Fire point parameter determination method and device and electronic equipment

Abstract

The invention discloses a fire point parameter determination method and device and electronic equipment. The method comprises the following steps: acquiring a target image corresponding to a target area; determining color images corresponding to the plurality of color channels according to the target image; determining transmission parameters respectively corresponding to the plurality of color images according to the wavelength parameters respectively corresponding to the plurality of color channels; determining a plurality of color intensity parameters respectively corresponding to a plurality of target pixel points in the target image according to the target image and the transmission parameters respectively corresponding to the plurality of color images; determining a defogged image according to a plurality of color intensity parameters corresponding to the target image and the plurality of target pixel points; and determining a fire point parameter corresponding to the target area according to the defogged image. According to the method and the device, the technical problem of inaccurate fire point parameter determination caused by poor image defogging effect during thermal power parameter determination according to the image is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to a fire point parameter determination method and device and electronic equipment. BACKGROUND

[0002] In related technologies, when fire points are identified according to photographed images, the influence factors such as clouds, fog and smoke in the natural environment will affect the definition and contrast of the images, resulting in that the fire point features become unclear, and thus the images need to be defogged. However, in related technologies, there is a technical problem that the fire point parameters are determined inaccurately due to poor defogging effect.

[0003] At present, no effective solution has been proposed for the above problems. SUMMARY

[0004] The embodiments of the present application provide a fire point parameter determination method and device and electronic equipment, to at least solve the technical problem that the fire point parameters are determined inaccurately due to poor image defogging effect when the fire point parameters are determined according to images.

[0005] According to an aspect of the embodiments of the present application, a fire point parameter determination method is provided, including: acquiring a target image corresponding to a target region; determining color images corresponding to a plurality of color channels respectively according to the target image; determining transmission parameters corresponding to the plurality of color images respectively according to wavelength parameters corresponding to the plurality of color channels respectively; determining a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image respectively according to the target image and the transmission parameters corresponding to the plurality of color images respectively, wherein the corresponding plurality of color intensity parameters include color intensity parameters corresponding to the corresponding pixel points under the plurality of color channels respectively; determining a defogged image according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively; and determining fire point parameters corresponding to the target region according to the defogged image.

[0006] Optionally, the determining the transmission parameters corresponding to the plurality of color images respectively according to the wavelength parameters corresponding to the plurality of color channels respectively includes: determining an environmental feature parameter corresponding to the target image; determining scattering coefficients corresponding to the plurality of color channels respectively according to the environmental feature parameter and the wavelength parameters corresponding to the plurality of color channels respectively; determining reference scattering coefficients corresponding to the plurality of color images according to the scattering coefficients corresponding to the plurality of color channels respectively; and determining the transmission parameters corresponding to the plurality of color images respectively according to the reference scattering coefficients and the wavelength parameters corresponding to the plurality of color channels respectively.

[0007] Optionally, the determining the fire point parameter corresponding to the target region according to the defogged image comprises: calling a target model, wherein the target model is obtained by training an initial model according to sample images and a joint loss function, the joint loss function comprises a first loss sub-function and a second loss sub-function, the first loss sub-function is used to determine a difference index between a sample defogged image and a sample fog-free image, so as to reduce the difference between the defogged image and the fog-free image, and the second loss sub-function is used to determine a difference index between a predicted sample fire point parameter and an actual sample fire point parameter, so as to predict the difference between the fire point parameter and the actual fire point parameter; and determining the fire point parameter corresponding to the target region according to the defogged image and the target model.

[0008] Optionally, before the obtaining the target image corresponding to the target region, the method further comprises: obtaining an initial image corresponding to the target region, wherein the initial image comprises a plurality of regions, and the plurality of regions comprises the target region; determining region brightness features and region color features corresponding to the plurality of regions respectively; determining the target region from the plurality of regions according to the region brightness features and the region color features corresponding to the plurality of regions respectively; and adjusting the initial image according to the target region to obtain the target image.

[0009] Optionally, the determining the defogged image according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively comprises: determining a plurality of intensity influence indexes corresponding to the plurality of target pixel points respectively according to the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively, wherein the corresponding plurality of intensity influence indexes represent the influence degree between each color intensity parameter in the plurality of color intensity parameters corresponding to the corresponding target pixel point and each color intensity parameter corresponding to other target pixel points respectively; and determining the defogged image according to the target image, the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively and the intensity influence indexes.

[0010] Optionally, the determining the defogged image according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively comprises: determining a color feature parameter corresponding to a fire point; and determining the defogged image according to the target image, the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively and the color feature parameter.

[0011] Optionally, before the determining, according to the target image and the transmission parameters corresponding to the plurality of color images respectively, a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image respectively, the method further comprises: determining a plurality of initial pixel points corresponding to the target image; determining a plurality of brightness parameters corresponding to the plurality of initial pixel points respectively; and determining the plurality of target pixel points from the plurality of initial pixel points according to the plurality of brightness parameters corresponding to the plurality of initial pixel points respectively, wherein the plurality of target pixel points are initial pixel points with brightness parameters greater than a brightness threshold.

[0012] According to an aspect of some embodiments of the present application, there is provided a fire point parameter determination apparatus, comprising: an obtaining module configured to obtain a target image corresponding to a target region; a first determining module configured to determine color images corresponding to a plurality of color channels according to the target image; a second determining module configured to determine transmission parameters corresponding to a plurality of color images according to wavelength parameters corresponding to the plurality of color channels respectively; a third determining module configured to determine a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image according to the target image and the transmission parameters corresponding to the plurality of color images respectively, wherein the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively comprise color intensity parameters corresponding to the plurality of target pixel points in the plurality of color channels respectively; a fourth determining module configured to determine a dehazed image according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively; and a fifth determining module configured to determine a fire point parameter corresponding to the target region according to the dehazed image.

[0013] According to an aspect of some embodiments of the present application, there is provided an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement any of the fire point parameter determination methods described above.

[0014] According to an aspect of some embodiments of the present application, there is provided a computer readable storage medium, when instructions in the computer readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement any of the fire point parameter determination methods described above.

[0015] In the embodiment of the present application, a target image corresponding to a target region is acquired; color images corresponding to a plurality of color channels are determined according to the target image; transmission parameters corresponding to the plurality of color images are determined according to wavelength parameters corresponding to the plurality of color channels; a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image are determined according to the target image and the transmission parameters corresponding to the plurality of color images, wherein the corresponding plurality of color intensity parameters include color intensity parameters corresponding to the corresponding pixel points under the plurality of color channels; a defogging image is determined according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points; and a fire point parameter corresponding to the target region is determined according to the defogging image. According to the acquired target image, the color images corresponding to the plurality of color channels are determined, and the transmission parameters of each color image are determined, so as to quantify the influence of cloud, fog, smoke and the like under different color channels on image definition. On this basis, the color intensity parameters of each pixel point under different color channels are determined according to the target image and the transmission parameters of each color image, so as to obtain a defogging image, so as to improve image definition. According to the defogging image, the fire point parameter can be accurately determined, thereby solving the technical problem that the fire point parameter is not accurately determined due to poor image defogging effect when the fire point parameter is determined according to an image. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are included to provide a further understanding of the present application and are incorporated in and constitute a part of this application, illustrate embodiments of the present application and serve to explain the present application. In the drawings:

[0017] Figure 1 FIG. 1 is a flowchart of a fire point parameter determination method according to an embodiment of the present application;

[0018] Figure 2 FIG. 2 is a structural block diagram of a fire point parameter determination device according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of the present application.

