WO2022012149A1 - 团雾识别方法、装置、电子设备、存储介质及计算机程序产品 - Google Patents

团雾识别方法、装置、电子设备、存储介质及计算机程序产品 Download PDF

Info

Publication number
WO2022012149A1
WO2022012149A1 PCT/CN2021/094385 CN2021094385W WO2022012149A1 WO 2022012149 A1 WO2022012149 A1 WO 2022012149A1 CN 2021094385 W CN2021094385 W CN 2021094385W WO 2022012149 A1 WO2022012149 A1 WO 2022012149A1
Authority
WO
WIPO (PCT)
Prior art keywords
fog
image
scene image
scene
level
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2021/094385
Other languages
English (en)
French (fr)
Inventor
赵永磊
朱铖恺
洪依君
李军
武伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Sensetime Intelligent Technology Co Ltd
Original Assignee
Shanghai Sensetime Intelligent Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Sensetime Intelligent Technology Co Ltd filed Critical Shanghai Sensetime Intelligent Technology Co Ltd
Priority to KR1020227006772A priority Critical patent/KR20220041892A/ko
Priority to JP2022513667A priority patent/JP2022545962A/ja
Publication of WO2022012149A1 publication Critical patent/WO2022012149A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/32—Normalisation of the pattern dimensions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING SYSTEMS, e.g. PERSONAL CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18—Status alarms
    • G08B21/182—Level alarms, e.g. alarms responsive to variables exceeding a threshold
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/048—Detecting movement of traffic to be counted or controlled with provision for compensation of environmental or other condition, e.g. snow, vehicle stopped at detector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Definitions

