Object-free image library optimization method and apparatus, electronic device, and storage medium

Abstract

Embodiments of the present application provide an object-free image library optimization method and apparatus, an electronic device, and a storage medium. The object-free image library optimization method comprises: determining a first object-free image and a plurality of second object-free images from an object-free image library; performing difference calculation on the first object-free image and each of the plurality of second object-free images, and obtaining a plurality of residual images on the basis of the calculation result; and when it is determined, on the basis of the plurality of residual images, that background texture interference has occurred in the first object-free image, deleting the first object-free image from the object-free image library. The object-free image library optimization method provided by the present application can improve the accuracy of fingerprint identification by fingerprint sensors.

Description

[0001] This application claims priority to patent application number 202410177012.5, filed February 8, 2024, entitled "Method, Apparatus, Electronic Device, and Storage Medium for Optimizing an Empty Image Library," the entire contents of which are incorporated herein by reference. Technical Field: Embodiments of this application relate to the field of data processing technology, and more particularly to a method, apparatus, electronic device, and storage medium for optimizing an empty image library. Background: With the development of the consumer electronics industry, finger-level recognition has become a mainstream security solution for electronic devices. When finger-level sensors capture finger-level images, fixed signal interference is present. Therefore, an empty image library needs to be created to cache real-time empty images for subsequent correction of captured finger-level images to remove fixed signal interference. However, when finger-level sensors capture empty images, low-level interference may be present in the empty images due to interference from foreign objects or other conditions. Currently, finger-level sensors capture air-sampled images, perform low-level interference detection on them, and then store them in an air-sampled image library. However, when performing low-level interference detection on air-sampled images, it is impossible to filter out all air-sampled images containing low-level interference. As a result, the air-sampled image library contains air-sampled images with low-level interference. This can cause the corrected finger-level images to be missing when the captured finger-level images are corrected, resulting in a low recognition rate for the finger-level sensor. In view of this, embodiments of the present application provide a method, apparatus, electronic device, and computer storage medium for optimizing an air-sampled image library to at least partially address the aforementioned issues. According to a first aspect of an embodiment of the present application, a method for optimizing a database of uncollected images is provided, comprising: determining a first uncollected image and multiple second uncollected images from the database of uncollected images, wherein the database of uncollected images includes multiple uncollected images, each of which is an image captured by a finger-level sensor when no detection object is present within a detection area, and the first uncollected image and the second uncollected images are different; performing difference calculations on the first uncollected image and the multiple second uncollected images, and obtaining multiple residual images based on the calculation results; and deleting the first uncollected image from the database of uncollected images when it is determined, based on the multiple residual images, that the first uncollected image has low-level interference. In one possible implementation, determining the first uncollected image and the multiple second uncollected images from the database of uncollected images includes: performing signal interference detection on the multiple uncollected images included in the database of uncollected images, determining an uncollected image whose signal interference detection result is greater than a signal interference threshold as the first uncollected image, and determining an uncollected image whose signal detection result is less than or equal to the signal interference threshold as the second uncollected image.In one possible implementation, performing difference calculations on the first unsampled image and the plurality of second unsampled images, and obtaining a plurality of residual images based on the calculation results, includes: performing difference calculations on the first unsampled image and the plurality of second unsampled images to obtain a plurality of first images, and determining each of the plurality of first images as the residual image; or selecting a plurality of second images from the plurality of first images, wherein the intensity or number of bar-level signals included in the second images is less than a preset noise threshold, and determining each of the plurality of second images as the residual image; or obtaining the plurality of residual images based on the plurality of second images. In one possible implementation, selecting a plurality of second images from the plurality of first images includes: calculating a mean pixel value of each column of pixels in the first image to obtain a column pixel mean corresponding to the first image; calculating a mean pixel value of each row of pixels in the first image to obtain a row pixel mean corresponding to the first image; and determining the first image in the plurality of first images for which both the column pixel mean and the row pixel mean are less than the noise threshold as the second image. In one possible implementation, obtaining the multiple residual images based on the second image includes: performing image denoising on each of the multiple second images to obtain the multiple residual images. In one possible implementation, the method further includes: normalizing the multiple residual images; performing feature extraction on each of the normalized residual images to obtain features corresponding to each residual image; predicting each residual image based on the features corresponding to the residual image using at least two fully connected layers to obtain a probability of the presence of bottom-level interference in each residual image; and performing bottom-level interference prediction on the first blank-sampled image based on the probability of the presence of bottom-level interference in each of the multiple residual images to obtain a prediction result. In one possible implementation, the bottom-level interference prediction is performed on the first air-sampled image based on the probability that each residual image in the multiple residual images has bottom-level interference, and a prediction result is obtained, including: calculating the average value of the probabilities that the multiple residual images have bottom-level interference; calculating the standard deviation of the probabilities that the multiple residual images have bottom-level interference based on the average value; determining the probability that the first air-sampled image has bottom-level interference based on the average value and the standard deviation; if the probability that the first air-sampled image has bottom-level interference is greater than a preset probability threshold, determining the first air-sampled image as having bottom-level interference.In one possible implementation, determining the probability that the first unsampled image contains underlying interference based on the mean value and the standard deviation includes: performing a weighted summation of the mean value and the standard deviation, and determining the weighted summation result as the probability that the first unsampled image contains underlying interference. In one possible implementation, when the number of unsampled images stored in the unsampled image library equals a preset image quantity threshold, and the unsampled image library receives at least one third unsampled image, the probabilities of the first unsampled images containing underlying interference are sorted from highest to lowest, the first unsampled images corresponding to the top N probabilities in the sorting are deleted, and the at least one third unsampled image is stored in the unsampled image library, where N is a positive integer greater than or equal to 1 and equal to the number of third unsampled images. In one possible implementation, the method further includes: inputting the multiple residual images into a pre-trained underlying interference detection network to obtain a prediction result output by the underlying interference detection network regarding whether the first unsampled image contains underlying interference. According to a second aspect of an embodiment of the present application, a device for optimizing a collection of uncollected images is provided, comprising: a determination unit configured to determine a first uncollected image and multiple second uncollected images from the collection of uncollected images, wherein the collection of uncollected images comprises multiple uncollected images, each of which is an image captured by a finger-level sensor when no detection object is present within a detection area, and the first uncollected image and the second uncollected images are different; a