map and a processed edge feature map.

[0020] In one embodiment, obtaining a gradient magnitude map according to the original edge feature map and the processed edge feature map includes:

[0021] The gradient amplitude map is calculated by the following formula:

[0022]

[0023] Among them, R represents the gradient amplitude map, m r Indicates the processing edge feature map, m o represents the original edge feature map, i represents the i-th pixel point of the processed edge feature map, j represents the j-th pixel point of the processed edge feature map, and c represents an arbitrary constant.

[0024] In one embodiment, obtaining the processing information loss value according to the gradient magnitude map includes:

[0025] The processing information loss value is calculated by the following formula:

[0026]

[0027] Among them, GRS R represents the processing information loss value, N represents the total pixel points of the gradient amplitude map, R represents the gradient amplitude map, k represents the kth pixel point of the gradient amplitude map, and R mean Represents the average value of the gradient magnitude map.

[0028] An evaluation system for effective information volume of industrial images, used to implement the evaluation method for effective information volume of industrial images as described above, comprising:

[0029] A thumbnail processing module, used to obtain an original image and a processed image, and obtain an original thumbnail and a processed thumbnail according to the original image and the processed image;

[0030] An edge processing module, used for extracting edge features from the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map;

[0031] A gradient calculation module, used for obtaining a gradient amplitude map according to the original edge feature map and the processed edge feature map;

[0032] A loss calculation module, used for obtaining a processing information loss value according to the gradient amplitude map;

[0033] A comprehensive evaluation module is used to evaluate the effective information volume of the industrial image according to the processing information loss value.

[0034] A device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of a method for evaluating the effective information volume of an industrial image described in each of the above embodiments are implemented.

[0035] Compared with the prior art, the advantages and beneficial effects of the present invention are: the evaluation method of the effective information volume of industrial images provided by the present invention has a strong correlation with human perception, can assist people to focus more on evaluating the image information loss in a specific area, can be applied to the field of industrial non-destructive testing, and can be used to evaluate whether sufficient information and edge features of defects are retained in the process of industrial image processing, that is, the effective information volume, so as to facilitate subsequent image recognition, target detection and other applications. It has positive significance for optimizing the pre-processing process of workpiece images, improving detection accuracy, and promoting the development of non-destructive testing technology. It not only helps to improve the existing low-precision model and eliminate variables that are not related to test accuracy, but also provides a more accurate benchmark and guidance for future research on high-precision models. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of a method for evaluating the effective information content of an industrial image in one embodiment;

[0037] Figure 2 A schematic diagram of the structure of an evaluation system for effective information volume of industrial images in one embodiment;

[0038] Figure 3 A schematic diagram of the internal structure of a device in one embodiment;

[0039] Figure 4 1 is a flow chart of a method for evaluating the effective information content of an industrial image using a weld image as an example in an embodiment;

[0040] Figure 5 A schematic diagram of a YOLO-v8 model training result obtained after processing an image of a small-diameter tube DR ray dataset using the M1 method in an embodiment;

[0041] Figure 6 A schematic diagram of a YOLO-v8 model training result obtained after processing an image of a small-diameter tube DR ray dataset using the M2 method in an embodiment;

[0042] Figure 7 A schematic diagram of a YOLO-v8 model training result obtained after processing images of a DR ray dataset of automotive parts casting using the M1 method in an embodiment;

[0043] Figure 8 This is a schematic diagram of the YOLO-v8 model training results obtained after processing images of an automobile parts casting DR ray dataset using the M2 method in an embodiment. DETAILED DESCRIPTION

[0044] Before describing the specific embodiments of the present invention, the overall concept of the present invention is described as follows:

[0045] The present invention is mainly developed for industrial nondestructive testing process. In the current field of industrial nondestructive testing, on the one hand, the digital output of nondestructive testing equipment is a 16-bit single-channel image, while the image format acceptable to the current mainstream detection model is an 8-bit 3-channel image. 16-bit single-channel images have a higher grayscale level than 8-bit 3-channel images, and can provide more detailed image information. However, in the process of information conversion, if we simply convert 16-bit single-channel images into 8-bit 3-channel images, a large amount of information will be truncated and distorted, which is crucial for the microstructure and details of the workpiece image; on the other hand, in order to improve the prediction performance of the model for tiny defects on the workpiece, the current common practice is to perform local contrast enhancement, histogram homogenization, sharpening and other data enhancement methods before the image is input into the model, highlight the features on the picture, and use the data set after data enhancement to train the model. However, when the data enhancement means are inappropriate or the data enhancement is excessive, the spatial information, color information, semantic information, and especially the significant loss of texture detail information in the picture will be caused.