[0020] It should be noted that the terms "first", "second", and the like in the description and claims of the application and the above drawings are used to distinguish similar objects and are not necessarily used to describe a particular sequential or chronological order. It should be understood that the data thus used can be interchanged, where appropriate, so that the embodiments of the application described herein can be carried out in sequences other than those illustrated or described herein. Furthermore, the terms "comprise" and "have", and any variations thereof, are intended to cover non-exclusive inclusion, for example, processes, methods, systems, products, or devices that include a list of steps or units not necessarily limited to those clearly listed, but can include other steps or units not clearly listed or inherent to such processes, methods, products, or devices.

[0021] First, some of the nouns or terms that appear in the description of the embodiments of the application are subject to the following explanations:

[0022] U-Net: U-Net is a convolutional neural network architecture for image segmentation tasks, consisting of an encoder (downsampling) and a decoder (upsampling), which can effectively extract image features and perform pixel-level segmentation.

[0023] L1 norm: L1 norm is a method for measuring the size of a vector or matrix, defined as the sum of the absolute values of all elements in the vector.

[0024] MobileNetV3: MobileNetV3 is a lightweight convolutional neural network architecture that significantly reduces computational complexity while maintaining high accuracy by introducing depthwise separable convolution and activation functions.

[0025] Farneback dense optical flow algorithm: Farneback dense optical flow algorithm is an algorithm for calculating the motion of pixels in a sequence of images, suitable for handling smooth motion between consecutive frames.

[0026] Kalman filter: Kalman filter is a recursive algorithm for estimating the state of a dynamic system, which minimizes the mean square error of the estimate by combining the predictive model of the system and the observation data.

[0027] Example 1

[0028] According to the embodiments of the application, an embodiment of a fire point parameter determination method is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0029] Figure 1is a flowchart of a fire point parameter determination method according to an embodiment of the present application, as shown in the figure, the method comprises the following steps: Figure 1

[0030] S102, obtaining a target image corresponding to a target region;

[0031] In step S102 provided in the present application, the target image corresponding to the target region is obtained.

[0032] Among them, the target region is the region for determining the fire parameter. For example, in the scene of fire point identification, there may be a geographical region of the fire point (such as a forest fire region).

[0033] Among them, the target image is an image containing the target region, which is used to identify the fire parameter of the target region. The target image can be an image containing the target region obtained by a drone, a monitoring camera or the like. For example, the target image may be blurred, contrast reduced and the like due to factors such as cloud, smoke and the like, which needs to be preprocessed by dehazing and the like to improve the image quality, so as to more accurately identify the fire point.

[0034] Obtaining a target image corresponding to a target region is to provide an image basis containing potential fire points, so as to accurately identify and analyze the fire point parameter subsequently.

[0035] S104, determining a color image corresponding to a plurality of color channels respectively according to the target image;

[0036] In step S104 provided in the present application, a color image corresponding to a plurality of color channels respectively is determined according to the target image.

[0037] Among them, the plurality of color channels are used to represent different color information. For example, the plurality of color channels include red (R) channel, green (G) channel and blue (B) channel, which are referred to as RGB color channels. In the RGB image, the color of each pixel point is composed of the intensity of the three channels.

[0038] Among them, the color image is an image of a specific color channel separated from the target image. For example, in the case of the target image being an RGB image, the red (R) channel image, the green (G) channel image and the blue (B) channel image are separated from the RGB image. Taking the red (R) channel image as an example, only the intensity information of the red channel is reserved in the image. By generating a color image for each color channel, the color characteristics of the flame can be more effectively analyzed. For example, the fire point may exhibit higher intensity in the red channel, and lower intensity in the blue channel.

[0039] ​By separating the image of each color channel according to the target image, the image features under different color channels can be analyzed and processed more meticulously.

[0040] S106, determine the transmission parameters corresponding to the plurality of color images according to the wavelength parameters corresponding to the plurality of color channels;

[0041] In step S106 provided in the present application, the transmission parameters corresponding to the plurality of color images are determined according to the wavelength parameters corresponding to the plurality of color channels.

[0042] The wavelength parameter is a physical quantity parameter (such as wavelength) related to the wavelength of light waves, which is used to describe the characteristics of different color light waves. For example, in the case of an RGB image, the wavelength of red light is about 650 nanometers, the wavelength of green light is about 550 nanometers, and the wavelength of blue light is about 450 nanometers. The wavelength affects the propagation and scattering characteristics of light in the atmosphere.

[0043] The transmission parameter is a parameter used to quantify the influence of clouds, smoke and the like on the image clarity under different color channels. The transmission parameter can be represented by a proportion parameter representing the intensity attenuation caused by scattering and absorption when light propagates in the atmosphere. The value of the transmission parameter is between 0 and 1. The closer the value is to 1, the higher the transmittance of light, and the clearer the objects in the image. The closer the value is to 0, the lower the transmittance of light, and the more blurred the objects in the image.

[0044] By determining the transmission parameters corresponding to the plurality of color images according to the wavelength parameters corresponding to the plurality of color channels, the propagation and scattering characteristics of color light waves corresponding to different color channels in the atmosphere can be quantified, which helps to quantify the influence of clouds, smoke and the like on the clarity of the target image.

[0045] S108, determine a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image according to the target image and the transmission parameters corresponding to the plurality of color images, wherein the corresponding plurality of color intensity parameters include color intensity parameters corresponding to the target pixel points under the plurality of color channels;

[0046] In step S108 provided in the present application, a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image are determined according to the target image and the transmission parameters corresponding to the plurality of color images.

[0047] The plurality of target pixel points are pixel points determined from the target image. In the scene of identifying mountain fire points, the target pixel points include pixel points with high brightness and color features similar to those of flames.

[0048] The color intensity parameters include color intensity of each pixel point in a corresponding color channel. For example, in an RGB color image, each pixel point has a corresponding color intensity value in a red, green, and blue channel. The value is usually an integer between 0 and 255 (for an 8-bit image). In fire point identification, the color intensity of the fire point in the red channel is usually high, and the color intensity in the blue channel is usually low. By analyzing the color intensity parameters, the fire point can be more accurately identified.

[0049] By using the transmission parameters corresponding to the target image and the plurality of color images, the scattering degree of color light in the corresponding color channel can be reflected, and the color intensity parameters of each target pixel point in each color channel can be quantified, thereby helping to determine the true color intensity of the corresponding target pixel point under the haze-free condition, and further helping to more accurately identify the fire point.