  • the present disclosure relates to the technical field of computer vision, and in particular, to a method, device, electronic device, storage medium, and computer program product for fog identification.
  • Mass fog also known as Tuo Tuo fog
  • Tuo Tuo fog is affected by the local microclimate environment. In the local area of tens of meters to hundreds of meters in heavy fog, fog with lower visibility appears. Fog has the characteristics of strong regionality and difficulty in forecasting and forecasting. On expressways, fog can lead to sudden changes in visibility, which is extremely harmful to expressway traffic safety and can easily lead to major traffic accidents.
  • special hardware such as photoelectric sensors or cameras can be used to identify fog clusters.
  • special hardware such as photoelectric sensors or cameras can be used to identify fog clusters.
  • the scheme of using the camera to realize the fog recognition scheme mostly adopts methods such as dark channel prior or manual selection of features to obtain the shallow features of the image from the image collected by the camera, which is easily affected by factors such as the angle of the light camera, resulting in a poor recognition effect. Difference.
  • Embodiments of the present disclosure are expected to provide a method, apparatus, electronic device, storage medium, and computer program product for fog identification.
  • An embodiment of the present disclosure provides a method for identifying a fog mass, and the method includes:
  • the global feature information is classified and processed to determine the fog in the scene image and the fog level corresponding to the fog.
  • the method further includes:
  • performing feature extraction on the scene image to obtain global feature information including:
  • Feature extraction is performed on the processed image to obtain the global feature information.
  • the image preprocessing is performed on the scene image to obtain a processed image, including:
  • the classification processing of the global feature information to determine the fog in the scene image and the fog level corresponding to the fog includes:
  • the global average feature information is classified to determine the fog and the fog level.
  • the method further includes:
  • the first visibility range is lower than the preset visibility range, acquiring first warning information corresponding to the fog mass level;
  • the first warning information is output.
  • the scene image includes multiple images at multiple times
  • the level of fog includes multiple levels corresponding to the multiple images
  • the determined image in the scene image includes multiple levels.
  • the second early warning information is transmitted to the early warning reminder device.
  • the scene image includes multiple images at multiple times
  • the level of fog includes multiple levels corresponding to the multiple images
  • the determined image in the scene image includes multiple levels.
  • the occurrence frequency of the fog mass in the target scene is determined.
  • An embodiment of the present disclosure provides a device for identifying a cloud of fog, and the device for identifying a cloud of cloud includes:
  • an image acquisition module configured to acquire a scene image of the target scene
  • a feature extraction module configured to perform feature extraction on the scene image to obtain global feature information
  • the classification processing module is configured to perform classification processing on the global feature information, and determine the fog in the scene image and the fog level corresponding to the fog.
  • the fog identification device also includes the image processing module,
  • the image processing module is configured to perform image preprocessing on the scene image to obtain a processed image
  • 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; perform normalization processing on the sampled image to obtain the processed image .
  • the classification processing module is specifically configured to perform average pooling on the global feature information to obtain global average feature information; classify the global average feature information to determine the fog mass and the fog level.
  • the fog identification device also includes an information output module,
  • the information output module is configured to obtain a first visibility range corresponding to the fog level according to the corresponding relationship between the preset fog level and the visibility range; when the first visibility range is lower than the preset visibility range , obtain the first warning information corresponding to the mass fog level; and output the first warning information.
  • the fog identification device further includes an information output module, the scene image includes multiple images at multiple times, and the fog level includes multiple levels corresponding to the multiple images,
  • the information output module is configured to obtain second early warning information under the condition that the multiple levels are positively correlated with the multiple times; obtain the early warning reminder device in the target scene; transmitted to the pre-warning reminder device.
  • the scene image includes multiple images at multiple times
  • the fog level includes multiple levels corresponding to the multiple images
  • the statistical processing module is configured to perform statistics on the multiple levels to obtain a statistical result; and determine the occurrence frequency of fog in the target scene according to the statistical result.
  • An embodiment of the present disclosure provides an electronic device, the electronic device includes: a processor, a memory, and a communication bus; wherein,
  • the communication bus configured to implement connection communication between the processor and the memory
  • the processor is configured to execute the fog identification program stored in the memory, so as to implement the above method for fog identification.
  • An embodiment of the present disclosure provides a computer-readable storage medium, where one or more programs are stored in the computer-readable storage medium, and the one or more programs can be executed by one or more processors, so as to realize the above group Fog identification method.
  • An embodiment of the present disclosure provides a computer program product, the computer program product includes a computer program or an instruction, and when the computer program or instruction runs on a computer, the computer is made to execute the above method for fog identification.
  • Embodiments of the present disclosure provide a method, device, electronic device, storage medium, and computer program product for fog identification.
  • the method includes: acquiring a scene image of a target scene, performing feature extraction on the scene image, and obtaining global feature information; The information is classified and processed to determine the fog in the scene image and the fog level corresponding to the fog.
  • the technical solution provided by the embodiments of the present disclosure extracts global deep-level features of the image, improves the effectiveness of the fog information representation, reduces the interference of information irrelevant to the fog recognition in the image, and improves the performance of the fog recognition. accuracy.
  • FIG. 1 is a schematic flowchart of a method for identifying a fog mass according to an embodiment of the present disclosure
  • FIG. 2 is a schematic diagram of an exemplary image processing process provided by an embodiment of the present disclosure
  • FIG. 3 is a schematic structural diagram 1 of a device for identifying a fog mass according to an embodiment of the present disclosure
  • FIG. 4 is a second schematic structural diagram of a device for identifying a fog mass according to an embodiment of the present disclosure
  • FIG. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.
  • the embodiments of the present disclosure provide a fog identification method, the execution body of which may be a fog identification device.
  • the fog identification method may be executed by a terminal device or a server or other electronic device, where the terminal device may be a user equipment ( User Equipment, UE), mobile devices, user terminals, terminals, cellular phones, cordless phones, Personal Digital Assistant (PDA), handheld devices, computing devices, in-vehicle devices, wearable devices, etc.
  • the method for identifying the fog cloud may be implemented by the processor calling computer-readable instructions stored in the memory.
  • FIG. 1 is a schematic flowchart of a method for identifying a fog mass according to an embodiment of the present disclosure. As shown in Figure 1, the fog identification method mainly includes the following steps:
  • the apparatus for identifying the fog mass may acquire the scene image of the target scene.
  • the scene image is the scene that needs to be identified by the fog, that is, the image corresponding to the target scene.
  • the fog recognition method is applied to the scene of expressway monitoring
  • the scene image is actually the expressway monitoring image
  • the subsequent fog fog recognition is not limited by special hardware, which is convenient for large-scale application.
  • the specific target scene and the scene image of the target scene are not limited in the embodiments of the present disclosure.
  • the scene image may be collected by the fog identification device, or may be obtained by an independent camera, server, or cloud device, and transmitted to the fog identification device.
  • the specific source of the scene image is not limited in this embodiment of the present disclosure.
  • the apparatus for fog identification can perform feature extraction on the scene image, thereby obtaining global feature information in the scene image.
  • the device for fog identification may further perform the following steps: performing image preprocessing on the scene image to obtain a processed image;
  • the fog identification device performs feature extraction on the scene image to obtain global feature information, including: performing feature extraction on the processed image to obtain global feature information.
  • the fog identification device performs image preprocessing on a scene image to obtain a processed image, including: performing pixel sampling on the scene image to obtain a sampled image of a target size; normalizing the sampled image processing to obtain a processed image.
  • the fog identification device performs pixel sampling on the scene image, which essentially reshapes the size of the scene image to a specific size, that is, the target size, so as to determine the obtained image as Sample image.
  • the fog identification device may use a bilinear interpolation algorithm to sample the pixels of the scene image.
  • other sampling algorithms may also be used to sample pixels to obtain a sampled image.
  • the specific target size and the pixel sampling method can be selected according to actual requirements, which are not limited in the embodiments of the present disclosure.
  • the fog identification device performs normalization processing on the sampled image, which essentially converts the pixel value of each pixel in the sampled image into a value between 0 and 1.
  • the pixel values of different pixel points are mapped to the same fixed range.
  • the fog identification device performs pixel sampling and normalization processing on the scene image, and the obtained processed image can meet certain standards in terms of size and pixel representation. , so as to facilitate subsequent feature extraction and improve the efficiency of feature extraction.
  • a preset deep neural network is stored in the fog identification device.
  • the preset deep neural network can realize the extraction of global deep-level features in the image, and the specific preset deep neural network can be obtained by using a large number of sample images for feature extraction training, which is not limited in the embodiment of the present application.
  • the preset deep neural network may include: multiple groups of convolutional layers connected in sequence from shallow to deep according to the depth of feature extraction, and the device for fog identification may use multiple groups of convolutional layers to The processed image is subjected to multiple iterative convolution processing to obtain feature images, and then the feature images are determined as global feature information.
  • FIG. 2 is a schematic diagram of an exemplary image processing process provided by an embodiment of the present disclosure.
  • the preset deep neural network includes 5 groups of convolutional layers, namely the first group of convolutional layers, the second group of convolutional layers, the third group of convolutional layers, the fourth group of convolutional layers and the fifth group of convolutional layers. Group convolutional layers.
  • the first group of convolution layers uses 7 ⁇ 7 filters, which can convolve the input 2-channel processed image to obtain a 64-channel image, and the image size is downsampled to 1/2;
  • Five groups of convolution layers each group contains 1-3 layers of convolution layers, the size of the convolution kernel is 3 ⁇ 3, and the number of convolution output channels is 64, 128, 256 and 512 in turn.
  • the result of the product that is, the output image can become 1/4, 1/8, 1/16 and 1/32 of the processed image.
  • the image size is gradually reduced. The deeper, the more advanced the feature level.
  • each group of convolutional layers of the multiple groups of convolutional layers may perform max-pooling on the images obtained by convolution processing before outputting the images to the next connected convolutional layer.
  • downsampling specifically, grouping every four pixels in the image into a group, and retaining only the pixel with the largest value, so that the length and width of the image obtained by downsampling are actually 1/2 of the input image.
  • the number of groups of convolutional layers included in the preset deep neural network, and the related parameters in each group of convolutional layers, for example, the size of the convolution kernel can be determined according to actual conditions. Requirement settings are not limited in the embodiments of the present disclosure.
  • the global feature extraction of the processed image can be realized by using the preset deep neural network, which can more effectively characterize the fog in the image compared with the traditional solution of extracting the local features of the image. information, thereby improving the accuracy of fog identification.
