An Automatic Discrimination Method and Device for Weak Texture Images Based on Spherical Measurement of Image Information

Abstract

The present invention proposes an automatic discrimination method and device for weakly textured images based on spherical measurement of image information. First, the present invention calculates features such as the mean texture gradient, the radius of the texture energy field, and the texture information entropy, and then analyzes the image texture from three perspectives: the degree of average gray change, the fineness of local gray jitter, and the degree of global gray-level chaos, to construct an image texture intensity vector field to describe the richness of the image texture. Then, an information sphere classifier is constructed, using a small amount of weakly textured image data obtained by expert visual interpretation as the prior criterion for weakly textured discrimination. Finally, an information sphere is constructed for the data to be measured in the vector space, and by calculating the occurrence probability of the event of the data to be measured in the prior weakly textured data, the three description components in the TRV of the image to be measured are generated as the information content of the data to be measured, thereby realizing the discrimination of weakly textured images.

Description

Technical Field

[0001] The present invention belongs to the field of remote sensing image processing methods, and particularly relates to a method and device for automatically discriminating weak texture images by extracting image texture richness and constructing an image information sphere. Background Art

[0002] Given the richness and diversity of geographical scenes, the aerial images usually obtained face complex scene information. The collected images often mix scenes such as houses, cities, rural areas, waters, snow fields, etc. These images can be roughly divided into conventional texture images and weak texture images. At present, a large number of studies are dedicated to conducting experiments on conventional images with clear geometric structures, and the technologies for analyzing and applying conventional texture objects such as buildings, roads, and vegetation have been relatively mature. However, when dealing with weak texture scenes in images, such as weak texture situations like lakes, deserts, and snow fields, the existing technologies still face a series of problems. Images in weak texture scenes have some unique characteristics, including low reflectivity, high neighborhood similarity, and insignificant regional differences. In the actual production process, due to the significant texture and geometric differences between the two types of images, it is often necessary to distinguish and process the two types of images. Therefore, in order to serve the full intelligent automation of the production process, it is necessary to classify and preprocess the initially collected images, and how to quantitatively extract the texture information of the images and construct a texture classifier is the key among them. Therefore, it is very necessary to conduct research on this.

[0003] Image texture discrimination and classification is a research hotspot in the field of computer vision and plays an important role in fields such as object detection and image segmentation. Existing methods for image texture classification can mostly be divided into two types: traditional algorithms and deep learning algorithms. Traditional texture discrimination and classification methods are mainly divided into four categories: filter-based, statistical feature-based, structure feature-based, and model feature-based. Filter methods such as Gabor filters can effectively extract image features but are highly dependent on parameters; statistical methods such as gray-level co-occurrence matrices are only suitable for small-scale images; model methods such as fractal geometry perform well in dealing with complex textures but are sensitive to image transformations. In recent years, some improved algorithms such as ILQP and NF-LBP, as well as multi-color texture descriptors, have proposed solutions to the limitations of these traditional methods, but still have difficulty effectively processing weak texture images. Deep learning-based texture classification methods have stronger feature abstraction capabilities than traditional methods and improve the classification accuracy. Autoencoders and convolutional neural networks (CNNs) were the focus of early research, and CapsNets and dual attention mechanism networks further improved the feature extraction and classification performance. In recent years, researchers have proposed a variety of innovative methods, such as dynamic texture analysis combining with the RNN model, fuzzy network models (FuzzyNet) based on gray-level co-occurrence matrices and particle swarm optimization, etc. These methods perform outstandingly in specific applications. However, deep learning methods still face challenges such as data requirements and label accuracy in practical applications.