calculation unit configured to perform difference calculations between the first uncollected image and the multiple second uncollected images, and obtain multiple residual images based on the calculation results; and a deletion unit configured to delete the first uncollected image from the collection of uncollected images when it is determined, based on the multiple residual images, that the first uncollected image contains bottom-level interference. According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; and the memory configured to store at least one executable instruction, wherein the executable instruction causes the processor to perform an operation corresponding to the method described in the first aspect. According to a fourth aspect of the embodiments of the present application, a computer storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method described in the first aspect is implemented.According to the optimization scheme for the empty-sampled image library provided in the embodiments of the present application, a first empty-sampled image and multiple second empty-sampled images are determined in the empty-sampled image library, and the difference between the first empty-sampled image and each second empty-sampled image is calculated to obtain multiple residual images. These multiple residual images are then predicted. This allows determining whether the first empty-sampled image has low-level interference. Low-level interference detection can be performed on the empty-sampled images in the empty-sampled image library, and the first empty-sampled image with low-level interference is removed from the empty-sampled image library. This ensures that the empty-sampled images in the empty-sampled image library are free of low-level interference. After the finger-level sensor acquires the finger-level image, it can use the empty-sampled images free of low-level interference to correct the finger-level image, ensuring the accuracy of the finger-level image and improving the recognition accuracy of the finger-level sensor. BRIEF DESCRIPTION OF THE DRAWINGS To more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly describes the drawings required in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some of the embodiments described in the embodiments of the present application, and those skilled in the art can also derive other drawings based on these drawings. Figure 1 is a flowchart of a method for optimizing a vacant image library provided in an embodiment of the present application; Figure 2 is a flowchart of a method for determining prediction results provided in an embodiment of the present application; Figure 3 is a schematic diagram of an apparatus for optimizing a vacant image library provided in an embodiment of the present application; and Figure 4 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS To help those skilled in the art better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only a portion of the embodiments of the present application, and not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present application should fall within the scope of protection of the embodiments of the present application. The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It should be understood that although the terms first, second, third, etc. may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other.For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the term "if" as used herein may be interpreted as "at the time of," "when," or "in response to a determination." As previously mentioned, with the development of the consumer electronics industry, finger-level recognition has become a mainstream security solution for electronic devices. Because finger-level sensors may experience fixed signal interference when capturing finger-level images, a database of unsampled images is required to cache real-time unsampled images for subsequent correction of captured finger-level images to remove fixed signal interference. However, when capturing unsampled images, finger-level sensors may also experience low-level interference due to interference from foreign objects or other conditions. For example, if there is dirt or an object in the acquisition area of ​​an ultrasonic finger-level sensor, low-level interference may appear in the unsampled image. Currently, air-sampled images are collected by finger-level sensors, and after bottom-level interference detection is performed on the air-sampled images, the air-sampled images are stored in an air-sampled image library. However, when bottom-level interference detection is performed on the air-sampled images, it is impossible to filter out all air-sampled images containing bottom-level interference, resulting in the air-sampled image library including air-sampled images with bottom-level interference. As a result, when the collected finger-level images are corrected, the corrected finger-level images are missing, resulting in a low recognition rate of the finger-level sensor. An embodiment of the present application provides a method for optimizing an empty-sampled image library. A first empty-sampled image and multiple second empty-sampled images are determined in the empty-sampled image library. The difference between the first empty-sampled image and each second empty-sampled image is calculated to obtain multiple residual images. These multiple residual images are predicted. This method can determine whether the first empty-sampled image has low-level interference. Low-level interference detection can be performed on the empty-sampled images in the empty-sampled image library, and the first empty-sampled image with low-level interference can be deleted from the empty-sampled image library. This ensures that the empty-sampled images in the empty-sampled image library are free of low-level interference. After a finger-level sensor acquires a finger-level image, it can use the empty-sampled images without low-level interference to correct the finger-level image, thereby ensuring the accuracy of the finger-level image and improving the recognition accuracy of the finger-level sensor. The following examples illustrate the method for optimizing the empty-sampled image library provided by the present application. FIG1 is a flow chart of a method for optimizing an air-collected image library provided in an embodiment of the present application. As shown in FIG1 , the method for optimizing the air-collected image library includes the following steps 101 to 103: Step 101: Determine a first air-collected image and multiple second air-collected images from the air-collected image library.The empty-sampled image library includes multiple empty-sampled images. Empty-sampled images are images captured by the finger-level sensor when there is no detection object within the detection area. For example, when there is no finger, the first empty-sampled image and the second empty-sampled image captured by the finger-level sensor are different. The first empty-sampled image and the second empty-sampled image are determined from the empty-sampled image library of the finger-level sensor. The first empty-sampled image is an empty-sampled image in the empty-sampled image library, and the second empty-sampled image is at least a portion of the empty-sampled images in the empty-sampled image library excluding the first empty-sampled image. For example, if there are 50 empty-sampled images in the empty-sampled image library, the second empty-sampled image can be the 49 empty-sampled images excluding the first empty-sampled image, or 24 random images from the 49 empty-sampled images excluding the first empty-sampled image. Step 102: Perform difference calculations on the first empty-sampled image and the multiple second empty-sampled images, and obtain multiple residual images based on the calculation results. The first air-sampled image is subtracted from each of the second air-sampled images. In one example, based on each of the second air-sampled images, images included in the first air-sampled image that are identical to the second air-sampled image can be removed. Specifically, images included in the first air-sampled image and in each of the second air-sampled images that are different from the first air-sampled image are determined. After calculating the difference between the first air-sampled image and each of the second air-sampled images, multiple residual images are determined based on the calculation results. It should be understood that the number of the multiple residual images is not fixed. For example, multiple images after the difference calculation can be determined as the multiple residual images. For example, if the difference between the first air-sampled image and five second air-sampled images is calculated, the number of residual images can be five. The multiple residual images can also be images obtained by processing the calculation results and including at least a portion of the calculation results, which is not limited in this embodiment of the present application. Step 103: If it is determined based on the multiple residual images that the first air-sampled image has bottom-level interference, the first air-sampled image is deleted from the air-sampled image library. The first uncollected image is predicted using multiple residual images to determine whether the first uncollected image has underlying interference, thereby obtaining a prediction result. It should be understood that the prediction result can be a binary result, i.e., a prediction result indicating whether the first uncollected image has underlying interference, or