[0046] Therefore, the present invention proposes a method for evaluating the effective information volume of industrial images. Edge feature images are extracted from the original image and the processed image after the original image is processed. The processing information loss value is calculated through the two edge feature images. The effective information volume of the industrial image is evaluated according to the processing information loss value. This can improve the detection accuracy while optimizing the pre-processing process of the workpiece image.

[0047] After introducing the overall concept of the present invention, in order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0048] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of this specification should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of this specification do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0049] In one embodiment, Figure 1 As shown, a method for evaluating the effective information content of an industrial image is provided, comprising the following steps:

[0050] Step S101, acquiring an original image and a processed image, and obtaining an original thumbnail and a processed thumbnail according to the original image and the processed image.

[0051] Specifically, we need to obtain two images, the original image (I o ) and the processed image after digital conversion and data enhancement (I r ), through the original image (I o ) and processed images (I r ) to obtain the original thumbnail and the processed thumbnail corresponding to the two.

[0052] On this basis, obtaining the original image and processing the image include:

[0053] An original image data set is acquired, and data enhancement and digital conversion are performed on the original image data set to obtain a processed image data set; wherein the original image data set includes a plurality of original images, and the processed image data set includes a plurality of processed images.

[0054] Specifically, obtain the original image dataset (S o ), for the original image dataset (S o ) to perform data enhancement and digital conversion to obtain the processed image dataset (S r ), original image dataset (S o ) includes several original images, processed image data set (S r ) includes several processed images, and the original images and processed images correspond to each other. From the two data sets, the corresponding images I are taken out respectively. o and I r .

[0055] On this basis, obtaining an original thumbnail and a processed thumbnail according to the original image and the processed image includes:

[0056] Acquire a Gaussian smoothing kernel according to the original image and the processed image, and perform a convolution operation on the original image and the processed image according to the Gaussian smoothing kernel to obtain an original denoised image and a processed denoised image;

[0057] Downsampling operations are performed on the original denoised image and the processed denoised image to obtain the original thumbnail image and the processed thumbnail image.

[0058] Specifically, according to the image features of the original image and the processed image, a suitable Gaussian smoothing kernel is selected, and the two images are convolved to remove background noise irrelevant to the features. The original denoised image and the processed denoised image are downsampled with a step size of 2 to generate two thumbnails (I o ’ , I r ’ ).

[0059] In this embodiment, downsampling is performed on the original denoised image and the processed denoised image after denoising, which can reduce the amount of algorithm calculation and improve the processing speed.

[0060] Step S102, edge feature extraction is performed on the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map.

[0061] On this basis, edge feature extraction is performed on the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map, including:

[0062] Obtain a convolution kernel containing a Sobel operator, and perform convolution operations in the horizontal direction and the vertical direction according to the convolution kernel to obtain a horizontal edge feature information map and a vertical edge feature information map of the original thumbnail and the processed thumbnail in the horizontal direction and the vertical direction;

[0063] The horizontal edge feature information map and the vertical edge feature information map are superimposed to obtain an original edge feature map and a processed edge feature map.

[0064] Specifically, a convolution kernel containing a Sobel operator is used to perform convolution operations in the x-axis and y-axis directions to obtain horizontal edge feature information maps and vertical edge feature information maps (M ox ,M oy ,M rx ,M rx ,). The horizontal edge feature information map and the vertical edge feature information map extracted from the original thumbnail and the processed thumbnail in the horizontal and vertical directions are superimposed respectively to obtain the edge feature maps corresponding to the original thumbnail and the processed thumbnail (M o ,M r ).

