Lampblack removal image acquisition method and device based on polarization imaging and electronic equipment

Abstract

The invention provides an oil smoke removal image acquisition method and device based on polarization imaging and electronic equipment, and the method comprises the steps: obtaining an oil smoke polarization image; obtaining a polarization degree image and a polarization angle image based on the lampblack polarization image; and according to the lampblack polarization image, the polarization degree image and the polarization angle image, removing lampblack based on a dictionary learning detection algorithm to obtain a clear lampblack-removed image. According to the lampblack removal image acquisition method and device based on polarization imaging and the electronic equipment disclosed by the invention, the polarization image is acquired by utilizing the polarization imaging system, the low-rank sparse matrix is obtained by utilizing the polarization image, and a relatively clean background is obtained by utilizing the low-rank sparse matrix decomposition algorithm; the background dictionary model is constructed from the background in a sparse expression mode, and finally the abnormal target is detected in a reconstruction error calculation mode, so that background noise is avoided, and the detection precision is improved.

Description

Technical Field

[0001] The present disclosure relates to an image acquisition method, and in particular to a method, device and electronic device for acquiring oil fume removal images for a range hood. Background Art

[0002] Traditional kitchen range hoods and stoves produce large amounts of oil smoke when cooking, which seriously affects the cook's vision, making it difficult to see the state of the food in the pot and hindering cooking operations. Although existing range hoods rely on the principle of negative pressure suction and have a certain oil smoke absorption function, the oil smoke concentration increases instantly when stir-frying, and the naked eye and ordinary cameras cannot penetrate the oil smoke, making it difficult to achieve ideal visual clarity, making it difficult for the cook to observe the state of the food, especially for elderly people with presbyopia. In addition, when using traditional range hood and stove systems, cooks need to lower their heads to stir-fry. The long-term lowering of the head, with an average lowering angle of more than 45 degrees, can also adversely affect the cook's cervical spine, especially for elderly people with poor health, increasing their risk of cervical spondylosis. In summary, current range hood and stove systems still have problems with oil smoke obstruction and poor ergonomic design.

[0003] Current solutions to the problem of oil fume obstruction are relatively limited. These solutions typically address the issue by increasing the hood's suction power to enhance its ability to absorb oil fumes, thereby reducing oil fume concentration and improving visual clarity. There is currently no clear solution to the ergonomic issue of chefs frequently having to lower their heads while cooking. When designing range hoods and stoves, most manufacturers only consider the ergonomics of the range hood and stove. Summary of the Invention

[0004] The technical problem to be solved by the present disclosure is to overcome the defects in the prior art and provide a method, device and electronic device for obtaining oil fume removal images based on polarization imaging that can be applied to range hoods.

[0005] The present disclosure solves the above technical problems through the following technical solutions: a method for obtaining a smoke removal image based on polarization imaging, comprising:

[0006] Acquire a fume polarization image, wherein the fume polarization image is an image of the fume area to be measured with polarization information;

[0007] Obtaining a polarization degree image and a polarization angle image based on the oil smoke polarization image;

[0008] According to the oil smoke polarization image, polarization degree image and polarization angle image, the oil smoke is removed based on a dictionary learning detection algorithm to obtain a clear oil smoke-removed image.

[0009] Preferably, the acquiring of the oil smoke polarization image includes acquiring polarization images with polarization angles of 0°, 45°, 90° and 135°.

[0010] Preferably, acquiring the oil smoke polarization image specifically includes:

[0011] Acquire original polarization images based on a split-focal-plane polarization imaging system;

[0012] The original polarization image is interpolated and the resolution is restored to obtain polarization images with polarization angles of 0°, 45°, 90° and 135°.

[0013] Preferably, before removing the oil smoke based on the oil smoke polarization image, polarization degree image and polarization angle image to obtain a clear oil smoke-free image, the step further includes:

[0014] The oil smoke polarization image, polarization degree image and polarization angle image are filtered to remove image noise.

