Image illumination intensity determination method and device, electronic equipment and storage medium

Abstract

The invention provides an image illumination intensity determination method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the preprocessing of a to-be-analyzed image, so as to obtain a preprocessed image; inputting the preprocessed image into a preset illumination intensity prediction model to obtain an illumination intensity result corresponding to the to-be-analyzed image output by the illumination intensity prediction model; wherein the illumination intensity prediction model is composed of a Conv module, a C2fWTConv module, a C3k2 module, an A2C2f module, a PfAAM module and a Classify module. The illumination intensity prediction model is built based on a deep learning mode, illumination intensity analysis can be performed on the to-be-analyzed image more simply and efficiently through the illumination intensity prediction model, and the method has certain practical application value in the fields of automatic driving and robot vision.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, in particular to an image illumination intensity determination method and device, electronic equipment and storage medium. BACKGROUND

[0002] The method for determining the illumination intensity of an image has important purposes and significance in many fields such as image processing, computer vision, photography art, etc. By evaluating the illumination intensity, color correction, contrast adjustment and other operations can be performed to improve the visual effect of the image, making it look more natural and more beautiful.

[0003] In applications such as autonomous driving and robot vision, illumination intensity information helps the system understand the lighting conditions of the surrounding environment, thereby making more reasonable decisions. Illumination intensity affects the extraction of image features, so accurate illumination intensity information is helpful to improve the accuracy and robustness of recognition when performing tasks such as target detection, face recognition, scene classification, etc. In building virtual scenes or augmented reality experiences, accurate simulation of illumination intensity is one of the key factors to achieve realism. SUMMARY

[0004] To solve the above technical problems, the technical solution adopted by the present application is: According to one aspect of the present application, an image illumination intensity determination method is provided, comprising: preprocessing the image to be analyzed to obtain a preprocessed image; inputting the preprocessed image into a preset illumination intensity prediction model to obtain an illumination intensity result corresponding to the image to be analyzed output by the illumination intensity prediction model; The illumination intensity prediction model is composed of a Conv module, a C2f_WTConv module, a C3k2 module, an A2C2f module, a PfAAM module, and a Classify module.

[0005] In an exemplary embodiment of the present application, preprocessing the image to be analyzed to obtain a preprocessed image comprises: obtaining the image to be analyzed; scaling the size of the image to be analyzed by an equal ratio to obtain a preprocessed image with a preset pixel size; the preset pixel size is a pixel size that can be processed by the illumination intensity prediction model.

[0006] In an exemplary embodiment of the present application, scaling the size of the image to be analyzed by an equal ratio to obtain a preprocessed image with a preset pixel size comprises: scaling the size of the image to be analyzed by an equal ratio to obtain an initial image with a preset pixel size; normalize pixel values of the initial image to obtain a first image; perform denoising processing on the first image by using a Gaussian filtering algorithm or a non-local mean algorithm to obtain a second image; in a case where an image scene corresponding to the second image is a first scene, sequentially perform histogram equalization processing, adaptive histogram equalization processing, and shadow compensation processing on the second image to obtain a preprocessed image; the first scene is a scene in which a light-dark contrast is greater than a preset contrast threshold; in a case where the image scene corresponding to the second image is a second scene, sequentially perform a retinal algorithm and an illumination mapping algorithm on the second image to obtain the preprocessed image; the second scene is a scene in which an illumination uniformity is less than a preset uniformity threshold; in a case where the image scene corresponding to the second image is a third scene, sequentially perform a low-light enhancement algorithm and a multi-exposure fusion algorithm on the second image to obtain the preprocessed image; the third scene is a scene in which a signal-to-noise ratio is less than a preset signal-to-noise ratio threshold.

