CN114332715A - Method, device and equipment for identifying snow through automatic meteorological observation and storage medium - Google Patents
Method, device and equipment for identifying snow through automatic meteorological observation and storage medium Download PDFInfo
- Publication number
- CN114332715A CN114332715A CN202111653424.4A CN202111653424A CN114332715A CN 114332715 A CN114332715 A CN 114332715A CN 202111653424 A CN202111653424 A CN 202111653424A CN 114332715 A CN114332715 A CN 114332715A
- Authority
- CN
- China
- Prior art keywords
- snow
- identification
- automatic
- channel
- features
- 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.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method, a device, equipment and a storage medium for identifying snow cover through automatic meteorological observation, wherein the method comprises the steps of acquiring current video data of an automatic meteorological station, and carrying out preliminary snow cover identification on the current video data to obtain a snow cover target area; extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics; the image features of the verification image are matched with the snow strengthening features, whether snow exists at the current moment or not is judged according to the matching result, detail information of the snow in the video data can be obtained, other useless information is restrained, the speed and the efficiency of snow identification are improved, misjudgment of snow weather is avoided, and the accuracy of snow identification is improved.
Description
Technical Field
The invention relates to the technical field of weather monitoring, in particular to a method, a device, equipment and a storage medium for identifying snow through automatic meteorological observation.
Background
The unattended automatic weather station is an important component of a weather monitoring automation project; the system has the advantages of high efficiency, convenience, accuracy and the like, can save labor cost, is beneficial to reducing adverse effects of human factors on data, has the characteristics of long distance between observation stations, short observation time interval, strong timeliness, high accuracy and the like, and has the advantage that in developed countries, an automatic station becomes an indispensable infrastructure for urban weather forecasting, disaster prevention and reduction.
The snow cover is a result of combined action of multiple meteorological elements on the ground, the meteorological data of different observation sources adopt more identification methods, the existing research has certain breakthrough on weather identification, but the research on snow cover identification is relatively less, and how to strengthen image interpretation and snow cover monitoring is better integrated with the deep learning model in the large background that machine learning and deep learning develop well; the snow cover identification is essentially image identification and feature expression, and the weather identification combining multi-source data angle classification can be divided into two directions of feature extraction and classifier classification; the feature extraction direction mainly extracts image features, and classification results can be obtained by inputting the extracted features into a trained or parameter-set classifier; representative methods are element-based methods such as gray level co-occurrence matrix; texture-based methods, such as structural methods; energy Analysis based methods such as spectroscopy, while classifier classification is most commonly used with supervised and unsupervised classification, such as K-means and Iterative Self-Organizing Data Analysis (ISODATA); the supervision classification method generally comprises a maximum likelihood method, a minimum distance method and a mahalanobis distance method; with the development of the subject, people also put forward clustering analysis, support vector machine algorithm, neural network and the like; the traditional algorithms have respective defects, the problems of deformation, illumination, blurring and the like exist in the image obtained by general weather, and the internal interference information is more.
Disclosure of Invention
The invention mainly aims to provide a method, a device, equipment and a storage medium for identifying snow cover through meteorological automatic observation, and aims to solve the technical problems of more snow cover identification interference information and lower identification accuracy in the prior art.
In a first aspect, the invention provides a method for identifying meteorological automatic observation snow, which comprises the following steps:
acquiring current video data of an automatic weather station, and carrying out snow preliminary identification on the current video data to obtain a snow target area;
extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics;
and matching the picture characteristics of the verification picture with the snow strengthening characteristics, and judging whether snow exists at the current moment according to a matching result.
Optionally, the acquiring current video data of the automatic weather station, performing preliminary snow identification on the current video data, and acquiring a snow target area includes:
acquiring current video data shot by each observation field lens of the automatic weather station;
unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area.
Optionally, the extracting, by a residual network model, the snow feature of the snow target area, and performing global integration and local focusing on the snow feature by a location attention module and a channel attention module to obtain a snow-enhanced feature includes:
extracting deep semantic features of the accumulated snow from the accumulated snow target area through a residual error network model to serve as accumulated snow features;
carrying out weighted summation on the features of each position in the accumulated snow features through a position attention module and carrying out local focusing to obtain position focusing features;
integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module and carrying out snow information channel mapping to obtain channel associated characteristics;
and taking the position focusing characteristic and the channel correlation characteristic as a snow enhancement characteristic.
Optionally, the obtaining, by the location attention module, location focus features by performing weighted summation and local focusing on features of each location in the snow-covered features includes:
acquiring a local feature map corresponding to the snow feature, inputting the local feature map into a position attention module, and acquiring a convolution feature map from the attention module through a convolution layer of a convolution neural network;
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain a space attention diagram;
and acquiring a preset scale parameter and a superposed pixel of the local feature map, and calculating to obtain a position focusing feature according to the space attention map, the preset size parameter and the superposed pixel.
Optionally, the dimension change of the convolution feature map and the matrix multiplication calculation of the feature map after the dimension change are performed to obtain the spatial attention map include:
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change through the following formula to obtain a space attention diagram:
wherein x isjiFor space purposes, C is the number of channels, Ai、AjThe element at the i-th and j-th positions of the local feature map A, exp (A)i·Aj) Is AiAnd AjThe product of (d) is an exponent to base e.