[0004] In summary, a large number of studies have been conducted on the discrimination and classification of image textures. However, whether traditional methods or deep learning methods are used, they all focus on conventional texture images and aim to discriminate a certain concrete object, and are unable to effectively discriminate the weak textures of images. Based on this, the present invention proposes a method for quantifying the richness of image textures and constructing an image information sphere classifier to achieve automatic discrimination of weak texture images. Summary of the Invention

[0005] The present invention proposes an automatic discrimination method and device for weak texture images based on image information sphere measurement to solve the discrimination problem of weak texture remote sensing images.

[0006] The technical solution adopted by the present invention is: an automatic discrimination method for weak texture images based on image information sphere measurement, including the following steps:

[0007] Step 1, initialize the input image, complete the initialization of parameters and discrimination thresholds respectively, and obtain prior weak texture image data through visual interpretation at the same time;

[0008] Step 2, construct a texture intensity vector TRV by calculating the mean value of the texture gradient, the radius of the texture energy field, and the texture contour information entropy of the image, and extract the quantization result of the strength of the image texture;

[0009] Step 3, input the prior weak texture image data obtained by visual interpretation, construct their respective TRV vectors, and obtain the prior knowledge of weak texture semantics;

[0010] Step 4, generate a vector space based on the TRV vectors in Step 3, construct an information sphere in the vector space for the data to be measured, and the radius of the information sphere is calculated by the minimum circumscribed sphere of the prior weak texture image data;

[0011] Step 5, construct an information sphere classifier, calculate the probability of the data to be measured occurring under the weak texture semantics in the information sphere, and calculate the amount of information of the data to be measured under the weak texture semantics according to Shannon's theorem, and finally compare it with the set threshold to realize the discrimination of the strength of the texture of the data to be measured.

[0012] Furthermore, in Step 2, the mean value of the image texture gradient is calculated based on the Scharr filtering result, which reflects the average change degree of the image gray level. The Scharr filtering formula is as shown in formula (1):

[0013]

[0014] In formula (1), G x and G t represent the gradients of the pixel in the x and y directions, and P1…9 represent the intensity magnitudes of the adjacent pixels in the 3×3 neighborhood of the image;

[0015] Calculate the mean value of the image texture gradient to evaluate the overall gradient change degree of the image. The relevant calculation formula is shown in Formula (2):

[0016]

[0017] In Formula (2), represents the mean value of the image texture gradient, w and h represent the width and length of the image gradient map, and G i represents the gradient intensity value of pixel i, and G x and G y represent the gradients of the pixel in the x and y directions.

[0018] Furthermore, in the second step, first use Laplace filtering to sharpen the edge features of the image and filter out noise and smoothly changing image information; then perform frequency domain conversion under the fast Fourier transform to obtain the frequency spectrum diagram of the image; then, through the analysis of the frequency spectrum diagram, calculate the intensity value range and the upper quartile intensity value; then find the limit distance point farthest from the center point according to the upper quartile intensity value; finally, construct a texture energy field with the limit distance point as the boundary to obtain the energy field radius;

[0019] The fast Fourier transform calculation is shown in Formula (3):

[0020]

[0021] In Formula (3), f is the function representation of the signal in the original spatial domain, F is the function representation of the signal after transformation in the frequency domain, the values of u and x both range from 0 to K - 1, and W 2K is the sine-cosine function, K is a positive integer, and F(u) is the Fourier transform, which is expressed as the sum of the products of the coefficients of a finite number of terms and the sine-cosine function;

[0022] Based on the frequency spectrum diagram, using the upper quartile intensity value of the frequency spectrum diagram as the threshold, calculate the high-frequency radiation range of the image. The upper quartile intensity value and the texture energy field radius are calculated as shown in Formulas (4) and (5):

[0023] Q up =(max(I)-min(I))·75% (4)

[0024] In Formula (4), Q up is the upper quartile intensity value in the frequency spectrum diagram, and max(I) and min(I) are operations to find the maximum and minimum values of the intensity values in the image I;

[0025]

[0026] In Formula (5), S pnt is all the intensity values in the frequency spectrum diagram that are Qup Set of point coordinates, Int i Is the spectrogram intensity value of pixel i, S dis Is S pnt Set of distances from the midpoint to the center point of the spectrum. x0 and y0 are the coordinates of the center point of the spectrogram, x i and y i Are the coordinates of pixel point i, R text Is the radius of the texture energy field of the image.