  • the device for identifying the fog may perform classification processing on the global feature information to determine the fog in the scene image and the fog level corresponding to the fog.
  • the fog identification device performs classification processing on the global feature information, and determines the fog in the scene image and the fog level corresponding to the fog, including: performing average pooling on the global feature information , obtain the global average feature information; classify the global average feature information to determine the fog and the fog level.
  • the fog identification device may store a global pooling layer and a preset classifier, and the fog identification device may utilize the global pooling layer to realize global features Average pooling of information.
  • the global pooling layer can be set in the preset classifier, and in fact, it can also be set in the preset deep neural network, and its function is unchanged.
  • the preset classifier may include a first fully connected layer, a second fully connected layer, and a normalization layer, which are used to realize the global average feature information Classification.
  • the number of fully connected layers may be two, of course, there may also be one or more than two layers.
  • the specific number of fully connected layers can be set according to actual requirements, which is not limited in the embodiment of the present disclosure.
  • the fog identification device uses the global pooling layer to perform global average pooling on the global feature information, that is, the feature Each channel in the image is averaged, so as to avoid the influence of local area interference factors in the image on the fog recognition due to poor local light during imaging, which can improve the accuracy of fog recognition.
  • the apparatus for identifying the fog may classify the global average feature information, thereby determining the fog in the scene image and the fog level corresponding to the fog.
  • the global average feature information represents the average feature of the scene image, and the fog level actually represents the fog situation of the scene image.
  • the fog identification device actually identifies the fog from the global average feature information, and classifies the fog level of the fog, so as to obtain different fog levels.
  • the global average feature information matches the most information with the third level in the fog level. Therefore, the final determined fog level is the third level.
  • the output fog level corresponds to the visibility. Compared with the current fog recognition scheme, the output is the presence or absence of fog, or the result of fog concentration, which can reflect the fog more clearly and flexibly. degree.
  • the visibility corresponding to the first fog level is 0-50 meters
  • the visibility corresponding to the second fog level is 50-100 meters
  • the visibility corresponding to fog level 3 is 100-200 meters
  • the visibility corresponding to fog level 4 is 200-500 meters
  • the visibility corresponding to fog level 5 is 500-1000 meters.
  • the visibility corresponding to no fog is more than 1000 meters.
  • the fog identification device may further perform the following steps: according to the preset correspondence between the fog grade and the visibility range, Obtain the first visibility range corresponding to the mass fog level; when the first visibility range is lower than the preset visibility range, obtain the first early warning information corresponding to the mass fog level; and output the first early warning information.
  • the correspondence between the preset fog level and the visibility range is stored, different fog levels correspond to different visibility ranges, and the specific fog level corresponds to the visibility range. It may be preset according to the actual situation, which is not limited in the embodiment of the present disclosure.
  • the device for identifying the fog mass may search for the corresponding visibility range from the correspondence between the preset fog mass level and the visibility range according to the obtained fog mass level, and use the obtained visibility range.
  • the range is determined to be the first visibility range.
  • a preset visibility range is stored in the fog identification device. For example, if the range is more than 500 meters, the fog level obtained by the fog identification device is between the fourth level and the first level. If the visibility is less than 500 meters, the corresponding early warning information needs to be determined.
  • the specific early warning information may be determined according to the corresponding relationship between the preset fog mass level and the early warning information, which is not limited in the 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
  • the fog recognition device determines the fog in the scene image and the fog corresponding to the fog.
  • the following steps can also be performed: in the case that multiple levels are positively correlated with multiple times, obtain second warning information; obtain the warning reminder device in the target scene; transmit the second warning information to the warning reminder equipment.
  • the fog identification device can obtain a large number of images of the target scene in a period of time, for example, multiple images on the highway in one day, and the obtained images are all scene images, that is, The scene image may include multiple images at multiple times.
  • the fog recognition device can perform fog recognition respectively, so as to obtain the corresponding fog level, that is, multiple levels. If as the time increases, multiple levels also increase, that is, multiple levels are positively correlated with multiple times, then the fog that characterizes the target scene is increasing.
  • the fog identification device can acquire the early warning and reminding equipment in the target scene, for example, a vehicle, so as to transmit the second early warning information indicating the increase of the fog to the early warning and reminding equipment.
  • the scene image includes multiple images at multiple times
  • the fog level includes multiple levels corresponding to the multiple images
  • the fog recognition device determines the fog in the scene image and the fog corresponding to the fog.
  • the following steps can also be performed: perform statistics on multiple levels to obtain statistical results; and determine the occurrence frequency of fog in the target scene according to the statistical results.
  • the fog identification device performs statistical analysis on multiple levels corresponding to the obtained multiple images, and can determine the fog occurrence frequency of the target scene according to the statistical results. It analyzes the frequent occurrence points and periods of cluster fog, summarizes the law of occurrence of cluster mist, and generates a cluster mist prevention map, so that prevention can be carried out in advance at the frequent occurrence points and multiple occurrence periods, and the loss caused by cluster mist can be reduced.
  • An embodiment of the present disclosure provides a method for identifying fog clusters, including: acquiring a scene image of a target scene; performing feature extraction on the scene image to obtain global feature information; and classifying the global feature information to determine the cluster in the scene image Fog and fog fog level corresponding to fog fog.
  • the fog fog identification method provided by the embodiments of the present disclosure adopts a deep neural network to extract global features of an image, which improves the effectiveness of fog fog information representation, reduces the interference of information irrelevant to fog fog recognition in the image, and improves the performance of the image. The accuracy of fog recognition.
  • FIG. 3 is a schematic structural diagram 1 of a device for identifying a fog mass according to an embodiment of the present disclosure.
  • the fog identification device includes:
  • 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
  • the classification processing module 203 is configured to perform classification processing on the global feature information, and determine the fog in the scene image and the fog level corresponding to the fog.
  • FIG. 4 is a second structural schematic diagram of a device for identifying a fog mass according to an embodiment of the present application.
  • the apparatus for identifying the fog mass further includes: an image processing module 204 ,
  • the image processing module 204 is configured to perform image preprocessing on the scene image to obtain a processed image
  • the feature extraction module 202 is specifically configured to perform feature extraction on the processed image to obtain the global feature information.
  • the image processing module 204 is specifically configured to perform pixel sampling on the scene image to obtain a sampled image of a target size; normalize the sampled image to obtain the processed image image.
  • the classification processing module 203 is specifically configured to perform average pooling on the global feature information to obtain global average feature information; classify the global average feature information to determine the cluster fog and the fog level.
  • the fog identification device further includes: an information output module 205.
  • the information output module 205 is configured to, according to the corresponding relationship between the preset fog level and the visibility range, Acquire a first visibility range corresponding to the mass fog level; when the first visibility range is lower than a preset visibility range, acquire first warning information corresponding to the mass fog level; output the first early warning information .
  • the scene image includes multiple images at multiple times
  • the mass fog level includes multiple levels corresponding to the multiple images
  • the information output module 205 is configured to obtain second warning information under the condition that the multiple levels are positively correlated with the multiple times; obtain the warning reminding device in the target scene; The information is transmitted to the early warning reminder device.
  • the fog identification device further includes: a statistical processing module 206, the scene image includes multiple images at multiple times, and the fog level includes a multiple levels corresponding to the multiple images,
  • the statistics processing module 206 is configured to perform statistics on the multiple levels to obtain a statistical result; and determine the occurrence frequency of fog in the target scene according to the statistical result.
  • the embodiments of the present disclosure provide a fog identification device, which acquires a scene image of a target scene; extracts and extracts the scene image to obtain global feature information; performs classification processing on the global feature information to determine the fog in the scene image and the corresponding the fog level.
  • the fog identification device provided by the embodiment of the present disclosure adopts a deep neural network to extract global features of an image, which improves the effectiveness of the information representation of the fog, reduces the interference of information irrelevant to the fog identification in the image, and improves the performance of the image. The accuracy of fog recognition.
  • FIG. 5 is a schematic structural diagram of an electronic device according to 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; wherein,
  • the communication bus 303 is configured to implement connection communication between the processor 301 and the memory 302;
  • the processor 301 is configured to execute the mass fog identification program stored in the memory 302, so as to realize the above-mentioned mass fog identification method.
  • An embodiment of the present disclosure provides a computer-readable storage medium, where one or more programs are stored in the computer-readable storage medium, and the one or more programs can be executed by one or more processors, so as to realize the above group Fog identification method.
  • the computer-readable storage medium may be a volatile memory (volatile memory), such as a random access memory (Random-Access Memory, RAM); or a non-volatile memory (non-volatile memory), such as a read-only memory (Read Only Memory) -Only Memory, ROM), flash memory (flash memory), hard disk (Hard Disk Drive, HDD) or solid-state drive (Solid-State Drive, SSD); it can also be a respective device including one or any combination of the above memories, Such as mobile phones, computers, tablet devices, personal digital assistants, etc.
  • An embodiment of the present disclosure provides a computer program product, the computer program product includes a computer program or an instruction, and when the computer program or instruction runs on a computer, the computer is made to execute the above method for fog identification.
  • 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 aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media having computer-usable program code embodied therein, including but not limited to disk storage, optical storage, and the like.
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable signal processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions
  • the apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.
  • Embodiments of the present disclosure provide a method, device, electronic device, storage medium, and computer program product for fog identification, wherein the method includes: acquiring a scene image of a target scene, performing feature extraction on the scene image, and obtaining global feature information; The global feature information is classified and processed to determine the fog in the scene image and the fog level corresponding to the fog.
  • the technical solution provided by the embodiments of the present disclosure extracts global deep-level features of the image, improves the effectiveness of the fog information representation, reduces the interference of information irrelevant to the fog recognition in the image, and improves the performance of the fog recognition. accuracy.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Emergency Management (AREA)
  • Business, Economics & Management (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Image Analysis (AREA)