Optionally, the obtaining of the superimposed pixel of the preset scale parameter and the local feature map and the calculating according to the space attention map, the preset size parameter and the superimposed pixel to obtain the position focusing feature includes:
acquiring a preset scale parameter and a superposed pixel of the local feature map;
calculating according to the space attention drawing, the preset size parameter and the superposed pixels to obtain a position focusing characteristic through the following formula:
wherein E isjFor the position-focusing feature, beta is a predetermined size parameter, xjiFor space purposes, C is the number of channels, Ai、AjThe elements at the i-th and j-th positions of the local feature map A are respectively.
Optionally, after the image feature of the verification image is matched with the snow enhancement feature and whether snow exists at the current moment is judged according to a matching result, the method for identifying weather automatic observation snow further includes:
mapping the snow features corresponding to the position attention module and the channel attention module into the same target feature map through a convolutional layer of the convolutional neural network;
overlapping the correlated channel characteristic pixels in the target characteristic diagram, and performing snow accumulation identification on the overlapped characteristics through a full-connection layer of the convolutional neural network according to the following formula to obtain a loss function:
where Loss is the Loss function, yiThe ith channel feature of the channel features after pixel superposition, N is the number of the channel features after pixel superposition, piThe prediction probability of the ith channel feature of the channel features after pixel superposition;
and evaluating a snow cover identification result according to the loss function.
In a second aspect, to achieve the above object, the present invention further provides a weather automatic snow observation device including:
the area determining module is used for acquiring current video data of the automatic weather station, and performing snow preliminary identification on the current video data to acquire a snow target area;
the focusing module is used for extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through the position attention module and the channel attention module to obtain snow strengthening characteristics;
and the matching module is used for matching the picture characteristics of the verification picture with the snow strengthening characteristics and judging whether snow exists at the current moment according to a matching result.
In a third aspect, to achieve the above object, the present invention further provides a weather automatic snow observation device including: a memory, a processor, and a weather-meteorological automatic observed snow identification program stored on the memory and executable on the processor, the weather-meteorological automatic observed snow identification program configured to implement the steps of the weather-meteorological automatic observed snow identification method as described above.
In a fourth aspect, to achieve the above object, the present invention further provides a storage medium having stored thereon a weather automatic observed snow identification program, which when executed by a processor, implements the steps of the weather automatic observed snow identification method as described above.
The invention provides a method for identifying snow cover for meteorological automatic observation, which comprises the steps of obtaining current video data of an automatic meteorological station, carrying out preliminary snow cover identification on the current video data, and obtaining a snow cover target area; extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics; the image features of the verification image are matched with the snow strengthening features, whether snow exists at the current moment or not is judged according to the matching result, detail information of the snow in the video data can be obtained, other useless information is restrained, the speed and the efficiency of snow identification are improved, misjudgment of snow weather is avoided, and the accuracy of snow identification is improved.
Drawings
FIG. 1 is a schematic diagram of an apparatus architecture of a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating a first embodiment of the automatic weather-observation snow-accumulation identification method according to the present invention;
FIG. 3 is a flowchart illustrating a snow identification method for automatic meteorological observation according to a second embodiment of the present invention;
FIG. 4 is a schematic flow chart illustrating a snow identification method for automatic meteorological observation according to a third embodiment of the present invention;
FIG. 5 is a schematic view illustrating a flow chart of a method for identifying snow cover for automatic meteorological observation according to a fourth embodiment of the present invention;
FIG. 6 is a schematic diagram of a location attention module in the automatic weather observation snow identification method according to the present invention;
FIG. 7 is a schematic diagram of a lane attention module in the automatic weather observation snow identification method according to the present invention;
FIG. 8 is a functional block diagram of the automatic weather-observation snow-covered recognition apparatus according to the first embodiment of the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The solution of the embodiment of the invention is mainly as follows: acquiring current video data of an automatic weather station, and carrying out snow preliminary identification on the current video data to obtain a snow target area; extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics; the picture characteristics of the verification picture are matched with the snow strengthening characteristics, whether snow exists at the current moment or not is judged according to the matching result, detail information of the snow in the video data can be obtained, other useless information is restrained, the speed and the efficiency of snow identification are improved, the misjudgment of snow weather is avoided, the accuracy of snow identification is improved, and the technical problems that in the prior art, the amount of snow identification interference information is large, and the identification accuracy is low are solved.
Referring to fig. 1, fig. 1 is a schematic device structure diagram of a hardware operating environment according to an embodiment of the present invention.
As shown in fig. 1, the apparatus may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Wherein a communication bus 1002 is used to enable connective communication between these components. The user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., a Wi-Fi interface). The Memory 1005 may be a high-speed RAM Memory or a Non-Volatile Memory (Non-Volatile Memory), such as a disk Memory. The memory 1005 may alternatively be a storage device separate from the processor 1001.
Those skilled in the art will appreciate that the configuration of the apparatus shown in fig. 1 is not intended to be limiting of the apparatus and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.
