A method for driving an image refresh of an electronic display device

Abstract

The present invention provides a method for driving an image refresh of an electronic display device, which relates to the technical field related to electronic displays, including: inputting a target image into an image detection model to obtain M pre-detection results and N results to be detected secondaryally; constructing a pre-detection detection sequence and a secondary recognition image set; obtaining first visual, second visual and third detection sequences; obtaining primary, secondary and third difference scores; pre-constructing a difference analysis model to obtain a difference analysis value of the secondary recognition image set, wherein the difference analysis value is greater than a preset second threshold, and P second detection results are obtained; generating a corresponding driving instruction based on the result of M+P and the result of the third threshold judgment, converting the control signal of the corresponding driving instruction into a corresponding voltage waveform, transmitting the voltage waveform to an electronic paper display screen, and refreshing the electronic paper display screen according to the corresponding waveform to update the display content; reducing energy consumption, improving the overall refresh efficiency, and enhancing the user experience.

Description

Technical Field

[0001] The present application relates to the technical field related to electronic display, and in particular to a method for driving an image refresh of an electronic display device. Background Art

[0002] Electronic paper displays are a special type of electronic display device with a unique display refresh mechanism. This passive refresh mechanism allows the screen to refresh to display a new image only when image data needs to be updated. Initially, electronic paper displays were limited to black and white. Due to the small amount of data transmitted, they can be globally refreshed to display new images, resulting in shorter image updates and less consuming transmission channel resources. Currently, electronic paper displays have evolved from their original black and white state to support multiple colors, such as three, four, and six. These displays are often used in scenarios such as e-books, book illustrations, and static advertisements, which require high-quality electronic paper images. This, in turn, increases the amount of image data transmitted and consumes significant transmission channel resources. Traditional electronic paper display hardware is not suitable for today's multi-color image transmission requirements. Currently, the main approach to improving global refresh efficiency is to upgrade and modify hardware such as the main control chip and display driver chip. This process involves complex hardware development, which not only increases costs but also may cause compatibility issues. Furthermore, it still consumes a large amount of transmission channel data, consumes a lot of energy, and can affect the user experience. Summary of the Invention

[0003] This application provides a method for driving an image refresh of an electronic display device, aiming to solve the problems of high cost and high energy consumption in the prior art when driving image refresh, and the impact on user experience due to compatibility issues.

[0004] A first aspect disclosed in the present application provides a method for driving an image refresh of an electronic display device, wherein the electronic display device includes an electronic paper display screen, and the method includes:

[0005] Acquire an image to be displayed, determine whether the image to be displayed is a non-black and white image, and obtain a target image, wherein the image to be displayed is a static image formed by a combination of text and / or pictures; input the target image into the image segmentation unit of the image detection model, obtain K target segmented images and image position identifiers, input the K target segmented images into the hash algorithm detection unit of the image detection model, and obtain hash values of the K target segmented images through the BMH hash algorithm; traverse the image position identifier, calculate the Hamming distance between the hash value of the target segmented image under each image position identifier and the hash value of the current segmented image, map the Hamming distance to the image position identifier one by one, perform a pre-inspection on the Hamming distance through a preset first threshold, and obtain M pre-inspection results. The method comprises the following steps: defining the image position identifiers of the N results to be detected twice and N results to be detected twice, wherein M and N are integers ≥ 0, and M + N = K; defining the image position identifiers of the N results to be detected twice as secondary detection position identifiers, arranging the secondary detection position identifiers in ascending order according to their corresponding Hamming distances, obtaining a pre-detection detection sequence, and collecting the current segmented images under the secondary detection position identifiers to form paired segmented images with the results to be detected twice, to obtain a secondary recognition image set; performing difference detection on each paired segmented image in the secondary recognition image set by using a histogram method, a structural similarity index, and a feature point matching method, respectively, and arranging the secondary detection position identifiers in ascending order according to their corresponding detection results, to obtain a first detection sequence, a second detection sequence, and a third detection sequence, respectively. column; taking the pre-detection detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the first detection sequence, and obtain a first difference score; taking the first detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the second detection sequence, and obtain a second difference score; taking the second detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the third detection sequence, and obtain a third difference score; synchronize the first difference score, the second difference score, and the third difference score to the difference analysis model to obtain a difference analysis value of the secondary recognition image set, wherein the difference analysis model is based on a multivariate regression model and uses three difference scores as as input and a difference analysis value as output; judging that the difference analysis value is greater than a preset second threshold, eliminating duplicate images, and obtaining P secondary detection results, P≤N; judging whether the result of M+P is less than a preset third threshold, if so, driving the master control chip to generate a local refresh drive instruction, driving the display driver chip to convert the control signal of the instruction into a corresponding voltage waveform, driving the electronic paper display to perform a local image refresh according to the voltage waveform corresponding to the instruction, and displaying the target image; if not, driving the master control chip to generate a global refresh drive instruction, driving the display driver chip to convert the control signal of the instruction into a corresponding voltage waveform, driving the electronic paper display to perform a global image refresh according to the voltage waveform corresponding to the instruction, and displaying the target image.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] Acquire the image to be displayed, determine whether the image to be displayed is a non-black and white image, and obtain the target image, which can avoid unnecessary image difference judgment, divert the execution of the refresh method of the image driving the electronic display device, and improve the overall refresh efficiency; input the target image into the image detection model, obtain K target segmentation images, perform pre-inspection, obtain M pre-inspection results and N secondary detection results, quickly identify the target segmentation image that needs to be subjected to multivariate visual detection, improve the accuracy of image detection and reduce the amount of calculation of the system, and improve the overall refresh efficiency; construct a pre-inspection detection sequence and a secondary recognition image set, and perform further image difference recognition in the secondary recognition atlas, with a clear target and a narrowed scope, thereby improving detection efficiency; obtain the first detection sequence, the second detection sequence, and the third detection sequence, and more comprehensively capture the differences between paired segmented images, thereby enhancing the comprehensiveness of detection; through the first difference score, the second difference score, and the third difference score , reflects the degree of difference between paired segmented images under different visual detection technologies, and helps to comprehensively evaluate the overall difference of paired segmented images in the secondary recognition image set; through the difference analysis model, a multivariate regression comprehensive judgment is performed on the first difference score, the second difference score, and the third difference score, and the scattered difference scores are combined to obtain the difference analysis value of the secondary recognition image set, thereby improving the judgment accuracy of the paired segmented images in the secondary recognition image set; it is judged that the difference analysis value is greater than a preset second threshold, and P second detection results are obtained; it is judged whether the result of M+P is less than the third threshold, and the judgment result is used to drive the main control chip to generate a corresponding drive instruction, drive the display driver chip to convert the control signal of the corresponding drive instruction into a corresponding voltage waveform, drive the electronic paper display to refresh the image according to the refresh strategy corresponding to the corresponding drive instruction, update the displayed image, and minimize unnecessary refresh operations. The present invention uses a simpler, more convenient and lower-cost method to minimize the occupation of transmission channel resources and thus reduce resource consumption, thereby improving the overall efficiency of driving the refresh of the electronic paper display image. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic flow chart of a method for driving an electronic display device to refresh an image provided in an embodiment of the present application.