Abstract

一种团雾识别方法、装置、电子设备、存储介质及计算机程序产品，该方法包括：获取目标场景的场景图像（S101）；对场景图像进行特征提取，得到全局特征信息（S102）；对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别（S103）。

Description

团雾识别方法、装置、电子设备、存储介质及计算机程序产品
相关申请的交叉引用
本申请基于申请号为202010699922.1、申请日为2020年07月17日，申请名称为“团雾识别方法、装置、电子设备及存储介质”的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此以引入方式结合在本申请中。
技术领域
本公开涉及计算机视觉技术领域，尤其涉及一种团雾识别方法、装置、电子设备、存储介质及计算机程序产品。
背景技术
“团雾”，又称坨坨雾，是受局部地区微气候环境的影响，在大雾中数十米到上百米的局部范围内，出现的能见度更低的雾。团雾具有区域性强，预测预报难度大的特点，在高速公路上，团雾会导致能见度的突然变化，对高速公路交通安全极具危害性，容易导致重大的交通事故。
目前，可以利用光电传感器等专用硬件，或者，摄像头进行团雾的识别。其中，相比于成本高昂，难以大规模推广的利用专用硬件的方案，利用摄像头获取图像信息，从而识别团雾，具有更高的可行性。
然而，利用摄像头实现团雾识别的方案，多采用暗通道先验或者人工选取特征等方法，从摄像头采集的图像中获取图像浅层特征，易受光线摄像头角度等因素影响，从而导致识别效果较差。
发明内容
本公开实施例期望提供一种团雾识别方法、装置、电子设备、存储介质及计算机程序产品。
本公开实施例的技术方案是这样实现的：
本公开实施例提供了一种团雾识别方法，所述方法包括：
获取目标场景的场景图像；
对所述场景图像进行特征提取，得到全局特征信息；
对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别。
在上述团雾识别方法中，所述对所述场景图像进行特征提取，得到全局特征信息之前，所述方法还包括：
对所述场景图像进行图像预处理，得到已处理图像；
相应的，所述对所述场景图像进行特征提取，得到全局特征信息，包括：
对所述已处理图像进行特征提取，得到所述全局特征信息。
在上述团雾识别方法中，所述对所述场景图像进行图像预处理，得到已处理图像，包括：
对所述场景图像进行像素采样，得到目标尺寸的采样图像；
对所述采样图像进行归一化处理，得到所述已处理图像。
在上述团雾识别方法中，所述对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别，包括：
对所述全局特征信息进行平均池化，得到全局平均特征信息；
对所述全局平均特征信息进行分类，确定出所述团雾和所述团雾级别。
在上述团雾识别方法中，所述确定出所述场景图像中的团雾以及所述团雾对应的团雾级别之后，所述方法还包括：
根据预设团雾等级与能见度范围的对应关系，获取所述团雾级别对应的第一能见度范围；
在所述第一能见度范围低于预设能见度范围的情况下，获取所述团雾级别对应的第一预警信息；
输出所述第一预警信息。
在上述团雾识别方法中，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，所述确定出所述场景图像中的团雾以及所述团雾对应的团雾级别之后，所述方法还包括：
在所述多个级别与所述多个时刻正相关的情况下，获取第二预警信息；
获取所述目标场景内的预警提醒设备；
将所述第二预警信息传输至所述预警提醒设备。
在上述团雾识别方法中，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，所述确定出所述场景图像中的团雾以及所述团雾对应的团雾级别之后，所述方法还包括：
对所述多个级别进行统计，得到统计结果；
根据所述统计结果，确定所述目标场景的团雾发生频率。
本公开实施例提供了一种团雾识别装置，所述团雾识别装置包括：
图像获取模块，配置为获取目标场景的场景图像；
特征提取模块，配置为对所述场景图像进行特征提取，得到全局特征信息；
分类处理模块，配置为对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别。
在上述团雾识别装置中，还包括所述图像处理模块，
所述图像处理模块，配置为对所述场景图像进行图像预处理，得到已处理图像；
相应的，所述特征提取模块，具体配置为对所述已处理图像进行特征提取，得到所述全局特征信息。
在上述团雾识别装置中，所述图像处理模块，具体配置为对所述场景图像进行像素采样，得到目标尺寸的采样图像；对所述采样图像进行归一化处理，得到所述已处理图像。
在上述团雾识别装置中，所述分类处理模块，具体配置为对所述全局特征信息进行平均池化，得到全局平均特征信息；对所述全局平均特征信息进行分类，确定出所述团雾和所述团雾级别。
在上述团雾识别装置中，还包括信息输出模块，
所述信息输出模块，配置为根据预设团雾等级与能见度范围的对应关系，获取所述团雾级别对应的第一能见度范围；在所述第一能见度范围低于预设能见度范围的情况下，获取所述团雾级别对应的第一预警信息；输出所述第一预警信息。
在上述团雾识别装置中，还包括信息输出模块，所述场景图像包括多个时刻下的多 个图像，所述团雾级别包括与所述多个图像对应的多个级别，
所述信息输出模块，配置为在所述多个级别与所述多个时刻正相关的情况下，获取第二预警信息；获取所述目标场景内的预警提醒设备；将所述第二预警信息传输至所述预警提醒设备。
在上述团雾识别装置中，还包括统计处理模块，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，
所述统计处理模块，配置为对所述多个级别进行统计，得到统计结果；根据所述统计结果，确定所述目标场景的团雾发生频率。
本公开实施例提供了一种电子设备，所述电子设备包括：处理器、存储器和通信总线；其中，
所述通信总线，配置为实现所述处理器和所述存储器之间的连接通信；
所述处理器，配置为执行所述存储器中存储的团雾识别程序，以实现上述团雾识别方法。
本公开实施例提供了一种计算机可读存储介质，所述计算机可读存储介质存储有一个或者多个程序，所述一个或者多个程序可以被一个或者多个处理器执行，以实现上述团雾识别方法。
本公开实施例提供了一种计算机程序产品，所述计算机程序产品包括计算机程序或指令，在所述计算机程序或指令在计算机上运行的情况下，使所述计算机执行上述团雾识别方法。
本公开实施例提供了一种团雾识别方法、装置、电子设备、存储介质及计算机程序产品，方法包括：获取目标场景的场景图像，对场景图像进行特征提取，得到全局特征信息；对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别。本公开实施例提供的技术方案，对图像进行全局深层次特征的提取，提高了团雾信息表征的有效性，降低了图像中与团雾识别无关的信息的干扰，从而提高了团雾识别的准确性。
附图说明
为了更清楚地说明本公开实施例或背景技术中的技术方案，下面将对本公开实施例或背景技术中所需要使用的附图进行说明。
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。
图1为本公开实施例提供的一种团雾识别方法的流程示意图；
图2为本公开实施例提供的一种示例性的图像处理过程示意图；
图3为本公开实施例提供的一种团雾识别装置的结构示意图一；
图4为本公开实施例提供的一种团雾识别装置的结构示意图二；
图5为本公开实施例提供的一种电子设备的结构示意图。
具体实施方式
下面将结合本公开实施例中的附图，对本公开实施例中的技术方案进行清楚、完整地描述。
本公开实施例提供了一种团雾识别方法，其执行主体可以是团雾识别装置，例如，团雾识别方法可以由终端设备或服务器或其它电子设备执行，其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、 个人数字处理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。在一些可能的实现方式中，该团雾识别方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。