As shown in fig. 1, a memory 1005, which is a kind of storage medium, may include therein an operating system, a network communication module, a user interface module, and a weather automatic snow-observation identifying program.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and performs the following operations:
acquiring current video data of an automatic weather station, and carrying out snow preliminary identification on the current video data to obtain a snow target area;
extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics;
and matching the picture characteristics of the verification picture with the snow strengthening characteristics, and judging whether snow exists at the current moment according to a matching result.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and also performs the following operations:
acquiring current video data shot by each observation field lens of the automatic weather station;
unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and also performs the following operations:
extracting deep semantic features of the accumulated snow from the accumulated snow target area through a residual error network model to serve as accumulated snow features;
carrying out weighted summation on the features of each position in the accumulated snow features through a position attention module and carrying out local focusing to obtain position focusing features;
integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module and carrying out snow information channel mapping to obtain channel associated characteristics;
and taking the position focusing characteristic and the channel correlation characteristic as a snow enhancement characteristic.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and also performs the following operations:
acquiring a local feature map corresponding to the snow feature, inputting the local feature map into a position attention module, and acquiring a convolution feature map from the attention module through a convolution layer of a convolution neural network;
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain a space attention diagram;
and acquiring a preset scale parameter and a superposed pixel of the local feature map, and calculating to obtain a position focusing feature according to the space attention map, the preset size parameter and the superposed pixel.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and also performs the following operations:
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change through the following formula to obtain a space attention diagram:
wherein x isjiFor space purposes, C is the number of channels, Ai、AjThe element at the i-th and j-th positions of the local feature map A, exp (A)i·Aj) Is AiAnd AjThe product of (d) is an exponent to base e.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and also performs the following operations:
acquiring a preset scale parameter and a superposed pixel of the local feature map;
calculating according to the space attention drawing, the preset size parameter and the superposed pixels to obtain a position focusing characteristic through the following formula:
wherein E isjFor the position-focusing feature, beta is a predetermined size parameter, xjiFor space purposes, C is the number of channels, Ai、AjThe elements at the i-th and j-th positions of the local feature map A are respectively.
The apparatus of the present invention calls the weather automatic snow observation snow identification program stored in the memory 1005 by the processor 1001, and also performs the following operations:
mapping the snow features corresponding to the position attention module and the channel attention module into the same target feature map through a convolutional layer of the convolutional neural network;
overlapping the correlated channel characteristic pixels in the target characteristic diagram, and performing snow accumulation identification on the overlapped characteristics through a full-connection layer of the convolutional neural network according to the following formula to obtain a loss function:
where Loss is the Loss function, yiThe ith channel feature of the channel features after pixel superposition, N is the number of the channel features after pixel superposition, piThe prediction probability of the ith channel feature of the channel features after pixel superposition;
and evaluating a snow cover identification result according to the loss function.
According to the scheme, the snow target area is obtained by acquiring the current video data of the automatic weather station and carrying out preliminary snow identification on the current video data; extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics; the image features of the verification image are matched with the snow strengthening features, whether snow exists at the current moment or not is judged according to the matching result, detail information of the snow in the video data can be obtained, other useless information is restrained, the speed and the efficiency of snow identification are improved, misjudgment of snow weather is avoided, and the accuracy of snow identification is improved.
Based on the hardware structure, the embodiment of the method for identifying the snow through automatic meteorological observation is provided.
Referring to fig. 2, fig. 2 is a schematic flow chart of a first embodiment of the method for identifying snow based on automatic weather observation according to the present invention.
In a first embodiment, the automatic meteorological observation snow identification method includes the steps of:
and step S10, acquiring current video data of the automatic weather station, and carrying out snow preliminary identification on the current video data to obtain a snow target area.
It should be noted that the current video data is data recorded by video shooting at an automatic weather station, and the snow cover of the current video data is preliminarily identified, so that a snow cover target area needing to be focused in the video data can be obtained.
And step S20, extracting the snow characteristics of the snow target area through a residual error network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow enhancement characteristics.
It can be understood that more detailed information of the snow can be extracted from the snow target area through the residual error network model, namely, the snow features can be globally integrated and locally focused through the position attention module and the channel attention module, so that the snow strengthening features are obtained, and finally, the accurate recognition of the snow weather is achieved.
And step S30, matching the picture characteristics of the verification picture with the snow strengthening characteristics, and judging whether snow exists at the current moment according to the matching result.
It should be understood that the verification picture is a preset snow verification picture for matching the snow enhancement features and the snow enhancement features, the picture features of the verification picture are matched with the snow enhancement features, a corresponding matching result can be generated, and whether snow exists at the current moment or not can be judged through the matching result.
Further, after the step S30, the method for identifying snow cover for meteorological automatic observation further includes the following steps:
mapping the snow features corresponding to the position attention module and the channel attention module into the same target feature map through a convolutional layer of the convolutional neural network;
overlapping the correlated channel characteristic pixels in the target characteristic diagram, and performing snow accumulation identification on the overlapped characteristics through a full-connection layer of the convolutional neural network according to the following formula to obtain a loss function:
where Loss is the Loss function, yiThe ith channel feature of the channel features after pixel superposition, N is the number of the channel features after pixel superposition, piThe prediction probability of the ith channel feature of the channel features after pixel superposition;
and evaluating a snow cover identification result according to the loss function.
It should be noted that, in order to fully combine spatial context information, local information, inter-class information, and the like, the features of the location and channel attention modules are aggregated herein; specifically, the outputs of the two modules are mapped into the same characteristic diagram through the convolution layer, and then pixel superposition is carried out; and finally, snow cover identification is carried out through the full connection layer.
It is understood that the position attention, i.e. the area corresponding to the length and width of the feature map, can be selectively focused; the attention of the channels corresponds to the number of the channels of the characteristic diagram, and the channels which are related to each other can be selectively emphasized; can obtain the loss function through above-mentioned formula, the loss function is the error between snow discernment predicted value and the true value, through the loss function can assess snow discernment result, and whether snow discernment effect is accurate promptly can be assessed through the loss function, and the loss function is big more, and snow discernment degree of accuracy is lower, and is corresponding, and the loss function is little, and snow discernment degree of accuracy is high more, can carry out snow discernment precision evaluation and analysis through the loss function.