[0009] Figure 2 This is a diagram of the internal structure of a system for driving an image refresh of an electronic display device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The embodiments of the present application provide a method for driving an image refresh of an electronic display device, thereby solving the problems of the prior art in driving image refresh, such as high cost and high energy consumption, and the impact on user experience due to compatibility issues.

[0011] After introducing the basic principles of this application, various non-limiting embodiments of this application will be specifically described below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described here are only used to explain this application and are not used to limit this application.

[0012] An electronic display device includes an electronic paper display screen. In applications for driving the image refresh of the electronic paper display screen, a main control chip and a display driver chip are integrated on a single circuit board and electrically connected to the electronic paper display screen for data transmission and control. The main control chip is responsible for processing input information and generating corresponding drive instructions based on the results of analyzing the input information. The display driver chip is responsible for generating a corresponding voltage waveform from the control signal of the drive instruction, transmitting the generated corresponding voltage waveform to the electronic paper display screen via a physical circuit, and driving the electronic paper display screen to refresh, thereby updating the displayed image. Improving refresh efficiency by upgrading and modifying hardware such as the main control chip and the display driver chip involves complex hardware development, higher costs, and possible compatibility issues. The method provided in this application aims to improve the overall efficiency of driving the image refresh of the electronic paper display screen by improving the efficiency of image difference detection, reducing resource consumption and minimizing the occupied transmission channel resources in a simpler, more convenient, and less expensive way.

[0013] like Figure 1 As shown, an embodiment of the present application provides a method for driving an image refresh of an electronic display device, wherein the electronic display device includes an electronic paper display screen, and the method includes:

[0014] Step S100: Acquire an image to be displayed, and determine whether the image to be displayed is a non-black and white image, wherein the image to be displayed is an image that has been pre-processed to meet the application requirements of an electronic paper display screen, the pre-processing including image denoising, image resolution adjustment, and image dithering. The image dithering includes a black and white mode and a non-black and white mode, and the dithering mode is marked. The non-black and white mode includes a three-color dithering mode, a four-color dithering mode, a five-color dithering mode, and a six-color dithering mode. The image to be displayed is a static image formed by a combination of text and / or pictures.

[0015] If so, determine that the image to be displayed is a non-black and white image, and obtain the target image.

[0016] If not, it is determined that the image to be displayed is a black and white image, the master control chip is driven to generate a global refresh drive instruction, the display driver chip is driven to convert the control signal of the instruction into a corresponding voltage waveform, and the electronic paper display screen is driven to perform global image refresh according to the voltage waveform corresponding to the instruction to display the target image.

[0017] In an embodiment of the present application, the image display refresh of the electronic paper display is jointly driven by the main control chip and the display driver chip. The acquired target image is transmitted to the main control chip, and a difference detection is performed between the target image and the current image. The preset refresh strategy is executed based on the detection result. The main control chip is driven to generate a drive instruction according to the corresponding refresh strategy. The display driver chip is driven to convert the control signal of the drive instruction into a corresponding voltage waveform. The corresponding voltage waveform is transmitted to the electronic paper display via a physical line. The electronic paper display is driven to refresh the corresponding area according to the voltage waveform and update the displayed image. The image to be displayed is obtained from multiple sources. The image can be uploaded from the web page, the stored image resource can be obtained from the server in the web page environment, or the image can be obtained through operations such as uploading. The image can be in a format such as JPEG or PNG and marked as the image to be displayed. In the e-paper display image processing tool on a webpage or mobile phone app, based on the application of the e-paper display, including the application scenario of the e-paper display, the resolution of the e-paper display, and the different color display modes of the e-paper display (supporting black and white, three-color, four-color, five-color, or six-color display), the image to be displayed is first denoised and the image resolution is adjusted to match the resolution of the e-paper display. The image is then dithered and can be transferred to the main control chip via USB. You can also choose to transfer the image from the WeChat mini-program, accessing the photo album or camera to obtain the image to be displayed. After the image dithering process, the main control chip supports Bluetooth function and communicates with the WeChat mini-program via the Bluetooth module. The main control chip and display driver chip are integrated on a circuit board, and the chip is usually installed on the device's motherboard. Obtain a pre-processed image to be displayed, and determine whether the image to be displayed is a black and white image or a multi-color image based on the jitter pattern marked on the display image. If it is a black and white image, execute a second refresh strategy, drive the master control chip to generate a second drive instruction according to the second refresh strategy, drive the display driver chip to convert the control signal of the second drive instruction into a corresponding voltage waveform, drive the electronic paper display to refresh the image according to the refresh strategy corresponding to the second drive instruction, and update the displayed image; determine whether it is a multi-color image, which is the target image, and input it into the image detection model.

[0018] Step S200: Input the target image into the image segmentation unit of the image detection model to obtain K target segmentation images and image position identifiers, input the K target segmentation images into the hash algorithm detection unit of the image detection model, and obtain the hash values of the K target segmentation images through the BMH hash algorithm.

[0019] In an embodiment of the present application, image difference detection is located at the front end of the entire refresh process. After acquiring the target image, the main control chip first performs difference detection. The detection result directly affects the subsequent generation of drive instructions. Efficiently performing difference detection between the target image and the current image can accurately find the parts of the image that have actually changed, avoiding subsequent driving and refreshing operations on most of the same areas, thereby directly reducing the overall refresh workload and improving work efficiency. Therefore, improving the efficiency of image difference detection is an efficient, simple and low-cost method for driving the refresh of the electronic paper display screen image. Specifically, the target image is divided into K target segmented images, each target segmented image carries an image position identifier, and prepares for the precise positioning of the position of each segmented image in subsequent image recognition and screen refresh operations. The BMH image hashing algorithm is used to calculate the average color value of each target segmented image. The BMH (Block Mean Hash) image hashing algorithm, also known as "block mean hashing" or "block average hashing," is an existing and low-cost image hashing technique. It creates a hash by dividing the image into multiple small blocks and then calculating the average color value of each block. The block size depends on the hash granularity required for the e-paper display application; smaller blocks provide finer hashes, while larger blocks produce coarser hashes. For each block, the average of the red, green, and blue (RGB) channels is calculated for all pixels within the area. Each block can be represented by a triplet (R, G, B) to represent its average color. When comparing two images, the Hamming distance between the two hashes can be calculated; a smaller distance indicates that the two images are similar. A hash value is calculated for each target segmented image and mapped to the image location identifier of the corresponding target segmented image. This mapping facilitates difference detection in subsequent secondary image recognition.

[0020] Furthermore, K target segmentation images and image position identifiers are obtained, including:

[0021] Obtaining a sample image, determining the number of pixels of a segmented image of the sample image based on the total number of pixels of the electronic paper display and a specific application scenario, calculating a ratio of the total number of pixels of the electronic paper display to the number of pixels of the segmented image of the sample image, and obtaining a number K of sample image segmentations;

[0022] Determine whether K is a perfect square number,

[0023] If so, the sample target image is divided into The grid,

[0024] If not, select the two closest factors r and c so that r×c=K, and divide the sample image into r×c grids;

[0025] Mark each grid with a location identifier to obtain an image segmentation unit;

[0026] The target image is input into an image segmentation unit to obtain K target segmented images and K image position identifiers.