[0027] Furthermore, in step two, the image contour information entropy is calculated based on the Canny-filtered image contour diagram, which reflects the global gray-level disorder. The calculation of the image information entropy is as shown in formula (6):

[0028]

[0029] In formula (6), H is the image information entropy, P i Is the occurrence probability that the gray value in the image is i, CNT i Is the number of pixels with gray level i. h and w are the height and width of the image.

[0030] Furthermore, in step four, the vector space uses the linear normalized Euclidean distance for position measurement, and the calculation is as shown in formula (7):

[0031]

[0032] In formula (7), S dis Is the set of similarity distance of the image, where N is the number of prior weak texture images, D j Represents the normalized Euclidean similarity distance between the data to be measured and the prior weak texture image j. n is the vector dimension, D uv Is the normalized Euclidean similarity distance between images u and v, x u 、x v Are the TRVs of images u and v, x u i Is the i-th component value of the TRV of image u; x u i ' Is the i-th component value of the normalized TRV of image u, x v i ' Is the i-th component value of the normalized TRV of image v, min(x i ) and max(x i ) Are the minimum and maximum values of the i-th component in all TRV data.

[0033] Furthermore, the calculation formula of the information sphere radius in step four is as shown in (8):

[0034]

[0035] In formula (8), Ten th is the spatial tension threshold, that is, the radius of the information sphere, is the tension threshold Ten th under the weak texture set, g n is a prior weak texture image in the weak texture set in the vector space, is the similarity distance set of the nth image.

[0036] Furthermore, in step five, the probability of the data to be measured in the information sphere occurring under the weak texture semantics is calculated as shown in formula (9):

[0037]

[0038] In formula (9), S weak represents the weak texture set, P(x|S weak ) is the conditional probability of the data to be measured x under the weak texture set S weak , m R (x) is the number of prior weak texture images contained in the information sphere of the data to be measured, and max represents taking the maximum value.

[0039] Furthermore, in step five, the information amount of the data to be measured is calculated through its occurrence probability, and the calculation formula is as shown in (10):

[0040] I(x|S weak ) = -log2(P(x|S weak )) (10)

[0041] In formula (10), I(x|S weak ) is the information amount of the data to be measured, and P(x|S weak ) is the conditional probability of the data to be measured x under the weak texture set S weak ;

[0042] Finally, an image information sphere classifier is constructed through the information amount of the data to be measured, and the formula is as shown in (11):

[0043]

[0044] In formula (11), B text is the final texture discrimination result of the image, 1 indicates discrimination as weak texture; I(x|S weak ) is the information amount of the data to be measured; σ th is the information amount threshold.

[0045] Furthermore, it also includes step six, using the confusion matrix, accuracy, precision, and recall rate to verify and analyze the discrimination result.

[0046] The present invention also provides an automatic discriminator for weakly textured images based on the spherical measure of image information, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the automatic discriminator method for weakly textured images based on the spherical measure of image information as described in the above solution.

[0047] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0048] The discriminator method for weakly textured remote sensing images proposed by the present invention is divided into two parts: extraction of the image texture richness vector and construction of the information sphere classifier. First, the texture of the image is analyzed from three aspects: the average degree of gray-scale change, the subtle degree of local gray-scale jitter, and the degree of global gray-scale chaos, respectively, through the mean of the texture gradient, the radius of the texture energy field, and the texture information entropy. Thus, a texture richness vector is constructed to analyze and describe the texture situation of the image. Then, an information sphere classifier is constructed. The sparse weakly textured image data obtained by visual interpretation is used as the evaluation criterion for weak texture discrimination. Then, an information sphere is constructed for the data to be measured in the vector space. By calculating the occurrence probability of the event of the data to be measured in the prior weak texture data, the three description components in the TRV of the image to be measured are condensed into the information content of the data to be measured, and finally the discrimination of weakly textured images is realized. The results show that the method proposed by the present invention can better realize the discrimination of weakly textured remote sensing images and is more stable than the traditional method. Description of the Drawings