According to the scheme, the snow target area is obtained by acquiring the current video data of the automatic weather station and carrying out preliminary snow identification on the current video data; extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics; the image features of the verification image are matched with the snow strengthening features, whether snow exists at the current moment or not is judged according to the matching result, detail information of the snow in the video data can be obtained, other useless information is restrained, the speed and the efficiency of snow identification are improved, misjudgment of snow weather is avoided, and the accuracy of snow identification is improved.
Further, fig. 3 is a schematic flow chart of a second embodiment of the method for identifying snow cover through automatic weather observation according to the present invention, and as shown in fig. 3, the second embodiment of the method for identifying snow cover through automatic weather observation according to the present invention is provided based on the first embodiment, and in this embodiment, the step S10 specifically includes the following steps:
and step S11, acquiring current video data shot by each observation field lens of the automatic weather station.
Note that the current video data is data observed by each observation field lens of the automated weather station.
And step S12, unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area.
It can be understood that, because reasons such as light, focusing, weather, the picture of shooing is not of uniform size, for calculation convenience, can be right current video data's image resolution ratio is unified, predetermines key regional information for the relevant information of the regional detail of the differentiation snow that sets up in advance, can carry out the preliminary discernment of snow to video data after unified resolution ratio according to predetermineeing key regional information, obtains the snow target area.
In a specific implementation, the numpy scientific calculation library of Python may be uniformly utilized to perform the same-specification pixel conversion on the picture into a picture with a corresponding resolution, for example, a resolution of 600 × 400 (width × height) pixels, or of course, other resolutions may be used, which is not limited in this embodiment.
According to the scheme, the current video data shot by the lenses of each observation field of the automatic weather station are acquired; unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area; corresponding snow areas can be accurately obtained, and the accuracy of snow identification is further improved.
Further, fig. 4 is a schematic flow chart of a third embodiment of the method for identifying snow cover through automatic weather observation according to the present invention, and as shown in fig. 4, the third embodiment of the method for identifying snow cover through automatic weather observation according to the present invention is provided based on the first embodiment, and in this embodiment, the step S20 specifically includes the following steps:
and step S21, extracting deep semantic features of the snow from the snow target area through a residual error network model to serve as snow features.
It should be noted that, deep semantic features of the snow can be extracted from the snow target region through the residual error network model, so as to extract the deep semantic features of the snow.
And step S22, carrying out weighted summation on the features of each position in the snow-covered features through a position attention module and carrying out local focusing to obtain position focusing features.
It is understood that the location attention module may selectively focus the feature of each location by weighted summation of all locations, i.e., the feature of each location in the snow-covered feature, and locally focus to obtain the location focus feature.
Step S23, integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module, and performing snow information channel mapping to obtain channel associated characteristics.
It should be understood that the channel attention module selectively emphasizes the correlated channel maps by integrating the correlated features in all the channel maps, and finally emphasizes the expression of the snow accumulation features by comparing the feature maps of the two attention modules, that is, integrates the correlated channel features in the snow accumulation features and performs snow accumulation information channel mapping to obtain the channel correlation features, so that the interdependent feature mapping can be emphasized to improve the feature representation of the original snow accumulation information.
And step S24, taking the position focusing characteristic and the channel correlation characteristic as a snow-cover strengthening characteristic.
It will be appreciated that after the location focus feature and the attention focus feature are obtained, the location focus feature and the channel correlation feature may be used as snow enhancement features.
According to the scheme, the deep semantic features of the accumulated snow are extracted from the accumulated snow target area through the residual error network model to serve as the accumulated snow features; carrying out weighted summation on the features of each position in the accumulated snow features through a position attention module and carrying out local focusing to obtain position focusing features; integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module and carrying out snow information channel mapping to obtain channel associated characteristics; the position focusing feature and the channel correlation feature are used as snow enhancement features, the snow enhancement features can be enhanced, detail information of snow in video data is obtained, other useless information is restrained, and speed and efficiency of snow identification are improved.
Further, fig. 5 is a schematic flow chart of a fourth embodiment of the method for identifying snow based on automatic weather observation according to the present invention, and as shown in fig. 5, the fourth embodiment of the method for identifying snow based on automatic weather observation according to the present invention is provided based on the third embodiment, in this embodiment, the step S22 specifically includes the following steps:
and step S221, acquiring a local feature map corresponding to the snow feature, inputting the local feature map into a position attention module, and acquiring a convolution feature map from the attention module through a convolution layer of a convolution neural network.
The snow feature corresponds to a local feature and a global feature, the snow feature corresponds to a local feature map of the local feature, and the convolution feature map is obtained from the attention module through a convolution layer of a convolution neural network.
In specific implementation, the distinguishing feature is a key for understanding an image scene, for snow identification, a large-range and local snow situation may occur, even though the convolution and pooling of the CNN can gradually improve the receptive field, the sensitivity detection on the local feature is lacked, and the location attention module can extract wider context information to encode the local feature, so that the network focuses on the local feature of the snow.