[0027] In an embodiment of the present application, an image segmentation unit is constructed to determine the number of pixels in each segmented image of a sample target image based on the total number of pixels of the electronic paper display and the specific application scenario of the electronic paper display. The number of pixels in each segmented image is consistent, and the ratio of the total number of pixels of the electronic paper display to the number of pixels in the segmented image of the sample target image is calculated to meet the visual granularity required by the electronic paper display in the specific application scenario, which is also the hash granularity of the BMH image hash algorithm. Specifically, the resolution of the electronic paper display is obtained. The resolution is the number of pixels in the width and height of the display, and the total number of pixels is the product of the width and height of the resolution. Based on the total number of pixels of the electronic paper display and the specific application scenario of the electronic paper display, the number of pixels of the segmented image of the sample target segmentation image can be manually determined, and the ratio of the total number of pixels of the electronic paper display to the number of pixels of the segmented image of the sample target image can be calculated to obtain the number of segmentation of the sample target image K. Alternatively, sample target images of the same application scenario can be obtained from multiple sources, and 80% of the sample target images can be used as a training set. Based on the total number of pixels of the electronic paper display and the specific application scenario of the electronic paper display, various numbers of pixels of the segmented image of the sample target image can be tried. Each solution is pre-cut and analyzed to see whether the cutting result meets the user's visual comfort threshold requirement when the image is refreshed according to the segmented image and the refresh efficiency threshold requirement of the specific application scenario. If the cutting result is greater than the visual comfort threshold and less than the refresh efficiency threshold, and meets both threshold requirements, then the number of pixels of the segmented image of the selected sample target image is reasonable. The remaining 20% of the sample target images are used as a validation set for the final evaluation of the accuracy of the selected segmentation scheme. The number of segmentation of the sample target image K is obtained, where K is an integer greater than 0. If K is a perfect square number, the sample target image is divided into If K is not a perfect square, the two closest factors r and c are selected such that r × c = K, and the sample target image is segmented into r × c grids. Each grid is labeled with an image location identifier, indicating the row and column of the sample target image. This is a key step in the image segmentation process, used to determine the location of each sample segmented image in the original image and obtain an image segmentation unit. The target image is input into the image segmentation unit to obtain K target segmented images and their corresponding K image location identifiers.

[0028] Step S300: traverse the image position identifiers, calculate the Hamming distance between the target segmented image hash value and the current segmented image hash value under each image position identifier, map the Hamming distance to the image position identifier one by one, perform a pre-check on the Hamming distance using a preset first threshold, and obtain M pre-check results and N secondary check results, where M and N are integers ≥ 0, and M + N = K.

[0029] Specifically, the algorithm traverses the image location identifiers, matches the hash values of the target segmented image and the hash values of the current segmented image under the same image location identifier, calculates the Hamming distance between the two, and establishes a one-to-one mapping relationship between the calculated Hamming distance and the corresponding image location identifier. The Hamming distance measures the distance between two hash values and is simple and fast to calculate. The more visually similar the two images, the smaller the Hamming distance.

[0030] Furthermore, before inputting the target image into the image segmentation unit of the image detection model, the method includes:

[0031] K segmented images of the current image displayed on the electronic paper display and their corresponding image position identifiers and hash values are extracted from the resident foreground image set, and the image position identifiers of the K current segmented images and their corresponding hash values are input into the hash algorithm detection unit.

[0032] Specifically, K current segmented images of the currently displayed image are selected from the resident foreground image set, along with their corresponding hash values and image location identifiers. The hash values and image location identifiers corresponding to each current segmented image are then input into a hash algorithm detection unit to prepare for identifying differences between the target image and the currently displayed image. The resident foreground image set can be stored on an external memory card. The specific interface and connection method between the external memory card and the main control chip can be based on an interface protocol to achieve data transmission.

[0033] Furthermore, the obtained Hamming distance is pre-checked using a preset first threshold to obtain M pre-check results and N results to be re-checked, including:

[0034] Input the K target segmentation images into a hash algorithm detection unit, obtain a hash value of each target segmentation image through a BMH hash algorithm, and store the K target segmentation images with the image position identifiers and their corresponding hash values in a foreground image set;

[0035] Traversing the image position identifiers, calculating the Hamming distance between the target segmented image hash value and the current segmented image hash value under the same image position identifier, and establishing a mapping relationship between the obtained Hamming distance and the corresponding position identifier;

[0036] A first threshold is preset, and the target segmentation images corresponding to the Hamming distance greater than the preset first threshold are added to the first image update list to obtain M preliminary inspection results. The target segmentation images corresponding to the Hamming distance less than the preset first threshold are N results to be tested again.

[0037] In an embodiment of the present application, the hash detection method is capable of quickly identifying image differences, and the BHM hash algorithm is used to calculate the hash value of each target segmented image. The calculated hash value of each target segmented image can be input into a hash value list, and the Hamming distance between the hash value of the target segmented image and the current segmented image under the same image position identifier is calculated and input into the hash value list. A first threshold is preset, and the K Hamming distances obtained are compared with the first threshold to obtain M pre-detection results with Hamming distances greater than the first threshold, and N results to be detected with Hamming distances less than the first threshold, which are the results to be detected. These N results to be detected may be due to limitations of the BMH hash detection method, or due to hash conflicts, and it is judged that the images are similar and do not need to be updated. Specifically, since multi-color displays on electronic paper displays have high requirements for image color depth and brightness, and pre-checking involves detecting differences between K paired segmented images, the pre-checking detection method needs to be sensitive to local color changes and details, as well as image edges and detail changes. The BMH hash algorithm is sensitive to changes in the overall brightness and color distribution of an image, is computationally simple, fast, and suitable for image segmentation, but is insensitive to local detail changes and cannot effectively capture local color changes and details. The BMH hash algorithm is embedded in a hash detection unit, which calculates a hash value for each target segmented image. The K target segmented images, each with an image position identifier, and their corresponding hash values are stored in a foreground image set for subsequent image difference detection. The K current segmented images of the currently displayed image and the K target segmented images of the target image in the foreground image set are each segmented by the image segmentation unit into K target segmented images with image position identifiers and K current segmented images with image position identifiers. Identical image position identifiers indicate that the target segmented image and the current segmented image are located at the same position in the original image. Traverse the image location identifiers, match the hash values of the target segmented image and the current segmented image corresponding to the same image location identifier, calculate the Hamming distance between the hash value of each target segmented image and the hash value of its corresponding current segmented image, obtain the Hamming distance of the paired segmented images under each image location identifier, and establish a mapping relationship between the obtained Hamming distance and the corresponding image location identifier. Perform a pre-check, preset a first threshold, compare K Hamming distances with the first threshold, obtain M Hamming distances greater than the first threshold, and according to the image location identifiers corresponding to the M Hamming distances, find the target segmented image with the same image location identifier from the K target segmented images, add the M target segmented images to the first image update list, and obtain M pre-check results. The M pre-check results include M target segmented images and their image location identifiers. These target segmented images can be quickly identified as different from the current segmented image through BMH hash detection and are images that must be updated.The remaining N target segmentation images corresponding to the Hamming distance less than the first threshold may not cause correct judgment of visual difference due to the limitations of the BMH hash detection method or hash conflicts. These N target segmentation images need to be recognized twice.