[0015] Preferably, the step of removing the oil smoke based on the dictionary learning detection algorithm according to the oil smoke polarization image, polarization degree image and polarization angle image to obtain a clear oil smoke-free image includes:

[0016] Arrange the polarization degree image, the polarization angle image, and the oil smoke polarization image into a low-rank sparse matrix;

[0017] Decomposing the low-rank sparse matrix to obtain a low-rank background matrix and a sparse anomaly matrix;

[0018] A background dictionary is constructed based on the low-rank background matrix;

[0019] The image of the oil smoke area to be detected is detected based on the background dictionary to obtain a clear oil smoke-free image.

[0020] Preferably, the method further comprises inputting the oil fume removal image into a display screen of the range hood for real-time display.

[0021] Another aspect of the present disclosure provides a device for obtaining an oil smoke removal image based on polarization imaging, comprising:

[0022] A first acquisition module is used to acquire a polarization image of oil smoke;

[0023] A second acquisition module is used to obtain a polarization degree image and a polarization angle image based on the oil smoke polarization image;

[0024] The oil smoke removal module is used to remove the oil smoke based on the oil smoke polarization image, polarization degree image and polarization angle image based on the dictionary learning detection algorithm to obtain a clear oil smoke removed image.

[0025] Another aspect of the present disclosure provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, wherein when the processor executes the computer program, the method for acquiring de-oiling smoke images based on polarization imaging as described above is implemented.

[0026] Another aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any one of the above-mentioned methods for acquiring de-oiling smoke images based on polarization imaging.

[0027] Another aspect of the present disclosure provides a computer program product, including a computer program, characterized in that when executed by a processor, the computer program implements any of the aforementioned methods for acquiring oil smoke removal images based on polarization imaging. The aforementioned preferred conditions may be arbitrarily combined, consistent with common knowledge in the art, to yield preferred embodiments of the present disclosure.

[0028] The positive progressive effects of the present disclosure are: the method, device, electronic device, medium and program product for obtaining de-oiling smoke images based on polarization imaging of the present disclosure utilize a polarization imaging system to obtain a polarization image, and then obtain a polarization angle image and a polarization degree image based on the polarization image, and set the polarization image, polarization angle image and polarization degree image into a low-rank sparse matrix, utilize a low-rank sparse matrix decomposition algorithm to obtain a relatively clean background, and then construct a background dictionary model from the background through sparse expression, and finally detect abnormal targets by calculating the reconstruction error, thereby avoiding background noise and improving detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flowchart of a method for acquiring a smoke removal image based on polarization imaging provided in Example 1 of the present disclosure;

[0030] Figure 2 A schematic structural diagram of a detector of a split-focal-plane polarization imaging system provided in Example 1 of the present disclosure;

[0031] Figure 3 A schematic flow chart of step S30 provided in Example 1 of the present disclosure;

[0032] Figure 4 A schematic diagram of decomposition of a low-rank sparse matrix provided in Example 1 of the present disclosure;

[0033] Figure 5 A schematic diagram of a device for acquiring a smoke removal image based on polarization imaging provided in Example 2 of the present disclosure;

[0034] Figure 6 This is a structural block diagram of an electronic device provided in Example 3 of the present disclosure. DETAILED DESCRIPTION

[0035] The present disclosure is further illustrated below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0036] In the embodiments of the present disclosure, prefixes such as "first" and "second" are used only to distinguish different description objects and have no limiting effect on the position, order, priority, quantity or content of the described objects. In the embodiments of the present disclosure, the use of prefixes such as ordinal numbers to distinguish description objects does not constitute a restriction on the described objects. For the statement of the described objects, please refer to the description in the context of the embodiments, and the use of such prefixes should not constitute an unnecessary restriction. In addition, in the description of this embodiment, unless otherwise specified, the meaning of "plurality" is two or more.

[0037] In the embodiments of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0038] Example 1

[0039] Figure 1 A schematic flow chart of a method for obtaining a smoke removal image based on polarization imaging provided in Example 1 of the present disclosure, the method comprising:

[0040] S10, obtaining a fume polarization image; the fume polarization image is an image of the fume area to be measured with polarization information;

[0041] S20, obtaining a polarization degree image and a polarization angle image based on the oil smoke polarization image;

[0042] S30 , removing the oil smoke based on the oil smoke polarization image, polarization degree image, and polarization angle image based on the dictionary learning detection algorithm to obtain a clear oil smoke removed image.