Referring to FIG. 6, FIG. 6 is a schematic diagram of the location attention module in the automatic weather-observing snow-accumulation recognition method of the present invention, as shown in FIG. 6, inputting a local feature map A ∈ RC*H*WC, H, W, which are the number of channels, height, and width of the feature map, first, B, C two feature maps will be generated by convolutional layers, in which B, C dimension is changed to RC*NWherein N is H W is the number of pixels, then B is processed by matrix multiplication with C after being processed by a device, and space attention S belongs to R by adopting softmax to calculate spaceN*N：
In the above formula, SjiDenotes the influence of the ith position on the jth position, BiFor the element at the i-th position of the matrix corresponding to the feature map B, CjFor the element at the j-th position of the matrix, exp (B) corresponds to the feature map Ci·Cj) Is BiAnd CjThe product of (a) is an index with e as the base; if the more similar the characteristics of the two positions areThe greater the correlation between them.
Meanwhile, the generation process of D is similar to B, C, and D is generated through convolution and dimension change, and D belongs to RC*H*W(ii) a Secondly, matrix multiplication and dimension change are carried out on the transpose of the D and the S, finally, the result is multiplied by a scale parameter alpha and is superposed with the characteristic A pixel, and the output is E ∈ RC*H*WNamely:
in the above formula, α is initialized to 0, and more weights are gradually learned and assigned, and as can be seen from the above formula, all feature points on the E feature map are weighted sums of features of all positions and original features, so that the E feature map has rich global information, and selectively focuses on local features according to spatial attention, that is, similar features express more sufficiently, and compactness and semantic consistency within a class are increased.
And step S222, carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain the space attention diagram.
It can be understood that dimension change is performed on the convolution feature map, a feature map after dimension change can be obtained, and the spatial attention map is obtained through calculation by performing matrix multiplication on the feature map after dimension change.
Further, the step S222 specifically includes the following steps:
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change through the following formula to obtain a space attention diagram:
wherein x isjiFor space purposes, C is the number of channels, Ai、AjThe element at the i-th and j-th positions of the local feature map A, exp (A)i·Aj) Is AiAnd AjThe product of (d) is an exponent to base e.
It is understood that, referring to FIG. 7, FIG. 7 is a schematic diagram of the lane attention module in the automatic weather observation snow cover identification method of the present invention, as shown in FIG. 7, unlike the location attention module, X ∈ RC*CFrom A e to RC*H*WAnd (3) directly calculating and generating, specifically, firstly carrying out dimension transformation on A, then carrying out matrix multiplication on the transpose of A and A, and finally generating a channel attention feature map X epsilon R by adopting softmaxC*C。
And S223, acquiring a preset scale parameter and a superposed pixel of the local feature map, and calculating to obtain a position focusing feature according to the space attention map, the preset size parameter and the superposed pixel.
It should be understood that the preset scale parameter is a preset scale parameter, a feature after pixel superposition is obtained by pixel superposition with a local feature map, and a position focusing feature can be obtained by calculation according to the space attention map, the preset size parameter and the superposed pixels.
Further, the step S223 includes the following steps:
acquiring a preset scale parameter and a superposed pixel of the local feature map;
calculating according to the space attention drawing, the preset size parameter and the superposed pixels to obtain a position focusing characteristic through the following formula:
wherein E isjFor the position-focusing feature, beta is a predetermined size parameter, xjiFor space purposes, C is the number of channels, Ai、AjThe elements at the i-th and j-th positions of the local feature map A are respectively.
It should be noted that this embodiment may continue to operate on X's device and a using matrix multiplication, and change the resulting dimension to RC*H*WSimilarly to the scale parameter α, the same scale parameter β is introduced here for the above resultsThe product operation is carried out, the result of the product operation and A are subjected to pixel superposition to generate a final feature map E epsilon RC*H*WIn the above formula, β gradually starts to learn from 0, and as can be seen from the above formula, the final feature of each channel is the weighted sum of all channel features and the original features, so that the rich semantic dependency relationship between different channel feature maps is strengthened, and the feature expression of snow cover discrimination is improved.
According to the scheme, the local feature map corresponding to the snow feature is obtained, the local feature map is input to the position attention module, and the convolution feature map is obtained from the attention module through the convolution layer of the convolution neural network; carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain a space attention diagram; the method comprises the steps of obtaining a preset scale parameter and overlapping pixels of the local feature map, calculating according to the space attention map, the preset size parameter and the overlapping pixels to obtain a position focusing feature, strengthening the snow feature, obtaining detail information of snow in video data, restraining other useless information and improving the speed and efficiency of snow identification.
Correspondingly, the invention further provides a recognition device for automatically observing the snow cover in the weather.
Referring to fig. 8, fig. 8 is a functional block diagram of the first embodiment of the automatic weather-observation snow-covered recognition apparatus according to the present invention.
In a first embodiment of the automatic weather-observation snow cover identification apparatus according to the present invention, the automatic weather-observation snow cover identification apparatus includes:
and the area determining module 10 is used for acquiring current video data of the automatic weather station, and performing snow preliminary identification on the current video data to acquire a snow target area.
And the focusing module 20 is used for extracting the snow characteristics of the snow target area through a residual error network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow enhancement characteristics.
And the matching module 30 is used for matching the picture characteristics of the verification picture with the snow strengthening characteristics and judging whether snow exists at the current moment according to a matching result.
The area determining module 10 is further configured to obtain current video data captured by lenses of each observation field of the automatic weather station; unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area.
The focusing module 20 is further configured to extract a deep semantic feature of the accumulated snow from the accumulated snow target region through a residual network model as an accumulated snow feature; carrying out weighted summation on the features of each position in the accumulated snow features through a position attention module and carrying out local focusing to obtain position focusing features; integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module and carrying out snow information channel mapping to obtain channel associated characteristics; and taking the position focusing characteristic and the channel correlation characteristic as a snow enhancement characteristic.