[0007] In an example embodiment of the present application, the illumination intensity prediction model is determined according to the following steps: obtain a plurality of sample images; each sample image corresponds to an image identifier; the image identifier corresponding to the sample image represents an illumination intensity corresponding to the sample image; randomly allocate the plurality of sample images to obtain a plurality of training sample images, a plurality of verification sample images, and a plurality of test sample images; input the plurality of training sample images into a preset neural network model for sample training to obtain an initial network model; input the plurality of verification sample images into the initial network model to obtain a verification result output by the initial network model; adjust parameters of the initial network model according to the verification result; input the plurality of test sample images into the initial network model to obtain a model test result corresponding to the initial network model; in a case where the model test result represents that the initial network model meets an identification accuracy requirement, determine the initial network model as the illumination intensity prediction model.

[0008] In an example embodiment of the present application, inputting the plurality of test sample images into the initial network model to obtain a model test result corresponding to the initial network model includes: input each test sample image into the initial network model to obtain a sample test result of each test sample image output by the initial network model; The sample test result of each test sample image is traversed, and in a case where the illumination intensity corresponding to the sample test result of the test sample image is the same as the illumination intensity corresponding to the image identifier of the test sample image, the test sample image is determined as a target test sample image. In a case where the ratio of the number of target test sample images to the number of test sample images is greater than a preset accuracy rate, the initial network model is determined as a model test result corresponding to the initial network model in a case where the initial network model meets the recognition accuracy requirement.

[0009] In an example embodiment of the present application, after obtaining a plurality of sample images, the image illumination intensity determination method further comprises: A preset proportion of sample images in the plurality of sample images are subjected to image processing such as flipping or symmetry.

[0010] In an example embodiment of the present application, the illumination intensity result corresponding to the image to be analyzed is one of low light, medium light, and high light.

[0011] In an example embodiment of the present application, after receiving the preprocessed image, the illumination intensity prediction model sequentially inputs the preprocessed image into a Conv module, a C2f_WTConv module, a Conv module, a C3k2 module, a Conv module, an A2C2f module, a Conv module, an A2C2f module, a PfAAM module, and a Classify module to obtain the illumination intensity result.

[0012] According to an aspect of the present application, an image illumination intensity determination device is provided, comprising: An image preprocessing module is configured to preprocess an image to be analyzed to obtain a preprocessed image. An image result determination module is configured to input the preprocessed image into a preset illumination intensity prediction model to obtain an illumination intensity result corresponding to the image to be analyzed output by the illumination intensity prediction model. The illumination intensity prediction model is composed of a Conv module, a C2f_WTConv module, a C3k2 module, an A2C2f module, a PfAAM module, and a Classify module.

[0013] According to an aspect of the present application, a non-transitory computer readable storage medium is provided, wherein the storage medium stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the image illumination intensity determination method described above.

[0014] According to an aspect of the present application, an electronic device is provided, comprising a processor and the non-transitory computer readable storage medium described above.

[0015] The present application has at least the following beneficial effects: The image illumination intensity determination method of the present application first pre-processes the image to be analyzed to obtain a pre-processed image, and then inputs the pre-processed image into an illumination intensity prediction model to obtain the illumination intensity result corresponding to the image to be analyzed output by the illumination intensity prediction model, wherein the illumination intensity prediction model is composed of a Conv module, a C2f_WTConv module, a C3k2 module, an A2C2f module, a PfAAM module, and a Classify module. The illumination intensity prediction model is built based on deep learning, and the illumination intensity of the image to be analyzed can be analyzed more efficiently through the illumination intensity prediction model. In the field of autonomous driving and robot vision, it has certain practical application value. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 The flow chart of the image illumination intensity determination method provided by the embodiment of the present application is shown in the figure. Figure 2 The module processing flow chart of the illumination intensity prediction model of the image illumination intensity determination method provided by the embodiment of the present application after receiving the pre-processed image is shown in the figure. Figure 3 The block diagram of the image illumination intensity determination device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings of the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0019] The illumination intensity of the current image is usually determined by a relatively complex calculation method (such as analyzing by writing specific code or calculating the illumination intensity by multiple formula theories). This method is too complex to analyze the illumination intensity, which occupies too much computing power of the system, affects the operation of the system, and takes too long to operate. When there are many images to be analyzed, the efficiency is low. Therefore, the present application provides an image illumination intensity determination method, as shown in Figure 1 which includes: Step S100, pre-processing the image to be analyzed to obtain a pre-processed image; Further, in step S100, the method for pre-processing the image to be analyzed to obtain a pre-processed image includes steps S110-S120: Step S110, obtaining the image to be analyzed; The image to be analyzed is an image for which the illumination intensity needs to be determined.