[0038] Step S400: Define the N image position identifiers of the results to be secondary detected as secondary detection position identifiers, arrange the secondary detection position identifiers in ascending order according to their corresponding Hamming distances to obtain a pre-detection sequence, collect the current segmented image under the secondary detection position identifier, and form a paired segmented image with the results to be secondary inspected to obtain a secondary recognition image set.

[0039] In an embodiment of the present application, the N results to be secondary detected include N target segmentation images whose Hamming distance is less than a first threshold and their corresponding image position identifiers. Specifically, the target image is pre-checked by the image detection model to obtain N target segmentation images that need to be secondary identified. The image position identifiers corresponding to the N target segmentation images that need to be secondary identified are defined as secondary detection position identifiers. The secondary detection position identifiers are arranged in ascending order of the corresponding Hamming distances to form a pre-check detection sequence. The secondary detection position identifiers are traversed. For each secondary detection position identifier, there is a corresponding paired segmentation image, including the target segmentation image and the current segmentation image. The paired segmentation images under the N secondary inspection position identifiers are grouped together to construct a secondary recognition image set.

[0040] Step S500: performing difference detection on each pair of segmented images in the secondary recognition image set by using the histogram method, the structural similarity index and the feature point matching method respectively, arranging the secondary detection position identifiers in ascending order according to their corresponding detection results, and obtaining a first detection sequence, a second detection sequence and a third detection sequence respectively.

[0041] In an embodiment of the present application, visual detection technologies are traversed to screen out visual detection technologies that are suitable for detecting static image differences and that can make up for the shortcomings of BMH hash detection, analyze the reasons for the errors in BMH hash detection, and determine the histogram method, structural similarity index and feature point matching method for image difference recognition of paired segmented images in the secondary recognition image set. Specifically, the BMH hash detection method may have misjudgments, such as hash conflicts, or be insensitive to changes in local details, such as fine-tuning of image color but the light and dark relationship and structure remain unchanged, or being insensitive to changes in image brightness and contrast. The histogram method directly focuses on the actual pixel value distribution characteristics of the image (such as the distribution of grayscale values or color values), and can detect segmented images that are misjudged due to hash conflicts; the histogram method is more sensitive to image color changes. Even if the light and dark relationship and structure of the image remain unchanged, as long as the color is fine-tuned, the histogram will be reflected, and it can detect segmented images that are misjudged by the hash detection method because the image color is fine-tuned but the light and dark relationship and structure remain unchanged. The secondary recognition image set is traversed and the histogram method is used to perform difference recognition on paired segmented images. This method can detect hash collisions or misclassifications caused by slight adjustments to image color while maintaining the same brightness and structure. The histogram detection method is used to detect the secondary recognition image set, mapping the calculated histogram distances to their corresponding secondary detection position markers. The secondary detection position markers are then sorted in ascending order according to the histogram detection results to obtain the first detection sequence. The structural similarity index has good detection capabilities for changes in image structure and can detect misclassified segmented images due to hash collisions. It can also compensate for areas that are insensitive to the hash algorithm and histogram detection methods. For example, the structural similarity index is very sensitive to changes in local details and can identify differences in segmented image boundaries and local details, as well as brightness differences between different image regions. The secondary recognition image set is traversed and the structural similarity index is used to perform difference recognition on paired segmented images. This method can detect misclassifications caused by hash collisions and types to which the hash detection and histogram detection methods are insensitive. The structural similarity index method detects the secondary recognition image set, performs a difference metric conversion on the calculated structural similarity index, establishes a mapping relationship between the converted structural similarity index and its corresponding secondary detection position identifier, and arranges the secondary detection position identifiers in ascending order of the converted structural similarity index to obtain a second detection sequence. The feature point matching method focuses on the feature points in the image (such as corner points, edge points, etc.) and can detect segmented images misjudged due to hash collisions and misjudgments caused by the insensitivity of the histogram method and the structural similarity index method. Compared with the structural similarity index, it pays more attention to the feature points in the image, and the granularity of local and detail detection is refined. The feature point matching method detects the secondary recognition image set, establishes a mapping relationship between the calculated feature point matching value and its corresponding secondary detection position identifier, and arranges the secondary detection position identifiers in ascending order of the feature point matching value to obtain a third detection sequence.The three aforementioned inspection methods offer diversified local and detailed detection. Combining these methods can significantly reduce missed detections. Each method has its strengths and limitations, and using one method alone may miss certain types of changes. When combined, these three methods can comprehensively detect image changes from three perspectives: pixel value distribution, structure, and feature points, thus reducing the likelihood of missed detections. The greater the detection data from the three visual inspection methods, the greater the image differences.

[0042] Furthermore, the first detection sequence, the second detection sequence, and the third detection sequence are obtained respectively, including:

[0043] Based on the secondary recognition image set, using the histogram method to perform difference detection on each pair of segmented images in the secondary recognition image set, establishing a mapping relationship between the obtained histogram distance and the secondary detection position identifier, and arranging the secondary detection position identifiers in ascending order according to their corresponding histogram distances to obtain a first detection sequence;

[0044] Based on the secondary recognition image set, using the structural similarity index method to perform difference detection on each pair of segmented images in the secondary recognition image set, performing difference metric conversion on the obtained structural similarity index, wherein the converted structural similarity index ranges from [0, 2], establishing a mapping relationship between the converted structural similarity index and the secondary detection position identifier, and arranging the secondary detection position identifiers in ascending order according to their corresponding converted structural similarity indexes to obtain a second detection sequence;

[0045] Based on the secondary recognition image set, the ORB feature point matching method is used to perform difference detection on each pair of segmented images in the secondary recognition image set, a mapping relationship is established between the obtained ORB difference degree and the secondary detection position identifier, and the secondary detection position identifiers are arranged in ascending order according to their corresponding ORB difference degrees to obtain a third detection sequence.

[0046] Specifically, based on the secondary recognition image set, the histogram method is used to detect the difference between paired segmented images. The distribution of pixel values in the image is statistically analyzed to characterize the image features. For paired segmented images, the degree of difference between the images can be detected by comparing their histograms. The histogram method can capture more information within the image by statistically analyzing the distribution of pixel values, thereby making up for the defect of the BMH hash algorithm in insufficient utilization of image details and internal structure information to a certain extent. The histogram distance reflects the degree of difference between paired images. The smaller the distance, the more similar the images are in terms of features. When calculating the histogram distance, the distance metric selected in this embodiment is the Bhattacharyya distance, and the formula is:

[0047]

[0048] Among them, H1 and H2 are the histograms of the target segmentation image and the current segmentation image, respectively. H1(i) and H2(i) represent the intervals of histogram H1 and H2 in the i-th histogram statistics, respectively. n represents the number of segmented images.

