JP2022545962A - Fog Recognition Method and Apparatus, Electronic Device, Storage Medium and Computer Program Product - Google Patents

Abstract

The present invention provides a fog cloud recognition method and apparatus, an electronic device, a storage medium, and a computer program product, the method comprising the steps of acquiring a scene image of a target scene, performing feature extraction on the scene image, and performing global feature and performing a classification process on the global feature information to determine agglomeration in the scene image and a corresponding agglomeration level.

Description

(Cross reference to related applications)
This application is filed based on a Chinese patent application with application number 202010699922.1, dated July 17, 2020, and titled "Fog Recognition Method and Apparatus, Electronic Device, Storage Medium" and claiming priority of the Chinese patent application, which is incorporated herein by reference in its entirety.

TECHNICAL FIELD The present disclosure relates to the technical field of computer vision, and in particular to fogging recognition methods and apparatus, electronic devices, storage media, and computer program products.

"Column fog", also called mass fog, is a fog with lower visibility that occurs within a local range of several tens to hundreds of meters in large fog under the influence of the microclimate environment of the local area. . Fog coagulation is characterized by strong locality and is difficult to predict and forecast. prone to traffic accidents.

At present, it is possible to recognize fog cloud using dedicated hardware such as photoelectric sensors and cameras. A solution that captures image information with a camera and recognizes fog clusters has a higher feasibility than dedicated hardware solutions that are expensive and difficult to deploy on a large scale.

However, the solution to realize fog cloud recognition with camera often adopts methods such as dark channel prior probability or manual feature selection, which is to obtain shallow layer features of the image from the image collected by the camera. is susceptible to factors such as light rays and camera angles, so its recognition effect is low.

Embodiments of the present disclosure aim to provide a method and apparatus for recognizing fog, an electronic device, a storage medium, and a computer program product.

The technical solutions of the embodiments of the present disclosure are implemented as follows.

Embodiments of the present disclosure provide a method for recognizing fog. The fog recognition method includes:
obtaining a scene image of a target scene;
performing feature extraction on the scene image to obtain global feature information;
performing a classification process on the global feature information to determine fog in the scene image and a fog level corresponding to the fog.

In the above fog cloud recognition method, before performing feature extraction on the scene image to obtain global feature information, the method includes:
further comprising performing image preprocessing on the scene image to obtain a processed image;
Correspondingly, the step of performing feature extraction on the scene image to obtain global feature information includes:
performing feature extraction on the processed image to obtain the global feature information.

In the above fog cloud recognition method, the step of performing image preprocessing on the scene image to obtain a processed image includes:
pixel sampling the scene image to obtain a sampled image of a target size;
normalizing the sampled image to obtain the processed image.

In the above method for recognizing fog, the step of performing a classification process on the feature information of all stations to determine fog in the scene image and a fog level corresponding to the fog,
performing average pooling on the all-station feature information to obtain all-station average feature information;
classifying the global mean feature information to determine the fog and the fog level.

In the above method for recognizing fog, after determining fog in the scene image and a fog level corresponding to the fog, the method includes:
obtaining a first visibility range corresponding to the fog level based on a preset correspondence relationship between the fog level and the visibility range;
obtaining first alarm information corresponding to the fog level if the first visibility range is lower than a preset visibility range;
and outputting the first alarm information.

In the above fog cloud recognition method, the scene image includes a plurality of images at a plurality of times, the fog cloud level includes a plurality of levels corresponding to the plurality of images, and the fog cloud in the scene image and the fog cloud After determining the fog level corresponding to the method, the method includes:
obtaining second alarm information if the plurality of levels and the plurality of times are positively correlated;
obtaining an alert reminder device within the target scene;
and sending the second alert information to the alert reminder device.

In the above fog cloud recognition method, the scene image includes a plurality of images at a plurality of times, the fog cloud level includes a plurality of levels corresponding to the plurality of images, and the fog cloud in the scene image and the fog cloud After determining the fog level corresponding to the method, the method includes:
Statisticalizing the plurality of levels to obtain statistical results;
and determining a fog aggregation frequency of the target scene based on the statistical results.