[0043] In step S10 , a fume polarization image is obtained, which is an image of the fume area to be measured with polarization information. For the fume area to be measured, polarization images with polarization angles of 0°, 45°, 90°, and 135° are obtained.

[0044] Specifically, the polarization imaging module, installed below the range hood, primarily comprises a multi-band polarized light source, using an LED array that outputs linearly polarized light with wavelengths of 450nm blue light and 850nm near-infrared light, covering the primary frequency band of fume scattering. It also includes a lens and a split-focal-plane polarization detector. The detector's focal plane is integrated with a micropolarizer array, enabling the acquisition of images in different polarization states in a single exposure.

[0045] Depending on its polarization state, light can be classified as natural light, linearly polarized light, partially polarized light, elliptically polarized light, and circularly polarized light. Light in which the electric vector of a light wave has a random orientation perpendicular to the direction of light propagation is called natural light or unpolarized light. Natural light has no fixed phase relationship between the vibrations of the electric vector in all directions, and the time-averaged values ​​of all vibrations are equal. When unpolarized light passes through a medium and is refracted, reflected, absorbed, or scattered, the vibrations in one direction become dominant over the others, and the vibration distribution becomes asymmetrical. This type of light is called partially polarized light. In this case, the vibrations of the electric vector in all directions still have no fixed phase relationship, but the time-averaged value of the electric vector in one direction is relatively dominant. If the magnitude and direction of the electric vector vibrations change regularly, and the trajectory of its endpoints forms an ellipse, this light is called elliptically polarized light. When viewed against the direction of light propagation, when the electric vector endpoint of a light wave circles around the ellipse clockwise, this elliptically polarized light is called right-handed polarized light. When the electric vector endpoint of a light wave circles around the ellipse counterclockwise, this elliptically polarized light is called left-handed polarized light. Circularly polarized light is a special type of elliptically polarized light. It appears circularly polarized when the two components of elliptically polarized light have the same amplitude and a phase difference of ±π / 2. If the trajectory of the endpoints of the electric vector of a light wave is a straight line, meaning that the electric vector of the light wave vibrates in only one direction, with its magnitude varying with phase but its direction remaining unchanged, this light is called linearly polarized light.

[0046] The human visual system and photodetectors cannot directly perceive the polarization information in light waves. Therefore, various polarization imaging detection systems are needed to convert polarization information into light intensity information in different ways, thereby perceiving, measuring, and analyzing the polarization information of a scene. In reality, to perceive the polarization information of a scene, researchers often use the following approach: first, use a polarization camera to obtain the original polarization image of the scene light wave; then use the matrix inversion method to obtain the Stokes component image of the target, and then obtain the polarization information image; finally, use image preprocessing, image enhancement, and image fusion techniques to process the Stokes component image and polarization information image to obtain a clear polarization characteristic image reflecting the target, thereby achieving the purpose of perceiving the scene's polarization information.

[0047] Therefore, preferably, as a preferred embodiment, the obtaining of the oil smoke polarization image includes obtaining the polarization image based on a split-focus plane polarization imaging system; interpolating the polarization image to restore the resolution to obtain the polarization images with polarization angles of 0°, 45°, 90° and 135°.

[0048] The focal plane polarization imaging system directly couples the micropolarizer array (MPA) to the focal plane array (FPA) of the detector. Each pixel on the focal plane corresponds to a micropolarizer element, such as Figure 2 As shown in the figure, it is a schematic diagram of the structure of the polarization split focal plane detector. As can be seen from the figure, the polarization split focal plane pixel array is arranged in a 2×2 array with four pixels forming a basic period, which is called a super pixel. In a super pixel, the polarization information obtained by the light of each pixel passing through the units with different polarization modulation directions on the micro polarizer is usually corresponding to the four pixels. 、 、 and Polarization information of these four different polarization modulation directions. During polarization imaging, incident light passes through array elements at different positions of the micro-polarization array and is incident on the photosensitive pixels in the detector focal plane, thereby acquiring intensity information of scene objects in different polarization directions. Based on this, polarization images of different polarization directions can be obtained through image interpolation algorithms, and then polarization characteristic images such as polarization degree and polarization angle can be obtained.