The focusing module 20 is further configured to obtain a local feature map corresponding to the feature of the snow, input the local feature map to a position attention module, and obtain a convolution feature map from the attention module through a convolution layer of a convolution neural network; carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain a space attention diagram; and acquiring a preset scale parameter and a superposed pixel of the local feature map, and calculating to obtain a position focusing feature according to the space attention map, the preset size parameter and the superposed pixel.
The focusing module 20 is further configured to perform dimension change on the convolution feature map, and perform matrix multiplication on the feature map after the dimension change according to the following formula to obtain a spatial attention map:
wherein x isjiFor space purposes, C is the number of channels, Ai、AjThe elements are respectively at the i th and j th positions of the local feature map AElement, exp (A)i·Aj) Is AiAnd AjThe product of (d) is an exponent to base e.
The focusing module 20 is further configured to obtain a preset scale parameter and a superimposed pixel of the local feature map;
calculating according to the space attention drawing, the preset size parameter and the superposed pixels to obtain a position focusing characteristic through the following formula:
wherein E isjFor the position-focusing feature, beta is a predetermined size parameter, xjiFor space purposes, C is the number of channels, Ai、AjThe elements at the i-th and j-th positions of the local feature map A are respectively.
The matching module 30 is further configured to map the snow features corresponding to the location attention module and the channel attention module into the same target feature map through a convolutional layer of the convolutional neural network;
overlapping the correlated channel characteristic pixels in the target characteristic diagram, and performing snow accumulation identification on the overlapped characteristics through a full-connection layer of the convolutional neural network according to the following formula to obtain a loss function:
where Loss is the Loss function, yiThe ith channel feature of the channel features after pixel superposition, N is the number of the channel features after pixel superposition, piThe prediction probability of the ith channel feature of the channel features after pixel superposition;
and evaluating a snow cover identification result according to the loss function.
The steps implemented by the functional modules of the automatic weather observation snow cover identification device can refer to the embodiments of the automatic weather observation snow cover identification method, and are not described herein again.
Furthermore, an embodiment of the present invention further provides a storage medium, in which a weather automatic snow observation identification program is stored, and when executed by a processor, the weather automatic snow observation identification program implements the following operations:
acquiring current video data of an automatic weather station, and carrying out snow preliminary identification on the current video data to obtain a snow target area;
extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics;
and matching the picture characteristics of the verification picture with the snow strengthening characteristics, and judging whether snow exists at the current moment according to a matching result.
Further, the automatic meteorological observation snow identification program when executed by the processor further implements the following operations:
acquiring current video data shot by each observation field lens of the automatic weather station;
unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area.
Further, the automatic meteorological observation snow identification program when executed by the processor further implements the following operations:
extracting deep semantic features of the accumulated snow from the accumulated snow target area through a residual error network model to serve as accumulated snow features;
carrying out weighted summation on the features of each position in the accumulated snow features through a position attention module and carrying out local focusing to obtain position focusing features;
integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module and carrying out snow information channel mapping to obtain channel associated characteristics;
and taking the position focusing characteristic and the channel correlation characteristic as a snow enhancement characteristic.
Further, the automatic meteorological observation snow identification program when executed by the processor further implements the following operations:
acquiring a local feature map corresponding to the snow feature, inputting the local feature map into a position attention module, and acquiring a convolution feature map from the attention module through a convolution layer of a convolution neural network;
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain a space attention diagram;
and acquiring a preset scale parameter and a superposed pixel of the local feature map, and calculating to obtain a position focusing feature according to the space attention map, the preset size parameter and the superposed pixel.
Further, the automatic meteorological observation snow identification program when executed by the processor further implements the following operations:
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change through the following formula to obtain a space attention diagram:
wherein x isjiFor space purposes, C is the number of channels, Ai、AjThe element at the i-th and j-th positions of the local feature map A, exp (A)i·Aj) Is AiAnd AjThe product of (d) is an exponent to base e.
Further, the automatic meteorological observation snow identification program when executed by the processor further implements the following operations:
acquiring a preset scale parameter and a superposed pixel of the local feature map;
calculating according to the space attention drawing, the preset size parameter and the superposed pixels to obtain a position focusing characteristic through the following formula:
wherein E isjFor the position-focusing feature, beta is a predetermined size parameter, xjiFor space purposes, C is the number of channels, Ai、AjThe elements at the i-th and j-th positions of the local feature map A are respectively.
Further, the automatic meteorological observation snow identification program when executed by the processor further implements the following operations:
mapping the snow features corresponding to the position attention module and the channel attention module into the same target feature map through a convolutional layer of the convolutional neural network;
overlapping the correlated channel characteristic pixels in the target characteristic diagram, and performing snow accumulation identification on the overlapped characteristics through a full-connection layer of the convolutional neural network according to the following formula to obtain a loss function:
where Loss is the Loss function, yiThe ith channel feature of the channel features after pixel superposition, N is the number of the channel features after pixel superposition, piThe prediction probability of the ith channel feature of the channel features after pixel superposition;
and evaluating a snow cover identification result according to the loss function.