[0065] Step S103, obtaining a gradient magnitude map according to the original edge feature map and the processed edge feature map.

[0066] Specifically, the original edge feature map and the processed edge feature map (M) are calculated pixel by pixel through the formula o ,M r ), and get Mo and M r Gradient magnitude map between (R).

[0067] On this basis, according to the original edge feature map and the processed edge feature map, a gradient amplitude map is obtained, including:

[0068] The gradient amplitude map is calculated by the following formula:

[0069]

[0070] Among them, R represents the gradient amplitude map, m r Indicates the processing edge feature map, m o represents the original edge feature map, i represents the i-th pixel of the processed edge feature map, j represents the j-th pixel of the processed edge feature map, and c represents an arbitrary constant, usually 1, to prevent the result from being infinite when the denominator in the gradient amplitude map calculation formula is 0.

[0071] Step S104: obtaining a processing information loss value according to the gradient magnitude map.

[0072] Specifically, the processing information loss value in the gradient amplitude map is calculated by the formula, that is, the amplitude standard deviation (GRS) in the gradient amplitude map is calculated. R ).

[0073] On this basis, according to the gradient amplitude map, the processing information loss value is obtained including:

[0074] The processing information loss value is calculated by the following formula:

[0075]

[0076] Among them, GRS R represents the processing information loss value, N represents the total pixel points of the gradient amplitude map, R represents the gradient amplitude map, k represents the kth pixel point of the gradient amplitude map, and R mean Represents the average value of the gradient magnitude map.

[0077] Step S105: evaluating the effective information amount of the industrial image according to the processing information loss value.

[0078] Specifically, the GRS values ​​of the corresponding images in the original image data set and the processed image data set are averaged to obtain the GRS average value. The smaller the GRS average value, the more effective information the industrial image has. The effective information content of the industrial image can be evaluated by processing the information loss value.

[0079] The present invention provides an evaluation method for the effective information volume of industrial images, which has a strong correlation with human perception and can help people focus more on evaluating the image information loss in a specific area. It can be applied to the field of industrial non-destructive testing and used to evaluate whether sufficient information and edge features of defects are retained during industrial image processing, that is, the effective information volume, for subsequent image recognition, target detection and other applications. It has positive significance for optimizing the pre-processing process of workpiece images, improving detection accuracy, and promoting the development of non-destructive testing technology. It not only helps to improve the existing low-precision model and eliminate variables that are not related to test accuracy, but also provides a more accurate benchmark and guidance for future research on high-precision models.

[0080] It should be noted that the method of the embodiment of the present invention can be performed by a single device, such as a computer or a server. The method of this embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present invention, and the multiple devices will interact with each other to complete the described method.

[0081] It should be noted that some embodiments of the present invention are described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present invention further provides an evaluation system for the effective information volume of industrial images.

[0083] refer to Figure 2 , the evaluation system of the effective information amount of industrial images comprises:

[0084] The thumbnail processing module 201 is used to obtain an original image and a processed image, and obtain an original thumbnail and a processed thumbnail according to the original image and the processed image;

[0085] The edge processing module 202 is used to extract edge features from the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map;

[0086] A gradient calculation module 203, used to obtain a gradient amplitude map according to the original edge feature map and the processed edge feature map;

[0087] A loss calculation module 204, used to obtain a processing information loss value according to the gradient amplitude map;

[0088] The comprehensive evaluation module 205 is used to evaluate the effective information volume of the industrial image according to the processing information loss value.

[0089] For the convenience of description, the above system is described as being divided into various modules according to their functions. Of course, when implementing the present invention, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0090] The system of the above embodiment is used to implement a corresponding method for evaluating the effective information volume of an industrial image in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0091] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an evaluation method for the effective information amount of an industrial image as described in any of the above embodiments is implemented.

[0092] Figure 3 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.

[0093] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0094] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0095] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0096] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0097] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0098] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0099] The electronic device of the above embodiment is used to implement a corresponding method for evaluating the effective information volume of an industrial image in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be described in detail here.