[0050] In step S110, a haze-removed image is determined according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points.

[0051] In step S110, a haze-removed image is determined according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points.

[0052] The haze-removed image is an image obtained by haze-removing the target image. By haze-removing the target image, the clarity and color accuracy of the image under the haze-free condition can be restored, thereby improving the accuracy and reliability of the fire point identification in the image.

[0053] According to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points, the haze-removed image is determined, which can accurately restore the clarity and color accuracy of the image under the haze-free condition, eliminate the influence of factors such as cloud, haze, and smoke in the atmosphere on the image quality, and improve the accuracy of subsequent determination of the fire point parameters.

[0054] In step S112, the fire point parameters corresponding to the target region are determined according to the haze-removed image.

[0055] In step S112, the fire point parameters corresponding to the target region are determined according to the haze-removed image.

[0056] The fire point parameters are parameters used to represent the characteristics of the fire point in the target region. The fire point parameters include the fire point position, the fire point intensity, and the fire point area.

[0057] The generation of the haze-removed image helps to more clearly determine the fire point parameters (such as the fire point position, the fire point intensity, and the fire point area), and effectively improves the accuracy and reliability of the fire point identification.

[0058] By the steps S102-S112, the target image corresponding to the target region is obtained; according to the target image, the color images corresponding to the plurality of color channels are determined; according to the wavelength parameters corresponding to the plurality of color channels, the transmission parameters corresponding to the plurality of color images are determined; according to the target image and the transmission parameters corresponding to the plurality of color images, the plurality of color intensity parameters corresponding to the plurality of target pixels in the target image are determined, wherein the corresponding plurality of color intensity parameters include the color intensity parameters corresponding to the corresponding pixels in the plurality of color channels; according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixels, the defogging image is determined; and according to the defogging image, the fire point parameter corresponding to the target region is determined. According to the obtained target image, the color images corresponding to the plurality of color channels are determined, and the projection parameters of each color image are determined, so as to quantify the influence of cloud, smoke and the like in different color channels on image clarity. On this basis, according to the target image and the projection parameters of each color image, the color intensity parameters of each pixel in different color channels are determined, and the defogging image is obtained, so as to improve the image clarity. According to the defogging image, the fire point parameter can be accurately determined, thereby solving the technical problem that the fire point parameter is not accurately determined due to poor image defogging effect when the fire point parameter is determined according to the image.

[0059] As an optional embodiment, according to the wavelength parameters corresponding to the plurality of color channels, the transmission parameters corresponding to the plurality of color images are determined, including: determining an environmental feature parameter corresponding to the target image; according to the environmental feature parameter and the wavelength parameters corresponding to the plurality of color channels, determining the scattering coefficients corresponding to the plurality of color channels; according to the scattering coefficients corresponding to the plurality of color channels, determining the reference scattering coefficients corresponding to the plurality of color images; and according to the reference scattering coefficients and the wavelength parameters corresponding to the plurality of color channels, determining the transmission parameters corresponding to the plurality of color images.

[0060] In this embodiment, the specific steps of determining the transmission parameters corresponding to the plurality of color images according to the wavelength parameters corresponding to the plurality of color channels are described.

[0061] The environmental feature parameter is related to a series of environmental features related to the shooting environment of the target image, such as the fog concentration, the particle number density, the refractive index and the like in the atmosphere.

[0062] The scattering coefficient is used to quantify the scattering degree of color light corresponding to different color channels in the atmosphere.

[0063] The reference scattering coefficient is determined by integrating the scattering coefficients of different color channels and is used as a reference.

[0064] In the steps involved in this embodiment, first, the environmental feature parameters corresponding to the target image are determined, and the scattering coefficients corresponding to the plurality of color channels are determined according to the environmental feature parameters and the wavelength parameters corresponding to the plurality of color channels, respectively. Then, the reference scattering coefficients corresponding to the plurality of color images are determined according to the scattering coefficients corresponding to the plurality of color channels, respectively. Finally, the transmission parameters corresponding to the plurality of color images are determined according to the reference scattering coefficients and the wavelength parameters corresponding to the plurality of color channels, respectively.

[0065] Through the above steps, the scattering coefficients of each color channel are determined by integrating the environmental feature parameters and the wavelength parameters corresponding to the plurality of color channels, which can accurately quantify the degree of influence of the atmosphere on different color channels under different environmental conditions. The reference scattering coefficients corresponding to the plurality of color images determined on this basis can adapt to the characteristics of each color channel, and the transmission parameters of each color image can be accurately determined by integrating the reference scattering coefficients and the wavelength parameters corresponding to the plurality of color channels.

[0066] As an optional embodiment, the fire point parameters corresponding to the target region are determined according to the defogged image, including: calling a target model, wherein the target model is obtained by training an initial model according to sample images and a joint loss function, the joint loss function includes a first loss sub-function and a second loss sub-function, the first loss sub-function is used to determine the difference index between the sample defogged image and the sample fog-free image, so as to reduce the difference between the defogged image and the error-free image, and the second loss sub-function is used to determine the difference index between the predicted sample fire point parameters and the actual sample fire point parameters, so as to predict the difference between the fire point parameters and the actual fire point parameters; determining the fire point parameters corresponding to the target region according to the defogged image and the target model.

[0067] In this embodiment, the specific steps of determining the fire point parameters corresponding to the target region according to the defogged image are illustrated.

[0068] The target model is involved, which is a model used to accurately identify and extract fire point parameters from a defogged image. The target model is obtained by training an initial model based on sample images and a joint loss function to improve the accuracy of the determination of the fire point parameters.

[0069] The sample image is an image data set used to train the target model, and contains images with labeled fire points and non-fire points. These images may come from different scenes and environmental conditions, helping the initial model learn the differences between fire points and backgrounds to improve the accuracy of fire point parameter determination.

[0070] The joint loss function is a comprehensive loss function used to train the target model, composed of a first loss sub-function and a second loss sub-function. The joint loss function improves the overall performance of the target model by considering both the dehazing effect and the accuracy of fire point parameter prediction.

[0071] The initial model is a preliminary model used to accurately identify and extract fire point parameters from dehazed images, including neural network models. The performance of the initial model is improved through optimization of its parameters during the training process.

[0072] The first loss sub-function is used to evaluate the difference between the dehazed image and the haze-free image.

[0073] The second loss sub-function is used to evaluate the difference between the predicted fire point parameters and the actual fire point parameters.

[0074] In the steps involved in this embodiment, the target model is invoked, and the fire point parameters corresponding to the target region are determined based on the dehazed image and the target model.