[0049] Figure 1 is the flow chart of the method of the present invention;

[0050] Figure 2 is the flow chart for calculating the radius of the texture energy field;

[0051] Figure 3 is the visualization diagram of the texture energy field result of the data to be measured;

[0052] Figure 4 is the construction result diagram of the TRV of the data to be measured;

[0053] Figure 5 is the schematic diagram for constructing the information sphere of the image of the data to be measured;

[0054] Figure 6 is the schematic diagram of the data to be measured with weak texture and the data to be measured without weak texture;

[0055] Figure 7 is the schematic diagram of the quantitative confusion matrix. Detailed Embodiments

[0056] To facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0057] Please refer to Figure 1 the flowchart. A method for discriminating weak-texture remote sensing images provided by the present invention includes the following steps:

[0058] Step 1: Initialize the input remote sensing image, complete the initialization of parameters and discrimination thresholds respectively, and at the same time calibrate a small number of weak-texture image samples through expert visual interpretation to obtain prior weak-texture image data.

[0059] Step 2: Analyze the texture situation of the image by calculating the texture richness vector (TRV) of the image, and extract the quantitative result of the strength of the image texture.

[0060] Preferably, the texture richness of the image in Step 1 is quantitatively described. The TRV of the present invention is composed of three texture statistical information components: the mean texture gradient, the radius of the texture energy field, and the texture contour information entropy, which are used to analyze the texture of the image from the average change degree of the gray level of the image, the subtle degree of local gray level jitter, and the global gray level chaos degree respectively. The texture distribution of the image is characterized by the refined TRV, and the construction of the fast, concise, and thorough texture richness vector provides a basic guarantee for the construction of the texture classifier below.

[0061] Step 3: Calculate the mean image gradient. Calculate the image gradient map through Scharr filtering, and then obtain the mean image gradient through image mean calculation. In the visual nervous system, the image texture is intuitively manifested as the intensity value difference between the image pixels and the surrounding pixels. When the change of the image pixels and the adjacent pixels is significant, the image is understood to have strong texture. In computer vision, in order to highlight the pixel value change between adjacent pixels, an image gradient map is often constructed for the image, and the change rate and change direction of the pixel values in the image are characterized by the gradient value. Therefore, by constructing the image gradient map, the pixel value change between adjacent pixels of the image can be characterized, and the texture information of the image can be effectively highlighted. Further, the image gradient map is condensed into a mean index to analyze the image texture from a quantitative perspective.

[0062] Therefore, the present invention calculates the image gradient map through Scharr filtering. Among them, the definition of Scharr filtering is shown in formula (1):

[0063]

[0064] In formula (1), G x and G yIndicates the gradients of the pixel in the x and y directions. P1…9 represent the magnitudes of the intensities of the adjacent pixels in the 3×3 neighborhood of the image.

[0065] Calculate the mean value of the image texture gradient to evaluate the overall gradient change degree of the image. The relevant calculation formula is shown in Formula (2).

[0066]

[0067] In Formula (2), represents the mean value of the image texture gradient, w and h represent the width and length of the image gradient map, G i represents the gradient intensity value of pixel i, G x and G y represent the gradients of the pixel in the x and y directions.