a prediction result indicating a probability of the first uncollected image having underlying interference. The specific form of the prediction result is not limited in this embodiment of the present application. If the first uncollected image has underlying interference, the first uncollected image is deleted from the uncollected image library. If the first uncollected image does not have underlying interference, the first uncollected image is retained in the uncollected image library. It should be understood that the first uncollected image without underlying interference can be stored in an uncollected image sub-library, and when the uncollected image library is updated, the uncollected image library can be updated to the uncollected image sub-library. Alternatively, the first uncollected image with underlying interference can be directly deleted from the uncollected image library.It should be noted that execution conditions may be set for steps 101 to 103. Once the execution conditions are met, steps 101 to 103 are executed. The execution conditions include, but are not limited to, a fixed period, the number of newly added uncollected images in the uncollected image library meeting a threshold, and the like. In an embodiment of the present application, a first uncollected image and multiple second uncollected images are determined in the uncollected image library, and the difference between the first uncollected image and each of the second uncollected images is calculated to obtain multiple residual images. These multiple residual images are then predicted to determine whether the first uncollected image contains low-level interference. Low-level interference detection can then be performed on the uncollected images in the uncollected image library, and the first uncollected image containing low-level interference can be deleted from the uncollected image library. This ensures that the uncollected images in the uncollected image library are free of low-level interference. After the finger-level sensor acquires the finger-level image, it can use the uncollected images free of low-level interference to correct the finger-level image, thereby ensuring the accuracy of the finger-level image and improving the recognition accuracy of the finger-level sensor. In one possible implementation, when determining a first unsampled image and multiple second unsampled images from an unsampled image library, a signal interference check can be performed on the multiple unsampled images included in the unsampled image library. Unsampled images whose signal interference check results are greater than a signal interference threshold are determined as first unsampled images, and unsampled images whose signal interference check results are less than or equal to the signal interference threshold are determined as second unsampled images. It should be understood that if low-level interference exists in an unsampled image, the intensity or amount of signal interference in that unsampled image is higher than that in an unsampled image without low-level interference. Therefore, a signal interference check can be performed on each unsampled image in the unsampled image library. Unsampled images whose signal interference intensity or amount is greater than the signal interference threshold are selected as first unsampled images, and unsampled images whose signal interference intensity or amount is less than or equal to the signal interference threshold are selected as second unsampled images.In an embodiment of the present application, signal interference detection is performed on multiple air-sampled images included in an air-sampled image library, and an air-sampled image with a signal interference detection result greater than a signal interference threshold is determined as a first air-sampled image. As a result, an air-sampled image with a high probability of bottom-level interference can be used as the first air-sampled image for bottom-level interference detection. Since signal interference detection on air-sampled images requires less computing power and time than bottom-level interference detection on air-sampled images, air-sampled images with a high probability of bottom-level interference are used as the first air-sampled image for bottom-level interference detection. Compared with using all air-sampled images as the first air-sampled image for bottom-level interference detection, fewer first air-sampled images require bottom-level interference detection, and there is no need to use all air-sampled images in the air-sampled image library as the first air-sampled image for bottom-level interference detection. Therefore, the efficiency of bottom-level interference detection on air-sampled images in the air-sampled image library can be improved. In one possible implementation, when performing difference calculations on a first air-sampled image and a plurality of second air-sampled images, and obtaining a plurality of residual images based on the calculation results, the difference calculations can be performed on the first air-sampled image and the plurality of second air-sampled images to obtain a plurality of first images, and each of the plurality of first images can be determined as a residual image. Alternatively, a plurality of second images can be selected from the plurality of first images, wherein the intensity or number of bar-level signals included in the second images is less than a preset noise threshold, and each of the plurality of second images can be determined as a residual image. Alternatively, the plurality of residual images can be obtained based on the plurality of second images. The plurality of first images can be determined by performing difference calculations on the first air-sampled image and each of the second air-sampled images, where the number of first images is the same as the number of second air-sampled images. For example, the plurality of first images can be determined by performing difference calculations on the first air-sampled image and five second air-sampled images. In an optional embodiment, the first image can be determined as a residual image. The purpose of performing difference calculation on the first air-sampled image and multiple second air-sampled images is to eliminate fixed signal interference. Determining the multiple first images obtained by the difference calculation as residual images can detect residual images between the first air-sampled image and all second air-sampled images. A large number of residual images are detected, which can improve the accuracy of low-level interference detection on the first air-sampled images.In an optional embodiment, stripe-level signal detection may be performed on multiple first images, and multiple second images whose stripe-level signal intensity or number is less than a preset noise threshold may be selected from the multiple first images. It should be understood that since fixed signals are stripe-level signals and are distributed linearly horizontally and / or vertically, such as stripe-level noise generated by acoustic wave diffraction at image edges, fixed bad pixels, and stripe-level noise generated by the finger-level sensor edge or the screen edge when the finger-level sensor acquires blank images, the purpose of performing difference calculations is to eliminate fixed signals. Therefore, first images whose stripe-level signal intensity or number is greater than the noise threshold are deleted to obtain multiple second images. Using the second images as residual images can reduce the impact of stripe-level signals on low-level interference detection and reduce the number of residual images, thereby reducing the number of residual images used for low-level interference detection and improving the efficiency of low-level interference detection. In an optional embodiment, multiple residual images can be determined based on the second image. The second image can be processed, for example, by signal enhancement or noise reduction. Determining multiple residual images based on the second image can ensure that the residual images more closely meet detection criteria, improve the efficiency of residual image detection, and enhance the efficiency of low-level interference detection. In this embodiment of the present application, a difference calculation is performed between the first sampled image and each of the second sampled images, and multiple residual images are determined based on the difference calculation results. Because the residual images are obtained by performing the difference calculation between the first sampled image and the second sampled image, fixed signals included in both the first sampled image and the second sampled image can be eliminated, resulting in less fixed signal interference in the residual images. Consequently, low-level interference is more pronounced, reducing the impact of fixed signal interference on low-level interference detection and improving the accuracy of low-level interference detection. In one possible implementation, when selecting multiple second images from multiple first images, the mean pixel value of each column of pixels in the first image may be calculated to obtain the column pixel mean corresponding to the first image. The mean pixel value of each row of pixels in the first image may be calculated to obtain the row pixel mean corresponding to the first image. The first image in which both the corresponding column pixel mean and the corresponding row pixel mean are less than a noise threshold among the multiple first images is determined as the second image.Fixed signals