[0075] Through the above steps, the target model is trained on the initial model using the joint loss function, where the first loss sub-function evaluates the difference between the dehazed image and the haze-free image, helping to learn the image dehazing situation and analyze the dehazing effect; the second loss sub-function evaluates the difference between the predicted fire point parameters and the actual fire point parameters, improving the accuracy of fire point parameter prediction. Based on the dehazed image and the target model, the fire point parameters can be predicted in combination with the dehazing effect, which can more accurately identify the fire point position, intensity and other parameters, thereby improving the accuracy and reliability of fire point parameter determination.

[0076] As an optional embodiment, before obtaining the target image corresponding to the target region, it further includes: obtaining an initial image corresponding to the target region, wherein the initial image includes a plurality of regions, and the plurality of regions include the target region; determining region brightness features and region color features corresponding to the plurality of regions respectively; determining the target region from the plurality of regions based on the region brightness features and the region color features corresponding to the plurality of regions respectively; and adjusting the initial image based on the target region to obtain the target image.

[0077] In this embodiment, specific steps before obtaining the target image corresponding to the target region are illustrated.

[0078] In this embodiment, the initial image is involved, which is a more extensive image containing the target region before obtaining the target image. The initial image contains multiple regions, of which only part is related to the target (such as the region where the fire point may exist). For example, the initial image can include a sky region and a ground region, of which the sky region is not the region for determining the fire point parameter, but will cause interference to the ground region dehazing and fire point parameter determination, and needs to be identified and removed.

[0079] In this embodiment, multiple regions are involved, which are different parts divided in the initial image, each part has different characteristics. The multiple regions include the target region (such as the region where the fire point may exist) and other non-target regions (such as background or irrelevant regions). By analyzing the characteristics of these regions, the location of the target region can be determined.

[0080] In this embodiment, the region brightness feature is involved, which is used to describe the brightness feature (such as brightness distribution, average brightness, maximum brightness, minimum brightness, etc.) of each region in the initial image. For example, taking the initial image including the sky region and the ground region (i.e. the target region) as an example, the brightness difference between the sky region and the ground region is large, which can be accurately identified by the region brightness feature which part is the sky region and which part is the ground region.

[0081] In this embodiment, the region color feature is involved, which is the color feature (such as color intensity, color distribution, average color value, etc.) of each region in the image.

[0082] In the steps involved in this embodiment, first, the initial image corresponding to the target region is obtained, and the region brightness feature and the region color feature corresponding to the multiple regions are determined. Then, according to the region brightness feature and the region color feature corresponding to the multiple regions, the target region is determined from the multiple regions. Finally, according to the target region, the initial image is adjusted to obtain the target image.

[0083] Through the above steps, the initial image corresponding to the target region is obtained and the brightness and color features of the multiple regions are analyzed, which can effectively identify and locate the target region (such as the region where the fire point may exist), while excluding the interference of non-target regions (such as the sky region). The role of this process is to accurately distinguish the target region and the non-target region through the analysis of the brightness and color features, so as to realize the targeted adjustment of the initial image and obtain a clearer and more accurate target image.

[0084] As an optional embodiment, the method for determining the defogged image according to the target image and the color intensity parameters corresponding to the target pixels comprises: determining the intensity influence indexes corresponding to the target pixels according to the color intensity parameters corresponding to the target pixels, wherein the intensity influence indexes represent the influence degree between each color intensity parameter corresponding to the target pixel and each color intensity parameter corresponding to the other target pixels; and determining the defogged image according to the target image, the color intensity parameters corresponding to the target pixels and the intensity influence indexes.

[0085] In this embodiment, the specific steps for determining the defogged image according to the target image and the color intensity parameters corresponding to the target pixels are illustrated.

[0086] In this embodiment, the intensity influence indexes are involved, which represent the influence degree of the color intensity parameter corresponding to each target pixel on the color intensity parameter of the other target pixels, so as to reflect the mutual influence between different pixels.

[0087] In the steps involved in this embodiment, the intensity influence indexes corresponding to the target pixels are determined according to the color intensity parameters corresponding to the target pixels. Then, the defogged image is determined according to the target image, the color intensity parameters corresponding to the target pixels and the intensity influence indexes.

[0088] Through the above steps, the color intensity parameters corresponding to the target pixels are determined, which can quantize the mutual influence between different pixels, help to evaluate the relative importance and contribution of each pixel in the image, and thus help to more accurately process the color information in the image, so as to better preserve and enhance the features of the target region in the defogging process and improve the quality of the defogged image.

[0089] As an optional embodiment, the method for determining the defogged image according to the target image and the color intensity parameters corresponding to the target pixels comprises: determining the color feature parameter corresponding to the fire point; and determining the defogged image according to the target image, the color intensity parameters corresponding to the target pixels and the color feature parameter.

[0090] In this embodiment, the specific steps for determining the defogged image according to the target image and the color intensity parameters corresponding to the target pixels are illustrated.

[0091] In this embodiment, the fire point is involved, which is the flame itself or the region point where the fire occurs.

[0092] The color feature parameter is used to describe the color characteristics of the fire point, and includes intensity values of the fire point in different color channels (such as RGB), color distribution, average color value, and the like. In the fire point recognition, the flame usually shows a higher intensity in the red channel, and a lower intensity in the blue channel.

[0093] In the steps involved in this embodiment, first, the color feature parameter corresponding to the fire point is determined. The dehazed image is determined according to the target image, the plurality of color intensity parameters respectively corresponding to the plurality of target pixel points, and the color feature parameter.

[0094] By analyzing the color features of the fire point, such as the intensity distribution in different color channels, it is helpful to quantify the interference of the fire point itself on the estimation of the atmospheric light, and then to more effectively remove the fog interference in the image while retaining and enhancing the key features of the fire point region.

[0095] As an optional embodiment, before the plurality of color intensity parameters respectively corresponding to the plurality of target pixel points in the target image are determined according to the target image and the transmission parameters respectively corresponding to the plurality of color images, it includes: determining a plurality of initial pixel points corresponding to the target image; determining a plurality of brightness parameters respectively corresponding to the plurality of initial pixel points; determining the plurality of target pixel points from the plurality of initial pixel points according to the plurality of brightness parameters respectively corresponding to the plurality of initial pixel points, wherein the plurality of target pixel points are initial pixel points with brightness parameters greater than a brightness threshold.

[0096] In this embodiment, specific steps before the plurality of color intensity parameters respectively corresponding to the plurality of target pixel points in the target image are determined according to the target image and the transmission parameters respectively corresponding to the plurality of color images are described.

[0097] The plurality of initial pixel points are all pixel points in the target image, and the plurality of initial pixel points cover the entire image region.

[0098] The brightness parameter is used to describe the brightness of each pixel point. The brightness parameter can be the brightness value of the pixel point in the gray image after the target image is converted into a gray image.

[0099] In the steps involved in this embodiment, first, the plurality of initial pixel points corresponding to the target image are determined, and the plurality of brightness parameters respectively corresponding to the plurality of initial pixel points are determined. Then, the plurality of target pixel points are determined from the plurality of initial pixel points according to the plurality of brightness parameters respectively corresponding to the plurality of initial pixel points.