[0068] Step 4: Calculate the radius of the image texture energy field. First, use Laplace filtering to sharpen the edge features of the image and filter out the noise and the image information with smooth changes; then perform frequency domain conversion under the fast Fourier transform to obtain the frequency spectrum diagram of the image; then, through the analysis of the frequency spectrum diagram, calculate the intensity value range and the upper quartile intensity value; after that, find the limit distance point farthest from the center point according to the upper quartile intensity value; finally, construct the texture energy field with the limit distance point as the boundary to obtain the energy field radius. As Figure 2 shown, it shows the schematic diagram of calculating the texture energy field of the remote sensing image. The fast Fourier transform calculation is shown in Formula (3).

[0069]

[0070] In Formula (3), f is the function representation of the signal in the original spatial domain, F is the function representation of the signal after transformation in the frequency domain (an image can be regarded as a kind of signal), the values of u and x both range from 0 to K - 1, W 2K is the sine and cosine function, K is a positive integer, and F(u) is the Fourier transform, which can be expressed as the sum of the products of a finite number of coefficients and the sine and cosine functions.

[0071] Based on the frequency spectrum diagram, with the upper quartile intensity value of the frequency spectrum diagram as the threshold, calculate the high-frequency radiation range of the image. The calculation of the upper quartile intensity value and the radius of the texture energy field are shown in Formulas (4) and (5).

[0072] Q up =(max(I) - min(I))·75% (4)

[0073] In Formula (4), Q up is the upper quartile intensity value in the frequency spectrum diagram, and max(I) and min(I) are the operations of finding the maximum and minimum values of the intensity values in the image I.

[0074]

[0075] In formula (5), S pnt is the set of coordinates of all points with intensity value Q up in the spectrogram, Int i is the spectrogram intensity value of pixel i, and S dis is the set of distances from the points in S pnt to the center point of the spectrogram. x0 and y0 are the coordinates of the center point of the spectrogram, x i and y i are the coordinates of pixel point i, and R text is the radius of the texture energy field of the image. Examples of the texture energy field radius calculated through the above steps are as shown in Figure 3 and Figure 3 shows the calculation results of 10 conventional images and 10 weakly textured images respectively.

[0076] Step 5: Calculate the image texture information entropy. Calculate the image contour information entropy based on the image contour map obtained by Canny filtering. Canny filtering extracts the image contour features through steps such as Gaussian filtering, gradient calculation, non-maximum gradient suppression, and double-threshold connection analysis, and can accurately extract the image edge contour. Information entropy is a statistical feature description of the information source and a measure of the degree of chaos of events in the system. As a digital signal, an image can use information entropy to describe the chaos of the distribution of each gray level in the image, reflecting the richness of colors in the image. The calculation of image information entropy is as shown in formula (6).

[0077]

[0078] In formula (6), H is the image information entropy, and P i is the occurrence probability of the gray value i in the image, CNT i is the number of occurrences of pixels with gray level i, and h and w are the height and width of the image. Thus, the quantification of the image texture richness is completed, and an example of the constructed TRV vector is as shown in Figure 4 and

[0079] Step 6: Thus, the quantification of the image texture richness is completed, and the abstract image texture information is reduced to a vector representation. Then, input the prior weakly textured image data obtained by visual interpretation, construct their respective TRV vectors for these data, thereby obtaining the prior knowledge of the weakly textured semantics, and through the quantified definition of the weakly textured, use it as the basis for subsequent image classification and discrimination.

[0080] Step 7: Construct the image information sphere. For the N input prior weak texture images and the unknown data to be measured, in the present invention, the value of N is taken as 8. By constructing the TRV of the above prior weak texture image data, the image texture information is expressed as spatial points in the vector space. Then, an information sphere in the space is constructed for the data to be measured, and the sphere radius is calculated through the minimum circumscribed sphere of the prior weak texture image data. Among them, the position calculation in the vector space adopts the normalized Euclidean distance, and its calculation formula is as follows:

[0081]

[0082] In formula (7), S dis is the set of similarity distance of the image, where N is the number of prior weak texture images, D j represents the normalized Euclidean similarity distance between the image to be measured and the prior weak texture image j; n is the vector dimension. In the present invention, since the TRV vector contains three components, the value of n is taken as 3, D uv is the normalized Euclidean similarity distance between images u and v, x u , x v are the TRVs of images u and v, x u i is the i-th component value of the TRV of image u. x u i ' is the i-th component value after normalization of the TRV of image u, x v i ' is the i-th component value after normalization of the TRV of image v, min(x i ) and max(x i ) are the minimum and maximum values of the i-th component in all TRV data.