[0020] Step S120, performing equal scaling processing on the size of the image to be analyzed to obtain a pre-processed image with a preset pixel size; The preset pixel size is a pixel size that can be processed by the illumination intensity prediction model. As an example, the preset pixel size is 640x640.

[0021] Further, in order to improve the detection accuracy of the illumination intensity of the image to be analyzed, a method for pre-processing the image to be analyzed according to the image environment of the image to be analyzed is provided. Specifically, step S120 includes steps S121-S126: Step S121, performing equal scaling processing on the size of the image to be analyzed to obtain an initial image with a preset pixel size; Step S122, performing normalization processing on the pixel values of the initial image to obtain a first image; The pixel values of the initial image are normalized (e.g., scaled to the range [0, 1] or [0, 255]) to eliminate the brightness differences of different image acquisition devices when acquiring the image to be analyzed.

[0022] Step S123, performing denoising processing on the first image by a Gaussian filtering algorithm or a non-local mean algorithm to obtain a second image; The Gaussian filtering algorithm or the non-local mean algorithm can reduce the interference of sensor noise on illumination analysis.

[0023] Step S124, in the case where the image scene corresponding to the second image is a first scene, sequentially performing histogram equalization processing, adaptive histogram equalization processing, and shadow compensation processing on the second image to obtain a pre-processed image; The first scene is a scene with a large contrast between light and dark, such as an outdoor natural light scene where the illumination is greatly affected by weather, time, and obstructions, and the contrast between light and dark is strong.

[0024] The histogram equalization processing can enhance the global contrast of the second image, and is suitable for overcast or backlight scenes; the adaptive histogram equalization processing (such as the CLAHE image processing algorithm) can avoid overexposure of the second image after global equalization, and is suitable for high dynamic range (HDR) scenes; the shadow compensation processing can use morphological operations (such as the top-hat transformation algorithm) to remove the influence of shadow areas.

[0025] In a case where the image scene corresponding to the second image is the second scene, the retinal algorithm and the light mapping algorithm are sequentially performed on the second image to obtain a preprocessed image. The second scene is a scene with a light uniformity less than a preset uniformity threshold, such as an indoor artificial light scene (an office, a shopping mall, or the like) or the like.

[0026] The retinal algorithm (a Retinex-based algorithm) can separate the reflection component and the light component in the second image to enhance the dark area details of the second image; the light mapping algorithm can balance the global light by estimating the light map of the second image and adjusting the intensity thereof.

[0027] In a case where the image scene corresponding to the second image is the third scene, the low-light enhancement algorithm and the multi-exposure fusion algorithm are sequentially performed on the second image to obtain a preprocessed image. The third scene is a scene with a signal-to-noise ratio less than a preset signal-to-noise ratio threshold, such as a night or low-light scene (a night scene, a dark room, or the like) or the like.

[0028] The low-light enhancement algorithm (such as the LLNet (Low-Light Image Enhancement Network) or Zero-DCE (Zero-Reference Deep Curve Estimation) algorithm) can use a deep learning model (such as Zero-DCE) to automatically learn a light enhancement curve to perform low-light enhancement processing on the second image; the multi-exposure fusion algorithm (such as the HDR module of OpenCV) can synthesize images with different exposure times to expand the image dynamic range of the second image.