[0049] The Structural Similarity Index (SSIM) is a metric used to assess the visual dissimilarity between two images, taking into account the image's brightness, contrast, and structural information. In paired segmented image recognition, when images exhibit structural variations, changes in local detail, or changes in contrast, the BMH hashing algorithm and histogram method may not accurately identify image similarity. SSIM can more accurately measure image similarity, thus addressing the shortcomings of the BMH hashing algorithm and histogram method in these areas. The SSIM calculation formula is as follows:

[0050]

[0051] C1=(K1L) 2

[0052] C2=(K2L) 2

[0053] Among them, x and y are the target segmentation image and the current segmentation image, μ x and μ y are the average values of the target segmentation image and the current segmentation image, σ x and σ y are the standard deviations of the target segmentation image and the current segmentation image, σ xy is the covariance of x and y, C1 and C2 are small constants used to prevent the denominator from being zero, K1 and K2 are small constants, and L is the maximum pixel value of the segmented image. The SSIM value ranges from [-1, 1]. The closer the SSIM value is to 1, the higher the structural similarity between the two images. To facilitate subsequent image recognition, a difference metric is used to transform y = 1 - x, where y ∈ [0, 2]. The larger the SSIM value parameter after the difference metric transformation, the greater the difference between the paired images, forming the second detection sequence.

[0054] ORB (Oriented FAST and Rotated BRIEF) is a feature point matching method that combines the advantages of the FAST keypoint detector (Oriented FAST) and the BRIEF descriptor (Rotated BRIEF). It is particularly suitable for image matching applications that require fast processing. Some literature and tutorials refer to ORB as "Oriented FAST corner detection with Rotated BRIEF descriptor" or "Oriented FAST feature point detection with Rotated BRIEF descriptor." The ORB algorithm determines the degree of difference between paired images through the following steps: feature point extraction, feature descriptor generation, and feature matching. First, the FAST (Features from Accelerated Segment Test) algorithm is used to detect keypoints in the image. This algorithm identifies potential keypoints by performing a rapid intensity test on the pixels in the image. For each detected keypoint, ORB uses the BRIEF (Binary Robust Independent Elementary Features) algorithm for feature description. BRIEF generates a binary descriptor, which determines the binary code by comparing the brightness of pairs of pixels surrounding the keypoint. The Hamming distance is used to measure the difference between two feature descriptors. The larger the Hamming distance, the greater the difference between the two feature descriptors.

[0055] In summary, the BMH hashing algorithm is a relatively simple image difference measurement method. The histogram method offers a more refined image difference detection granularity compared to the BMH hashing algorithm. The histogram method generates one or more color histograms by statistically analyzing the pixel value distribution of each color channel (e.g., the red, green, and blue channels in the RGB space). These histograms reflect the color composition and hue distribution of the segmented image across each color channel, providing richer information. This not only preserves the color information of the original image but also better reflects the color features and visual effects within the image, resulting in a more accurate and comprehensive image description. Compared to the histogram method, which only describes the distribution of color intensity, SSIM focuses on the structural information of the image, going beyond the distribution of pixel values. It measures differences from the perspective of the image's physical structure, taking into account multiple aspects such as brightness, contrast, and structure, providing a more in-depth analysis. Compared to SSIM, ORB focuses on comparing overall structure, while SSIM delves deeper into local feature points and local feature descriptions, capturing more subtle local changes. This progresses from measuring overall structural differences to accurately capturing and describing local features. From the BMH hash algorithm to the histogram method, to SSIM, and finally to ORB image difference detection, each step adds new dimensions and depth to the previous one, making image analysis and matching more and more sophisticated and accurate.

[0056] Step S600: Taking the pre-check detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the first detection sequence, and obtain a first difference score; taking the first detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the second detection sequence, and obtain a second difference score; taking the second detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the third detection sequence, and obtain a third difference score.

[0057] Furthermore, the first difference score, the second difference score and the third difference score include:

[0058] Determine the number of parameters in the pre-detection sequence, the first detection sequence, the second detection sequence, and the third detection sequence as N, and assign a constraint coefficient w=1 / N to each detection sequence, where N is the number of results to be re-detected;

[0059] Taking the pre-flight detection sequence as a reference, traverse the secondary detection position identifiers, identify their ranking in the first detection sequence that is higher than their ranking in the pre-flight detection sequence, calculate the total number of steps from the lower ranking to the higher ranking, and combine the obtained total number of steps with the constraint coefficient to obtain a first difference score;

[0060] Using the first detection sequence as a reference, traverse the secondary detection position identifiers, identify their ranking in the second detection sequence as higher than their ranking in the first detection sequence, calculate the total number of steps from the lower ranking to the higher ranking, and combine the obtained total number of steps with the constraint coefficient to obtain a second difference score;

[0061] Taking the second detection sequence as a benchmark, traverse the secondary detection position identifiers, identify that their ranking in the third detection sequence is higher than their ranking in the second detection sequence, calculate the total number of steps from the low ranking to the high ranking, combine the obtained total number of steps with the constraint coefficient, and obtain a third difference score.

[0062] In an embodiment of the present application, three levels of difference scores are obtained through four different visual detection technologies: hash value, histogram method, SSIM index and ORB feature points. Each technology focuses on different image characteristics, such as overall similarity, color distribution, structural information and local features, thereby providing a comprehensive assessment of image differences. The calculation process from the first difference score to the third difference score is progressive, and the difference is quantified by calculating the number of steps of parameter sorting adjustment. It not only reflects the degree of difference between images, but also provides an intuitive way to measure the magnitude of the change, which helps to quickly locate paired segmented images with large image differences. The three levels of difference scores constitute a systematic evaluation framework, which enables different types of paired segmented images to be compared on the same scale, which provides a guarantee for subsequent comprehensive data analysis and refresh strategy formulation. Specifically, the number of parameters in the pre-detection detection sequence, the first detection sequence, the second detection sequence and the third detection sequence is first determined to be N, and the number of parameters in each detection sequence is consistent with the number of results to be detected twice. The number of results awaiting secondary inspection is affected by factors such as the diversity of the target image and the currently displayed image. This means there's no guarantee that the number of results awaiting secondary inspection will be consistent during each image refresh. Furthermore, the number of parameters in the pre-inspection sequence and the subsequent multidimensional inspection sequence is related to the number of results awaiting secondary inspection. This results in a lack of comparability in the difference scores across different batches of image refreshes, making it difficult to use a unified standard to reflect image differences. For example, during the current image refresh, each inspection sequence has five parameters arranged in ascending order; during the previous image refresh, each inspection sequence had seven parameters arranged in ascending order. When calculating the number of steps required to adjust a parameter from a low ranking to a high ranking, the number of steps required to adjust a parameter from the lowest ranking to the highest ranking is four for the current image refresh, while it was six for the previous image refresh. This demonstrates the impact of the number of parameters on the difference score. To ensure that the subsequent comprehensive difference scores and thresholds accurately reflect image differences, a constraint coefficient is applied to the total number of steps obtained by comparing the pre-test sequence with the first test sequence, the total number of steps obtained by comparing the first test sequence with the second test sequence, and the total number of steps obtained by comparing the second test sequence with the third test sequence. The constraint coefficient is S' = S × 1 / N, where S' represents the difference score, S represents the total number of steps, and 1 / N is the constraint coefficient. Applying the constraint coefficient eliminates the influence of the number of parameters on the multi-level difference score, making the multi-level difference scoring system more adaptable. Regardless of changes in external factors, the consistency and rationality of the scores are maintained to the greatest extent possible, thus establishing a stable image difference assessment mechanism.