本公开实施例提供了一种团雾识别方法。图1为本公开实施例提供的一种团雾识别方法的流程示意图。如图1所示，团雾识别方法主要包括以下步骤：
S101、获取目标场景的场景图像。
在本公开的实施例中，团雾识别装置可以获取到目标场景的场景图像。
需要说明的是，在本公开的实施例中，场景图像为需要进行团雾识别的场景，即目标场景所对应的图像。其中，在将团雾识别方法应用于高速公路监控的场景中的情况下，场景图像实际上就是高速公路监控图像，进行后续团雾识别不受专用硬件的限制，便于大规模的应用。具体的目标场景和目标场景的场景图像本公开实施例不作限定。
需要说明的是，在本公开的实施例中，场景图像可以是团雾识别装置采集到的，也可以是独立的摄像头、服务器或者云端等设备获取到，并传输给团雾识别装置的图像。具体的场景图像的来源本公开实施例不作限定。
S102、对场景图像进行特征提取，得到全局特征信息。
在本公开的实施例中，团雾识别装置在获得场景图像之后，即可对场景图像进行特征提取，从而得到场景图像中的全局特征信息。
需要说明的是，在本公开的实施例中，团雾识别装置在对场景图像进行特征提取，得到全局特征信息之前，还可以执行以下步骤：对场景图像进行图像预处理，得到已处理图像；相应的，团雾识别装置对场景图像进行特征提取，得到全局特征信息，包括：对已处理图像进行特征提取，得到全局特征信息。
具体地，在本公开的实施例中，团雾识别装置对场景图像进行图像预处理，得到已处理图像，包括：对场景图像进行像素采样，得到目标尺寸的采样图像；对采样图像进行归一化处理，得到已处理图像。
需要说明的是，在本公开的实施例中，团雾识别装置对场景图像进行像素采样，其实质是将场景图像的尺寸重塑到特定的大小，即目标尺寸，从而将得到的图像确定为采样图像。其中，团雾识别装置可以采用双线性插值算法对场景图像的像素进行采样，当然，也可以采用其它的采样算法实现像素的采样，以得到采样图像。具体的目标尺寸，以及像素采样的方式可以根据实际需求选择，本公开实施例不作限定。
需要说明的是，在本公开的实施例中，团雾识别装置对采样图像进行归一化处理，其实质是将采样图像中每个像素点的像素值转换成0-1之间的数值，从而将不同像素点的像素值映射到同一固定范围内。
可以理解的是，在本公开的实施例中，团雾识别装置对场景图像进行像素采样和归一化处理，得到的已处理图像，其在尺寸以及像素的表征方式上都能够符合一定的标准，从而能够便于后续进行特征提取，提高特征提取的效率。
需要说明的是，在本公开的实施例中，团雾识别装置中存储有预设深度神经网络。预设深度神经网络能够实现图像中全局深层次特征的提取，具体的预设深度神经网络可以为利用大量的样本图像进行特征提取训练得到的，本申请实施例不作限定。
需要说明的是，在本公开的实施例中，预设深度神经网络可以包括：按照特征提取深度由浅至深依次连接的多组卷积层，团雾识别装置可以利用多组卷积层，对已处理图像进行多次迭代卷积处理，得到特征图像，之后，将特征图像确定为全局特征信息。
图2为本公开实施例提供的一种示例性的图像处理过程示意图。如图2所示，预设深度神经网络包括5组卷积层，分别为第一组卷积层、第二组卷积层、第三组卷积层、第四组卷积层和第五组卷积层。其中，第一组卷积层采用7×7大小的滤波器，可以将 输入的2通道的已处理图像卷积得到64通道的图像，同时图像尺寸下采样到1/2；第二组至第五组卷积层，每组包含1-3层卷积层，卷积核尺寸为3×3，且卷积输出通道数依次为64、128、256和512，每组卷积层输出的卷积结果，即输出的图像可以变为已处理图像的1/4、1/8、1/16和1/32，已处理图像经过5组卷积层处理之后，图像尺寸逐渐缩小，通到逐步加深，特征层次愈加高级。
需要说明的是，在本公开的实施例中，多组卷积层的每组卷积层，在输出卷积处理得到的图像至相连的下一组卷积层之前，可以对其进行最大池化，即下采样，具体的，将图像中每四个像素为一组，仅将其中像素值最大的保留，从而下采样得到的图像的长和宽实际上为输入图像的1/2。
需要说明的是，在本公开的实施例中，预设深度神经网络中包括的卷积层的组数，以及每组卷积层中的相关参数，例如，卷积核的尺寸，可以根据实际需求设置，本公开实施例不作限定。
可以理解的是，在本公开的实施例中，利用预设深度神经网络可以实现已处理图像的全局特征提取，相比于传统提取图像局部特征的方案，能够更有效的表征图像中的团雾信息，从而提高团雾识别的准确性。
S103、对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别。
在本公开的实施例中，团雾识别装置在场景图像的全局特征信息之后，可以对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别。
具体地，在本公开的实施例中，团雾识别装置对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别，包括：对全局特征信息进行平均池化，得到全局平均特征信息；对全局平均特征信息进行分类，确定出团雾和团雾级别。
需要说明的是，在本公开的实施例中，如图2所示，团雾识别装置中可以存储有全局池化层和预设分类器，团雾识别装置可以利用全局池化层实现全局特征信息的平均池化。全局池化层可以设置在预设分类器中，实际上也可以设置在预设深度神经网络中，其实现的功能是不变的。
需要说明的是，在本公开的实施例中，如图2所示，预设分类器可以包括第一全连接层、第二全连接层和归一化层，用于实现全局平均特征信息的分类。其中，全连接层可以是两个，当然，也可以是一个或者两个以上。具体的全连接层的数量可以根据实际需求设置，本公开实施例不作限定。
可以理解的是，在本公开的实施例中，由于全局特征信息实际上是卷积处理得到的特征图像，团雾识别装置利用全局池化层对全局特征信息进行全局平均池化，即对特征图像中每个通道取平均，从而避免由于成像时局部光线不佳，导致图像中局部区域干扰因素对团雾识别的影响，能够提高团雾识别的准确性。
需要说明的是，在本公开的实施例中，团雾识别装置可以对全局平均特征信息进行分类，从而确定出场景图像中的团雾以及团雾对应的团雾级别。全局平均特征信息表征了场景图像的平均特征，团雾级别实际上就是表征场景图像的团雾情况。
需要说明的是，在本公开的实施例中，团雾识别装置实际上是对全局平均特征信息识别出团雾，并对团雾进行了团雾级别的分类，从而得到了不同团雾级别的占比，如图2所示，全局平均特征信息与团雾等级中的三级匹配的信息最多，因此，最终确定的团雾级别即为三级。此外，输出的团雾级别与能见度对应，相比于目前的团雾识别方案中，输出的均为有无团雾，或者团雾浓度的结果，可以更明显和灵活的的反映出团雾的程度。
示例性的，在本公开的实施例中，在预设团雾等级与能见度范围的对应关系中，团雾级别为一级对应的能见度为0-50米，团雾级别二级对应的能见度为50-100米，团雾 级别三级对应的能见度为100-200米，团雾级别四级对应的能见度为200-500米，团雾级别五级对应的能见度为500-1000米，团雾级别为无雾对应的能见度为1000米以上。
在本公开的实施例中，团雾识别装置在确定出场景图像中的团雾以及团雾对应的团雾级别之后，还可以执行以下步骤：根据预设团雾等级与能见度范围的对应关系，获取团雾级别对应的第一能见度范围；在第一能见度范围低于预设能见度范围的情况下，获取团雾级别对应的第一预警信息；输出第一预警信息。
需要说明的是，在本公开的实施例中，存储有预设团雾等级与能见度范围的对应关系，不同的团雾级别与不同的能见度范围对应，具体的团雾等级与能见度范围的对应关系可以根据实际情况预先设置，本公开实施例不作限定。
可以理解的是，在本公开的实施例中，团雾识别装置可以根据得到的团雾级别，从预设团雾等级与能见度范围的对应关系中查找相对应的能见度范围，将查找到的能见度范围确定为第一能见度范围。
需要说明的是，在本公开的实施例中，团雾识别装置中存储有预设能见度范围，例如，500米以上，在团雾识别装置得到的团雾级别，处于四级至一级之间的情况下，则其表征的能见度低于500米，因此，需要确定相应的预警信息。具体的预警信息，可以是根据预设的团雾等级与预警信息之间的对应关系确定，本公开实施例不作限定。
需要说明的是，在本公开的实施例中，不同的团雾级别与不同的能见度对应，因此，在团雾级别对应的能见度范围低于预设能见度范围的情况下，可以及时确定相应的预警信息，并输出预警信息以进行提示，为团雾的处理提供预警和科学依据。
在本公开的实施例中，场景图像包括多个时刻下的多个图像，团雾级别包括与多个图像对应的多个级别，团雾识别装置确定出场景图像中的团雾以及团雾对应的团雾级别之后，还可以执行以下步骤：在多个级别与多个时刻正相关的情况下，获取第二预警信息；获取目标场景内的预警提醒设备；将第二预警信息传输至预警提醒设备。
可以理解的是，在本公开的实施例中，团雾识别装置可以获取目标场景在一段时间内大量的图像，例如，一天内高速公路上的多张图像，得到的图像均为场景图像，即场景图像可以包括多个时刻下的多个图像。针对每个图像，团雾识别装置可以分别进行团雾识别，从而得到相应的团雾级别，即多个级别。如果随着时刻的递增，多个级别也呈现提升的情况，也就是多个级别与多个时刻正相关，则表征目标场景的团雾在加重。团雾识别装置可以获取目标场景内的预警提醒设备，例如，车辆，从而将提示团雾加重的第二预警信息传输至预警提醒设备。
在本公开的实施例中，场景图像包括多个时刻下的多个图像，团雾级别包括与多个图像对应的多个级别，团雾识别装置确定出场景图像中的团雾以及团雾对应的团雾级别之后，还可以执行以下步骤：对多个级别进行统计，得到统计结果；根据统计结果，确定目标场景的团雾发生频率。
可以理解的是，在本公开的实施例中，团雾识别装置对得到的多个图像对应的多个级别进行统计分析，可以根据统计结果确定出目标场景的团雾发生频率，此外，还可以分析团雾多发点和多发时段，总结归纳团雾发生的规律，生成团雾预防图，从而可以在多发点和多发时段提前进行预防，减少团雾造成的损失。
本公开实施例提供了一种团雾识别方法，包括：获取目标场景的场景图像，；对场景图像进行特征提取，得到全局特征信息；对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别。本公开实施例提供的团雾识别方法，采用深度神经网络对图像进行全局特征的提取，提高了团雾信息表征的有效性，降低了图像中与团雾识别无关的信息的干扰，从而提高了团雾识别的准确性。
本公开实施例提供了一种团雾识别装置，图3为本公开实施例提供的一种团雾识别 装置的结构示意图一。如图3所示，团雾识别装置包括：
图像获取模块201，配置为获取目标场景的场景图像；
特征提取模块202，配置为对所述场景图像进行特征提取，得到全局特征信息；
分类处理模块203，配置为对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别。
图4为本申请实施例提供的一种团雾识别装置的结构示意图二。如图4所示，在本公开一实施例中，所述团雾识别装置还包括：图像处理模块204，
所述图像处理模块204，配置为对所述场景图像进行图像预处理，得到已处理图像；
相应的，所述特征提取模块202，具体配置为对所述已处理图像进行特征提取，得到所述全局特征信息。
在本公开一实施例中，所述图像处理模块204，具体配置为对所述场景图像进行像素采样，得到目标尺寸的采样图像；对所述采样图像进行归一化处理，得到所述已处理图像。
在本公开一实施例中，所述分类处理模块203，具体配置为对所述全局特征信息进行平均池化，得到全局平均特征信息；对所述全局平均特征信息进行分类，确定出所述团雾和所述团雾级别。
在本公开一实施例中，如图4所示，所述团雾识别装置还包括：信息输出模块205，所述信息输出模块205，配置为根据预设团雾等级与能见度范围的对应关系，获取所述团雾级别对应的第一能见度范围；在所述第一能见度范围低于预设能见度范围的情况下，获取所述团雾级别对应的第一预警信息；输出所述第一预警信息。
在本公开一实施例中，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，
所述信息输出模块205，配置为在所述多个级别与所述多个时刻正相关的情况下，获取第二预警信息；获取所述目标场景内的预警提醒设备；将所述第二预警信息传输至所述预警提醒设备。
在本公开一实施例中，如图4所示，所述团雾识别装置还包括：统计处理模块206，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，
所述统计处理模块206，配置为对所述多个级别进行统计，得到统计结果；根据所述统计结果，确定所述目标场景的团雾发生频率。
本公开实施例提供了一种团雾识别装置，获取目标场景的场景图像；对场景图像进行提取提取，得到全局特征信息；对全局特征信息进行分类处理，确定出场景图像中的团雾以及对应的团雾级别。本公开实施例提供的团雾识别装置，采用深度神经网络对图像进行全局特征的提取，提高了团雾信息表征的有效性，降低了图像中与团雾识别无关的信息的干扰，从而提高了团雾识别的准确性。
本公开实施例提供了一种电子设备。图5为本公开实施例提供的一种电子设备的结构示意图。如图5所示，电子设备包括：处理器301、存储器302和通信总线303；其中，
所述通信总线303，配置为实现所述处理器301和所述存储器302之间的连接通信；
所述处理器301，配置为执行所述存储器302中存储的团雾识别程序，以实现上述团雾识别方法。
本公开实施例提供了一种计算机可读存储介质，所述计算机可读存储介质存储有一个或者多个程序，所述一个或者多个程序可以被一个或者多个处理器执行，以实现上述团雾识别方法。计算机可读存储介质可以是是易失性存储器(volatile memory)，例如随 机存取存储器(Random-Access Memory，RAM)；或者非易失性存储器(non-volatile memory)，例如只读存储器(Read-Only Memory，ROM)，快闪存储器(flash memory)，硬盘(Hard Disk Drive，HDD)或固态硬盘(Solid-State Drive，SSD)；也可以是包括上述存储器之一或任意组合的各自设备，如移动电话、计算机、平板设备、个人数字助理等。
本公开实施例提供了一种计算机程序产品，所述计算机程序产品包括计算机程序或指令，在所述计算机程序或指令在计算机上运行的情况下，使所述计算机执行上述团雾识别方法。
本领域内的技术人员应明白，本公开的实施例可提供为方法、系统、或计算机程序产品。因此，本公开可采用硬件实施例、软件实施例、或结合软件和硬件方面的实施例的形式。而且，本公开可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器和光学存储器等)上实施的计算机程序产品的形式。
本公开是参照根据本公开实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程信号处理设备的处理器以产生一个机器，使得通过计算机或其他可编程信号处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程信号处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程信号处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
以上所述，仅为本公开的较佳实施例而已，并非用于限定本公开的保护范围。
工业实用性
本公开实施例提供了一种团雾识别方法、装置、电子设备、存储介质及计算机程序产品，其中，方法包括：获取目标场景的场景图像，对场景图像进行特征提取，得到全局特征信息；对全局特征信息进行分类处理，确定出场景图像中的团雾以及团雾对应的团雾级别。本公开实施例提供的技术方案，对图像进行全局深层次特征的提取，提高了团雾信息表征的有效性，降低了图像中与团雾识别无关的信息的干扰，从而提高了团雾识别的准确性。