[0029] In a case where the image scene corresponding to the second image is the second scene, the retinal algorithm and the light mapping algorithm are sequentially performed on the second image to obtain a preprocessed image. The illumination intensity result corresponding to the image to be analyzed is one of low illumination, medium illumination, and high illumination. Furthermore, the illumination intensity result corresponding to each image scene is also one of these three categories. Specifically, when the image scene to be analyzed is the first scene, the illumination intensity result is one of these three categories; when the image scene to be analyzed is the second scene, the illumination intensity result is one of these three categories; and when the image scene to be analyzed is the third scene, the illumination intensity result is one of these three categories. By classifying the illumination intensity results of these three different image scenes, nine illumination intensity results can be obtained, demonstrating the influence of different image scenes on the illumination intensity results. This further improves the classification accuracy of the illumination intensity results and enhances the detection accuracy of subsequent target detection or target tracking models, resulting in better target detection performance.

[0030] Low light indicates low light intensity, which may affect image visibility and detail, such as at dusk or in a dark indoor environment; medium light indicates moderate light, which can ensure image sharpness and color reproduction, suitable for most everyday shooting conditions; high light indicates high light intensity, which may cause images to be overexposed or have high contrast, such as in an outdoor environment under direct sunlight.

[0031] Furthermore, the illumination intensity results output by the illumination intensity prediction model can be further refined according to actual training requirements. For example, several illumination intensity results can be determined based on the pixels of the image to be analyzed, and each illumination intensity result corresponds to a scene to meet the actual needs of users.

[0032] The illumination intensity prediction model consists of the Conv module, C2f_WTConv module, C3k2 module, A2C2f module, PfAAM module, and Classify module, such as... Figure 2 As shown, the light intensity prediction model receives a preprocessed image (such as...) Figure 2 After obtaining a 640×640 color image, the preprocessed image is sequentially input into the Conv module, Conv module, C2f_WTConv module, Conv module, C3k2 module, Conv module, A2C2f module, Conv module, A2C2f module, PfAAM module, and Classify module to obtain the illumination intensity result from the pixel values ​​of the preprocessed image.

[0033] The following sections will describe each module that makes up the light intensity prediction model: The Conv module is a convolution module, which includes a convolution layer, a batch normalization BN layer, and a sigmoid linear unit (SILU) activation function. The C2f_WTConv module is a module improved according to a WTConv layer and a C2f module. The WTConv layer (a wavelet convolution layer, which is a convolution layer implemented by using a wavelet transform (WT), can effectively increase the receptive field of convolution) can be used to replace the existing downlink in the architecture, produce an effective multi-frequency response, and scale with the size of the receptive field. Therefore, the WTConv layer has a beneficial effect on the performance of image classification. The WTConv layer is used to replace the second convolution layer in the bottleneck structure in the C2f module, thereby obtaining the improved C2f_WTConv module, which can focus on a larger receptive field. The C3k2 module is used for auxiliary feature extraction and is an improvement of the traditional C3 module. By combining a variable convolution kernel (for example, 3x3, 5x5, etc.) and a channel separation strategy, the C3k2 module provides stronger feature extraction capability. The A2C2f module is a structure combining area attention (A2) and an improved feature extraction network (C2f) and is mainly used for feature extraction and processing. The PfAAM module (Parameter-free Average Attention Module) is a module that applies a parameter-free attention mechanism. With small computational overhead, the PfAAM module can improve the performance of the model without affecting the model size. The Classify module is mainly used for converting the features extracted from the backbone network into classification results (such as class prediction). The Classify module is a classification module used for class prediction.

[0034] The structures of the Conv module, the C3k2 module, the A2C2f module, the PfAAM module, and the Classify module mentioned in the present application are structures in the existing target detection model YOLO (You Only Look Once). Therefore, the composition architecture of the above modules will not be described here.