[0100] By determining a plurality of initial pixel points corresponding to the target image and brightness parameters thereof, and screening out target pixel points with brightness greater than a threshold according to the brightness parameters, a region possibly containing a fire point can be preliminarily screened out, so as to reduce the data amount for subsequent processing, improve the calculation efficiency, focus on a region with higher brightness, quickly locate a potential fire point, and provide a more accurate analysis basis for subsequent fire point identification and parameter determination.

[0101] Based on the above embodiments and optional embodiments, an optional implementation is provided, which is specifically described below.

[0102] In the related art, when a fire point is identified according to a photographed image, the influence factors such as cloud, fog and smoke in the natural environment will affect the definition and contrast of the image, so that the fire point features become blurred, and thus the image needs to be de-fogged. However, in the related art, there is a technical problem that the fire point parameters are not accurately determined due to poor de-fogging effect.

[0103] At present, no effective solution has been proposed for the above problems.

[0104] In view of this, the optional implementation of the present application provides a fire point parameter determination method, which can also be referred to as a cloud and fog interference identification method for mountain fire point identification, which can effectively solve the technical problem that the fire point parameters are not accurately determined due to poor de-fogging effect when the fire point parameters are determined according to the image in the related art.

[0105] A cloud and fog interference identification method for mountain fire point identification, which improves the color attenuation prior model, mainly including two parts of wavelength-dependent transmittance model and robustness optimization of atmospheric light estimation, then constructs a fog-flame joint segmentation framework, and optimizes and spatiotemporal checks the dynamic threshold. The method comprises the following steps:

[0106] S1, improved model of color attenuation prior;

[0107] Step S1 distinguishes the scattering characteristics of fire (long wavelength) and fog (short wavelength) by introducing wavelength dependence, to avoid spectral confusion. The improved model mainly includes two parts: wavelength-dependent transmittance model and robustness optimization of atmospheric light estimation, which specifically comprises the following steps:

[0108] S11, wavelength-dependent transmittance model.

[0109] According to the Rayleigh scattering theory, the relationship between the scattering coefficient β c and the wavelength λ c is:

[0110]

[0111] wherein:

[0112] β c denotes the scattering coefficient corresponding to color channel c;

[0113] λ c denotes the wavelength (may be the central wavelength) corresponding to color channel c;

[0114] n denotes the refractive index of the medium;

[0115] N denotes the particle number density;

[0116] ρ is the polarization factor.

[0117] The magnitude of the scattering coefficient increases sharply with the decrease of the wavelength. Light rays of different wavelengths are scattered to different degrees, which has an important influence on the propagation of light in the atmosphere, and further affects the visibility in the image. The target image corresponding to the target region is obtained; according to the target image, the color images corresponding to the plurality of color channels are determined respectively in the color image processing corresponding to the plurality of color channels. In the color image processing corresponding to the plurality of color channels, the red-green-blue (RGB) three-channel color model can be used.

[0118] For the RGB three-channel (the same as the above-mentioned plurality of color channels), λ R = 650 nm, λ G = 550 nm, and λ B = 450 nm, the transmittance satisfies:

[0119] t R (x) < t G (x) < t B (x)

[0120] Wherein:

[0121] λ R denotes the wavelength corresponding to the red (R) channel in the RGB channel;

[0122] λ G denotes the wavelength corresponding to the green (G) channel in the RGB channel;

[0123] λ B denotes the wavelength corresponding to the blue (B) channel in the RGB channel;

[0124] t R (x) denotes the transmittance of pixel point x in the red (R) channel image in the RGB channel;

[0125] t G (x) denotes the transmittance of pixel point x in the green (G) channel image in the RGB channel;

[0126] t B(x) represents the transmittance of pixel x in the image of the blue (B) channel in the RGB channel.

[0127] Transmittance t of the RGB channel c (x) (same as the transmittance parameter above) can be calculated by the following expression:

[0128]

[0129] Where:

[0130] t c (x) represents the transmittance of pixel x in the image corresponding to color channel c;

[0131] β base is the reference scattering coefficient (same as the reference scattering coefficient above) which can be determined by calibration experiments;

[0132] d(x) is the scene depth (meters) which is obtained by monocular depth estimation network (MiDaS, etc.).

[0133] In practical applications, the reference scattering coefficient β base needs to be calibrated first. The calibration experiment can be completed by shooting a standard object under known distance and known atmospheric conditions. By calculating the transmittance of the standard object at different distances, the value of the reference scattering coefficient β base can be obtained. Then, using this reference scattering coefficient, combined with the wavelength-dependent transmittance model, the transmittance of each pixel in the haze image can be more accurately estimated.

[0134] S12, robustness optimization of atmospheric light estimation.

[0135] The main purpose of robustness optimization of atmospheric light estimation is to exclude the interference of flame pixels and accurately estimate the global atmospheric light:

[0136] A = (A R , A G , A B )

[0137] Where:

[0138] A represents the atmospheric light;

[0139] A R represents the atmospheric light corresponding to the red (R) channel in the RGB channel;

[0140] A G represents the atmospheric light corresponding to the green (G) channel in the RGB channel;

[0141] A B represents the atmospheric light corresponding to the blue (B) channel in the RGB channel.

[0142] Atmospheric light estimation is a key step in the defogging algorithm. In traditional defogging algorithms, the atmospheric light is often estimated by selecting the brightest region in the image. Due to the existence of extreme brightness regions such as flames and bright spots, this method may lead to estimation errors. Therefore, a robust optimization method is proposed to improve the estimation of atmospheric light by combining brightness screening, flame mask filtering, and median filtering. The method mainly includes the following three steps:

[0143] A1, brightness screening:

[0144] Assuming that the atmospheric light usually appears in the brightest region of the image, the brightness (same as the brightness parameter described above) is screened, the brightness L(x) of each pixel is calculated, and the top 0.1% of the brightness is selected as the candidate set (same as the above multiple initial pixel points):

[0145] L(x) = 0.299R(x) + 0.587G(x) + 0.114B(x)

[0146]

[0147] Where:

[0148] L 0.999 represents the brightness threshold.

[0149] A2, flame mask filtering:

[0150] The purpose of this step is to remove the interference of the flame region on the estimation of the atmospheric light, because the brightness of the flame region is very high, but it is not the atmospheric light. Define the flame color criterion to obtain the modified candidate set (same as the above multiple target pixel points).

[0151]

[0152] Where:

[0153] M fire (x) is a flag indicating whether pixel point x belongs to the flame region. If pixel point x satisfies a specific color condition, it is considered to belong to the flame region, and the value is 1, otherwise, the value is 0;

[0154] R(x) represents the color intensity parameter of pixel point x in the red channel of the RGB channel;

[0155] G(x) represents the color intensity parameter of pixel point x in the green channel of the RGB channel;

[0156] B(x) represents the color intensity parameter of pixel point x in the blue channel of the RGB channel.