[0100] The technical solutions in the present invention will be described clearly and completely below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0101] Embodiment 1

[0102] like Figure 4 As shown in the figure, taking the weld image as an example, this method requires two sets of images: the original weld image dataset (S o ) and the weld image dataset after digital conversion and data enhancement (S r ). The implementation steps are as follows:

[0103] The first step is to extract the corresponding images I from the two data sets.o and I r .

[0104] The second step is to analyze the original weld image (I o ) and process weld images (I r ) The two images are convolved using an appropriately selected Gaussian smoothing kernel to remove background noise that is irrelevant to the defect features on the weld. Then, the denoised image is downsampled to generate a thumbnail (I o ',I r ').

[0105] The third step is to generate thumbnails (I o ',I r ') to perform edge extraction operations. Using the convolution kernel containing the Sobel operator to perform convolution operations in the horizontal and vertical directions, the edge feature information maps (M) of the two images in the horizontal and vertical directions are obtained respectively. ox ,M oy ,M rx ,M ry The edge feature information maps extracted in the horizontal and vertical directions are superimposed to obtain the edge feature maps corresponding to the original weld image and the processed weld image (M o ,M r ).

[0106] Step 4: Use the formula Calculate edge feature map pixel by pixel (M o ,M r ), and get M o and M r Gradient magnitude map between (R).

[0107] Step 5: Use the formula The standard deviation (GRS) of the gradient magnitude map was calculated to quantify the information loss caused by the single image conversion method.

[0108] In the sixth step, the GRS values ​​of the corresponding images in the two datasets are averaged to quantify the difference between the original weld image dataset (S o ) and the image dataset after digital conversion and data enhancement (S r ) information loss.

[0109] Through the above steps, the information loss of weld radiographic images for nondestructive testing after preprocessing is evaluated. By comparing multiple preprocessing methods, the most suitable preprocessing method for the group of weld images can be selected, which can effectively improve the accuracy and recall rate of the weld defect detection model.

[0110] Embodiment 2

[0111] Two image preprocessing methods were evaluated using this method, namely: 1. First convert the high-resolution image to a low-resolution image, and then perform adaptive contrast adjustment (M1); 2. First perform contrast adjustment on the high-resolution image, and then convert it to a low-resolution image (M2).

[0112] In the boiler small-diameter tube DR ray data set, Figure 5 As shown in the figure, the training result of the YOLO-v8 model is obtained after processing the image of the small-diameter tube DR ray dataset using the M1 method. The GRS value measured by M1 is 0.028592. Figure 6 The following figure shows the training results of the YOLO-v8 model after processing the image using the M2 method on the small-diameter tube DR ray dataset. The GRS value measured by M2 is 0.026804. In comparison, the information loss of the original image after M1 processing is more, with a loss of 0.001788, about 6.67%. The YOLO-v8 target detection model was trained using the datasets processed by M1 and M2, and the mAPs of the two models were 0.790 and 0.727 respectively. After adjusting the order of image preprocessing, the model performance was improved by 8.6%.

[0113] In the automotive parts casting DR ray dataset, Figure 7 As shown in the figure, the YOLO-v8 model training results are obtained after the automotive parts casting DR ray data set is processed using the M1 method. The GRS value measured by M1 is 0.238710. Figure 8 The figure shows the training results of the YOLO-v8 model after processing the images of the automotive parts casting DR ray dataset using the M2 method. The GRS value measured by M2 is 0.238698. In comparison, the information loss of the original image after M1 processing is more, which is 0.000012, about 0.05%. The YOLO-v8 target detection model is trained using the two processed datasets of M1 and M2. The mAP does not change significantly, but the AP for small target defects such as pores is increased from 0.865 to 0.883. After adjusting the image preprocessing order, the performance of the model for small targets is improved by 2%. The training results of the above two datasets prove the effectiveness of this method in evaluating information loss.