[0035] Further, the illumination intensity prediction model determines according to steps S210-S280: In step S210, a plurality of sample images are obtained. The sample images can be images for which the user has performed light intensity analysis in a historical period and obtained light intensity results. Each sample image corresponds to an image identifier, and the image identifier of the sample image represents the light intensity corresponding to the sample image. That is, each sample image corresponds to a light intensity result (one of low light, medium light, and high light), and the light intensity result corresponding to each sample image is verified by the user to ensure the accuracy of the light intensity result of the sample image and improve the accuracy of the subsequently determined light intensity prediction model.

[0036] In step S220, a preset proportion of the sample images are subjected to flipping or symmetrical image processing. The flipping or symmetrical operation is performed on the preset proportion (for example, 30%) of the sample images, and the light intensity prediction model is determined based on the sample images subjected to the flipping or symmetrical operation (the image identifier of the sample image subjected to the flipping or symmetrical operation is the same as that of the sample image before the flipping or symmetrical operation), which can improve the determination accuracy of the light intensity prediction model for images of different angles.

[0037] In step S230, the sample images are randomly allocated to obtain training sample images, verification sample images, and test sample images. The sample images are randomly allocated to obtain the training sample images of the training set, the verification sample images of the verification set, and the test sample images of the test set, so that the processing result of the obtained light intensity prediction model is more accurate.

[0038] In step S240, the training sample images are input into a preset neural network model for sample training to obtain an initial network model. The training sample images are used to train the preset neural network model.

[0039] In step S250, the verification sample images are input into the initial network model to obtain a verification result output by the initial network model. In step S260, the parameters of the initial network model are adjusted based on the verification result. The verification sample images are used to verify the processing result of the initial network model, and the parameters of the initial network model are adjusted based on the verification result to improve the image processing accuracy of the initial network model.

[0040] In step S270, the test sample images are input into the initial network model to obtain a model test result corresponding to the initial network model. The test sample images are used to test the processing result of the initial network model to verify the recognition accuracy of the initial network model.

[0041] In step S270, several test sample images are input into the initial network model to obtain the model test results corresponding to the initial network model, including steps S271-S273: Step S271: Input each test sample image into the initial network model to obtain the sample test result of each test sample image output by the initial network model; Step S272: Traverse the sample test results of several test sample images. If the light intensity corresponding to the sample test result of the test sample image is the same as the light intensity corresponding to the image identifier of the test sample image, then determine the test sample image as the target test sample image. If the light intensity corresponding to the test result of the test sample image is the same as the light intensity corresponding to the image identifier of the test sample image, it means that the result obtained by the initial network model from testing the test sample image is accurate, and the test sample image is determined as the target test sample image.

[0042] Step S273: If the ratio of the number of target test sample images to the number of test sample images is greater than the preset accuracy, the initial network model that meets the recognition accuracy requirement is determined as the model test result corresponding to the initial network model.

[0043] Step S280: If the model test results indicate that the initial network model meets the recognition accuracy requirements, the initial network model is determined as the light intensity prediction model.

[0044] The image illumination intensity determination method of the present invention first preprocesses the image to be analyzed to obtain a preprocessed image, and then inputs the preprocessed image into an illumination intensity prediction model to obtain the illumination intensity result corresponding to the image to be analyzed output by the illumination intensity prediction model. The illumination intensity prediction model consists of a Conv module, a C2f_WTConv module, a C3k2 module, an A2C2f module, a PfAAM module, and a Classify module. The illumination intensity prediction model is built based on deep learning. The illumination intensity prediction model can perform illumination intensity analysis on the image to be analyzed in a more concise and efficient manner, and has certain practical application value in the fields of autonomous driving and robot vision.