[0049] Compared to amplitude- and aperture-based polarization imaging systems, the split-focal-plane polarization imaging system requires only a micropolarizer integrated with a focal-plane detector, eliminating the need for multiple polarizers or lenses, resulting in a simpler structure. Furthermore, it can simultaneously acquire polarization information in four directions, breaking the limitation that the camera and target must not shift during capture. This makes it suitable not only for experimental detection research in static scenes, but also for experimental research in dynamic scenes, particularly in field experiments where the relative position of the imaging target and the imager varies over time. Although the split-focal-plane polarization imaging system suffers from a loss of spatial resolution, it does not affect the ability to acquire polarization information of the target scene over a wide viewing angle through wide-view imaging in field experiments.

[0050] Of course, those skilled in the art will also understand that other methods can be used to obtain polarization images at 0°, 45°, 90°, and 135°, such as using polarizers for time-sharing acquisition. Polarization images at angles other than 0°, 45°, 90°, and 135° can also be obtained.

[0051] Specifically, the focal plane polarization imaging system integrates the micro-polarization array on the focal plane of the camera, and every four pixels form a super pixel and follow 、 、 and It can simultaneously obtain the light intensity responses in four polarization directions in one imaging, and has the advantages of high integration and strong real-time performance.

[0052] However, due to the structure of the focal plane, the focal plane polarization imaging system will suffer from the problem of loss of spatial resolution; at the same time, the existence of instantaneous field of view error will also affect the accuracy of the reconstructed polarization information. In order to restore the spatial resolution and reduce the impact of the instantaneous field of view error on the accuracy of the reconstructed polarization information, an interpolation algorithm is often used to restore the spatial resolution and fill in the missing polarization information. This is the process of restoring spatial resolution using interpolation technology. Therefore, it is necessary to obtain a polarization image based on a focal plane polarization imaging system; interpolate the polarization image to restore the resolution to obtain the polarization image with polarization angles of 0°, 45°, 90°, and 135°. There can be many difference algorithms, such as bilinear difference, bicubic difference, etc.

[0053] In step S20, a polarization degree image and a polarization angle image are obtained based on the oil smoke polarization image. Specifically, the polarized light is represented by using the Stokes vector method, that is, the vector representation of the light on each pixel detected by the detector is:

[0054]

[0055] The Stokes vector is represented by S, which consists of four parameters S0, S1, S2, and S3. 45 , I 90 , I 135 The light intensities obtained from the four polarization angles of 0°, 45°, 90°, and 135°, respectively. L and I R The intensities of left-handed and right-handed polarized light, respectively. S0 represents the total light intensity received by the detector, which can also be calculated by dividing the sum of the light intensities at the four polarization angles by 2. S1 represents the intensity difference between linearly polarized light at 0° and 90° (i.e., parallel and perpendicular directions), and linearly polarized light at 45° and 135°. S3 is related to circularly polarized light and can be ignored in linear polarization imaging detection systems.

[0056] The degree of polarization (DoLP) refers to the ratio of linear polarization intensity to total light intensity, and the angle of polarization (AOP) is used to describe the shape of polarized light. The calculation formula is as follows:

[0057]

[0058]

[0059] In the actual calculation process, whether it is a time-sharing polarization imaging system or a focal plane polarization imaging system, calculation can be performed using the acquired 0°, 45°, 90°, and 135°, where the grayscale value of the pixel point in the grayscale image can be defaulted to the light intensity.

[0060] Therefore, the above polarization angle (AOP) image and degree of polarization (DoLP) image can be obtained through the polarization images at four angles.

[0061] However, polarization imaging technology is very susceptible to noise, resulting in a decrease in image quality. In particular, when calculating the degree of polarization and polarization angle, nonlinear calculations can amplify noise, thereby calculating incorrect polarization information. Effective filtering technology can improve the quality of polarization images and obtain more accurate polarization information from polarization patterns. Therefore, the method, before step S30, further includes:

[0062] The oil smoke polarization image, polarization degree image, and polarization angle image are filtered to remove image noise. Specifically, a spatial domain image filtering algorithm, a transform domain image filtering algorithm, etc. can be used. In this embodiment, a three-dimensional block matching collaborative filtering (BM3D) filtering algorithm is used.