Embodiments of the present disclosure provide a fog recognition system. The fog recognition device includes:
an image capture module configured to capture a scene image of a target scene;
a feature extraction module configured to perform feature extraction on the scene image to obtain global feature information;
a classification processing module configured to perform a classification process on the global feature information to determine fog in the scene image and a fog level corresponding to the fog.

The fog recognition device further includes an image processing module,
the image processing module is configured to perform image pre-processing on the scene image to obtain a processed image;
Correspondingly, the feature extraction module is specifically configured to perform feature extraction on the processed image to obtain the global feature information.

In the above apparatus for recognizing fog, the image processing module specifically performs pixel sampling on the scene image to obtain a sampled image of a target size, normalizes the sampled image, and processes the configured to obtain a finished image.

In the above fog cluster recognition apparatus, the classification processing module specifically performs average pooling on the all-station feature information to obtain all-station average feature information, classifies the all-station average feature information, and classifies the group. configured to determine fog and said fog level;

The fog recognition device further includes an information output module,
The information output module obtains a first visibility range corresponding to the fog level based on a correspondence relationship between a preset fog level and a visibility range, and the first visibility range is the preset visibility range. If it is lower, it is configured to obtain first warning information corresponding to said fog level and output said first warning information.

The apparatus for recognizing fog fog further comprises an information output module, wherein the scene image comprises a plurality of images at a plurality of times, the fog level comprises a plurality of levels corresponding to the plurality of images,
The information output module obtains second alarm information when the plurality of levels and the plurality of times are positively correlated, obtains alarm alerting devices in the target scene, and obtains the second alarm. configured to transmit information to the alarm reminder device;

The fog cloud recognition apparatus further includes a statistical processing module, the scene image includes a plurality of images at a plurality of times, the fog cloud level includes a plurality of levels corresponding to the plurality of images,
The statistic processing module is configured to statisticize the plurality of levels, obtain a statistic result, and determine a fogging frequency of the target scene based on the statistic result.

An embodiment of the disclosure provides an electronic device. the electronic device includes a processor, memory and a communication bus;
the communication bus is configured to provide connection communication between the processor and the memory;
The processor is configured to implement the above method for recognizing fog by executing a fog confluence recognition program stored in the memory.

Embodiments of the disclosure provide a computer-readable storage medium. The computer-readable storage medium stores one or more programs capable of realizing the above-described fog swarm recognition method by being executed by one or more processors.

An embodiment of the disclosure provides a computer program product. The computer program product comprises a computer program or instructions which, when executed on a computer, cause the computer to perform the above method for recognizing fog.

Embodiments of the present disclosure provide a fog cloud recognition method and apparatus, an electronic device, a storage medium and a computer program product, the method comprising obtaining a scene image of a target scene and performing feature extraction on the scene image. , obtaining global feature information, and performing a classification process on the global feature information to determine fog and fog level corresponding to the fog in the scene image. The technical solution provided by the embodiments of the present disclosure is to extract the global depth features of the image, improve the effectiveness of the information representing fog coagulation, and prevent the interference of information irrelevant to fog coagulation recognition in the image. is reduced and the accuracy of cloud cluster recognition is improved.

4 is a flow chart of a fog cloud recognition method provided by an embodiment of the present disclosure; 1 is a schematic diagram of an exemplary image processing process provided by embodiments of the present disclosure; FIG. FIG. 1 is a block diagram 1 of a fog cloud recognition device provided by an embodiment of the present disclosure; FIG. 2 is a block diagram 2 of a fog cloud recognition device provided by an embodiment of the present disclosure; 1 is a configuration diagram of an electronic device provided by an embodiment of the present disclosure; FIG.

In the above, the drawings used in the embodiments of the present disclosure or the background art have been described in order to more clearly describe the technical solutions in the embodiments of the present disclosure or the background art.