[0157] A3, median filtering:

[0158] To avoid high-brightness pixels from flames contaminating atmospheric light estimation and to improve the color fidelity of the dehazed image, it is necessary to... The median RGB value of the middle pixel is taken to obtain atmospheric light A = (A R A G A B Median filtering is a classic denoising method that effectively removes the influence of outliers on atmospheric light estimation. During the estimation process, applying median filtering to the RGB values ​​yields a more robust atmospheric light estimate.

[0159] S2, a joint segmentation framework for haze and flame;

[0160] Step S2 specifically includes the following steps:

[0161] S21, construct a dual-branch network structure.

[0162] The dual-branch network structure in step 21 consists of two parts: a defogging branch and a fire point splitting branch, as follows:

[0163] B1, Defogging Branch:

[0164] The dehazing branch uses a lightweight U-Net structure, with the input fog map I. hazy The output transmittance map t(x) (same as the transmittance parameters mentioned above, which may include transmittance and the transmittance map) and atmospheric light A are calculated. The network loss function (same as the first loss function mentioned above) is:

[0165]

[0166] in:

[0167] This represents the loss value of the dehazing network, used to measure the predicted dehazed image J. pred Compared to actual fog-free images J GT The differences between them;

[0168] α represents the contribution of SSIM and L1 norm to the loss function, and can be 0.85;

[0169] SSIM(J pred J GT ) indicates J pred With J GT Structural similarity index;

[0170] J pred This represents the dehazed image predicted by the dehazing network;

[0171] J GT This represents the actual fog-free image.

[0172] SSIM (Structural SIMilarity) is defined as:

[0173]

[0174] wherein:

[0175] SSIM(p,q) represents the structural similarity index between image p and image q;

[0176] μ p represents the mean value of image p;

[0177] μ q represents the mean value of image q;

[0178] C1 represents a stabilization constant, C1 = (0.01·255) 2 ;

[0179] σ xy represents the covariance of image p and image q;

[0180] C2 represents a stabilization constant, C2 = (0.03·255) 2 ;

[0181] represents the variance of image p;

[0182] represents the variance of image q.

[0183] U-Net is a classic convolutional neural network structure, widely used in image segmentation tasks. Its characteristics are the encoder-decoder structure, which can effectively extract the features of the image and perform upsampling. In the dehazing task, U-Net can learn the mapping relationship between the haze image and the clear image, so as to restore the clear image.

[0184] B2, fire point segmentation branch:

[0185] The fire point segmentation branch is based on the encoder-decoder structure of MobileNetV3, and the input is the dehazing image J pred , and the output is the fire point probability map P fire (x). The loss function (same as the second loss sub-function described above) is weighted cross entropy:

[0186]

[0187] wherein:

[0188] represents the loss value of fire point segmentation, used to measure the predicted fire point probability map P fire (x) and the actual fire point label map Y labthe difference between (x) and (x) ;

[0189] w(x) represents the weight used to balance positive and negative samples in the weighted cross-entropy, w(x) = 1 + 4·Y lab (x) and (x) ;

[0190] Y lab (x) represents the actual fire point label at the pixel point x;

[0191] x represents the pixel position in the image.

[0192] MobileNetV3 is a lightweight convolutional neural network suitable for mobile devices and embedded systems. By introducing depth separable convolution and efficient activation function, it can significantly reduce the computational complexity while maintaining high accuracy. In the fire point segmentation task, MobileNetV3 can effectively extract flame features in the defogged image, thereby realizing accurate segmentation of fire points.

[0193] S22, generating a joint training strategy.

[0194] Step 22: Joint training strategy:

[0195] In order to make the defogging process directly serve the fire point segmentation and improve the overall system robustness, the present application adopts a joint training strategy. Its total loss function (same as the above-mentioned joint loss function) is:

[0196]

[0197] Among them:

[0198] Total loss value represents the total loss value, which is used to comprehensively evaluate the performance of the defogging and fire point segmentation network.

[0199] Total variation loss represents the total variation loss, which is used to constrain the smoothness of the transmittance map.

[0200] Among them, the defogging loss (same as the first loss sub-function above) formula is:

[0201]

[0202] The segmentation loss (focal loss, which alleviates class imbalance, same as the second loss sub-function above) formula is:

[0203]

[0204] Among them:

[0205] α w Loss parameter α represents the loss parameter, αw = 0.25, for balancing sample weights, suppressing easy-to-classify samples;

[0206] γ represents a loss parameter, γ = 2, for balancing sample weights, suppressing easy-to-classify samples.

[0207] TV regular term, i.e., constraint the smoothness of transmittance:

[0208]

[0209] wherein:

[0210] represents the gradient of the transmittance map in the horizontal direction of the coordinate axis (i.e., the horizontal direction);

[0211] represents the gradient of the transmittance map in the vertical direction of the coordinate axis (i.e., the vertical direction);

[0212] represents a normalization factor, for averaging the loss value to each pixel;

[0213] t(x) represents the transmittance at the position of pixel x.

[0214] S3, dynamic threshold optimization and space-time verification.

[0215] Step S3 specifically comprises the following steps:

[0216] S31, adaptive color space conversion;

[0217] Convert the defogged image to the color (YCbCr) space, so as to better separate the brightness information and the chrominance information. Through this conversion, the brightness and color contrast of the image can be adjusted more accurately:

[0218] Y = 0.299R + 0.587G + 0.114B

[0219] Cb = 128 - 0.168736R - 0.331264G + 0.5B

[0220] Cr = 128 + 0.5R - 0.418688G - 0.081312B

[0221] wherein:

[0222] Y represents the brightness component, reflecting the light and dark degree of the image;

[0223] Cb represents the chrominance component, reflecting the difference between blue and green, and is used for processing color information related to blue;

[0224] Cr represents the chroma component, reflecting the difference between red and blue, and is used to process color information related to red;

[0225] R represents the intensity value of the red channel (same as the color intensity parameter described above);

[0226] G represents the intensity value of the green channel (same as the color intensity parameter described above);

[0227] B represents the intensity value of the blue channel (same as the color intensity parameter described above).

[0228] S32, dynamic threshold calculation.

[0229] When processing haze images, due to the large difference in brightness between the sky area and the ground area, the influence of the sky area needs to be excluded. By setting a dynamic threshold to detect the brightness distribution of the non-sky area, the segmentation effect of the image is optimized. Exclude the sky area (detected by Cb<120 and Cr>150), calculate the statistics of the non-sky area:

[0230] μ Y = mean(Y non-sky ), σ Y = std(Y non-sky )

[0231] μ Cr = mean(Cr non-sky ), σ Cr = std(Cr non-sky )

[0232] Where:

[0233] μ Y represents the mean of the brightness component Y of the non-sky area;

[0234] σ Y represents the standard deviation of the brightness component Y of the non-sky area;

[0235] μ Cr represents the mean of the red chroma component Cr of the non-sky area;

[0236] σ Cr represents the standard deviation of the red chroma component Cr of the non-sky area;

[0237] Y non-sky represents the brightness component of the non-sky area;

[0238] Cr non-sky represents the chroma component of the non-sky area.

[0239] Set the dynamic threshold:

[0240] Y th= μ Y + k w · σ Y

[0241] Cr th = μ Cr - k w · σ Cr

[0242] wherein:

[0243] Y th denotes a dynamic threshold value for the luminance component Y;

[0244] Cr th denotes a dynamic threshold value for the chrominance component Cr;

[0245] k w denotes a sensitivity for adjusting the dynamic threshold value.

[0246] k w is adaptively adjusted according to the fog density τ:

[0247]

[0248] τ>0.7 can be regarded as thick fog. In the thick fog scene, the threshold value is increased to suppress false alarms, and in the thin fog scene, the threshold value is decreased to improve recall rate.

[0249] S33, timing consistency verification.

[0250] In order to ensure the stability of the fire point in the continuous frames, the video timing information is used to exclude transient interference (such as reflection, flying birds) and the like, the optical flow field calculation method is adopted, the motion vector between adjacent frames is calculated through the Farneback algorithm. The trajectory filtering is performed through Kalman filtering, and the detection accuracy of the fire point is further improved. If the motion trajectory meets the stability requirement, it is determined as a stable fire point.

[0251] Optical flow field calculation: Farneback dense optical flow algorithm is adopted to calculate the motion vector field between adjacent frames The formula is:

[0252]

[0253] wherein:

[0254] denotes a spatial gradient of the ti-th frame (i.e., time);

[0255] denotes a motion vector component in the horizontal direction of the coordinate axis (i.e., horizontal direction); denotes a motion vector component in the vertical direction of the coordinate axis (i.e., vertical direction);I ti+1an image at time ti+1;

[0256] I ti an image at time ti.

[0257] Motion trajectory filtering is performed using Kalman filtering with state equations:

[0258]

[0259] wherein:

[0260] SX ti a state vector at time ti;

[0261] SX ti-1 a state vector at time ti-1.

[0262] Δt represents a time step;

[0263] w eti a process noise vector at time ti, which can be considered as Gaussian white noise.

[0264] The state vector SX is represented as:

[0265]

[0266] wherein:

[0267] aixs x represents a position in the direction of the horizontal axis of the coordinate system (i.e., the horizontal direction);

[0268] aixs y represents a position in the direction of the vertical axis of the coordinate system (i.e., the vertical direction);

[0269] represents a velocity component in the direction of the horizontal axis of the coordinate system (i.e., the horizontal direction);

[0270] represents a velocity component in the direction of the vertical axis of the coordinate system (i.e., the vertical direction).

[0271] The observation matrix VH is represented as:

[0272]

[0273] Trajectory matching is performed and is represented as:

[0274]

[0275] wherein:

[0276] S track is a trajectory similarity;

[0277] T is the total time;

[0278] P t is the actual observed position;

[0279] is the Kalman filter predicted position;

[0280] σ P is the Gaussian kernel bandwidth.

[0281] If S track > 0.7, it is determined that the stable fire point.

[0282] Through the above optional implementation, at least the following beneficial effects can be achieved:

[0283] (1) Compared with the related art, the present application determines the color image corresponding to each color channel according to the obtained target image, and determines the projection parameter of each color image, to quantify the influence of cloud, smoke and the like under different color channels on image clarity. On this basis, the color intensity parameter of each pixel point under different color channels is determined according to the target image and the projection parameter of each color image, to obtain a dehazed image, so as to improve the image clarity. According to the dehazed image, the fire point parameter can be accurately determined, thereby solving the technical problem that the fire point parameter is not accurately determined due to poor dehazing effect when the fire point parameter is determined according to the image.

[0284] (2) Compared with the related art, the present application determines the scattering coefficient of each color channel by comprehensively considering the environmental characteristic parameter and the wavelength parameter corresponding to each color channel, so that the influence degree of the atmosphere on different color channels under different environmental conditions can be accurately quantified. On this basis, the reference scattering coefficient corresponding to the multiple color images can adapt to the characteristics of each color channel, and the transmission parameter of each color image can be accurately determined by comprehensively considering the reference scattering coefficient and the wavelength parameter corresponding to each color channel.

[0285] (3) Compared with the related art, the present application trains the initial model by using the joint loss function to obtain the target model. The first loss sub-function evaluates the difference between the dehazed image and the haze-free image, which is helpful for learning the image dehazing condition and analyzing the dehazing effect. The second loss sub-function evaluates the difference between the predicted fire point parameter and the actual fire point parameter, which improves the accuracy of the fire point parameter prediction. The fire point parameter is determined based on the dehazed image and the target model, which can predict the fire point parameter in combination with the dehazing effect, so that the fire point position, intensity and other parameters can be more accurately recognized, thereby improving the accuracy and reliability of the fire point parameter determination.

[0286] (4) Compared with the related art, the present application can quantize the mutual influence between different pixel points by determining the multiple color intensity parameters corresponding to the multiple target pixel points, which helps to evaluate the relative importance and contribution of each pixel point in the image, thereby helping to more accurately process the color information in the image to better preserve and enhance the features of the target region in the defogging process, and improve the quality of the defogged image.

[0287] (5) Compared with the related art, the present application can preliminarily filter out the region that may contain the fire point as the data amount for subsequent reduction processing by determining the multiple initial pixel points and their brightness parameters corresponding to the target image, and filtering out the target pixel points with brightness greater than the threshold according to the brightness parameters, thereby improving the calculation efficiency, focusing on the region with higher brightness, and quickly locating the potential fire point to provide a more accurate analysis basis for subsequent fire point identification and parameter determination.

[0288] (6) Compared with the related art, the present application can significantly improve the quality of the defogged image by proposing an improved wavelength-dependent transmittance model and a robustly optimized atmospheric light estimation method. In the atmospheric light estimation process, the traditional method is avoided due to the error estimation caused by factors such as flame and extreme bright spots, ensuring the accuracy and stability of the defogging effect. Moreover, the proposed fog and fire point joint segmentation framework can simultaneously process the fog and fire point tasks in the image. Through the dual-branch network structure, defogging and fire point segmentation can be cooperatively optimized in a unified framework, avoiding the cumbersome of separately processing these two problems in the traditional method, which is suitable for images with multiple interference sources in complex environments, and has strong practicality and flexibility. In addition, the lightweight network structure is adopted, which not only ensures efficient processing performance, but also reduces the consumption of computing resources, and compared with the traditional complex network structure, it can achieve better defogging and segmentation effect at lower computational cost.

[0289] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily necessary for the present application.

[0290] Those skilled in the art can clearly understand the method according to the above-mentioned embodiments can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a plurality of instructions to make a terminal device (may be a mobile phone, computer, server, or network device, etc.) execute the method of each embodiment of the present application.