[0114] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention (including the claims) is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0115] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0116] In addition, to simplify the description and discussion, and in order not to obscure the embodiments of the present invention, known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, the devices may be shown in the form of block diagrams to avoid obscuring the embodiments of the present invention, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present invention will be implemented (i.e., these details should be fully within the scope of understanding of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present invention, it will be apparent to those skilled in the art that embodiments of the present invention may be implemented without these specific details or with variations in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0117] Although the invention has been described in conjunction with specific embodiments of the invention, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0118] The embodiments of the present invention are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the effective information content of industrial images, characterized in that: include: Acquire an original image and a processed image, and obtain an original thumbnail and a processed thumbnail according to the original image and the processed image; Performing edge feature extraction on the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map; Obtaining a gradient magnitude map according to the original edge feature map and the processed edge feature map; According to the gradient amplitude map, a processing information loss value is obtained; The effective information amount of the industrial image is evaluated according to the processing information loss value.

2. The method for evaluating the effective information content of industrial images according to claim 1, characterized in that: The obtaining of the original image and the processing of the image include: An original image data set is acquired, and data enhancement and digital conversion are performed on the original image data set to obtain a processed image data set; wherein the original image data set includes a plurality of original images, and the processed image data set includes a plurality of processed images.

3. The method for evaluating the effective information content of industrial images according to claim 1, characterized in that: The obtaining of the original thumbnail and the processed thumbnail according to the original image and the processed image comprises: Acquire a Gaussian smoothing kernel according to the original image and the processed image, and perform a convolution operation on the original image and the processed image according to the Gaussian smoothing kernel to obtain an original denoised image and a processed denoised image; Downsampling operations are performed on the original denoised image and the processed denoised image to obtain the original thumbnail image and the processed thumbnail image.

4. The method for evaluating the effective information content of industrial images according to claim 1, characterized in that: The extracting edge features of the original thumbnail and the processed thumbnail to obtain the original edge feature map and the processed edge feature map comprises: Obtain a convolution kernel containing a Sobel operator, and perform convolution operations in the horizontal direction and the vertical direction according to the convolution kernel to obtain a horizontal edge feature information map and a vertical edge feature information map of the original thumbnail and the processed thumbnail in the horizontal direction and the vertical direction; The horizontal edge feature information map and the vertical edge feature information map are superimposed to obtain an original edge feature map and a processed edge feature map.

5. The method for evaluating the effective information content of industrial images according to claim 1, characterized in that: The step of obtaining a gradient magnitude map according to the original edge feature map and the processed edge feature map comprises: The gradient amplitude map is calculated by the following formula: Among them, R represents the gradient amplitude map, m r Indicates the processing edge feature map, m o represents the original edge feature map, i represents the i-th pixel point of the processed edge feature map, j represents the j-th pixel point of the processed edge feature map, and c represents an arbitrary constant.

6. The method for evaluating the effective information content of industrial images according to claim 1, characterized in that: The obtaining of the processing information loss value according to the gradient magnitude map comprises: The processing information loss value is calculated by the following formula: Among them, GRS R represents the processing information loss value, N represents the total pixel points of the gradient amplitude map, R represents the gradient amplitude map, k represents the kth pixel point of the gradient amplitude map, and R mean Represents the average value of the gradient magnitude map.

7. An evaluation system for the effective information content of industrial images, characterized in that: A method for evaluating the effective information volume of an industrial image according to claims 1 to 6, comprising: A thumbnail processing module, used to obtain an original image and a processed image, and obtain an original thumbnail and a processed thumbnail according to the original image and the processed image; An edge processing module, used for extracting edge features from the original thumbnail and the processed thumbnail to obtain an original edge feature map and a processed edge feature map; A gradient calculation module, used for obtaining a gradient amplitude map according to the original edge feature map and the processed edge feature map; A loss calculation module, used to obtain a processing information loss value according to the gradient amplitude map; A comprehensive evaluation module is used to evaluate the effective information volume of the industrial image according to the processing information loss value.

8. A device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

Citation Information

Cited By

  • Printed matter printing quality detection and analysis system based on machine vision

    CN119904425A

  • Industrial robot efficient communication method and system based on edge computing

    CN121864918A