[0063] Step S700: Synchronize the first difference score, the second difference score and the third difference score to a difference analysis model to obtain a difference analysis value of the secondary recognition image set, wherein the difference analysis model is based on a multivariate regression model, with three difference scores as input and a difference analysis value as output.

[0064] Furthermore, the difference analysis model includes:

[0065] Difference analysis model, including:

[0066] Obtain sample images from multiple sources, randomly form sample paired images in pairs, input the sample images into the image detection model, and obtain a sample secondary recognition image set after pre-checking;

[0067] Obtaining a sample first difference score, a sample second difference score, and a sample third difference score of the sample paired segmented images in the sample secondary recognition image set, as well as a sample actual difference value of the sample secondary recognition image set, wherein the sample actual difference value is obtained by manual annotation;

[0068] A multiple regression model is used as the framework, with the first difference score, second difference score, and third difference score of the sample as inputs and the sample difference analysis value of the sample secondary recognition image set as output to construct a difference analysis model. The multiple regression model is as follows:

[0069]

[0070] in, is the sample difference analysis value of the sample secondary recognition image set, β0 is the intercept term, β1, β2, and β3 are the sample regression coefficients corresponding to the first, second, and third difference scores of the sample, respectively, and ε is the error term;

[0071] The optimization is performed in the sample multidimensional coefficient space so that the sum of square errors between the sample difference analysis value of the sample secondary recognition image set output by the multivariate regression model and the actual sample difference value is minimized. The multivariate regression model obtained by the optimization is a difference analysis model, wherein the sample multidimensional coefficients include β0, β1, β2 and β3, and the formula for the sum of square errors is as follows:

[0072]

[0073] Where S is with y i The sum of squared errors between them, n is the number of sample paired segmentation images, y i is the actual difference value of the sample of the i-th sample paired segmentation image, is the sample difference analysis value of the i-th sample paired segmentation image.

[0074] In an embodiment of the present application, a difference analysis model is constructed to improve the generalization and accuracy of image difference judgment. Specifically, sample images are acquired from multiple sources, ranging from highly similar to significantly different, including natural images, text images, and graphic images. These images are then jittered, varying in color from black and white to six colors. The diversity and variability of the sample images ensure that the model can be trained and tested in a wide range of scenarios, improving its generalization and accuracy. Sample paired images are randomly formed into pairs, and one of the sample paired images is selected as the sample target image. The sample paired images are pre-inspected to obtain N secondary recognition results for the sample paired images. Based on the N secondary recognition results for the sample paired images, a sample pre-inspection detection sequence and a sample secondary recognition image set are constructed. The sample paired segmented images in the sample secondary recognition image set are subjected to diversified visual inspection using the histogram method, structural similarity index, and feature point matching method to obtain the sample first detection sequence, sample second detection sequence, and sample third detection sequence. Difference analysis is then performed to obtain the sample first difference score, sample second difference score, and sample third difference score. The actual difference values of the N sample paired segmented images in the sample secondary recognition image set are obtained through manual annotation methods. The actual sample difference values of the N sample paired segmented images are summed to obtain the actual sample difference values of the sample secondary recognition image set. The manual annotation method needs to determine the scoring dimensions for evaluating the differences between the sample segmented images. Experts or trained annotation personnel can be selected to use special annotation tools to annotate, such as color, texture, shape and content dimensional differences, establish corresponding difference scoring rules, select a suitable scoring scale for each dimension, which can be a continuous value or a discrete level, introduce a weighting mechanism, and obtain the actual sample difference value of each sample paired segmented image in the sample secondary recognition image set. Then construct a multivariate regression model as follows:

[0075]

[0076] in, is the sample difference analysis value of the sample secondary recognition image set, β0 is the intercept term, β 1、 β 2、 β3 are the sample regression coefficients corresponding to the first, second and third sample difference scores, and ε is the error term. The multivariate regression model is trained using the sample paired segmentation images of the sample image. The sample regression coefficients β0, β 1、 β2 and β3 are optimized in the sample multidimensional coefficient space so that the sample difference analysis value of the sample secondary recognition image set output by the model is The sum of squared errors between the actual difference values y of the sample in the sample secondary recognition image set is minimized, that is, the sum of squared vertical distances is minimized. The sum of squared errors is as follows:

[0077]

[0078] Where S is with y i The sum of squared errors between them, n is the number of sample paired segmentation images, y i is the actual difference value of the sample of the i-th sample paired segmentation image, Is the sample difference analysis value of the i-th sample paired segmentation image. 80% of the sample paired segmentation images are used for training set and 20% for validation set. The sample regression coefficient is estimated using the least squares method on the training set. The sample regression coefficient includes β0, β 1、 β2 and β3, the sample difference analysis values of the sample secondary recognition image set The sum of squared errors between the actual difference value y and the sample is minimized, and the least squares method is used on the validation set to find the optimal sample regression coefficients β0, β 1、 β 2、 β3, the sample difference analysis value of the sample secondary recognition image set The sum of squared errors between the actual difference value y of the sample is minimized to verify the model performance, converge the model, and obtain the difference analysis model.

[0079] S800: Determine if the difference analysis value is greater than a preset second threshold, remove duplicate images, and obtain P secondary detection results, where P≤N.

[0080] Obtain P second detection results, including:

[0081] Obtaining a difference analysis value of the secondary recognition image set;

[0082] If it is determined that the difference analysis value of the secondary recognition image set is greater than a preset second threshold, the target segmentation images corresponding to the secondary detection position identifiers adjusted from low ranking to high ranking in the first inspection sequence, the second inspection sequence and the third inspection sequence during the three difference scoring processes are extracted, the duplicate images are eliminated, and the retained target segmentation images are added to the second image update list to obtain P second detection results.

[0083] Specifically, a second threshold is preset. When the difference analysis value of the secondary recognition image set is greater than the preset second threshold, it means that the N paired segmented images in the secondary recognition image set have a high degree of difference as a whole. The target segmented images corresponding to the secondary detection position identifiers adjusted from low ranking to high ranking in the first inspection sequence, the second inspection sequence, and the third inspection sequence are extracted. There may be some duplicate images in the extracted target segmented images. Removing these duplicate images can avoid redundant information. The retained target segmented images are added to the second image update list to obtain P second detection results. The second detection results include P target segmented images and their image position identifiers, indicating that after multivariate visual detection, P target images with update value are screened out from the N paired segmented images in the secondary recognition image set, and P≤N.