are stripe-level signals, such as stripe-level noise generated by acoustic wave diffraction at image edges, fixed bad pixels, and stripe-level noise generated by the edge of a finger-level sensor or the screen edge when acquiring an empty image. Since stripe-level signals are linearly distributed horizontally and / or vertically and are regular linear signals, the mean pixel value of each column and the mean pixel value of each row can be calculated for each of the multiple first images to determine the column pixel mean and row pixel mean. For example, if the first image is a 20*20 image, the mean pixel value of each of the 20 rows and the mean pixel value of each of the 20 columns can be calculated. After calculating the column pixel mean and row pixel mean, they are compared with a preset noise threshold. The noise threshold can be set as needed. It should be understood that since the bottom-level signal is an irregular signal, it has little impact on the mean of the row or column pixels. However, the stripe-level signal is a linear regular signal distributed vertically or horizontally, and therefore has a greater impact on the mean of the row or column pixels. When the mean pixel value of the row or column containing the stripe-level signal in the first image is greater than the noise threshold, it proves that the stripe-level signal is distributed in the row or column. Therefore, the first image in which the corresponding column pixel mean and row pixel mean are both less than the noise threshold among the multiple first images is determined as the second image. It should be noted that, when calculating the mean pixel value of each column of pixels in the first image and calculating the mean pixel value of each row of pixels in the first image, the pixels in the first image can be arrayed and converted into an array. For example, a color image is a three-dimensional array, and a grayscale image is a two-dimensional array. After converting each pixel in the first image into an array, the mean pixel value of each column of pixels in the first image and the mean pixel value of each row of pixels in the first image are calculated. In an embodiment of the present application, the pixel value averages for each row and column of pixels in a first image are calculated. The first image whose row and column pixel averages are both less than a noise threshold is determined as the second image. This allows multiple second images to be selected from multiple first images, enabling stripe-level signal detection in the first image. Because stripe-level signals are regular, linear signals distributed vertically or horizontally, they significantly impact the pixel value average. Therefore, calculating the row and column pixel averages allows for stripe-level signal detection. This reduces the number of stripe-level signals in the second image, resulting in less fixed signal interference and, consequently, less pronounced background noise. This reduces the impact of fixed signal interference on background noise detection, improving the accuracy of background noise detection. In one possible implementation, when obtaining multiple residual images based on the second image, image noise reduction processing can be performed on each of the multiple second images to obtain multiple residual images.It should be understood that multiple targeted image noise reduction algorithms can be used to perform image noise reduction processing on the second image. Multiple image noise reduction methods can be used simultaneously, or only one image noise reduction method can be used. The specific image noise reduction method is not limited in this embodiment of the present application. It should be noted that since bottom-level interference is generally an irregular signal, and fixed signal interference is generally a regular signal, such as a bar-level signal, when performing noise reduction processing on the second image, the image noise reduction algorithm is used to reduce the regular signal in the second image, thereby reducing the intensity of the fixed signal interference in the obtained residual image and making the bottom-level interference more obvious. This can improve the accuracy of bottom-level interference detection in the residual image. In this embodiment of the present application, image noise reduction processing is performed on each of the multiple second images to obtain multiple residual images. This can reduce the fixed signal interference and other regular signal interference in the residual image. Therefore, if bottom-level interference exists in the residual image, the bottom-level interference can be made more obvious, thereby improving the accuracy of bottom-level interference detection, the speed of identifying the bottom-level interference, and the efficiency of bottom-level interference detection. In one possible implementation, multiple residual images can be normalized, and then feature extraction can be performed on each normalized residual image to obtain features corresponding to each residual image. Prediction can then be performed on each residual image using at least two fully connected layers based on the features corresponding to the residual image to obtain a probability of the presence of bottom-level interference in each residual image. Based on the probability of the presence of bottom-level interference in each residual image among the multiple residual images, bottom-level interference prediction can be performed on the first blank image to obtain a prediction result. When performing bottom-level interference prediction on the first blank image, each residual image is normalized, for example, by converting the residual image format into an image format that can be input into a prediction model. A structure of stacked convolutional and pooling layers is used to extract features from each residual image. Specifically, at least two convolutional and pooling layers are used to extract features from each normalized residual image to obtain features corresponding to each residual image. A classification prediction result is then obtained by stacking fully connected layers. Specifically, at least two fully connected layers are used to predict each residual image based on its features to obtain the probability of each residual image containing underlying noise. It should be understood that the features of the residual image can be, for example, a feature matrix or feature vector corresponding to the residual image obtained through convolution and pooling. Based on the probability of each residual image containing underlying noise, a prediction result for the first blank sampled image is determined, i.e., whether the first blank sampled image contains underlying noise.It should be understood that the multiple residual images are obtained by performing difference calculation based on the first air-sampled image and the multiple second air-sampled images. Since there are fewer air-sampled images with low-level interference in the air-sampled image library, there are fewer second air-sampled images with low-level interference. Therefore, when the first air-sampled image includes low-level interference, most of the multiple residual images include low-level interference. In an embodiment of the present application, a small convolutional neural network is used to perform image detection. Residual images are normalized so that they can be input into a prediction model. Feature extraction is performed on the residual images to determine features corresponding to the residual images. The features corresponding to the residual images can then be used to predict the probability of the residual images containing bottom-level interference, thereby determining a prediction result for the first blank image. Because the prediction model is used to predict the first blank image, prediction efficiency is high. Furthermore, the probability that the first blank image contains bottom-level interference is determined using multiple residual images related to the first blank image. Because multiple residual images are determined from the first blank image and multiple second blank images, if the first blank image contains bottom-level interference, most residual images also contain it. Furthermore, the fixed signal interference in the residual images is relatively low, preventing the fixed signal from affecting the detection result. Therefore, compared with directly detecting bottom-level interference on the first blank image, determining whether the first blank image contains bottom-level interference by determining whether multiple residual images contain bottom-level interference is more accurate. FIG2 is a flow chart of a method for determining a prediction result provided by an embodiment of the present application. As shown in FIG2 , the prediction result of the first blank sampled image can be determined by the following steps 201 to 204: Step 201: Calculate the average value of the probability that bottom-level interference exists in multiple residual images. Multiple residual mean values ​​can be calculated using the following formula (1). Wherein, mean (Q) represents the average value, y represents the probability that the i-th residual image has low-level interference, and n represents the number of residual images. Step 202: Calculate the standard deviation of the probability that the multiple residual images have low-level interference based on the average value. The standard deviation of the probability that the multiple residual images have low-level interference is calculated using the following formula (2).