The above drawings are incorporated into and constitute a part of the specification, and these drawings show embodiments consistent with the present disclosure and are used together with the specification to explain the technical solutions of the present disclosure.

The following clearly and completely describes the technical solutions in the embodiments of the present disclosure with reference to the drawings in the embodiments of the present disclosure.

Embodiments of the present disclosure provide a method for recognizing fog, the execution body of which can be a device for recognizing fog, and the method for recognizing fog can be applied to, for example, User Equipment (UE), mobile devices , user terminals, terminals, cellular telephones, cordless telephones, personal digital assistants (abbreviated as PDA), portable devices, computing devices, in-vehicle devices, wearable devices, or other terminal devices or servers or other electronic devices You may In some possible embodiments, the method of recognizing fog may be implemented by a processor invoking computer readable instructions stored in memory.

Embodiments of the present disclosure provide a method for recognizing fog. FIG. 1 is a flow chart of a fog cloud recognition method provided by an embodiment of the present disclosure. As shown in FIG. 1, the fog cluster recognition method mainly includes the following steps.

At S101, a scene image of a target scene is acquired.

In embodiments of the present disclosure, the fog cluster recognizer can acquire a scene image of the target scene.

By way of explanation, in the embodiments of the present disclosure, the scene image is the image corresponding to the scene requiring fog cluster recognition, ie, the target scene. When the fog coagulation method is applied to the highway surveillance scene, the scene image is actually the highway surveillance image, and the subsequent fog coagulation recognition is not limited to dedicated hardware, and can be applied on a large scale. be. A specific target scene and a scene image of the target scene are not limited in the embodiments of the present disclosure.

By way of illustration, in embodiments of the present disclosure, scene images may be collected by a fog confluence recognizer, captured by a separate camera, server, cloud, or other device and sent to the confluence recognizer. It may be an image. The specific scene image supplier is not limited in the embodiments of the present disclosure.

In S102, feature extraction is performed on the scene image to obtain all-station feature information.

In an embodiment of the present disclosure, after obtaining a scene image, the fog cluster recognizer can perform feature extraction on the scene image to obtain global feature information in the scene image.

By way of illustration, in embodiments of the present disclosure, the fog cloud recognizer performs image pre-processing on the scene image to perform feature extraction on the scene image and before obtaining global feature information, the processed image Correspondingly, the step of the fog cluster recognizer performing feature extraction on the scene image and obtaining global feature information includes performing feature extraction on the processed image and obtaining global feature information. Obtaining feature information.

Specifically, in an embodiment of the present disclosure, the step of performing image preprocessing on the scene image by the fog recognition device to obtain the processed image includes pixel sampling the scene image and sampling to a target size. obtaining an image; and normalizing the sampled image to obtain a processed image.

By way of illustration, in embodiments of the present disclosure, the pixel sampling of the scene image by the fog confluence recognizer essentially resizes the scene image to a specific size, i.e., a target size. and the resulting image is used as a sampling image. The fog recognizer may use a bilinear interpolation algorithm to sample the pixels of the scene image, or of course other sampling algorithms may be used to implement the pixel sampling to obtain the sampled image. The specific target size and pixel sampling method can be selected according to actual needs, and are not limited in the embodiments of the present disclosure.

By way of illustration, in embodiments of the present disclosure, the normalization process of the sampled image by the fog confluence recognizer essentially translates the pixel value of each pixel point in the sampled image into a range of 0-1. By converting into values, the pixel values of different pixel points are mapped within the same predetermined range.

In embodiments of the present disclosure, the processed image obtained by subjecting the scene image to pixel sampling and normalization by the fog cluster recognizer can meet certain standards in terms of size and pixel representation. , it will be appreciated that this can facilitate subsequent feature extraction and improve the efficiency of feature extraction.

By way of illustration, in embodiments of the present disclosure, the fog confluence recognizer is stored with a preconfigured deep neural network. The preset deep neural network can realize the extraction of global deep features in the image. may be obtained, and the examples herein do not limit this.