Claims (17)

  1. 一种团雾识别方法，所述方法包括：
    获取目标场景的场景图像；
    对所述场景图像进行特征提取，得到全局特征信息；
    对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别。
  2. 根据权利要求1所述的方法，其中，所述对所述场景图像进行特征提取，得到全局特征信息之前，所述方法还包括：
    对所述场景图像进行图像预处理，得到已处理图像；
    相应的，所述对所述场景图像进行特征提取，得到全局特征信息，包括：
    对所述已处理图像进行特征提取，得到所述全局特征信息。
  3. 根据权利要求2所述的团雾识别方法，其中，所述对所述场景图像进行图像预处理，得到已处理图像，包括：
    对所述场景图像进行像素采样，得到目标尺寸的采样图像；
    对所述采样图像进行归一化处理，得到所述已处理图像。
  4. 根据权利要求1至3任一项所述的团雾识别方法，其中，所述对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别，包括：
    对所述全局特征信息进行平均池化，得到全局平均特征信息；
    对所述全局平均特征信息进行分类，确定出所述团雾和所述团雾级别。
  5. 根据权利要求1至4任一项所述的团雾识别方法，其中，所述确定出所述场景图像中的团雾以及所述团雾对应的团雾级别之后，所述方法还包括：
    根据预设团雾等级与能见度范围的对应关系，获取所述团雾级别对应的第一能见度范围；
    在所述第一能见度范围低于预设能见度范围的情况下，获取所述团雾级别对应的第一预警信息；
    输出所述第一预警信息。
  6. 根据权利要求1至5任一项所述的团雾识别方法，其中，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，所述确定出所述场景图像中的团雾以及所述团雾对应的团雾级别之后，所述方法还包括：
    在所述多个级别与所述多个时刻正相关的情况下，获取第二预警信息；
    获取所述目标场景内的预警提醒设备；
    将所述第二预警信息传输至所述预警提醒设备。
  7. 根据权利要求1至5任一项所述的团雾识别方法，其中，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，所述确定出所述场景图像中的团雾以及所述团雾对应的团雾级别之后，所述方法还包括：
    对所述多个级别进行统计，得到统计结果；
    根据所述统计结果，确定所述目标场景的团雾发生频率。
  8. 一种团雾识别装置，所述团雾识别装置包括：
    图像获取模块，配置为获取目标场景的场景图像；
    特征提取模块，配置为对所述场景图像进行特征提取，得到全局特征信息；
    分类处理模块，配置为对所述全局特征信息进行分类处理，确定出所述场景图像中的团雾以及所述团雾对应的团雾级别。
  9. 根据权利要求8所述的团雾识别装置，其中，还包括：图像处理模块，
    所述图像处理模块，配置为对所述场景图像进行图像预处理，得到已处理图像；
    相应的，所述特征提取模块，具体配置为对所述已处理图像进行特征提取，得到所述全局特征信息。
  10. 根据权利要求9所述的团雾识别装置，其中，
    所述图像处理模块，具体配置为对所述场景图像进行像素采样，得到目标尺寸的采样图像；对所述采样图像进行归一化处理，得到所述已处理图像。
  11. 根据权利要求8至10任一项所述的团雾识别装置，其中，
    所述分类处理模块，具体配置为对所述全局特征信息进行平均池化，得到全局平均特征信息；对所述全局平均特征信息进行分类，确定出所述团雾和所述团雾级别。
  12. 根据权利要求8至11任一项所述的团雾识别装置，其中，还包括：信息输出模块，
    所述信息输出模块，配置为根据预设团雾等级与能见度范围的对应关系，获取所述团雾级别对应的第一能见度范围；在所述第一能见度范围低于预设能见度范围的情况下，获取所述团雾级别对应的第一预警信息；输出所述第一预警信息。
  13. 根据权利要求8至12任一项所述的团雾识别装置，其中，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，
    所述信息输出模块，配置为在所述多个级别与所述多个时刻正相关的情况下，获取第二预警信息；获取所述目标场景内的预警提醒设备；将所述第二预警信息传输至所述预警提醒设备。
  14. 根据权利要求8至12任一项所述的团雾识别装置，其中，还包括：统计处理模块，所述场景图像包括多个时刻下的多个图像，所述团雾级别包括与所述多个图像对应的多个级别，
    所述统计处理模块，配置为对所述多个级别进行统计，得到统计结果；根据所述统计结果，确定所述目标场景的团雾发生频率。
  15. 一种电子设备，所述电子设备包括：处理器、存储器和通信总线；其中，
    所述通信总线，配置为实现所述处理器和所述存储器之间的连接通信；
    所述处理器，配置为执行所述存储器中存储的团雾识别程序，以实现权利要求1-7任一项所述的团雾识别方法。
  16. 一种计算机可读存储介质，所述计算机可读存储介质存储有一个或者多个程序，所述一个或者多个程序可以被一个或者多个处理器执行，以实现权利要求1-7任一项所述的团雾识别方法。
  17. 一种计算机程序产品，所述计算机程序产品包括计算机程序或指令，在所述计算机程序或指令在计算机上运行的情况下，使所述计算机执行权利要求1-7中任一项所述的方法。
PCT/CN2021/094385 2020-07-17 2021-05-18 团雾识别方法、装置、电子设备、存储介质及计算机程序产品 Ceased WO2022012149A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
KR1020227006772A KR20220041892A (ko) 2020-07-17 2021-05-18 덩어리안개 식별 방법, 장치, 전자 기기, 저장 매체 및 컴퓨터 프로그램 제품
JP2022513667A JP2022545962A (ja) 2020-07-17 2021-05-18 団霧認識方法及び装置、電子機器、記憶媒体並びにコンピュータプログラム製品

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010699922.1A CN111783732B (zh) 2020-07-17 2020-07-17 团雾识别方法、装置、电子设备及存储介质
CN202010699922.1 2020-07-17

Publications (1)

Publication Number Publication Date
WO2022012149A1 true WO2022012149A1 (zh) 2022-01-20

Family

ID=72763612

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/094385 Ceased WO2022012149A1 (zh) 2020-07-17 2021-05-18 团雾识别方法、装置、电子设备、存储介质及计算机程序产品

Country Status (4)