[0083] The set of similarity distances shows the spatial semantic gap between a certain image and the rest of the images. The smaller the distance, the closer the semantics. Then, calculate the maximum semantic distance in the input prior weak texture data and use it as the sphere radius of the subsequent data to be measured to ensure that under this distance, the semantic information of the N input prior weak texture data in the vector space is consistent. Its calculation formula is shown in (8).

[0084]

[0085] In formula (8), Ten th is the spatial tension threshold, that is, the information sphere radius, is the weak texture set under the tension threshold Ten th , is the set of similarity distances of the n-th image, g nIt is a prior weak texture image in the weak texture set in the vector space. Ten refers to the tension magnitude when constructing an information sphere in the space, which is the sphere radius in a mathematical sense. Figure 5 It shows the information sphere of the data to be measured constructed through the above steps.

[0086] Step 8: Construct an information sphere classifier. Analyze the semantics of the data to be measured by the number of weak texture samples contained in the information sphere to be measured. First, calculate the occurrence probability of the data to be measured under the weak texture semantics through the information sphere. The calculation of the image occurrence probability is shown in formula (9).

[0087]

[0088] In formula (9), S weak represents the weak texture set, P(x|S weak ) is the conditional probability of the data to be measured x under S weak , m R (x) is the number of prior weak texture images contained in the information sphere of the data to be measured, is the maximum m R (x) existing in the prior weak texture image data. In this invention, since 8 prior weak texture data are input, this value is taken as 8.

[0089] After that, according to Shannon's theorem, calculate the information amount of the data to be measured through its occurrence probability. The calculation formula is shown in (10).

[0090] I(x|S weak )=-log2(P(x|S weak )) (10)

[0091] In formula (10), I(x|S weak ) is the information amount of the data to be measured, and P(x|S weak ) is the conditional probability of the data to be measured x under the prior data weak texture set S weak .

[0092] Finally, construct an information sphere classifier through the information amount of the data to be measured. The formula is shown in (11).

[0093]

[0094] In formula (11), B text is the final texture discrimination result of the image. 1 indicates discrimination as weak texture; I(x|S weak ) is the information amount of the data to be measured; σ th is the information amount threshold, and this value is not a fixed value. In this invention, this value is taken as 0.5 for experimental verification.

[0095] Step 9: Use the confusion matrix, accuracy, precision, and recall to verify and analyze the discrimination results of the present invention. The present invention uses 80 remotely sensed images to be tested to evaluate the performance of the algorithm, including 40 conventional texture images and 40 weak texture images. Part of the dataset is shown in Figure 6 . The present invention names the proposed weak texture remotely sensed image discrimination method as MBISC and compares it with several optimal image classification and discrimination methods (Ori-KMC, Ori-HC, LBP-KMC, and LBP-HC). The comparison results are shown in Table 1.

[0096] Table 1 Index Analysis Table

[0097]

[0098] Figure 7 The confusion matrix results of 80 images under different discrimination methods are shown. All methods can accurately discriminate non-weak texture images, but there are differences in the discrimination of weak texture images. The Ori-KMC and Ori-HC methods have a low success rate in discriminating weak textures, mainly due to noise and environmental factor interference. The LBP-KMC and LBP-HC algorithms improve the discrimination success rate through filtering, but there is still about 1 / 4 misjudgment. Table 1 shows that although various methods have a high precision in the classification of non-weak texture images, the problem of missed detection of weak textures is still prominent. In contrast, the MBISC method of the present invention can accurately discriminate 40 conventional non-weak texture images and 40 weak texture images, and reaches 100% in all three evaluation indicators, showing the best performance and proving its effectiveness in weak texture image discrimination.