[0045] On the other hand, this application also proposes an image illumination intensity determination device 100, such as... Figure 3 As shown, it includes: Image preprocessing module 110 is used to preprocess the image to be analyzed to obtain a preprocessed image; The image preprocessing module 110 preprocesses the image to be analyzed to obtain a preprocessed image, including the following methods: acquire an image to be analyzed; perform equal proportion scaling processing on the size of the image to be analyzed to obtain a preprocessed image with a preset pixel size; and the preset pixel size is a pixel size that can be processed by the illumination intensity prediction model.

[0046] an image result determination module 120 configured to input the preprocessed image into the preset illumination intensity prediction model to obtain an illumination intensity result corresponding to the image to be analyzed output by the illumination intensity prediction model; and the illumination intensity result corresponding to the image to be analyzed is one of low light, medium light, and high light. The illumination intensity prediction model is composed of a Conv module, a C2f WTConv module, a C3k2 module, an A2C2f module, a PfAAM module, and a Classify module, and the illumination intensity prediction model is determined according to the following method: acquire a plurality of sample images; each sample image corresponds to an image identifier; and the image identifier corresponding to the sample image represents the illumination intensity corresponding to the sample image; perform flip or symmetric image processing on a preset proportion of the plurality of sample images; randomly allocate the plurality of sample images to obtain a plurality of training sample images, a plurality of verification sample images, and a plurality of test sample images; input the plurality of training sample images into the preset neural network model for sample training to obtain an initial network model; input the plurality of verification sample images into the initial network model to obtain a verification result output by the initial network model; adjust the parameters of the initial network model according to the verification result; input the plurality of test sample images into the initial network model to obtain a model test result corresponding to the initial network model; In a case where the model test result indicates that the initial network model meets the recognition accuracy requirement, the initial network model is determined as the illumination intensity prediction model.

[0047] Further, the plurality of test sample images are input into the initial network model to obtain a model test result corresponding to the initial network model, including: input each test sample image into the initial network model to obtain a sample test result of each test sample image output by the initial network model; traverse the sample test results of the plurality of test sample images, and in a case where the illumination intensity corresponding to the sample test result of a test sample image is the same as the illumination intensity corresponding to the image identifier of the test sample image, the test sample image is determined as a target test sample image; In a case where the ratio of the number of target test sample images to the number of test sample images is greater than the preset accuracy, the initial network model satisfying the recognition accuracy requirement is determined as the model test result corresponding to the initial network model.

[0048] Further, after receiving the preprocessed image, the light intensity prediction model sequentially inputs the preprocessed image into the Conv module, the C2f WTConv module, the C3k2 module, the A2C2f module, the PfAAM module, and the Classify module to obtain the light intensity result.

[0049] Embodiments of the application also provide a computer program product comprising program code for causing an electronic device to perform the steps of the methods according to the various exemplary embodiments of the application described above when the program product is run on the electronic device.

[0050] In addition, although the various steps of the methods of the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in this particular order, or that all of the steps shown must be performed to achieve the desired result. Additionally or alternatively, certain steps can be omitted, multiple steps can be combined into one step, one step can be divided into multiple steps, etc.

[0051] From the above description of the embodiments, those skilled in the art will readily appreciate that the example embodiments described herein can be implemented by software and / or by hardware coupled with software. Accordingly, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or a network, and includes a number of instructions to make a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) execute the methods according to the embodiments of the present disclosure.

[0052] In the exemplary embodiments of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0053] Those skilled in the art can understand that each aspect of the present application can be implemented as a system, a method or a program product. Therefore, each aspect of the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects, which can be collectively referred to as "circuitry", "module" or "system".

[0054] The electronic device according to this embodiment of the present application. The electronic device is merely an example and should not bring any limitation to the function and usage range of the embodiments of the present application.

[0055] The electronic device is in the form of a general computing device. The components of the electronic device can include, but are not limited to, the at least one processor described above, the at least one memory described above, and a bus connecting different system components, including the memory and the processor.

[0056] The memory stores program codes which can be executed by the processor, so that the processor executes the steps described in the above "Exemplary Method" section according to various exemplary embodiments of the present application.