[0002] E' S — mean (4))2# (2) std (0) = n, wherein mean (4) is used to characterize the mean value, s is used to characterize the probability that the i-th residual image has low-level interference, n is used to characterize the number of residual images, and std (0) is used to characterize the standard deviation. Step 203: Determine the probability that the first air-sampled image has low-level interference based on the mean value and the standard deviation. When the first air-sampled image carries low-level interference, most residual images will carry low-level interference. The probability that the first air-sampled image has low-level interference is determined based on the mean value and the standard deviation of the probabilities of the multiple residual images having low-level interference. For example, the probability that the first air-sampled image has low-level interference can be determined based on the mean value, the standard deviation, a preset mean value threshold, and a preset standard deviation threshold. When the mean value and the standard deviation are both greater than different mean value thresholds and standard deviation thresholds, the probability of the low-level interference corresponding to different first air-sampled images exists. Step 204: If the probability of the presence of bottom-level interference in the first sampled image is greater than a preset probability threshold, the first sampled image is determined to contain bottom-level interference. The relationship between the probability of the presence of bottom-level interference in the first sampled image and the preset probability threshold is determined. When the probability of the presence of bottom-level interference in the first sampled image is greater than the preset probability threshold, the first sampled image is determined to contain bottom-level interference. For example, if the probability threshold is 60%, then when the probability of the presence of bottom-level interference in the first sampled image is greater than 60%, a signal indicating the presence of bottom-level interference in the first sampled image is output. In this embodiment of the present application, the average and standard deviation of the probabilities of the presence of bottom-level interference in multiple residual images are calculated, and then the probability of the presence of bottom-level interference in the first sampled image is determined based on the average and standard deviation. Since the average value indicates the probability of the presence of bottom-level interference in the multiple residual images, and the standard deviation indicates the stability of the presence of bottom-level interference in the multiple residual images, the average and standard deviation of the probabilities of the presence of bottom-level interference in the multiple residual images can be used to accurately predict whether the first sampled image contains bottom-level interference, thereby improving the accuracy of bottom-level interference detection. In one possible implementation, when determining the probability of the presence of low-level interference in the first air-sampled image based on the mean and standard deviation, a weighted sum of the mean and standard deviation may be performed, and the summed result may be determined as the probability of the presence of low-level interference in the first air-sampled image. The mean may represent the probability of the presence of low-level interference in multiple residual images, and the standard deviation may represent the stability of the presence of low-level interference in the multiple residual images. The mean and standard deviation may be weightedly summed. For example, the following formula (3) may be used to perform the weighted sum to calculate the probability of the presence of low-level interference in the first air-sampled image.