[0063] Step S30 , removing the oil smoke based on a dictionary learning detection algorithm according to the oil smoke polarization image, polarization degree image and polarization angle image to obtain a clear oil smoke removed image.

[0064] Anomaly detection algorithms based on sparse representation assume that the polarization angle spectrum of any pixel in a polarization image can be represented as a linear combination of a small number of dictionary atoms. In this sparse representation model, no assumptions about the statistical distribution of the data are made, nor does it require that dictionary atoms be independent and identically distributed. Detection is accomplished by weighting the dictionary atoms, overcoming the low accuracy of statistical anomaly detection algorithms, which often suffer from an inability to accurately estimate background information.

[0065] Assuming polarization data The polarization angle spectrum vector of the pixel to be detected is , Can be approximately represented as a dictionary The linear combination of the corresponding training samples, that is:

[0066]

[0067] in, For a complete dictionary, is the number of atoms in the dictionary, is the weight vector of sparse representation. According to the sparse representation theory, It is sparse, that is, most of the elements in the weight vector are zero. The sparse representation detection (SRD) algorithm is to find a set of the sparsest weight vectors to represent the pixels to be detected.

[0068] Generally, abnormal target detection in polarization images can be regarded as a special binary classification problem, that is, the pixels in the polarization image are divided into two categories: target and background. The SRD target detection algorithm uses a linear combination of dictionary atoms to represent the pixels to be detected, and requires the reconstruction error to be as small as possible. Therefore, the type of pixel to be detected can be determined based on the dictionary type to which the atoms used in the linear combination belong. It can be approximately represented by a linear combination of the corresponding atoms in the background dictionary and the target dictionary, that is:

[0069]

[0070]

[0071]

[0072] in, Represented by the background sub-dictionary and the target sub-dictionary Composed dictionary; and Represent the number of atoms in the background sub-dictionary and the target sub-dictionary respectively; Represented by the background coefficient weight vector and the target coefficient weight vector The weight vector composed of . is a background pixel, then is a sparse vector, Is a zero vector; if the pixel to be detected is the target pixel, then is a sparse vector, is the zero vector, The solution can be transformed into the following optimization problem:

[0073]

[0074] in, It means finding the zero norm, that is, the number of non-zero elements in the vector, but this is a non-deterministic polynomial problem. is sparse, so it can be solved Norm, that is, the sum of the absolute values ​​of the elements in the vector, The minimum value of the norm is used to approximate the minimization norm. At this time, The solution can be transformed into the following optimization problem:

[0075]

[0076] The weight vector can be solved by the tracking algorithm from the above formula . According to the obtained weight vector The reconstructed pixel can be obtained . Reconstructed pixels The difference between the pixel to be measured is called the residual. Since the residuals of different materials are different, the background and the target can be distinguished by comparing the reconstructed residuals: ;

[0077] ;

[0078] ;

[0079] if Greater than threshold , then it means the pixel to be tested is an abnormal pixel; otherwise, the pixel to be tested is the background pixel.

[0080] The SRD target detection algorithm solves a constrained The minimum flower norm algorithm achieves a sparse representation of the pixels to be detected, but it is computationally intensive and can have a high error rate when the background is noisy or contaminated by abnormal targets. Therefore, through low-rank sparse matrix factorization, the matrix can be decomposed into a low-rank matrix, a sparse matrix, and a noise matrix. The low-rank matrix corresponds to the background matrix, while the sparse matrix can be used to detect abnormal targets. This method can significantly reduce background noise, prevent abnormal target contamination, and improve detection accuracy.

[0081] Therefore, the above step S30, such as Figure 3 Shown, including,

[0082] S31, arranging the polarization degree image, the polarization angle image, and the oil smoke polarization image into a low-rank sparse matrix;

[0083] S32, decomposing the low-rank sparse matrix to obtain a low-rank background matrix L and a sparse anomaly matrix S;

[0084] S33, constructing a background dictionary based on the low-rank background matrix L;

[0085] S34, detecting an image of the oil smoke area to be detected based on the background dictionary to obtain a clear oil smoke-free image.