[0057] The memory can include a readable medium in the form of a volatile memory, such as a random access memory (RAM) and / or a cache memory, and can further include a read-only memory (ROM).

[0058] The memory can further include programs / utilities with a set of (at least one) program modules, such as an operating system, one or more application programs, other program modules, and program data, each of which or some combination of which can include implementation of a network environment.

[0059] The bus can be one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures.

[0060] The electronic device can also communicate with one or more external devices (such as a keyboard or a pointing device, a Bluetooth device, etc.) and can also communicate with one or more devices that enable a user to interact with the electronic device (and / or one or more input / output (I / O) devices 620) and / or with any devices (such as a router, a modem, etc.) that enable the electronic device to communicate with one or more other computing devices. Such communication can occur via an I / O interface. Also, the electronic device can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter.

[0061] In exemplary embodiments of the present disclosure, a computer readable storage medium is also provided, having stored thereon a program product capable of implementing the method described above. In some possible implementations, various aspects of the present application can also be implemented as a program product in the form of a computer readable medium having computer readable program code embodied in the medium. The computer readable program code comprises instructions configured to cause an end device to perform the steps described in the "Exemplary Method" section above according to various exemplary embodiments of the present application when the program product is run on the end device.

[0062] The program product can take any combination of one or more computer readable media. The computer readable media can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of the computer readable storage medium include an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0063] The computer readable signal medium can include a data signal propagated in baseband or propagated as a carrier wave in a propagated data signal, in which the readable program code is embodied. Such propagated data signal can take various forms, including but not limited to electro-magnetic, optical or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium other than the computer readable storage medium that can send, receive, or transport program for use by or in connection with an instruction execution system, apparatus or device.

[0064] The program code contained on the computer readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0065] The program code may be executed by one or more programmable processing devices, which can include processors, microprocessor, microcomputer or microcontrollers, as well as other internal components such as volatile and non-volatile memory. The program code can be stored in any type of computer readable medium, for example, random access memory (RAM), for example, static random access memory (SRAM), dynamic random access memory (DRAM), or other volatile memory, for example, a variant of DRAM such as synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), SynchBurst DRAM (SLDRAM), fast page mode DRAM (FPM DRAM), or a variant of DRAM such as RAMBUS DRAM (RDRAM), etc., read only memory (ROM), programmable ROM (PROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), electrically erasable programmable ROM (EEPROM), flash memory, or any other memory or data storage technology, either now known or developed in the future.

[0066] Furthermore, the above-described diagrams are merely schematic illustrations of the processes included in the method according to the exemplary embodiments of the present application, and are not intended for limiting purposes. It is readily understood that the processes illustrated in the above-described diagrams do not indicate or limit the time sequence of the processes. In addition, it is readily understood that the processes can be executed synchronously or asynchronously, for example, in a plurality of modules.

[0067] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, such a division is not mandatory. Indeed, according to embodiments of the present disclosure, the features and functionalities of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functionalities of one module or unit described above can be further divided into embodied by a plurality of modules or units.

[0068] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed by the present application can be easily thought by those skilled in the art, and should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for determining image illumination intensity, characterized in that, The method includes: The image to be analyzed is preprocessed to obtain a preprocessed image; The preprocessed image is input into a preset illumination intensity prediction model to obtain the illumination intensity result corresponding to the image to be analyzed, output by the illumination intensity prediction model. The light intensity prediction model consists of the Conv module, the C2f_WTConv module, the C3k2 module, the A2C2f module, the PfAAM module, and the Classify module.

2. The method according to claim 1, characterized in that, The preprocessing of the image to be analyzed to obtain a preprocessed image includes: Acquire the image to be analyzed; The size of the image to be analyzed is scaled proportionally to obtain a preprocessed image with a preset pixel size; the preset pixel size is the pixel size that can be processed by the illumination intensity prediction model.