[0291] Embodiment 2

[0292] According to the embodiments of the present application, a device for implementing the above-mentioned fire point parameter determination method is further provided, Figure 2 is a structural block diagram of the fire point parameter determination device of the embodiments of the present application, as Figure 2 shown, the device comprises: an acquisition module 202, a first determination module 204, a second determination module 206, a third determination module 208, a fourth determination module 210 and a fifth determination module 212, which will be described in detail below.

[0293] The acquisition module 202 is configured to acquire a target image corresponding to a target region; the first determination module 204 is connected to the acquisition module 202 and configured to determine, according to the target image, color images corresponding to a plurality of color channels respectively; the second determination module 206 is connected to the first determination module 204 and configured to determine, according to the plurality of color channels respectively corresponding wavelength parameters, transmission parameters corresponding to the plurality of color images respectively; the third determination module 208 is connected to the second determination module 206 and configured to determine, according to the target image and the transmission parameters corresponding to the plurality of color images respectively, a plurality of color intensity parameters corresponding to a plurality of target pixel points in the target image, wherein the corresponding plurality of color intensity parameters comprise color intensity parameters corresponding to the corresponding pixel points in the plurality of color channels respectively; the fourth determination module 210 is connected to the third determination module 208 and configured to determine, according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixel points respectively, a defogging image; and the fifth determination module 212 is connected to the fourth determination module 210 and configured to determine, according to the defogging image, a fire point parameter corresponding to the target region.

[0294] It should be noted that the above obtaining module 202, the first determining module 204, the second determining module 206, the third determining module 208, the fourth determining module 210 and the fifth determining module 212 correspond to steps S102 to S112 in the fire point parameter determination method, and the modules and the corresponding steps have the same instances and application scenarios, but are not limited to the above embodiment 1.

[0295] Embodiment 3

[0296] According to another aspect of the embodiments of the present application, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions, wherein the processor is configured to execute the instructions to implement the fire point parameter determination method of any of the above.

[0297] Embodiment 4

[0298] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, when the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the fire point parameter determination method of any of the above.

[0299] The above-mentioned embodiment numbers of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0300] In the above embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0301] In several embodiments provided in the present application, it should be understood that the disclosed technical contents can be implemented by other ways. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the units can be a logical function division, and actual implementation can have another division way, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed units can be indirect coupling or communication connection through some interfaces, units or modules, which can be electrical or other forms.

[0302] The units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed to multiple units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment scheme.

[0303] In addition, each function unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software function unit.

[0304] When the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0305] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A method of determining a fire point parameter, characterized by, The method comprises: acquiring a target image corresponding to a target region; determining color images corresponding to a plurality of color channels according to the target image; determining transmission parameters corresponding to the plurality of color images according to wavelength parameters corresponding to the plurality of color channels; determining a plurality of color intensity parameters corresponding to a plurality of target pixels in the target image according to the target image and the transmission parameters corresponding to the plurality of color images, wherein the corresponding plurality of color intensity parameters comprise color intensity parameters corresponding to the corresponding pixels under the plurality of color channels; determining a defogging image according to the target image and the plurality of color intensity parameters corresponding to the plurality of target pixels; determining a fire point parameter corresponding to the target region according to the defogging image.

2. The method of claim 1, wherein, The method comprises: determining an environmental feature parameter corresponding to the target image; determining scattering coefficients corresponding to the plurality of color channels according to the environmental feature parameter and wavelength parameters corresponding to the plurality of color channels; determining reference scattering coefficients corresponding to the plurality of color images according to the scattering coefficients corresponding to the plurality of color channels; determining transmission parameters corresponding to the plurality of color images according to the reference scattering coefficients and wavelength parameters corresponding to the plurality of color channels.

3. The method of claim 1, wherein, The method comprises: calling a target model, wherein the target model is obtained by training an initial model according to a sample image and a joint loss function, the joint loss function comprises a first loss sub-function and a second loss sub-function, the first loss sub-function is used to determine a difference index between a sample defogging image and a sample fog-free image to reduce the difference between the defogging image and the error-free image, and the second loss sub-function is used to determine a difference index between a predicted sample fire point parameter and an actual sample fire point parameter to predict the difference between the fire point parameter and the actual fire point parameter; determining a fire point parameter corresponding to the target region according to the defogging image and the target model.

4. The method of claim 1, wherein, Before the step of acquiring a target image corresponding to a target region, the method further comprises: acquiring an initial image corresponding to the target region, wherein the initial image comprises a plurality of regions, and the plurality of regions comprise the target region; determining region brightness features and region color features corresponding to the plurality of regions; determining the target region from the plurality of regions according to the region brightness features and the region color features corresponding to the plurality of regions; adjusting the initial image according to the target region to obtain the target image.

5. The method of claim 1, wherein, The method comprises: Determine, according to the color intensity parameters corresponding to the target pixels, intensity influence indexes corresponding to the target pixels, wherein the intensity influence indexes represent the influence degree between each color intensity parameter of the corresponding target pixel and each color intensity parameter of other target pixels; Determine, according to the target image, the color intensity parameters corresponding to the target pixels and the intensity influence indexes, a defogging image.

6. The method of claim 1, wherein, The method for determining the defogging image according to the target image and the color intensity parameters corresponding to the target pixels comprises: Determine a color feature parameter corresponding to the fire point; Determine, according to the target image, the color intensity parameters corresponding to the target pixels, the intensity influence indexes and the color feature parameter, a defogging image.

7. The method according to any one of claims 1 to 6, characterized in that, Before determining the color intensity parameters corresponding to the target pixels according to the target image and the transmission parameters corresponding to the color images, the method comprises: Determine a plurality of initial pixels corresponding to the target image; Determine brightness parameters corresponding to the initial pixels; Determine, according to the brightness parameters corresponding to the initial pixels, a plurality of target pixels from the initial pixels, wherein the target pixels are the initial pixels with brightness parameters greater than a brightness threshold.

8. An ignition point parameter determination apparatus characterized by comprising: The method comprises: An acquisition module configured to acquire a target image corresponding to a target region; A first determination module configured to determine color images corresponding to a plurality of color channels according to the target image; A second determination module configured to determine transmission parameters corresponding to the color images according to wavelength parameters corresponding to the color channels; A third determination module configured to determine color intensity parameters corresponding to a plurality of target pixels in the target image according to the target image and the transmission parameters corresponding to the color images, wherein the color intensity parameters comprise color intensity parameters corresponding to the target pixels in the color channels; A fourth determination module configured to determine a defogging image according to the target image and the color intensity parameters corresponding to the target pixels; A fifth determination module configured to determine a fire point parameter corresponding to the target region according to the defogging image.

9. An electronic device, comprising: The method comprises: A processor; A memory configured to store instructions executable by the processor; The processor is configured to execute the instructions to implement the method for determining the fire point parameter according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, When the instructions in the computer readable storage medium are executed by the processor of the electronic device, the electronic device can execute the method for determining the fire point parameter according to any one of claims 1 to 7.

Images