By way of illustration, in embodiments of the present disclosure, the preconfigured deep neural network may include sets of convolutional layers connected sequentially in order of decreasing depth of feature extraction, wherein the fog recognizer: Multiple sets of convolution layers may be used to perform multiple iterative convolutions on the processed image to obtain feature images, which may then be determined as global feature information.

FIG. 2 is a schematic diagram of an exemplary image processing process provided by embodiments of the present disclosure. As shown in FIG. 2, the preset deep neural network includes five sets of convolutional layers, respectively a first set of convolutional layers, a second set of convolutional layers, a third set of convolutional layers and a fourth set of convolutional layers. layers and a fifth set of convolutional layers. Among them, the first set of convolution layers uses a 7×7 size filter to convolve the input 2-channel processed image to obtain a 64-channel image and reduce the image size by half. The second set to the fifth set of convolution layers each include 1-3 convolution layers, the size of the convolution kernel is 3×3, and the number of convolution output channels is sequentially 64, 128, 256 and 512, and the output images, which are the convolution results output from each set of convolution layers, can be changed to 1/4, 1/8, 1/16 and 1/32 of the processed image. The processed image is processed by 5 sets of convolutional layers, after which the image size becomes progressively smaller, the channel becomes progressively deeper, and the feature level becomes higher.

By way of illustration, in embodiments of the present disclosure, each set of sets of convolutional layers maximally pools the images resulting from the convolution process before outputting them to the next set of convolutional layers to which they are connected. , that is, downsampling may be performed, specifically, each set of four pixels in the image is left, and only the pixel with the maximum pixel value is left, so that the length of the image obtained by downsampling is and the width is actually 1/2 of the input image.

By way of explanation, in the embodiments of the present disclosure, the number of sets of convolutional layers included in a preset deep neural network, and the relevant parameters of each set of convolutional layers, such as the size of the convolution kernel, are determined according to the actual needs. , and is not limited in the embodiments of the present disclosure.

In the embodiments of the present disclosure, the extraction of global features of the processed image can be realized by a preset deep neural network, and compared to the conventional solution of extracting local features of the image, It will be appreciated that the fog information can be represented more effectively and the accuracy of fog recognition will be improved.

In S103, classification processing is performed on the all-station feature information to determine the fog in the scene image and the fog level corresponding to the fog.

In an embodiment of the present disclosure, the apparatus for recognizing fog clusters obtains the global feature information of the scene image, and then performs classification processing on the global feature information to determine the fog clusters in the scene image and the fog levels corresponding to the fog clusters. may be determined.

Specifically, in an embodiment of the present disclosure, the step of the fog recognition device performing a classification process on the global feature information to determine the fog in the scene image and the fog level corresponding to the fog performing average pooling on the feature information to obtain global average feature information; and classifying the global average feature information to determine fog and fog level.

By way of explanation, in an embodiment of the present disclosure, the fog recognizer may be stored with global pooling layers and preset classifiers, as shown in FIG. A pooling layer may be used to achieve average pooling of all-station feature information. The global pooling layer may be placed in a preset classifier, and in implementation, may be placed in a preset deep neural network, and the function to be implemented is unchanged.

By way of explanation, in an embodiment of the present disclosure, as shown in FIG. 2, the preset classifier includes a first fully connected layer, a second fully connected layer, and a A normalization layer may be included. There may be two fully bonded layers, and of course there may be one or more. The specific number of all-bonded layers can be set according to actual needs, and is not limited in the embodiments of the present disclosure.

In the embodiment of the present disclosure, since the global feature information is substantially a feature image obtained by convolution processing, the fog cloud recognition apparatus performs global average pooling on the global feature information using the global pooling layer. What it does is average each channel in the feature image so that local light deficiencies during imaging avoid affecting fog perception due to local area interference factors in the image. It will be appreciated that the accuracy of fog cloud recognition can be improved.

By way of illustration, in embodiments of the present disclosure, the fog recognizer may classify global average feature information to determine fog and fog levels corresponding to the fog in the scene image. The all-station average feature information represents the average feature of the scene image, and the fog level substantially represents the fog aggregation state of the scene image.

By way of illustration, in embodiments of the present disclosure, the fog recognizer essentially recognizes fog with respect to global average feature information, classifies the fog by fog levels, and ratio of As shown in FIG. 2, since most of the average feature information of all stations matches level 3 within the fog level, level 3 is the finally determined fog level. In addition, compared to the conventional fog recognition solution that outputs the result of the presence or absence of fog and fog concentration, the output fog level corresponds to the visibility, making the degree of fog more clear and flexible. can be expressed as

Illustratively, in the embodiment of the present disclosure, in the preset correspondence relationship between the fog level and the visibility range, the visibility corresponding to the fog level of level 1 is 0-50 meters, and the fog level of level 2 is 0-50 meters. The visibility corresponding to level 3 fog coagulation shall be 50-100 meters, the visibility corresponding to level 3 fog coagulation shall be 100-200 meters, the visibility corresponding to level 4 fog coagulation shall be 200-500 meters, and the visibility corresponding to level 5 fog coagulation shall be 200-500 meters. The visibility corresponding to the level shall be 500-1000 meters, and the visibility corresponding to the fog-free fog level shall be 1000 meters or more.

In an embodiment of the present disclosure, after determining the fog in the scene image and the fog level corresponding to the fog in the scene image, the fog coagulation recognition device, based on the correspondence relationship between the preset fog level and the visibility range, obtaining a first visibility range corresponding to the fog level; obtaining first warning information corresponding to the fog level if the first visibility range is lower than a preset visibility range; and obtaining a first warning. and outputting the information.

Illustratively, in embodiments of the present disclosure, preset correspondences between fog levels and visibility ranges are stored, different fog levels corresponding to different visibility ranges, and can be set in advance according to the actual situation, and is not limited in the embodiments of the present disclosure.

In the embodiment of the present disclosure, the fog accumulator recognizing device searches for the corresponding visibility range from the corresponding relationship between the fog confluence level and the visibility range set in advance according to the obtained fog accumulation level, It is understood that there may be one line of sight range.

By way of illustration, in embodiments of the present disclosure, the fog confluence recognizer is stored with a preset visibility range, such as, for example, 500 meters or more, and the fog level obtained by the fog confluence recognizer is the level If it is between 4 and level 1, it represents a visibility of less than 500 meters, so the corresponding warning information needs to be determined. The specific warning information may be determined based on a preset correspondence relationship between the fog level and the warning information, and is not limited in the embodiments of the present disclosure.

It should be noted that in the embodiments of the present disclosure, different fog levels correspond to different visibility, so if the visibility range corresponding to the fog level is lower than the preset visibility range, the corresponding alarm information is immediately displayed. It may determine and output warning information to alert, providing warning and scientific reliance on dealing with fog clusters.

In an embodiment of the present disclosure, the scene image includes multiple images at multiple times, the fog level includes multiple levels corresponding to the multiple images, and the fog coagulation recognizer recognizes the fog and fog within the scene image. After determining the fog level corresponding to the fog, if the plurality of levels and the plurality of times are positively correlated, obtaining second alarm information, and obtaining an alarm alerting device in the target scene. and sending the second alert information to the alert alert device.

In embodiments of the present disclosure, the fog cluster recognizer may acquire a large number of images of the target scene over a period of time, e.g. It is understood that a scene image, that is, a scene image may include multiple images at multiple times. The fog recognition device may perform fog recognition for each image and obtain a corresponding fog level, ie obtain a plurality of levels. If the levels increase over time, that is, if there is a positive correlation between the levels and the times, it indicates that the fog in the target scene is getting thicker. The fog accumulator recognizer may acquire an alert alert device, eg, a car, in the target scene and send second alert information to the alert alert device indicating that the fog ensemble is increasing.

In an embodiment of the present disclosure, the scene image includes multiple images at multiple times, the fog level includes multiple levels corresponding to the multiple images, and the fog coagulation recognizer recognizes the fog and fog within the scene image. After determining the fog level corresponding to the fog, the steps of statistically obtaining a statistical result for the plurality of levels and determining the fogging frequency of the target scene based on the statistical result may be further performed. good.

In embodiments of the present disclosure, the fog accumulator recognizer may statistically analyze multiple levels corresponding to the multiple images obtained, and determine the frequency of fog occurrence for the target scene based on the statistical results. , It may also analyze the area and time period where fog occurs frequently, summarize the rules of fog generation, and generate a fog prevention map, so as to prevent in advance in the area and time period where fog occurs frequently, It will be appreciated that losses due to fogging can be reduced.

An embodiment of the present disclosure provides a method for recognizing fog cluster, obtaining a scene image of a target scene; performing feature extraction on the scene image to obtain global feature information; processing to determine fog in the scene image and a fog level corresponding to the fog. The method for recognizing fog coagulation provided by the embodiments of the present disclosure uses deep neural networks to extract global features of an image, improve the effectiveness of information representing It reduces information interference and improves the accuracy of cloud cluster recognition.

Embodiments of the present disclosure provide a fog recognition system. FIG. 3 is a block diagram 1 of a fog cloud recognition device provided by an embodiment of the present disclosure. As shown in FIG. 3, the mist cluster recognition device
an image acquisition module 201 configured to acquire a scene image of a target scene;
a feature extraction module 202 configured to perform feature extraction on the scene image to obtain global feature information;
a classification processing module 203 configured to perform a classification process on the global feature information to determine fog in the scene image and a fog level corresponding to the fog.

FIG. 4 is a structural diagram 2 of a fog recognizing device provided by an embodiment of the present application. As shown in FIG. 4, in one embodiment of the present disclosure, the fog cloud recognition device further includes an image processing module 204,
the image processing module 204 is configured to perform image pre-processing on the scene image to obtain a processed image;
Correspondingly, the feature extraction module 202 is specifically configured to perform feature extraction on the processed image to obtain the global feature information.

In one embodiment of the present disclosure, the image processing module 204 specifically performs pixel sampling on the scene image to obtain a target size sampled image, and normalizes the sampled image, configured to obtain said processed image;

In one embodiment of the present disclosure, the classification processing module 203 specifically performs average pooling on the all-stations feature information to obtain all-stations average feature information, and classifies the all-stations average feature information, configured to determine the fog and the fog level;

In one embodiment of the present disclosure, as shown in FIG. 4 , the apparatus for recognizing fog accumulation further includes an information output module 205, wherein the information output module 205 is configured to determine the correspondence between a preset fog level and visibility range. obtaining a first visibility range corresponding to the fog liquefaction level according to the relationship, and obtaining first warning information corresponding to the fog confluence level if the first visibility range is lower than a preset visibility range; And it is configured to output the first alarm information.

In one embodiment of the present disclosure, the scene images include multiple images at multiple times, the fog level includes multiple levels corresponding to the multiple images, and
The information output module 205 obtains second alarm information when the plurality of levels and the plurality of times are positively correlated, obtains alarm reminder devices in the target scene, It is configured to send alarm information to the alarm reminder device.

In one embodiment of the present disclosure, as shown in FIG. 4, the apparatus for recognizing fog further includes a statistical processing module 206, the images of the scene include a plurality of images at a plurality of times, and the level of fog is the Contains multiple levels corresponding to multiple images,
The statistic processing module 206 is configured to statisticize the plurality of levels, obtain a statistic result, and determine a fogging frequency of the target scene based on the statistic result.

An embodiment of the present disclosure provides an apparatus for recognizing fog cluster, which acquires a scene image of a target scene, performs feature extraction on the scene image, obtains global feature information, and to determine the fog in the scene image and the fog level corresponding to the fog in the scene image. The apparatus for recognizing fog coagulation provided by embodiments of the present disclosure uses deep neural networks to extract global features of an image to improve the effectiveness of information representative of fog It reduces information interference and improves the accuracy of cloud cluster recognition.

An embodiment of the disclosure provides an electronic device. FIG. 5 is a configuration diagram of an electronic device provided by an embodiment of the present disclosure. As shown in FIG. 5, the electronic device includes a processor 301, a memory 302 and a communication bus 303;
the communication bus 303 is configured to provide connection communication between the processor 301 and the memory 302;
The processor 301 is configured to implement the fog confluence recognition method by executing a fog confluence recognition program stored in the memory 302 .

An embodiment of the present disclosure provides a computer-readable storage medium, in which one or more programs capable of being executed by one or more processors to implement the above fogging recognition method are provided. remembered. The computer-readable storage medium may be, for example, a volatile memory, such as a Random-Access Memory (abbreviated as RAM), or a Read-Only Memory, for example. non-volatile memory such as ROM), flash memory, hard disk drive (abbreviated as HDD) or solid-state disk (abbreviated as SSD) and various devices including one or any combination of the above memories, such as mobile phones, computers, tablet devices, personal digital assistants, and the like.

An embodiment of the present disclosure provides a computer program product, said computer program product comprising computer programs or instructions which, when executed on a computer, cause said computer to perform the above-described fogging recognition method.

A person skilled in the art should understand that the embodiments of the present disclosure may be provided as a method, system or computer program product. Accordingly, the present disclosure may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware. The present disclosure also includes a computer program embodied in one or more computer-usable storage media (including, but not limited to, disk memory, optical memory, etc.) containing computer-usable program code. The form of the product may be adopted.

The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the disclosure. It is to be understood that each flow and/or block in the flowchart and/or block diagrams, and combinations of the flows and/or blocks in the flowchart and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a common computer, special purpose computer, embedded processor or other programmable signal processing device to produce the machine, thereby rendering the computer or other programmable signal processing device The instructions executed by the processor create the means to perform the functions specified in one or more of the flows in the flowcharts and/or one or more blocks of the block diagrams.

These computer program instructions may be stored in a computer readable memory capable of directing a computer or other programmable signal processing apparatus to operate in a particular manner, whereby the instructions stored in the computer readable memory are Create an article of manufacture that includes instruction means that implement the functions specified in one or more of the flows in the flowcharts and/or in one or more of the blocks in the block diagrams.

These computer program instructions may be loaded into a computer or other programmable signal processing device to cause the computer or other programmable device to perform a series of operational steps to produce a computer-implemented process. Accordingly, instructions executed on a computer or other programmable device provide steps for implementing the functionality specified in one or more flows of the flowcharts and/or one or more blocks of the block diagrams.

The above are only preferred embodiments of the present disclosure, and are not intended to limit the protection scope of the present disclosure.

Embodiments of the present disclosure provide a fog cloud recognition method and apparatus, an electronic device, a storage medium and a computer program product, the method comprising obtaining a scene image of a target scene and performing feature extraction on the scene image. , obtaining global feature information, and performing a classification process on the global feature information to determine fog and fog level corresponding to the fog in the scene image. The technical solution provided by the embodiments of the present disclosure is to extract the global depth features of the image, improve the effectiveness of the information representing fog coagulation, and prevent the interference of information irrelevant to fog coagulation recognition in the image. is reduced and the accuracy of cloud cluster recognition is improved.

Claims (17)

obtaining a scene image of a target scene;
performing feature extraction on the scene image to obtain global feature information;
performing a classification process on the global feature information to determine fog in the scene image and a fog level corresponding to the fog. Before performing feature extraction on the scene image to obtain global feature information,
further comprising performing image preprocessing on the scene image to obtain a processed image;
Correspondingly, the step of performing feature extraction on the scene image to obtain global feature information includes:
performing feature extraction on the processed image to obtain the global feature information;
The method of claim 1. performing image preprocessing on the scene image to obtain a processed image,
pixel sampling the scene image to obtain a sampled image of a target size;
normalizing the sampled image to obtain the processed image;
The method for recognizing fog cloud according to claim 2. performing a classification process on the global feature information to determine fog in the scene image and a fog level corresponding to the fog;
performing average pooling on the all-station feature information to obtain all-station average feature information;
classifying the global mean feature information to determine the fog and the fog level;
The fog cloud recognition method according to any one of claims 1 to 3. After determining a fog in the scene image and a fog level corresponding to the fog,
obtaining a first visibility range corresponding to the fog level based on a preset correspondence relationship between the fog level and the visibility range;
obtaining first alarm information corresponding to the fog level if the first visibility range is lower than a preset visibility range;
and outputting the first alarm information.
The fog cloud recognition method according to any one of claims 1 to 4. the scene image includes a plurality of images at a plurality of times; the fog level includes a plurality of levels corresponding to the plurality of images; After deciding
obtaining second alarm information if the plurality of levels and the plurality of times are positively correlated;
obtaining an alert reminder device within the target scene;
sending the second alert information to the alert reminder device;
The fog cloud recognition method according to any one of claims 1 to 5. the scene image includes a plurality of images at a plurality of times; the fog level includes a plurality of levels corresponding to the plurality of images; After deciding
Statisticalizing the plurality of levels to obtain statistical results;
6. The method according to any one of claims 1 to 5, further comprising: determining the frequency of occurrence of fog for the target scene based on the statistical results. an image capture module configured to capture a scene image of a target scene;
a feature extraction module configured to perform feature extraction on the scene image to obtain global feature information;
a classification processing module configured to perform a classification process on the global feature information to determine fog in the scene image and a fog level corresponding to the fog. further comprising an image processing module;
the image processing module is configured to perform image pre-processing on the scene image to obtain a processed image;
Correspondingly, the feature extraction module is specifically configured to perform feature extraction on the processed image to obtain the global feature information. The image processing module is specifically configured to perform pixel sampling on the scene image to obtain a sampled image of a target size, and normalize the sampled image to obtain the processed image. The fog recognition device according to claim 9. Specifically, the classification processing module performs average pooling on the all-station feature information to obtain all-station average feature information, classifies the all-station average feature information, and determines the fog cloud and the fog cloud level. 11. A fog recognition device according to any one of claims 8 to 10, configured to: further comprising an information output module;
The information output module obtains a first visibility range corresponding to the fog level based on a correspondence relationship between a preset fog level and a visibility range, and the first visibility range is the preset visibility range. 12. A fog integrator recognition device according to any one of claims 8 to 11, configured to obtain first alarm information corresponding to said fog integrator level if lower, and to output said first alarm information. the scene image includes a plurality of images at a plurality of times, the fog level includes a plurality of levels corresponding to the plurality of images;
The information output module obtains second alarm information when the plurality of levels and the plurality of times are positively correlated, obtains an alarm alerting device in the target scene, and obtains the second alarm information to the alerting device. further comprising a statistical processing module, wherein the scene images include a plurality of images at a plurality of times, the fog level includes a plurality of levels corresponding to the plurality of images;
13. The statistic processing module is configured to statisticize the plurality of levels, obtain a statistic result, and determine an agglomeration frequency of the target scene based on the statistic result. 2. The fog recognition device according to item 1. An electronic device comprising a processor, memory and a communication bus,
the communication bus is configured to provide connection communication between the processor and the memory;
An electronic device, wherein the processor is configured to implement the method according to any one of claims 1-7, by executing a fog confluence recognition program stored in the memory. A computer-readable storage medium storing one or more programs capable of implementing the method for recognizing fog condensation according to any one of claims 1 to 7 by being executed by one or more processors. . A computer program product comprising a computer program or instructions which, when run on a computer, causes said computer to perform the method of any one of claims 1-7.

Images