3. The method according to claim 2, characterized in that, The image to be analyzed is scaled proportionally to obtain a preprocessed image with a preset pixel size, including: The size of the image to be analyzed is scaled proportionally to obtain an initial image with a preset pixel size; The pixel values ​​of the initial image are normalized to obtain the first image; The first image is denoised using a Gaussian filtering algorithm or a nonlocal mean algorithm to obtain the second image. When the image scene corresponding to the second image is the first scene, the second image is sequentially subjected to histogram equalization, adaptive histogram equalization, and shadow compensation to obtain a preprocessed image; the first scene is a scene where the brightness contrast is greater than a preset contrast threshold. When the image scene corresponding to the second image is the second scene, the second image is sequentially processed by the retina algorithm and the illumination mapping algorithm to obtain a preprocessed image; the second scene is a scene where the illumination uniformity is less than a preset uniformity threshold. When the image scene corresponding to the second image is the third scene, the second image is subjected to low light enhancement algorithm and multi-exposure fusion algorithm in sequence to obtain a preprocessed image; the third scene is a scene with a signal-to-noise ratio less than a preset signal-to-noise ratio threshold.

4. The method according to claim 1, characterized in that, The light intensity prediction model is determined according to the following steps: Acquire a number of sample images; each sample image corresponds to an image identifier; the image identifier corresponding to the sample image represents the light intensity corresponding to the sample image. The sample images are randomly assigned to obtain a number of training sample images, a number of verification sample images, and a number of test sample images. Several training sample images are input into a preset neural network model for sample training to obtain an initial network model; Several of the aforementioned verification sample images are input into the initial network model to obtain the verification results output by the initial network model; Based on the verification results, the parameters of the initial network model are adjusted; Several test sample images are input into the initial network model to obtain the model test results corresponding to the initial network model; If the model test results indicate that the initial network model meets the recognition accuracy requirements, the initial network model is determined as the light intensity prediction model.

5. The method according to claim 4, characterized in that, The step of inputting a plurality of the test sample images into the initial network model to obtain the model test results corresponding to the initial network model includes: Each of the test sample images is input into the initial network model to obtain the sample test result of each of the test sample images output by the initial network model; The test results of several test sample images are traversed. If the light intensity corresponding to the test result of the test sample image is the same as the light intensity corresponding to the image identifier of the test sample image, the test sample image is determined as the target test sample image. If the ratio of the number of target test sample images to the number of test sample images is greater than a preset accuracy rate, the initial network model that meets the recognition accuracy requirement is determined as the model test result corresponding to the initial network model.

6. The method according to claim 5, characterized in that, After acquiring several sample images, the method further includes: Image processing is performed on a sample image of a predetermined ratio among a plurality of the sample images, either by flipping or by making it symmetrical.

7. The method according to any one of claims 1-6, characterized in that, After receiving the preprocessed image, the illumination intensity prediction model sequentially inputs the preprocessed image into the Conv module, the C2f_WTConv module, the C3k2 module, the A2C2f module, the PfAAM module, and the Classify module to obtain the illumination intensity result; the illumination intensity result corresponding to each image scene is one of low illumination, medium illumination, and high illumination.

8. An image illumination intensity determination device, characterized in that, include: The image preprocessing module is used to preprocess the image to be analyzed to obtain a preprocessed image; The image result determination module is used to input the preprocessed image into the preset illumination intensity prediction model in order to obtain the illumination intensity result corresponding to the image to be analyzed output by the illumination intensity prediction model. The light intensity prediction model consists of the Conv module, the C2f_WTConv module, the C3k2 module, the A2C2f module, the PfAAM module, and the Classify module.

9. A non-transitory computer-readable storage medium, characterized in that, The storage medium stores at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the method as described in any one of claims 1-7.

10. An electronic device, characterized in that, Includes a processor and the non-transitory computer-readable storage medium as described in claim 9.

Images