[0099] On the other hand, the embodiment of the present invention also provides an automatic weak texture image discrimination device based on the spherical metric of image information, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the automatic weak texture image discrimination method based on the spherical metric of image information as described in the above solution.

[0100] It should be understood that the parts not elaborated in detail in this specification belong to the prior art.

[0101] It should be understood that the above description of the preferred embodiment is relatively detailed, and it should not be considered as a limitation to the protection scope of the present invention patent. Under the inspiration of the present invention, those of ordinary skill in the art can also make substitutions or deformations without departing from the protection scope defined by the claims of the present invention, and all fall within the protection scope of the present invention. The scope of the present invention claimed should be subject to the appended claims.

Claims

1. An automatic discrimination method for weakly textured images based on spherical measurement of image information, characterized in that, It includes the following steps: Step 1: Initialize the input image, initialize the parameters and discrimination thresholds respectively, and obtain prior weak texture image data through visual interpretation at the same time; Step 2: Construct a texture intensity vector TRV by calculating the mean value of the texture gradient, the radius of the texture energy field, and the texture contour information entropy of the image, and extract the quantization result of the texture strength and weakness of the image; In Step 2, first use Laplace filtering to sharpen the edge features of the image and filter out the noise and smoothly changing image information; then perform frequency domain conversion under the fast Fourier transform to obtain the spectrogram of the image; then through the analysis of the spectrogram, calculate the intensity value range and the upper quartile intensity value; then find the limit distance point farthest from the center point according to the upper quartile intensity value; finally, construct a texture energy field with the limit distance point as the boundary to obtain the energy field radius; Step 3: Input the prior weak texture image data obtained by visual interpretation, construct their respective TRV vectors, and obtain the prior knowledge of weak texture semantics; Step 4: Generate a vector space based on the TRV vectors in Step 3, construct an information sphere in the vector space for the data to be measured, and the radius of the information sphere is calculated through the minimum circumscribed sphere of the prior weak texture image data; In Step 4, the vector space uses the linear normalized Euclidean distance for position measurement, and the calculation is shown in formula (7): In formula (7), S dis is the similarity distance set of the images, where N is the number of prior weak texture images, and D j represents the normalized Euclidean similarity distance between the data to be measured and the prior weak texture image j, n is the vector dimension, and D uv is the normalized Euclidean similarity distance between images u and v, x u , x v are the TRVs of images u and v, and x u i is the i-th component value of the TRV of image u; x u i ' is the i-th component value after normalization of the TRV of image u, and x v i ' is the i-th component value after normalization of the TRV of image v. min(x i ) and max(x i ) are the minimum and maximum values of the i-th component in all TRV data; The calculation formula of the radius of the information sphere in Step 4 is shown in (8): In formula (8), Ten th is the spatial tension threshold, that is, the information sphere radius, is the tension threshold Ten th under the weak texture set, g n is a prior weak texture image in the weak texture set in the vector space, is the similarity distance set of the nth image; Step 5: Construct an information sphere classifier, calculate the probability of the data to be measured occurring under weak texture semantics in the information sphere, and calculate the amount of information of the data to be measured under weak texture semantics according to Shannon's theorem, and finally compare it with the set threshold to realize the discrimination of the texture strength and weakness of the data to be measured.