[0003] P(A) = mean(A) * 80% + std(A) * 20%#(3) In an embodiment of the present application, a weighted sum is performed on the average value and the standard deviation of the probability that bottom-level interference exists in multiple residual images, and the sum result is determined as the probability that bottom-level interference exists in the first air-sampled image. Since the average value can represent the probability of bottom-level interference occurring in multiple residual images, and the standard deviation can represent the stability of bottom-level interference occurring in multiple residual images, the weighted summation can simultaneously consider the average value and the standard deviation. Therefore, the average value and the standard deviation of the probability that bottom-level interference exists in multiple residual images can be used to accurately predict whether the first air-sampled image has bottom-level interference, thereby improving the accuracy of bottom-level interference detection. In one possible implementation, when the number of unsampled images stored in the unsampled image library equals a preset image quantity threshold and at least one third unsampled image is received, the probabilities of the first unsampled images containing bottom-level interference are sorted in descending order, the first unsampled images corresponding to the top N probabilities are deleted, and the at least one third unsampled image is saved in the unsampled image library, where N is a positive integer greater than or equal to 1 and is equal to the number of third unsampled images. When the probability of the presence of low-level interference in the first uncollected image is less than or equal to a probability threshold, the first uncollected image is associated with the probability of the presence of low-level interference in the first uncollected image and returned to the uncollected image library. When the number of uncollected images stored in the uncollected image library reaches an upper limit (i.e., the number of uncollected images reaches a preset threshold for the number of uncollected images), if multiple third uncollected images (i.e., multiple new uncollected images) are received, the first uncollected images are sorted in descending order by the probability of the presence of low-level interference. Since the first uncollected images are associated with the probability of the presence of low-level interference in the first uncollected images, this is equivalent to sorting the first uncollected images according to the probability of the presence of low-level interference in the first uncollected images. The top N first uncollected images in the sorting order (i.e., the N first uncollected images with a higher probability of the presence of low-level interference in the first uncollected images) are deleted to store the new third uncollected images. Therefore, N is a positive integer greater than or equal to 1, and N is equal to the number of third uncollected images.When the number of newly added third uncollected images reaches a certain number, or a certain amount of time has passed since the last low-level interference detection, the uncollected image library optimization method described in any of the above embodiments is re-executed to delete uncollected images from the uncollected image library that have a high probability of containing low-level interference. For example, if low-level interference is detected in the newly added third uncollected image, the third uncollected image is deleted. If low-level interference is not detected in the newly added third uncollected image, but the number of uncollected images stored in the uncollected image library exceeds a threshold, and a new fourth uncollected image is received, the third uncollected image is deleted if it has a high probability of containing low-level interference, making way for the new fourth uncollected image. In other words, the uncollected image library optimization method described in any of the above embodiments is an iterative optimization method. In an embodiment of the present application, when the number of empty-sampled images stored in the empty-sampled image library exceeds a threshold and a new empty-sampled image is received, the first empty-sampled image with a high probability of containing low-level interference is deleted. This ensures that the probability of low-level interference in the empty-sampled images in the empty-sampled image library is low, ensuring that the captured finger-level image is free of low-level interference when correcting the captured finger-level image. Therefore, the impact of low-level interference can be avoided during correction of the captured finger-level image, improving the accuracy of finger-level recognition. In one possible implementation, multiple residual images can be input into a pre-trained low-level interference detection network to obtain a prediction result output by the low-level interference detection network regarding whether the first empty-sampled image contains low-level interference. Low-level interference detection can be performed on the first empty-sampled image using a pre-trained neural network model. Specifically, multiple residual images can be input into the low-level detection network, for example, by inputting multiple residual images into a recurrent neural network. The bottom-level interference detection network can directly predict multiple residual images to determine whether the first blank image contains bottom-level interference. The network model can directly output a prediction result regarding whether the first blank image contains bottom-level interference. In an embodiment of the present application, all residual images are directly detected using a pre-trained bottom-level interference detection network, and the presence of bottom-level interference in the first blank image is output. This achieves bottom-level interference prediction for the first blank image. Compared to the bottom-level interference prediction in the previous embodiment, there is no need to perform predictions for each residual image separately, resulting in higher prediction efficiency. Figure 3 is a schematic diagram of an apparatus for optimizing a blank image library provided in an embodiment of the present application. As shown in Figure 3, the apparatus 300 includes a determination unit 301 for determining a first blank image and multiple second blank images from a blank image library. The blank image library includes multiple blank images. Blank images are images captured by a finger-level sensor when no detection object is present within a detection area. The first blank image and the second blank images are different.A calculation unit 302 is configured to perform difference calculations between the first unsampled image and multiple second unsampled images, and obtain multiple residual images based on the calculation results. A deletion unit 303 is configured to delete the first unsampled image from the unsampled image library when it is determined, based on the multiple residual images, that the first unsampled image has bottom-level interference. In an embodiment of the present application, the determination unit 301 may be configured to perform step 101 in the above-described method embodiment, the calculation unit 302 may be configured to perform step 102 in the above-described method embodiment, and the deletion unit 303 may be configured to perform step 103 in the above-described method embodiment. In one possible implementation, the determination unit 301 may also be configured to perform signal interference detection on multiple unsampled images included in the unsampled image library, determining an unsampled image whose signal interference detection result is greater than a signal interference threshold as a first unsampled image, and determining an unsampled image whose signal detection result is less than or equal to the signal interference threshold as a second unsampled image. In one possible implementation, the calculation unit 302 may be further configured to perform difference calculations between the first blank-sampled image and multiple second blank-sampled images to obtain multiple first images, and to determine each of the multiple first images as a residual image; or to select multiple second images from the multiple first images, wherein the intensity or number of bar-level signals included in the second images is less than a preset noise threshold, and to determine each of the multiple second images as a residual image; or to obtain multiple residual images based on the multiple second images. In one possible implementation, the calculation unit 302 may be further configured to calculate the mean pixel value of each column of pixels in the first image to obtain the column mean pixel value corresponding to the first