According to the scheme, the snow target area is obtained by acquiring the current video data of the automatic weather station and carrying out preliminary snow identification on the current video data; extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics; the image features of the verification image are matched with the snow strengthening features, whether snow exists at the current moment or not is judged according to the matching result, detail information of the snow in the video data can be obtained, other useless information is restrained, the speed and the efficiency of snow identification are improved, misjudgment of snow weather is avoided, and the accuracy of snow identification is improved.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. The method for identifying the meteorological automatic observation snow cover is characterized by comprising the following steps of:
acquiring current video data of an automatic weather station, and carrying out snow preliminary identification on the current video data to obtain a snow target area;
extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through a position attention module and a channel attention module to obtain snow strengthening characteristics;
and matching the picture characteristics of the verification picture with the snow strengthening characteristics, and judging whether snow exists at the current moment according to a matching result.
2. The weather automatic snow observation and identification method according to claim 1, wherein the acquiring current video data of an automatic weather station, performing snow preliminary identification on the current video data, and acquiring a snow target area comprises:
acquiring current video data shot by each observation field lens of the automatic weather station;
unifying the image resolution of the current video data, and carrying out snow initial identification on the video data with unified resolution according to preset key area information to obtain a snow target area.
3. The method for weather automatic observation snow identification as claimed in claim 1, wherein the extracting the snow feature of the snow target area through a residual network model, and the global integration and local focusing of the snow feature through a location attention module and a channel attention module to obtain the snow enhancement feature comprises:
extracting deep semantic features of the accumulated snow from the accumulated snow target area through a residual error network model to serve as accumulated snow features;
carrying out weighted summation on the features of each position in the accumulated snow features through a position attention module and carrying out local focusing to obtain position focusing features;
integrating the mutually-associated channel characteristics in the snow characteristics through a channel attention module and carrying out snow information channel mapping to obtain channel associated characteristics;
and taking the position focusing characteristic and the channel correlation characteristic as a snow enhancement characteristic.
4. The weather-meteorological automatic observed snow identification method according to claim 3, wherein the obtaining, by the location attention module, the location focus feature by performing weighted summation and local focusing on the features of each of the snow features comprises:
acquiring a local feature map corresponding to the snow feature, inputting the local feature map into a position attention module, and acquiring a convolution feature map from the attention module through a convolution layer of a convolution neural network;
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change to obtain a space attention diagram;
and acquiring a preset scale parameter and a superposed pixel of the local feature map, and calculating to obtain a position focusing feature according to the space attention map, the preset size parameter and the superposed pixel.
5. The method for identifying snow cover through automatic meteorological observation according to claim 4, wherein the step of performing dimension change on the convolution feature map, and performing matrix multiplication calculation on the feature map after dimension change to obtain the spatial attention map comprises the following steps:
carrying out dimension change on the convolution characteristic diagram, and carrying out matrix multiplication calculation on the characteristic diagram after the dimension change through the following formula to obtain a space attention diagram:
wherein x isjiFor space purposes, C is the number of channels, Ai、AjThe element at the i-th and j-th positions of the local feature map A, exp (A)i·Aj) Is AiAnd AjThe product of (d) is an exponent to base e.
6. The method for identifying snow cover through meteorological automatic observation according to claim 4, wherein the obtaining of the superimposed pixels of the preset scale parameter and the local feature map, and the obtaining of the position focusing feature through calculation according to the space attention map, the preset size parameter and the superimposed pixels comprise:
acquiring a preset scale parameter and a superposed pixel of the local feature map;
calculating according to the space attention drawing, the preset size parameter and the superposed pixels to obtain a position focusing characteristic through the following formula:
wherein E isjFor the position-focusing feature, beta is a predetermined size parameter, xjiFor space purposes, C is the number of channels, Ai、AjThe elements at the i-th and j-th positions of the local feature map A are respectively.
7. The method for identifying snow cover through automatic weather observation according to claim 1, wherein the matching of the picture feature of the verification picture with the snow cover strengthening feature is performed, and after determining whether snow cover exists at the current time according to the matching result, the method for identifying snow cover through automatic weather observation further comprises:
mapping the snow features corresponding to the position attention module and the channel attention module into the same target feature map through a convolutional layer of the convolutional neural network;
overlapping the correlated channel characteristic pixels in the target characteristic diagram, and performing snow accumulation identification on the overlapped characteristics through a full-connection layer of the convolutional neural network according to the following formula to obtain a loss function:
where Loss is the Loss function, yiThe ith channel feature of the channel features after pixel superposition, N is the number of the channel features after pixel superposition, piThe prediction probability of the ith channel feature of the channel features after pixel superposition;
and evaluating a snow cover identification result according to the loss function.
8. The utility model provides a meteorological automatic observation snow recognition device which characterized in that, meteorological automatic observation snow recognition device includes:
the area determining module is used for acquiring current video data of the automatic weather station, and performing snow preliminary identification on the current video data to acquire a snow target area;
the focusing module is used for extracting the snow characteristics of the snow target area through a residual network model, and performing global integration and local focusing on the snow characteristics through the position attention module and the channel attention module to obtain snow strengthening characteristics;
and the matching module is used for matching the picture characteristics of the verification picture with the snow strengthening characteristics and judging whether snow exists at the current moment according to a matching result.
9. The utility model provides a meteorological automatic observation snow identification equipment which characterized in that, meteorological automatic observation snow identification equipment includes: a memory, a processor, and a weather-meteorological automatic observed snow identification program stored on the memory and executable on the processor, the weather-meteorological automatic observed snow identification program configured to implement the steps of the weather-meteorological automatic observed snow identification method of any one of claims 1 to 7.