[0086] The polarization degree image, the polarization angle image, and the oil smoke polarization image are set to train a low-rank sparse matrix, specifically, at least two images, the polarization angle image and the polarization degree image, are selected and set to train a low-rank sparse matrix. , X is the low-rank sparse matrix, Represents the total number of pixels in the image, Represents the number of images. For example, if the polarization angle image and the polarization degree image are selected, P is 2. If the polarization angle image, the polarization degree image, and the polarization images at angles of 0°, 45°, 90°, and 135° are selected, P is 6.

[0087] S32, decomposing the training low-rank sparse matrix to obtain a low-rank background matrix L and a sparse abnormality matrix S; specifically, using a low-rank sparse matrix decomposition algorithm (Low-Rank and Matrix Decomposition, LRaSMD), decomposing it into: ;like Figure 4 Specifically, the LRaSMD algorithm can obtain a relatively clean background, which effectively solves the problem that statistical-based abnormal target detection cannot accurately estimate background information. When there is noise in the background or the abnormal target is contaminated, the expression-based abnormal target detection has the problem of low detection accuracy. Therefore, preferably, the GoDec algorithm is used to solve the above low-rank background matrix components and sparse abnormal matrix components.

[0088] The GoDec algorithm is a fast approximation algorithm that constrains the low-rank background matrix The rank and sparse anomaly matrix of The sparsity of is used to control the complexity of the reconstruction model. The GoDec algorithm solves the low-rank background matrix by minimizing the decomposition error in the formula and the sparse anomaly matrix .

[0089]

[0090] in, represents the norm; represents the matrix rank, Indicates the number of matrix components; Represents the low-rank background matrix Maximum value of rank; Represents a sparse anomaly matrix The sparsity reflects the sparse components in the image, which is usually defined as a sparse anomaly matrix of norm. The background can be approximately represented as a linear combination of several basis vectors, where the number of basis vectors is equal to the rank of the background matrix. Generally, the decomposition error decreases monotonically with increasing iterations. Therefore, the optimization problem in the above equation can be transformed into a sub-problem of alternately solving the following two parameters:

[0091]

[0092]

[0093] In the formula is the matrix dimension, and the low-rank approximation theory of the bilateral projection algorithm (BRP) is used to solve the equation, assuming that:

[0094]

[0095] in, and are all random matrices, Based on the BRP algorithm, we can get:

[0096]

[0097] middle depending on Hard thresholding:

[0098]

[0099] in, Is the matrix in the input set The projection on yes Before A non-zero subset of the largest terms.

[0100] Therefore, in the GoDec algorithm, the input data matrix , fault tolerance coefficient , maximum number of iterations ;

[0101] initialization , , , ;

[0102] Repeat times, until

[0103] (1) ;

[0104] (2) , , ;

[0105] (3) If ,but , repeat step (2);

[0106] (4) ;

[0107] (5)

[0108] The GoDec algorithm can generate a low-rank background matrix and a sparse anomaly matrix. However, if low-rank sparse matrix decomposition is directly used for polarization anomaly target detection, since its detection decision relies on sparse components, which may contain some non-anomalous pixels at large values ​​of the sparse parameters, the detection performance will deteriorate and there will still be a high number of false positives.

[0109] S33, constructing a background dictionary based on the low-rank background matrix L; the present application proposes using a low-rank sparse matrix decomposition and sparse dictionary expression algorithm to detect polarization anomaly targets in complex background environments. First, a relatively clean background is obtained through a low-rank sparse matrix decomposition algorithm. Then, a sparse dictionary expression model is constructed from the background through a sparse expression method. Finally, abnormal targets are detected by calculating the reconstruction error.