image; calculate the mean pixel value of each row of pixels in the first image to obtain the row mean pixel value corresponding to the first image; and determine as the second image a first image in the multiple first images whose corresponding column mean pixel value and row mean pixel value are both less than the noise threshold pixel value. In one possible implementation, the calculation unit 302 may be further configured to perform image denoising on each of the multiple second images to obtain multiple residual images. In one possible implementation, the deletion unit 303 may also be configured to perform normalization processing on multiple residual images; perform feature extraction on each normalized residual image to obtain features corresponding to each residual image; predict each residual image based on the features corresponding to the residual image using at least two fully connected layers to obtain a probability that each residual image has bottom-level interference; and perform bottom-level interference prediction on the first blank image based on the probability that each residual image in the multiple residual images has bottom-level interference to obtain a prediction result.In one possible implementation, the deletion unit 303 may further be configured to calculate an average of the probabilities of the presence of bottom-level interference in multiple residual images; calculate the standard deviation of the probabilities of the presence of bottom-level interference in the multiple residual images based on the average; determine the probability of the presence of bottom-level interference in the first air-sampled image based on the average and standard deviation; and determine that the first air-sampled image contains bottom-level interference if the probability of the presence of bottom-level interference in the first air-sampled image is greater than a preset probability threshold. In another possible implementation, the deletion unit 303 may further be configured to perform a weighted sum of the average and standard deviation, and determine the weighted sum result as the probability of the presence of bottom-level interference in the first air-sampled image. In one possible implementation, the deletion unit 303 may also be configured to, when the number of empty images stored in the empty image library reaches a preset image number threshold and the empty image library receives at least one third empty image, sort the probabilities of the presence of low-level interference in the first empty images in descending order, delete the first empty images corresponding to the top N probabilities, and save the at least one third empty image to the empty image library, where N is a positive integer greater than or equal to 1 and equal to the number of third empty images. In one possible implementation, the deletion unit 303 may also be configured to input multiple residual images into a pre-trained low-level interference detection network to obtain a prediction result output by the low-level interference detection network regarding whether the first empty image contains low-level interference. It should be noted that the information exchange and execution process between the various modules within the above-described empty image library optimization device are based on the same concept as the aforementioned empty image library optimization method embodiment. For details, please refer to the description of the aforementioned empty image library optimization method embodiment and will not be repeated here. Referring to Figure 4 , a schematic diagram of the structure of an electronic device according to an embodiment of the present application is shown. The specific embodiments of the present application do not limit the specific implementation of the electronic device. As shown in Figure 4 , the electronic device may include a processor 402, a communications interface 404, a memory 406, and a communication bus 408. Processor 402, communications interface 404, and memory 406 communicate with each other via communication bus 408. Communication interface 404 is used to communicate with other electronic devices or servers. Processor 402 is configured to execute program 410, which may specifically perform the relevant steps of the above-mentioned embodiment of the method for optimizing the air-sampled image library. Specifically, program 410 may include program code, which includes computer operating instructions.Processor 402 may be a central processing unit (CPU), a graphics processing unit (GPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present application. The one or more processors included in the smart device may be of the same type, such as one or more CPUs, one or more GPUs, or different types, such as one or more CPUs, one or more GPUs, and one or more ASICs. Memory 406 is used to store program 410. Memory 406 may include high-speed RAM or non-volatile memory, such as at least one disk drive. Program 410 may specifically cause processor 402 to execute the method for optimizing an empty image library described in any of the aforementioned embodiments. The specific implementation of each step in program 410 can be found in the corresponding descriptions of the steps and units in any of the aforementioned embodiments of the method for optimizing an empty image library, and will not be repeated here. Those skilled in the art will clearly understand that, for ease of description and brevity, the specific operating processes of the devices and modules described above can refer to the corresponding process descriptions in the aforementioned method embodiments, and will not be repeated here. In an embodiment of the present application, a first unsampled image and multiple second unsampled images are determined in an unsampled image library. The difference between the first unsampled image and each second unsampled image is calculated to obtain multiple residual images. These residual images are then predicted. This allows determination of whether the first unsampled image contains underlying interference. Unsampled images in the unsampled image library can then be tested for underlying interference, and the first unsampled image containing underlying interference can be removed from the library. This ensures that the unsampled images in the library are free of underlying interference. After a finger-level sensor acquires a finger-level image, it can use the unsampled images free of underlying interference to correct the finger-level image, ensuring the accuracy of the finger-level image and improving the recognition accuracy of the finger-level sensor. This embodiment of the present application also provides a computer program product comprising computer instructions that instruct a computing device to perform operations corresponding to any of the aforementioned method embodiments. It should be noted that, according to implementation needs, the various components / steps described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.The methods described above according to the embodiments of the present application can be implemented in hardware or firmware, or as software or computer code that can be stored on a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or non-transitory machine-readable medium downloaded over a network and then stored on a local recording medium. Thus, the methods described herein can be processed by software stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It will be understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the method for optimizing the air-collected image library described herein is implemented. Furthermore, when a general-purpose computer accesses the code for implementing the method for optimizing the air-collected image library described herein, the execution of the code transforms the general-purpose computer into a dedicated computer for executing the method for optimizing the air-collected image library described herein. Those skilled in the art will appreciate that the various exemplary units and method steps described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed using hardware or software depends on the specific application and design constraints of the technical solution. Professionals skilled in the art may use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the embodiments of this application. The above embodiments are intended only to illustrate the embodiments of this application and are not intended to limit them. Persons skilled in the relevant technical fields may make various changes and modifications without departing from the spirit and scope of the embodiments of this application. Therefore, all equivalent technical solutions are also within the scope of the embodiments of this application, and the scope of patent protection for the embodiments of this application shall be defined by the claims.