2. The automatic discrimination method for weakly textured images based on spherical measurement of image information according to claim 1, characterized in that: In Step 2, calculate the mean value of the texture gradient of the image based on the Scharr filtering result, which reflects the average change degree of the image gray level. The Scharr filtering formula is shown in formula (1): In formula (1), G x and G y represent the gradients of the pixel in the x and y directions, and P1…P9 represent the intensity magnitudes of the adjacent pixels in the 3×3 neighborhood of the image; Calculate the mean value of the texture gradient of the image to evaluate the overall gradient change degree of the image. The relevant calculation formula is shown in formula (2): In formula (2), represents the mean texture gradient of the image, w and h represent the width and length of the image gradient map, and G i represents the gradient intensity value of pixel i, and G x and G y represent the gradients of the pixel in the x and y directions.

3. The automatic discrimination method for weakly textured images based on spherical measurement of image information according to claim 1, characterized in that: The fast Fourier transform calculation in Step 2 is shown in formula (3): In Equation (3), f is the function representation of the signal in the original spatial domain, F is the function representation of the signal in the frequency domain after transformation, both u and x take values from 0 to K - 1, and W 2K is a sine and cosine function, K is a positive integer, and F(u) is the Fourier transform, expressed as the sum of the products of the coefficients of a finite number of terms and sine and cosine functions; Based on the spectrogram, use the upper quartile intensity value of the spectrogram as the threshold to calculate the high-frequency radiation range of the image. The upper quartile intensity value and the radius of the texture energy field are calculated as shown in formulas (4) and (5): Q up = (max(I) - min(I)) · 75% (4) In formula (4), Q up is the upper quartile intensity value in the spectrogram, and max(I) and min(I) are operations to find the maximum and minimum values of the intensity values in image I; In formula (5), S pnt is the set of coordinates of all points with intensity value Q up in the spectrogram, Int i is the spectrogram intensity value of pixel i, S dis is the set of distances from the points in S pnt to the center point of the spectrogram. x0 and y0 are the coordinates of the center point of the spectrogram, x i and y i are the coordinates of pixel point i, and R text is the radius of the texture energy field of the image.

4. The automatic discrimination method for weakly textured images based on spherical measurement of image information according to claim 1, characterized in that: In Step 2, calculate the texture contour information entropy of the image based on the Canny-filtered image contour map, which reflects the global gray level confusion degree. The calculation of the image information entropy is shown in formula (6): In Equation (6), H is the image information entropy, and P i is the occurrence probability of the gray value i in the image, and CNT i is the number of occurrences of the pixel with the gray level i, and h and w are the height and width of the image.

5. The automatic discrimination method for weakly textured images based on spherical measurement of image information according to claim 1, characterized in that: In Step 5, the calculation of the probability of the data to be measured occurring under weak texture semantics in the information sphere is shown in formula (9): In formula (9), S weak represents the weak texture set, and P(x|S weak ) is the conditional probability of the data x to be measured under the weak texture set S weak , m R (x) is the number of prior weak texture images contained in the information sphere of the data x to be measured, and max represents taking the maximum value.

6. The automatic discrimination method for weakly textured images based on spherical measurement of image information according to claim 1, wherein: In Step 5, calculate the amount of information of the data to be measured through the occurrence probability of the data to be measured. The calculation formula is shown in (10): I(x|S weak ) = -log2(P(x|S weak )) (10) In formula (10), I(x|S weak ) is the amount of information of the data to be measured, and P(x|S weak ) is the conditional probability of the data x to be measured under the weak texture set S weak ; Finally, construct an image information sphere classifier through the amount of information of the data to be measured, and the formula is shown in (11): In formula (11), B text is the final texture discrimination result of the image, where 1 indicates weak texture discrimination; I(x|S weak ) is the amount of information of the data to be measured; σ th is the information amount threshold.

7. The automatic discrimination method for weakly textured images based on spherical measurement of image information according to claim 1, characterized in that: It also includes Step 6: Use the confusion matrix, accuracy, precision, and recall rate to verify and analyze the discrimination results.

8. An automatic discriminator for weakly textured images based on spherical measurement of image information, characterized in that: It includes a processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the automatic discrimination method for weak texture images based on image information sphere measurement described in any one of claims 1-7.

Images