[0084] Step S900: Determine whether the result of M+P is less than a preset third threshold.

[0085] If so, the master control chip generates a local refresh drive instruction, which drives the display driver chip to convert the control signal of the instruction into a corresponding voltage waveform, and drives the electronic paper display to perform a local image refresh according to the voltage waveform corresponding to the instruction to display the target image;

[0086] If not, the driver control chip generates a global refresh drive instruction, drives the display driver chip to convert the control signal of the instruction into a corresponding voltage waveform, drives the electronic paper display to perform global image refresh according to the voltage waveform corresponding to the instruction, and displays the target image.

[0087] In this embodiment, a third threshold is preset, which is the number K of target segmented images obtained by the image segmentation unit. The entire method for driving the image refresh of the electronic display device determines different refresh strategies based on the comparison of the M+P result with the third threshold. A determination is made as to whether the M+P result is less than the third threshold. If so, the current state is suitable for a local refresh strategy. The driver control chip generates a first drive instruction, indicating that the current image-related state has been detected to be suitable for local processing relative to the segmentation of the target image. The display driver chip converts the control signal of the first drive instruction into a corresponding voltage waveform, customized according to the local refresh strategy and affecting only the area on the electronic paper display requiring local update. The second list of images and their position identifiers are updated locally based on the first update list and image update, and the corresponding position of the currently displayed image is updated. If not, the current state is suitable for a global refresh strategy. The driver control chip generates a second drive instruction, indicating that the current image-related state has been detected to require comprehensive and holistic processing relative to the segmentation of the target image. The display driver chip converts the control signal of the second drive instruction into a corresponding voltage waveform, designed for a global refresh strategy, and drives the electronic paper display to refresh the image according to the global refresh strategy corresponding to the second drive instruction.

[0088] In summary, the method for driving an electronic display device to refresh an image provided by the embodiments of the present application has the following technical effects:

[0089] 1. Determine whether the image to be displayed is a non-black and white image and obtain the target image, which can avoid unnecessary image difference judgment, divert the execution of the refresh method of the image driving the electronic display device, and improve the overall refresh efficiency. The target image is input into the image detection model to obtain K target segmentation images, perform pre-inspection, obtain M pre-inspection results and N secondary detection results, and quickly identify the target segmentation images that need to be subjected to multivariate visual inspection, thereby improving the accuracy of image detection, reducing the amount of calculation of the system, and improving the overall refresh efficiency.

[0090] 2. Obtaining the first detection sequence, the second detection sequence, and the third detection sequence can more comprehensively capture the differences between paired segmented images, thereby enhancing the comprehensiveness of the detection;

[0091] 3. Obtaining the first difference score, the second difference score, and the third difference score reflects the degree of difference between the paired segmented images under different visual detection technologies, which helps to comprehensively evaluate the overall difference between the paired segmented images in the secondary recognition image set;

[0092] 4. Through the difference analysis model, a multivariate regression comprehensive judgment is performed on the first difference score, the second difference score, and the third difference score. The scattered difference scores are combined to obtain the difference analysis value of the secondary recognition image set, which improves the judgment accuracy of the paired segmented images in the secondary recognition image set;

[0093] 5. A difference analysis value obtained by the difference analysis model is greater than a preset second threshold value, and P second detection results are obtained. It is judged whether the result of M+P is less than a third threshold value. The master control chip is driven to generate a corresponding drive instruction based on the judgment result, and the display driver chip is driven to convert the control signal of the corresponding drive instruction into a corresponding voltage waveform, and the electronic paper display is driven to refresh the image according to the waveform corresponding to the corresponding drive instruction, update the displayed image, minimize unnecessary refresh operations, reduce energy consumption, improve the overall refresh efficiency, and enhance the user experience.

[0094] In one embodiment, a system for driving an image refresh of an electronic display device is provided, the internal structure of which is Figure 2 As shown. The drive refresh system includes a main control chip module, a display driver chip module, and a storage chip module. The main control chip module, display driver chip module, and storage chip module are electrically connected via a communication line. The storage chip module of the drive refresh system is externally connected to the main control chip module and is used to store a foreground image set, receive images from the main control chip, and assist the main control chip in obtaining the current display image in the foreground image set. The main control chip module of the drive refresh system is used to receive the current display image stored in the storage chip module, receive the image to be displayed, and generate corresponding drive instructions by comparing the target image with the current display image to determine the difference area. The display driver chip module of the drive refresh system converts the control signal of the corresponding drive instruction into a corresponding voltage waveform, and drives the corresponding area of the electronic paper display screen to refresh according to the corresponding voltage waveform. The drive refresh system executes a method for driving the image refresh of an electronic display device.

[0095] Those skilled in the art will understand that Figure 2 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0096] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for driving an image refresh of an electronic display device, characterized in that: The electronic display device includes an electronic paper display screen, and the method includes: Acquire an image to be displayed, determine whether the image to be displayed is a non-black and white image, and obtain a target image, wherein the image to be displayed is a static image formed by a combination of text and / or pictures; Input the target image into the image segmentation unit of the image detection model to obtain K target segmentation images and image position identifiers, input the K target segmentation images into the hash algorithm detection unit of the image detection model, and obtain hash values of the K target segmentation images through the BMH hash algorithm; Traversing the image position identifiers, calculating the Hamming distance between the target segmented image hash value and the current segmented image hash value under each image position identifier, mapping the Hamming distance to the image position identifier one by one, performing a pre-inspection on the Hamming distance using a preset first threshold, and obtaining M pre-inspection results and N secondary inspection results, where M and N are integers ≥ 0, and M + N = K; The N image position identifiers of the results to be secondary detected are defined as secondary detection position identifiers, the secondary detection position identifiers are arranged in ascending order according to their corresponding Hamming distances to obtain a pre-detection sequence, and the current segmented image under the secondary detection position identifier is combined with the results to be secondary detected to form a paired segmented image to obtain a secondary recognition image set; Performing difference detection on each pair of segmented images in the secondary recognition image set by using a histogram method, a structural similarity index, and a feature point matching method, respectively, and arranging the secondary detection position identifiers in ascending order according to their corresponding detection results to obtain a first detection sequence, a second detection sequence, and a third detection sequence, respectively; Taking the pre-flight detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the first detection sequence, and obtain a first difference score; taking the first detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the second detection sequence, and obtain a second difference score; taking the second detection sequence as a benchmark, traverse the secondary detection position identifier, calculate the total number of steps of its ranking improvement in the third detection sequence, and obtain a third difference score; Synchronizing the first difference score, the second difference score, and the third difference score to a difference analysis model to obtain a difference analysis value of the secondary recognition image set, wherein the difference analysis model is based on a multivariate regression model, takes the three difference scores as input, and outputs the difference analysis value; Determine that the difference analysis value is greater than a preset second threshold, eliminate duplicate images, and obtain P secondary detection results, where P≤N; Determine whether the result of M+P is less than a preset third threshold, If so, the master control chip generates a local refresh drive instruction, which drives the display driver chip to convert the control signal of the instruction into a corresponding voltage waveform, and drives the electronic paper display to perform a local image refresh according to the voltage waveform corresponding to the instruction to display the target image; If not, the driver control chip generates a global refresh drive instruction, drives the display driver chip to convert the control signal of the instruction into a corresponding voltage waveform, drives the electronic paper display to perform global image refresh according to the voltage waveform corresponding to the instruction, and displays the target image.