[0110] In the case of polarization anomaly detection in a complex background environment, due to the diversity of abnormal situations, it is difficult to learn the dictionary of abnormal targets and it is often difficult to achieve. Therefore, the background dictionary is used to reconstruct the target. The sparse dictionary expression model believes that a pixel to be detected can be approximately linearly expressed by background training samples and target training samples. The matrix composed of background training samples and target training samples is regarded as a dictionary. , the detected pixels can be modeled as follows based on the training samples of the target and background categories:

[0111]

[0112] in, is a pixel to be measured with P polarization angle spectra; A dictionary consisting of background and target training samples; is a sparse component, which corresponds to the dictionary The weight of each sample in . Sparse component It can be expressed as:

[0113]

[0114] in, is the balance parameter between the reconstruction error term and the regularization term. is the background dictionary. If the pixel to be tested cannot be expressed by the background dictionary, then the pixel to be tested is an anomaly; if it can be expressed by the background dictionary, then it is a background pixel. Specifically, abnormal target detection can be performed through reconstruction error. The calculation formula of reconstruction error is:

[0115]

[0116] After removing most of the noise and anomalies in the image through low-rank sparse matrix decomposition, a relatively clean background is obtained, and the background dictionary is learned. The specific background dictionary learning algorithm is as follows:

[0117] Input: Data matrix , low-rank background matrix , the number of dictionary bases , number of dictionaries ;

[0118] Output: Learned background dictionary ;

[0119] initialization: , , , randomly from the low-rank background matrix Select samples as the initial background dictionary , ;

[0120] for

[0121] Solve for the sparse coefficient:

[0122] Update the dictionary:

[0123] for

[0124] Update List:

[0125]

[0126]

[0127] end

[0128] end

[0129] S34, detecting an image of the oil smoke area to be detected based on the background dictionary to obtain a clear oil smoke-free image.

[0130] Using the background dictionary obtained in the above steps, an image of the oil fume area to be measured is used. The image can be any of the above-mentioned oil fume polarization images, or it can be the above-mentioned polarization image obtained by the split-focal plane polarization imaging system, or it can be an ordinary image of the oil fume area to be measured. Of course, an image with polarization information is preferred. Using this image and the background dictionary, a clear oil fume-free image can be obtained. The background dictionary is used to detect each pixel of the image of the oil fume area to be measured, and after judging whether it is background or target, a clear oil fume-free image can be obtained after removing the target.

[0131] The method disclosed herein further includes step S40 of inputting the oil fume removal image to a display screen of the range hood for real-time display, thereby avoiding the problem of the cook frequently lowering his head and providing a clear oil fume removal image for the cook to use.

[0132] The disclosed method for obtaining smoke removal images based on polarization imaging utilizes a polarization imaging system to obtain a polarization image, and then obtains a polarization angle image and a polarization degree image based on the polarization image, and obtains a low-rank sparse matrix based on the polarization image. A low-rank sparse matrix decomposition algorithm is used to obtain a relatively clean background, and then a background dictionary model is constructed from the background through sparse expression. Finally, abnormal targets are detected by calculating the reconstruction error, thereby avoiding background noise and improving detection accuracy.

[0133] Example 2

[0134] Corresponding to the aforementioned embodiment of the oil fume removal method based on polarization imaging, the present disclosure also provides an embodiment of an oil fume removal device based on polarization imaging.

[0135] Figure 5 A schematic diagram of a module of a smoke removal image acquisition device based on polarization imaging provided by an exemplary embodiment of the present disclosure, the system comprising:

[0136] The first acquisition module 1 is used to acquire the oil smoke polarization image;

[0137] A second acquisition module 2 is configured to obtain a polarization degree image and a polarization angle image based on the oil smoke polarization image;

[0138] The oil smoke removal module 3 is configured to remove the oil smoke based on the oil smoke polarization image, polarization degree image and polarization angle image and the dictionary learning detection algorithm to obtain a clear oil smoke removed image.

[0139] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components of the units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution.

[0140] Example 3

[0141] Figure 6 This is a structural schematic diagram of an electronic device showing an example embodiment of the present disclosure, wherein the electronic device includes a memory, a processor, and a computer program stored in the memory and for running on the processor. When the processor executes the computer program, the oil smoke removal method based on polarization imaging described in any of the above embodiments is implemented. Figure 6 The electronic device 60 shown is only an example and should not limit the functionality and scope of use of the embodiments of the present disclosure.

[0142] like Figure 6 As shown, the electronic device 60 may be a general-purpose computing device, such as a server device. Components of the electronic device 60 may include, but are not limited to, the at least one processor 61, the at least one memory 62, and a bus 63 connecting different system components (including the memory 62 and the processor 61).

[0143] The bus 63 includes a data bus, an address bus, and a control bus.