Country Link
JP (1) JP2022545962A (zh)
KR (1) KR20220041892A (zh)
CN (1) CN111783732B (zh)
WO (1) WO2022012149A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111783732B (zh) * 2020-07-17 2025-02-21 上海商汤智能科技有限公司 团雾识别方法、装置、电子设备及存储介质
CN112686105B (zh) * 2020-12-18 2021-11-02 云南省交通规划设计研究院有限公司 一种基于视频图像多特征融合的雾浓度等级识别方法
CN112816483A (zh) * 2021-03-05 2021-05-18 北京文安智能技术股份有限公司 基于雾值分析的团雾识别预警方法、系统和电子设备

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779349A (zh) * 2012-06-30 2012-11-14 东南大学 一种基于图像颜色空间特征的雾天检测方法
US20160005152A1 (en) * 2014-07-01 2016-01-07 Adobe Systems Incorporated Multi-Feature Image Haze Removal
CN105303844A (zh) * 2015-10-26 2016-02-03 南京本来信息技术有限公司 基于激光的夜间高速公路团雾自动检测装置及其检测方法
CN109683209A (zh) * 2018-12-27 2019-04-26 吉林省气象服务中心 团雾的监测方法和监测系统
CN109961070A (zh) * 2019-03-22 2019-07-02 国网河北省电力有限公司电力科学研究院 一种输电线路智能化图像监测中区分雾体浓度的方法
CN110378865A (zh) * 2019-04-28 2019-10-25 湖南师范大学 一种复杂背景下的雾天能见度智能分级识别方法及系统
CN111783732A (zh) * 2020-07-17 2020-10-16 上海商汤智能科技有限公司 团雾识别方法、装置、电子设备及存储介质

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101219659B1 (ko) * 2012-07-27 2013-01-09 한국건설기술연구원 Ｃｃｔｖ 영상을 이용한 안개 검지 시스템 및 그 방법
JP2014164375A (ja) * 2013-02-22 2014-09-08 Pioneer Electronic Corp 情報処理装置、情報処理システム並びに情報処理方法及び情報処理装置用プログラム
CN203825644U (zh) * 2014-03-26 2014-09-10 国家电网公司 一种基于图像识别输电线路的雾等级的系统
CN107194924A (zh) * 2017-05-23 2017-09-22 重庆大学 基于暗通道先验和深度学习的高速公路雾天能见度检测方法
KR101874968B1 (ko) * 2017-07-03 2018-08-02 (주)시정 이미지 정보 기반의 시정거리 측정 시스템 및 이를 이용한 시정거리 측정 방법
KR101927227B1 (ko) * 2017-10-27 2018-12-10 (주)시정 인공지능을 적용하는 시정계
CN108710847B (zh) * 2018-05-15 2020-11-27 北京旷视科技有限公司 场景识别方法、装置及电子设备
CN108846324B (zh) * 2018-05-28 2020-06-30 中用科技有限公司 一种基于图像的道路团雾检测方法
KR102016262B1 (ko) * 2018-05-30 2019-08-29 세종대학교산학협력단 연기를 포함하는 환경에 대한 영상 분석 방법 및 그 시스템
CN109241831B (zh) * 2018-07-26 2021-11-26 东南大学 一种基于图像分析的夜间雾天能见度分类方法
CN109886920A (zh) * 2019-01-16 2019-06-14 安徽谛听信息科技有限公司 一种雾天分级方法、雾天分级系统
CN110222559A (zh) * 2019-04-24 2019-09-10 深圳市微纳集成电路与系统应用研究院 基于卷积神经网络的烟雾图像检测方法及装置
CN110598613B (zh) * 2019-09-03 2022-10-25 长安大学 一种高速公路团雾监测方法
CN110705619B (zh) * 2019-09-25 2023-06-06 南方电网科学研究院有限责任公司 雾浓度等级判别方法及装置

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779349A (zh) * 2012-06-30 2012-11-14 东南大学 一种基于图像颜色空间特征的雾天检测方法
US20160005152A1 (en) * 2014-07-01 2016-01-07 Adobe Systems Incorporated Multi-Feature Image Haze Removal
CN105303844A (zh) * 2015-10-26 2016-02-03 南京本来信息技术有限公司 基于激光的夜间高速公路团雾自动检测装置及其检测方法
CN109683209A (zh) * 2018-12-27 2019-04-26 吉林省气象服务中心 团雾的监测方法和监测系统
CN109961070A (zh) * 2019-03-22 2019-07-02 国网河北省电力有限公司电力科学研究院 一种输电线路智能化图像监测中区分雾体浓度的方法
CN110378865A (zh) * 2019-04-28 2019-10-25 湖南师范大学 一种复杂背景下的雾天能见度智能分级识别方法及系统
CN111783732A (zh) * 2020-07-17 2020-10-16 上海商汤智能科技有限公司 团雾识别方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN111783732A (zh) 2020-10-16
CN111783732B (zh) 2025-02-21
JP2022545962A (ja) 2022-11-01
KR20220041892A (ko) 2022-04-01

Similar Documents

Publication Publication Date Title
CN107609485B (zh) 交通标志的识别方法、存储介质、处理设备
CN109086656A (zh) 机场异物检测方法、装置、计算机设备及存储介质
CN105550701A (zh) 实时图像提取识别方法及装置
CN111783732A (zh) 团雾识别方法、装置、电子设备及存储介质
CN111627057B (zh) 一种距离测量方法、装置及服务器
CN108268867A (zh) 一种车牌定位方法及装置
CN112101114A (zh) 一种视频目标检测方法、装置、设备以及存储介质
Srinivasan et al. Real-time vehicle classification and license plate recognition via deformable convolution-based yolo v8 network
WO2022217551A1 (zh) 目标检测方法和装置
CN113887472A (zh) 基于级联颜色及纹理特征注意力的遥感图像云检测方法
CN113781510A (zh) 边缘检测方法、装置及电子设备
CN110688926A (zh) 主体检测方法和装置、电子设备、计算机可读存储介质
CN115311581A (zh) 一种基于无人机高光谱的红树林生态健康评估系统
WO2024174726A1 (zh) 基于深度学习的手写及打印文本检测方法和装置
CN110428402B (zh) 图像篡改识别方法、装置、计算机设备及存储介质
Öztürk et al. Computer vision-based lane detection and detection of vehicle, traffic sign, pedestrian using YOLOv5
CN113723176B (zh) 一种目标对象确定方法、装置、存储介质和电子装置
CN113469070A (zh) 车载人员数量检测方法、装置和服务器
CN117596058A (zh) 网络信息的安全保护系统及方法
CN117095198A (zh) 遥感图像检索网络训练方法、应用方法、电子设备及介质
CN114419460B (zh) 水体区域提取方法及装置
CN110852300A (zh) 地物分类的方法、地图的绘制方法、装置及电子设备
CN116012731B (zh) 一种星上在轨云检测的方法、系统及设备
CN118865171A (zh) 图像中线状结构识别分割的深度学习模型、方法、存储介质和装置
HK40039167A (zh) 团雾识别方法、装置、电子设备及存储介质

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022513667

Country of ref document: JP

Kind code of ref document: A

Ref document number: 20227006772

Country of ref document: KR

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21843372

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21843372

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 522432623

Country of ref document: SA

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10.07.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21843372

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 522432623

Country of ref document: SA

WWR Wipo information: refused in national office

Ref document number: 522432623

Country of ref document: SA