2. The method for driving an image refresh of an electronic display device according to claim 1, wherein: Obtain K target segmentation images and image location identifiers, including: Obtaining a sample image, determining the number of pixels of a segmented image of the sample image based on the total number of pixels of the electronic paper display and a specific application scenario, calculating a ratio of the total number of pixels of the electronic paper display to the number of pixels of the segmented image of the sample image, and obtaining a number K of sample image segmentations; Determine whether K is a perfect square number, If so, the sample target image is divided into The grid, If not, select the two closest factors r and c so that r×c=K, and divide the sample image into r×c grids; Mark each grid with a location identifier to obtain an image segmentation unit; The target image is input into an image segmentation unit to obtain K target segmented images and K image position identifiers.

3. The method for driving an image refresh of an electronic display device according to claim 1, wherein: Before inputting the target image into the image segmentation unit of the image detection model, the method includes: K segmented images of the current image displayed on the electronic paper display and their corresponding image position identifiers and hash values are extracted from the resident foreground image set, and the image position identifiers of the K current segmented images and their corresponding hash values are input into the hash algorithm detection unit.

4. The method for driving an image refresh of an electronic display device according to claim 3, wherein: The Hamming distance is pre-checked using a preset first threshold, and M pre-check results and N secondary check results are obtained, including: Input the K target segmentation images into a hash algorithm detection unit, obtain a hash value of each target segmentation image through a BMH hash algorithm, and store the K target segmentation images with the image position identifiers and their corresponding hash values in a foreground image set; Traversing the image position identifiers, calculating the Hamming distance between the target segmented image hash value and the current segmented image hash value under the same image position identifier, and establishing a mapping relationship between the obtained Hamming distance and the corresponding position identifier; A first threshold is preset, and the target segmentation images corresponding to the Hamming distance greater than the preset first threshold are added to the first image update list to obtain M preliminary inspection results. The target segmentation images corresponding to the Hamming distance less than the preset first threshold are N results to be tested again.

5. The method for driving an electronic display device to refresh an image according to claim 1, wherein: Obtaining a first detection sequence, a second detection sequence, and a third detection sequence respectively, including: Based on the secondary recognition image set, using the histogram method to perform difference detection on each pair of segmented images in the secondary recognition image set, establishing a mapping relationship between the obtained histogram distance and the secondary detection position identifier, and arranging the secondary detection position identifiers in ascending order according to their corresponding histogram distances to obtain a first detection sequence; Based on the secondary recognition image set, using the structural similarity index method to perform difference detection on each pair of segmented images in the secondary recognition image set, performing difference metric conversion on the obtained structural similarity index, wherein the converted structural similarity index ranges from [0, 2], establishing a mapping relationship between the converted structural similarity index and the secondary detection position identifier, and arranging the secondary detection position identifiers in ascending order according to their corresponding converted structural similarity indexes to obtain a second detection sequence; Based on the secondary recognition image set, the ORB feature point matching method is used to perform difference detection on each pair of segmented images in the secondary recognition image set, a mapping relationship is established between the obtained ORB difference degree and the secondary detection position identifier, and the secondary detection position identifiers are arranged in ascending order according to their corresponding ORB difference degrees to obtain a third detection sequence.

6. The method for driving an image refresh of an electronic display device according to claim 1, wherein: The first difference score, the second difference score and the third difference score include: Determine the number of parameters in the pre-detection sequence, the first detection sequence, the second detection sequence, and the third detection sequence as N, and assign a constraint coefficient w=1 / N to each detection sequence, where N is the number of results to be re-detected; Taking the pre-flight detection sequence as a reference, traverse the secondary detection position identifiers, identify their ranking in the first detection sequence that is higher than their ranking in the pre-flight detection sequence, calculate the total number of steps from the lower ranking to the higher ranking, and combine the obtained total number of steps with the constraint coefficient to obtain a first difference score; Using the first detection sequence as a benchmark, traverse the secondary detection position identifiers, identify their ranking in the second detection sequence as higher than their ranking in the first detection sequence, calculate the total number of steps from the lower ranking to the higher ranking, and combine the obtained total number of steps with the constraint coefficient to obtain a second difference score; Taking the second detection sequence as a benchmark, traverse the secondary detection position identifiers, identify that their ranking in the third detection sequence is higher than their ranking in the second detection sequence, calculate the total number of steps from the low ranking to the high ranking, combine the obtained total number of steps with the constraint coefficient, and obtain a third difference score.

7. The method for driving an image refresh of an electronic display device according to claim 1, wherein: Difference analysis model, including: Obtain sample images from multiple sources, randomly form sample paired images in pairs, input the sample images into the image detection model, and obtain a sample secondary recognition image set after pre-checking; Obtaining a sample first difference score, a sample second difference score, and a sample third difference score of the sample paired segmented images in the sample secondary recognition image set, as well as a sample actual difference value of the sample secondary recognition image set, wherein the sample actual difference value is obtained by manual annotation; A multiple regression model is used as the framework, with the first difference score, second difference score, and third difference score of the sample as inputs and the sample difference analysis value of the sample secondary recognition image set as output to construct a difference analysis model. The multiple regression model is as follows: in, is the sample difference analysis value of the sample secondary recognition image set, β0 is the intercept term, β1, β2, and β3 are the sample regression coefficients corresponding to the first, second, and third difference scores of the sample, respectively, and ε is the error term; The optimization is performed in the sample multidimensional coefficient space so that the sum of square errors between the sample difference analysis value of the sample secondary recognition image set output by the multivariate regression model and the actual sample difference value is minimized. The multivariate regression model obtained by the optimization is a difference analysis model, wherein the sample multidimensional coefficients include β0, β1, β2 and β3, and the formula for the sum of square errors is as follows: Where S is with y i The sum of squared errors between them, n is the number of sample paired segmentation images, y i is the actual difference value of the sample of the i-th sample paired segmentation image, is the sample difference analysis value of the i-th sample paired segmentation image.

8. The method for driving an image refresh of an electronic display device according to claim 1, wherein: Obtain P secondary detection results, including: Obtaining a difference analysis value of the secondary recognition image set; If it is determined that the difference analysis value of the secondary recognition image set is greater than a preset second threshold, the target segmentation images corresponding to the secondary detection position identifiers adjusted from low ranking to high ranking in the first detection sequence, the second detection sequence, and the third detection sequence during the three difference scoring processes are extracted, the duplicate images are eliminated, and the retained target segmentation images are added to the second image update list to obtain P secondary detection results.

Images