Claims

Claims 1. A method for optimizing an empty image library, characterized in that: include: Determining a first air-collected image and a plurality of second air-collected images from an air-collected image library, wherein the air-collected image library includes a plurality of air-collected images, each of the air-collected images being images captured by the finger-level sensor when there is no detection object within a detection area; performing difference calculations on the first air-collected image and the plurality of second air-collected images, and obtaining a plurality of residual images based on the calculation results; When it is determined according to the multiple residual images that the first blank-sampled image has bottom-level interference, the first blank-sampled image is deleted from the blank-sampled image library.

2. The method according to claim 1, characterized in that: The method of determining the first air-sampled image and multiple second air-sampled images from the air-sampled image library includes: performing signal interference detection on the multiple air-sampled images included in the air-sampled image library, determining the air-sampled image with a signal interference detection result greater than a signal interference threshold as the first air-sampled image, and determining the air-sampled image with a signal detection result less than or equal to the signal interference threshold as the second air-sampled image.

3. The method according to claim 1, characterized in that: The step of performing difference calculations on the first air-sampled image and the second air-sampled images, and obtaining a plurality of residual images based on the calculation results, includes: performing difference calculations on the first air-sampled image and the plurality of second air-sampled images to obtain a plurality of first images, and determining the plurality of first images as the residual images; or selecting a plurality of second images from the plurality of first images, wherein the degree or number of strip-level signals included in the second images is less than a preset noise threshold, and determining the plurality of second images as the residual images; or obtaining the plurality of residual images based on the plurality of second images.

4. The method according to claim 3, characterized in that: The selecting of multiple second images from the multiple first images includes: calculating the mean pixel value of each column of pixels in the first image to obtain the column pixel mean corresponding to the first image; calculating the mean pixel value of each row of pixels in the first image to obtain the row pixel mean corresponding to the first image; and determining the first image in the multiple first images, for which both the column pixel mean and the row pixel mean are smaller than the noise threshold, as the second image.

5. The method according to claim 3, characterized in that: The obtaining of the plurality of residual images according to the plurality of second images includes: performing image noise reduction processing on each of the plurality of second images to obtain the plurality of residual images.

6. The method according to claim 1, characterized in that: The method further includes: normalizing the plurality of residual images; extracting features from each of the normalized residual images to obtain features corresponding to each residual image; predicting the residual image according to the features corresponding to each residual image through at least two fully connected layers to obtain features corresponding to each residual image; A probability that the residual image has bottom-level interference, and based on the probability that each residual image in the plurality of residual images has bottom-level interference, performing bottom-level interference prediction on the first blank-sampled image to obtain a prediction result.

7. The method according to claim 6, characterized in that: The performing of bottom-level interference prediction on the first blank image based on the probability of bottom-level interference existing in each residual image among the multiple residual images to obtain a prediction result includes: calculating an average value of the probabilities of bottom-level interference existing in the multiple residual images; calculating a standard deviation of the probabilities of bottom-level interference existing in the multiple residual images based on the average value; determining a probability of bottom-level interference existing in the first blank image based on the average value and the standard deviation; and determining that the first blank image has bottom-level interference if the probability of bottom-level interference existing in the first blank image is greater than a preset probability threshold.

8. The method according to claim 7, characterized in that: Determining the probability that the first air-sampled image has bottom-level interference based on the average value and the standard deviation includes: performing a weighted summation on the average value and the standard deviation, and determining the weighted summation result as the probability that the first air-sampled image has bottom-level interference.

9. The method according to claim 7, characterized in that: Also includes: When the number of empty-sampled images stored in the empty-sampled image library equals a preset image quantity threshold, and the empty-sampled image library receives at least one third empty-sampled image, the probabilities of the presence of bottom-level interference in the first empty-sampled images are sorted in descending order, the first empty-sampled images corresponding to the top N probabilities in the sorting are deleted, and the at least one third empty-sampled image is saved in the empty-sampled image library, where N is a positive integer greater than or equal to 1, and N is equal to the number of third empty images.

10. The method according to claim 1, characterized in that The method further includes: inputting the plurality of residual images into a pre-trained bottom-level interference detection network to obtain a prediction result output by the bottom-level interference detection network on whether the first air-sampled image has bottom-level interference.

11. An optimization device for an empty sampling image library, characterized in that: include: A determination unit is configured to determine a first air-sampled image and multiple second air-sampled images from an air-sampled image library, wherein the image library includes multiple air-sampled images, the air-sampled images are images collected when the finger-level sensor has no detection object in the detection area, and the first air-sampled image and the second air-sampled images are different; a calculation unit is configured to perform difference calculations on the first air-sampled image and the multiple second air-sampled images, and obtain multiple residual images based on the calculation results; and a deletion unit is configured to delete the first air-sampled image from the air-sampled image library when it is determined based on the multiple residual images that the first air-sampled image has bottom-level interference.

12. An electronic device, comprising: Processor, memory, communication interface and communication bus, the processor, the The memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute the method for optimizing the air sampling image library as described in any one of claims 1-10.

13. A computer storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for optimizing the air sampling image library according to any one of claims 1 to 10 is implemented. 17

Images