[0144] The memory 62 may include a volatile memory, such as a random access memory (RAM) 621 and / or a cache memory 622 , and may further include a read-only memory (ROM) 623 .

[0145] The memory 62 may also include a program tool 625 (or utility) having a set (at least one) of program modules 624, such program modules 624 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include an implementation of a network environment.

[0146] The processor 61 executes various functional applications and data processing by running the computer program stored in the memory 62, such as the oil smoke removal image acquisition method based on polarization imaging provided in any of the above embodiments.

[0147] The electronic device 60 can also communicate with one or more external devices 64 (e.g., a keyboard, pointing device, etc.). This communication can be performed via an input / output (I / O) interface 65. Furthermore, the electronic device 60 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 66. As shown, the network adapter 66 communicates with other modules of the electronic device 60 via a bus 63. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 60, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0148] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0149] Example 4

[0150] The embodiments of the present disclosure further provide a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for acquiring a de-oiling fume image based on polarization imaging provided in any of the above embodiments is implemented.

[0151] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0152] Example 5

[0153] An embodiment of the present disclosure further provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for acquiring de-oiling smoke images based on polarization imaging.

[0154] The program code for executing the computer program product of the present disclosure may be written in any combination of one or more programming languages, and the program code may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0155] While specific embodiments of the present disclosure have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of protection of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications are intended to fall within the scope of protection of the present disclosure.

Claims

1. A method for obtaining a smoke removal image based on polarization imaging, characterized in that: include, Acquire a fume polarization image, wherein the fume polarization image is an image of the fume area to be measured with polarization information; Obtaining a polarization degree image and a polarization angle image based on the oil smoke polarization image; According to the oil smoke polarization image, polarization degree image and polarization angle image, the oil smoke is removed based on a dictionary learning detection algorithm to obtain a clear oil smoke-removed image.

2. The method for obtaining a smoke removal image according to claim 1, wherein: The obtaining of the oil smoke polarization image, Including obtaining polarization images with polarization angles of 0°, 45°, 90° and 135°.

3. The method for obtaining a smoke removal image according to claim 2, wherein: The obtaining of the oil smoke polarization image specifically includes: Acquire original polarization images based on a split-focal-plane polarization imaging system; The original polarization image is interpolated and the resolution is restored to obtain polarization images with polarization angles of 0°, 45°, 90° and 135°.

4. The method for obtaining a smoke removal image according to claim 1, wherein: Before removing the oil smoke based on the oil smoke polarization image, polarization degree image and polarization angle image to obtain a clear oil smoke-free image based on a dictionary learning detection algorithm, the step further includes: The oil smoke polarization image, polarization degree image and polarization angle image are filtered to remove image noise.

5. The method for obtaining a smoke removal image according to claim 1, wherein: The step of removing the oil smoke based on the oil smoke polarization image, polarization degree image and polarization angle image to obtain a clear oil smoke-free image based on the dictionary learning detection algorithm includes: Arrange the polarization degree image, the polarization angle image, and the oil smoke polarization image into a low-rank sparse matrix; Decomposing the low-rank sparse matrix to obtain a low-rank background matrix and a sparse anomaly matrix; A background dictionary is constructed based on the low-rank background matrix; The image of the oil smoke area to be detected is detected based on the background dictionary to obtain a clear oil smoke-free image.

6. The method for obtaining a smoke removal image according to claim 1, wherein: The method further comprises, The oil fume removal image is input to a display screen of the range hood for real-time display.

7. A device for obtaining oil smoke removal images based on polarization imaging, characterized in that: include, A first acquisition module is used to acquire a polarization image of oil smoke; A second acquisition module is used to obtain a polarization degree image and a polarization angle image based on the oil smoke polarization image; The oil smoke removal module is used to remove the oil smoke based on the oil smoke polarization image, polarization degree image and polarization angle image based on a dictionary learning detection algorithm to obtain a clear oil smoke removed image.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and configured to run on the processor, wherein: When the processor executes the computer program, the method for acquiring a smoke removal image based on polarization imaging according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for acquiring a smoke removal image based on polarization imaging according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for acquiring a smoke removal image based on polarization imaging according to any one of claims 1 to 6 is implemented.

Images