10. A storage medium characterized in that the storage medium has stored thereon a weather automatic observed snow identification program which, when executed by a processor, implements the steps of the weather automatic observed snow identification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111653424.4A CN114332715A (en) | 2021-12-30 | 2021-12-30 | Method, device and equipment for identifying snow through automatic meteorological observation and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111653424.4A CN114332715A (en) | 2021-12-30 | 2021-12-30 | Method, device and equipment for identifying snow through automatic meteorological observation and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114332715A true CN114332715A (en) | 2022-04-12 |
Family
ID=81019123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111653424.4A Pending CN114332715A (en) | 2021-12-30 | 2021-12-30 | Method, device and equipment for identifying snow through automatic meteorological observation and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114332715A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116437121A (en) * | 2023-04-28 | 2023-07-14 | 华风气象传媒集团有限责任公司 | Meteorological video processing method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978418A (en) * | 2015-06-29 | 2015-10-14 | 杨昕吉 | Intelligent weather image acquisition and identification method and system |
| CN106453523A (en) * | 2016-09-27 | 2017-02-22 | 北海益生源农贸有限责任公司 | Intelligent weather identification system and method |
| CN112131943A (en) * | 2020-08-20 | 2020-12-25 | 深圳大学 | Video behavior identification method and system based on dual attention model |
| CN112329533A (en) * | 2020-10-07 | 2021-02-05 | 江苏大学 | Local pavement adhesion coefficient estimation method based on image segmentation |
| CN113421269A (en) * | 2021-06-09 | 2021-09-21 | 南京瑞易智能科技有限公司 | Real-time semantic segmentation method based on double-branch deep convolutional neural network |
| CN113700231A (en) * | 2021-08-31 | 2021-11-26 | 南京邮电大学盐城大数据研究院有限公司 | Highway photovoltaic ceiling system of intelligence snow removing |
-
2021
- 2021-12-30 CN CN202111653424.4A patent/CN114332715A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978418A (en) * | 2015-06-29 | 2015-10-14 | 杨昕吉 | Intelligent weather image acquisition and identification method and system |
| CN106453523A (en) * | 2016-09-27 | 2017-02-22 | 北海益生源农贸有限责任公司 | Intelligent weather identification system and method |
| CN112131943A (en) * | 2020-08-20 | 2020-12-25 | 深圳大学 | Video behavior identification method and system based on dual attention model |
| CN112329533A (en) * | 2020-10-07 | 2021-02-05 | 江苏大学 | Local pavement adhesion coefficient estimation method based on image segmentation |
| CN113421269A (en) * | 2021-06-09 | 2021-09-21 | 南京瑞易智能科技有限公司 | Real-time semantic segmentation method based on double-branch deep convolutional neural network |
| CN113700231A (en) * | 2021-08-31 | 2021-11-26 | 南京邮电大学盐城大数据研究院有限公司 | Highway photovoltaic ceiling system of intelligence snow removing |
Non-Patent Citations (1)
| Title |
|---|
| JUN FU ET AL.: "Dual Attention Network for Scene Segmentation", 《COMPUTER VISION AND PATTERN RECOGNITION》, pages 3 - 4 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116437121A (en) * | 2023-04-28 | 2023-07-14 | 华风气象传媒集团有限责任公司 | Meteorological video processing method and system |
| CN116437121B (en) * | 2023-04-28 | 2023-10-13 | 华风气象传媒集团有限责任公司 | Meteorological video processing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111259786B (en) | Pedestrian re-identification method based on synchronous enhancement of appearance and motion information of video | |
| CN114067160B (en) | Small sample remote sensing image scene classification method based on embedded smooth graph neural network | |
| US11475660B2 (en) | Method and system for facilitating recognition of vehicle parts based on a neural network | |
| CN111931684B (en) | Weak and small target detection method based on video satellite data identification features | |
| CN111563508A (en) | Semantic segmentation method based on spatial information fusion | |
| CN113657425B (en) | Multi-label image classification method based on multi-scale and cross-modal attention mechanism | |
| CN114936623B (en) | Aspect-level emotion analysis method integrating multi-mode data | |
| CN111461039B (en) | Landmark identification method based on multi-scale feature fusion | |
| CN112861970B (en) | Fine-grained image classification method based on feature fusion | |
| CN115496928A (en) | Multi-modal image feature matching method based on multi-feature matching | |
| Zhou et al. | Algorithm of Helmet Wearing Detection Based on AT-YOLO Deep Mode. | |
| CN117152416A (en) | Sparse attention target detection method based on DETR improved model | |
| CN114972316A (en) | Battery case end surface defect real-time detection method based on improved YOLOv5 | |
| CN116091946A (en) | Yolov 5-based unmanned aerial vehicle aerial image target detection method | |
| CN117197763A (en) | Road crack detection method and system based on cross attention guide feature alignment network | |
| CN111882000A (en) | Network structure and method applied to small sample fine-grained learning | |
| CN111523586A (en) | Noise-aware-based full-network supervision target detection method | |
| CN114926826A (en) | Scene text detection system | |
| CN112668662B (en) | Outdoor mountain forest environment target detection method based on improved YOLOv3 network | |
| CN114332715A (en) | Method, device and equipment for identifying snow through automatic meteorological observation and storage medium | |
| CN112418207B (en) | Weak supervision character detection method based on self-attention distillation | |
| CN114038007A (en) | Pedestrian re-recognition method combining style transformation and attitude generation | |
| CN112269892B (en) | Based on multi-mode is unified at many levels Interactive phrase positioning and identifying method | |
| CN111178370B (en) | Vehicle searching method and related device | |
| CN108960005B (en) | Method and system for establishing and displaying object visual label in intelligent visual Internet of things |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |