CN117809230A - Water flow velocity identification method based on image identification and related products - Google Patents
Water flow velocity identification method based on image identification and related products Download PDFInfo
- Publication number
- CN117809230A CN117809230A CN202410225742.8A CN202410225742A CN117809230A CN 117809230 A CN117809230 A CN 117809230A CN 202410225742 A CN202410225742 A CN 202410225742A CN 117809230 A CN117809230 A CN 117809230A
- Authority
- CN
- China
- Prior art keywords
- matrix
- water flow
- hidden layer
- image
- identification
- 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
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 91
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 28
- 238000003062 neural network model Methods 0.000 claims abstract description 26
- 230000004927 fusion Effects 0.000 claims abstract description 22
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 15
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 87
- 239000013598 vector Substances 0.000 claims description 43
- 230000006870 function Effects 0.000 claims description 29
- 238000004590 computer program Methods 0.000 claims description 20
- 238000000354 decomposition reaction Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 9
- 238000004891 communication Methods 0.000 claims description 8
- RZVHIXYEVGDQDX-UHFFFAOYSA-N 9,10-anthraquinone Chemical compound C1=CC=C2C(=O)C3=CC=CC=C3C(=O)C2=C1 RZVHIXYEVGDQDX-UHFFFAOYSA-N 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 230000002123 temporal effect Effects 0.000 claims description 3
- 239000012530 fluid Substances 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 11
- 238000005259 measurement Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000000691 measurement method Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000013515 script Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- 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
- Y02A90/30—Assessment of water resources
Abstract
The invention relates to the field of image processing and fluid dynamics, in particular to a water flow velocity identification method based on image identification and a related product, wherein the method comprises the steps of obtaining a video sample and a sample picture; NSST processing is carried out on the sample picture; obtaining image characteristics through a CNN neural network; extracting time-frequency dynamic characteristics through an LBP-TOP algorithm; obtaining fusion characteristics; training to obtain a water flow velocity identification model; acquiring a water flow real-time video of a water area to be identified, processing the water flow real-time video to acquire fusion characteristics for identification, and inputting the fusion characteristics into a water flow velocity identification model to acquire the water flow velocity of the current water area; according to the invention, video samples under different water quality and flow rates are collected, images and time-frequency dynamic characteristics are extracted by combining CNN and LBP-TOP algorithms, and then high-precision identification of water flow rates is realized through characteristic fusion and establishment of a kernel random weight neural network model.
Description
Technical Field
The invention relates to the field of image processing and fluid dynamics, in particular to a water flow velocity identification method based on image identification and a related product.
Background
Accurate measurement of water flow velocity is important to fields such as hydraulic engineering, environmental monitoring and disaster prevention. Conventional flow rate measurement methods typically rely on contact measurement tools such as flowmeters, buoys, and the like. These methods are not only costly to install and maintain, but may not work properly under harsh environmental conditions.
With the development of image processing technology, image-based flow rate measurement methods are attracting attention. The method utilizes video or image data to analyze the water flow characteristics, and realizes non-contact flow velocity measurement. However, the accuracy and stability of existing image recognition techniques under conditions of complex water quality and variable flow rates are still to be improved.
Therefore, a new water flow speed identification method based on image identification is developed, accurate and stable flow speed measurement can be provided under different water quality and flow speed conditions, and the method has important significance in the fields of hydrology, environmental monitoring, hydraulic engineering and the like.
Disclosure of Invention
In order to solve the technical problems, the invention provides a water flow velocity identification method based on image identification and a related product, which realize non-contact measurement of water flow velocity.
The invention is realized by the following technical scheme:
a water flow rate identification method based on image identification, comprising:
acquiring video samples of different water qualities at different flow rates, and taking frame images in the video samples as sample pictures;
NSST processing is carried out on the sample picture, and the sample picture is divided into a training sample and a test sample;
performing feature extraction on the training sample and the test sample through a CNN neural network to obtain image features;
processing the training sample and the test sample by using an LBP-TOP algorithm, and extracting time-frequency dynamic characteristics;
carrying out feature fusion on the image features and the time-frequency dynamic features to obtain fusion features;
constructing a nuclear random weight neural network model, and training through fusion characteristics to obtain a water flow velocity identification model;
and acquiring a water flow real-time video of the water area to be identified, processing the water flow real-time video to acquire fusion characteristics for identification, and inputting the fusion characteristics into a water flow velocity identification model to acquire the current water flow velocity of the water area.
Optionally, the method for performing NSST processing on the sample picture includes:
performing multi-scale decomposition on the sample picture through a non-downsampled pyramid filter bank to obtainLow frequency image and->A layer high frequency subband image;
high frequency subband images by shear wave filtersStage multidirectional decomposition to obtain->A plurality of multidirectional subbands; the size of the multidirectional subband is the same as the size of the sample picture.
Optionally, the method for extracting the time-frequency dynamic characteristics comprises the following steps:
orthogonalization segmentation of sequences of video into spatial and temporal relationshipsPlane, & gt>Plane, & gt>A plane;
for a pair ofPlane, & gt>Plane, & gt>The plane is respectively provided with a radius and a field point;
determining a central pixel, obtaining LBP values of the central pixel on three planes, generating a characteristic vector of the central pixel on each plane, and normalizing the characteristic vector;
the within-class divergence matrix is calculated and,wherein->For category->Mean vector of>For category->Feature vector set of>Is the total number of categories;
the inter-class divergence matrix is calculated and,wherein->For the total mean vector of all feature vectors, +.>For category->Is a feature vector number of (a);
solving eigenvalues of inter-class divergence matrix and intra-class divergence matrixAnd feature vector->，/>Selecting the best discrimination feature vector corresponding to the largest feature value;
and projecting the original feature vector to the optimal discrimination feature vector to obtain the time-frequency dynamic feature.
Specifically, the method for obtaining the water flow velocity identification model comprises the following steps:
constructing a random weight neural network model and defining a loss function;
calculating hidden layer output matrix by using double hidden layer self-coding random weight neural network；
Obtaining a weight matrixWherein->For the target value matrix +.>Is a unitary matrix->Is a punishment parameter;
selecting RBF kernel function as kernel function of kernel random weight neural network modelAnd introducing a kernel matrix into the random weight neural network model>Constructing a nuclear random weight neural network model, wherein,input +_for hidden layer pair>Output of->Input +_for hidden layer pair>An output of (2);
updating a target value matrix based on a kernel function and a kernel matrixAnd weight matrix->；
Determining hidden layer outputAnd updated weight matrix +.>Obtaining the output of the kernel random weight neural network model。
Optionally, the loss function is:wherein->For training error +.>Is Lagrangian multiplier +.>For training sample total number>For outputting the number of categories +.>Input +_for hidden layer pair>Output of->For samples in the target value matrix->In category->Is (are) true tags->For sample->In category->Training error of->Is->Weight value of category corresponding in weight matrix, < ->For sample->In category->Lagrangian multiplier of (c).
Optionally, the calculating method of the hidden layer output matrix includes:
the first hidden layer and the second hidden layer of the random weight neural network model are replaced by constructing a double hidden layer self-coding random weight neural network;
randomly generating an input weight vector for a first hidden layer nodeAnd bias->The method comprises the steps of carrying out a first treatment on the surface of the Randomly generating an input weight vector for a second hidden layer node>And bias->；
Input device、/>、/>Calculation ofAn output matrix of the first hidden layer;
through the output matrix of the first hidden layer,、/>Calculating a second hidden layer output matrix +.>；
Will be from encoderCalculating +.>Calculating an output weight matrix of the self-encoder>Wherein->The number of layers for the self-encoder;
calculating an output matrix for each hidden node；
And taking the output matrix of the last hidden layer as the output matrix of the hidden layer.
Optionally, updating the target value matrix based on the kernel functionThe formula of (2) is: />，Wherein->，/>，/>Width parameters that are kernel functions;
the method for updating the weight matrix comprises the steps of updating the target value matrixSubstituted into->Obtaining an updated weight matrix +.>。
A water flow rate identification system based on image identification, comprising:
unmanned aerial vehicle and flight control system thereof;
the high-definition video shooting device is installed on the unmanned aerial vehicle;
the wireless communication assembly is arranged on the unmanned aerial vehicle, and the high-definition video shooting device is communicated with the main control computer through the wireless communication assembly;
the main control computer comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the water flow rate identification method based on image identification when executing the computer program.
A computer readable storage medium storing a computer program which when executed by a processor implements a water flow rate identification method based on image identification as described above.
A computer program product comprising computer programs/instructions which when executed by a processor implement a water flow rate identification method based on image identification as described above.
Compared with the prior art, the invention has the following advantages and beneficial effects:
according to the invention, video samples under different water quality and flow rates are collected, NSST (non-downsampling pyramid filter bank) is used for carrying out multi-scale decomposition and multi-directional subband processing on sample pictures, the feature extraction capability of images is improved, and the images and time-frequency dynamic features can be effectively extracted by combining CNN (convolutional neural network) and LBP-TOP (local binary pattern-space-time relation) algorithms, and then high-precision identification of water flow rates is realized through feature fusion and establishment of a kernel random weight neural network model.
According to the invention, through combining NSST processing and CNN feature extraction, details and features of the water flow image can be more effectively captured, and accuracy of water flow speed identification is improved under the conditions of complex water quality and variable flow speeds.
And the LBP-TOP algorithm is utilized to process the video sample, the dynamic change characteristic of the water flow is extracted, and the sensitivity and the response capability of the system to the flow speed change are improved.
Compared with the traditional contact type flow velocity measurement method, the non-contact type flow velocity measurement method provided by the invention has the advantages that equipment loss, maintenance cost and use limitation in a severe environment in the traditional method are avoided.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the principles of the invention.
Fig. 1 is a schematic flow chart of a water flow velocity identification method based on image identification according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and embodiments, for the purpose of making the objects, technical solutions and advantages of the present invention more apparent. It is to be understood that the specific embodiments described herein are merely illustrative of the substances, and not restrictive of the invention.
It should be further noted that, for convenience of description, only the portions related to the present invention are shown in the drawings.
Embodiments of the present invention and features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Example 1
As shown in fig. 1, a method for identifying water flow velocity based on image identification is provided, which comprises the following steps:
firstly, acquiring video samples of different water qualities at different flow rates, and taking frame images in the video samples as sample pictures; the purpose of this step is: sufficiently diverse data is collected to train and test the model. Different water quality and flow rate conditions can make the model more generalized, thereby providing more accurate identification in practical applications. The conversion of the video samples into frame images is to facilitate subsequent image processing and feature extraction.
Secondly, NSST processing is carried out on the sample picture, and the sample picture is divided into a training sample and a test sample; NSST (non-downsampled Shearlet transform) is an efficient image processing method for extracting multi-scale and multi-directional information of an image. The features of the image are enhanced through NSST, so that richer input data is provided for a subsequent machine learning model.
The processed image is divided into training and test samples for training and validating the model, typically at a ratio of 7:3.
Thirdly, extracting features of the training sample and the test sample through a CNN neural network to obtain image features; CNN (convolutional neural network) is a deep learning model widely used in the field of image recognition. Important features of the image are automatically extracted by the convolution layer, including but not limited to edges, shapes, and textures. CNNs can learn deep abstract features from complex image data, providing powerful data support for flow rate recognition.
Fourthly, processing the training sample and the test sample through an LBP-TOP algorithm, and extracting time-frequency dynamic characteristics; the LBP-TOP (local binary pattern-spatiotemporal relation) algorithm is a method for extracting dynamic features of a video sequence, and considers the spatiotemporal relation in an image sequence, so that dynamically changing features, such as a motion pattern of an object, can be extracted from the video. In water flow rate identification, these dynamic features are critical to understanding the state of motion of the water flow.
Fifthly, carrying out feature fusion on the image features and the time-frequency dynamic features to obtain fusion features; feature fusion refers to combining features extracted from different methods (e.g., CNN and LBP-TOP) to form a more comprehensive representation of the features. The method aims to comprehensively utilize static image features and dynamic time-frequency features, so that the accuracy of the model in identifying the water flow velocity is improved.
Step six, constructing a nuclear random weight neural network model, and training through fusion characteristics to obtain a water flow velocity identification model; the Kernel random weighted neural network (Kernel-based Extreme Learning Machine, K-ELM for short) is suitable for dealing with complex nonlinear problems. Nonlinear data is processed by introducing core skills based on the concept of an Extreme Learning Machine (ELM). The model is trained by means of fused features, i.e. features extracted by the aforementioned CNN and LBP-TOP algorithms. During the training process, the model learns how to predict the flow rate of the water flow according to the input image characteristics. The kernel functions here play a role in mapping data into a high-dimensional space, so that the model can better process complex data structures and improve the accuracy of prediction.
And seventh, acquiring a water flow real-time video of the water area to be identified, processing the water flow real-time video to obtain fusion characteristics for identification, and inputting the fusion characteristics into a water flow velocity identification model to obtain the current water flow velocity of the water area. In practical application, the technical scheme firstly needs to capture real-time video of the water area to be identified. These video data undergo the same process flows, including NSST processing, CNN feature extraction, and LBP-TOP dynamic feature extraction, to generate fusion features for recognition. These fused features are then input into a pre-trained kernel stochastic weight neural network model. The model predicts the current water flow rate according to the input characteristics. This process can provide instant flow rate information to the user in real time.
Example two
The method for performing NSST processing on the sample picture in the second step comprises the following steps:
performing multi-scale decomposition on the sample picture through a non-downsampled pyramid filter bank to obtainLow frequency image and->A layer high frequency subband image; NSST (non-downsampled Shearlet transform) begins with multi-scale decomposition of an image. The non-downsampled pyramid filter bank is capable of decomposing the image into different scales without degrading the image resolution. Multiscale decomposition means that an image is decomposed into different levels of sub-images, each representing information of the original image at a different scale (or resolution). For example, a->The low frequency images capture the general outline and basic shape of the image, with details omitted.
In the multi-scale decomposition process, in addition to the low-frequency image, the low-frequency image is obtainedLayer high frequency subband images. The high frequency subband images contain detailed information of the image such as edges, textures, etc.
High frequency subband images by shear wave filtersStage multidirectional decomposition to obtain->A plurality of multidirectional subbands; the size of the multidirectional subband is the same as the size of the sample picture.
The high frequency subband image of each layer is further decomposed using a shear wave Filter (Shearlet Filter) to extract direction information of the image. In this context,the level multi-directional decomposition means that each layer of high frequency subband image is further subdivided into subbands of different directions.
The end result is a decomposition of each layer of high frequency subband images intoSub-bands in each direction. These multidirectional subbands are capable of capturing features of the image in different directions, such as the flow direction of the water flow, the direction of the waves, etc. Notably, the size of these subbands remains the same as the original sample picture, ensuring the integrity of the information and preservation of detail.
Through NSST processing, the sample picture is converted into a series of sub-images containing rich space and direction information, which provides a detailed and comprehensive data basis for subsequent image feature extraction and analysis. The processing method is particularly suitable for complex and detail-rich images, such as water flow images, and can effectively improve the accuracy and reliability of the follow-up flow speed recognition model.
In the fourth step, the method for extracting the time-frequency dynamic characteristics comprises the following steps:
orthogonalization segmentation of sequences of video into spatial and temporal relationshipsPlane, & gt>Plane, & gt>A plane; the video sequence is converted into three different planes for analysis. />The plane represents the traditional spatial relationship, namely the spatial distribution of the pixel points in a fixed moment; />Plane sum->Plane rule meterShowing time relationship, i.e. change with time along the horizontal or vertical direction. Such segmentation helps to capture the dynamics of the water flow from different dimensions.
For a pair ofPlane, & gt>Plane, & gt>The plane is respectively provided with a radius and a field point; certain radius and field points are selected on each plane for determining the characteristics of the local area.
Determining a central pixel, obtaining LBP values of the central pixel on three planes, generating a characteristic vector of the central pixel on each plane, and normalizing the characteristic vector; the local binary pattern (Local Binary Patterns, LBP) is a texture descriptor that can generate a feature vector describing the local texture characteristics of a point by calculating the relative intensities of the center pixel and its neighborhood pixels on each plane. The LBP value of the center pixel on each plane constitutes a feature vector. And the influence of different scales or under the illumination condition is eliminated through standardization, so that the consistency of the data is ensured.
The within-class divergence matrix is calculated and,wherein->For category->Mean vector of>For category->Feature vector set of>Is the total number of categories; calculating an inter-class divergence matrix,>wherein->For the total mean vector of all feature vectors, +.>For category->Is a feature vector number of (a);
the intra-class and inter-class divergence matrices are used to measure the similarity inside the feature vector and the difference between classes. The intra-class divergence matrix describes the degree of dispersion of feature vectors within the same class, while the inter-class divergence matrix describes the discrimination of feature vectors between different classes.
Solving eigenvalues of inter-class divergence matrix and intra-class divergence matrixAnd feature vector->，/>Selecting the best discrimination feature vector corresponding to the largest feature value; by calculating->The feature value and feature vector of (2) can find the direction which can distinguish different categories.
And projecting the original feature vector to the optimal discrimination feature vector to obtain the time-frequency dynamic feature.
Example III
In the sixth step, the method for obtaining the water flow velocity identification model comprises the following steps:
constructing a random weight neural network model and defining a loss function; the random weight neural network is characterized in that the weights and the biases of the hidden layer nodes are randomly generated, and are not required to be adjusted in the training process. The loss function defines the difference between the model output and the actual label and is the key to optimizing the model. Typically, the loss function is intended to minimize prediction errors.
Calculating hidden layer output matrix by using double hidden layer self-coding random weight neural networkThe method comprises the steps of carrying out a first treatment on the surface of the A self-encoder is a special neural network that learns the effective representation (i.e., encoding) of data. The double hidden layer self-encoder in this embodiment is used to extract deep features of input data and output a representation of the hidden layer, i.e. an output matrix.
Obtaining a weight matrixWherein->For a matrix of target values (i.e. labels of training data),is a unitary matrix->Is a penalty parameter (for preventing overfitting); the weight matrix is a parameter learned by the model in the training process and is used for connecting the hidden layer and the output layer.
Selecting RBF kernel function as kernel function of kernel random weight neural network modelAnd introducing a kernel matrix into the random weight neural network model>Constructing a nuclear random weight neural network model, wherein,input +_for hidden layer pair>Output of->Input +_for hidden layer pair>An output of (2); RBF (radial basis function) kernels are a common kernel function used to map data into higher dimensional space to solve the problem of nonlinearities.
Updating a target value matrix based on a kernel function and a kernel matrixAnd weight matrix->The method comprises the steps of carrying out a first treatment on the surface of the The kernel function is introduced into the random weighted neural network model, and the performance of the model is optimized by updating the target value matrix and the weight matrix.
Wherein the target value matrix is updated based on a kernel functionThe formula of (2) is: />，Wherein->，/>，/>Width parameters that are kernel functions; the updating of the target value matrix T is based on a kernel function that is used to map the original input data to a higher dimensional space.
The method for updating the weight matrix comprises the steps of updating the target value matrixSubstituted into->Obtaining an updated weight matrix +.>The calculation process involves the calculation of an inverse matrix in such a way that the model can learn how to adjust its weights according to the input data X and the kernel-mapped features.
Determining hidden layer outputAnd updated weight matrix +.>Obtaining the output of the kernel random weight neural network modelI.e. to obtain a flow rate prediction of the model for a given input X.
The loss function in the above method is:wherein->For training error +.>Is Lagrangian multiplier +.>For training sample total number>For outputting the number of categories +.>Input +_for hidden layer pair>Output of->For samples in the target value matrix->In category->Is (are) true tags->For sample->In category->Training error of->Is->Weight value of category corresponding in weight matrix, < ->For sample->In category->Lagrangian multiplier of (c).
By minimizing the loss function, the kernel stochastic weight neural network is able to learn more accurately the mapping from input data to output predictions (e.g., water flow rate).
The calculation method of the hidden layer output matrix comprises the following steps:
the first hidden layer and the second hidden layer of the random weight neural network model are replaced by constructing a double hidden layer self-coding random weight neural network; the double hidden layer self-coding network is used for extracting deep features of input data. Self-encoders are an unsupervised learning model, typically used for dimension reduction or feature learning of data. In this embodiment, the first two hidden layers used to replace the traditional pre-training model are used to more effectively capture the intrinsic structure and pattern of the input data.
Randomly generating an input weight vector for a first hidden layer nodeAnd bias->The method comprises the steps of carrying out a first treatment on the surface of the Randomly generating an input weight vector for a second hidden layer node>And bias->The method comprises the steps of carrying out a first treatment on the surface of the For converting input data in hidden layers.
Input device、/>、/>Calculating an output matrix of the first hidden layer; output matrix through first hidden layer, +.>、/>Calculating a second hidden layer output matrix +.>The method comprises the steps of carrying out a first treatment on the surface of the An output matrix for each hidden layer is calculated. The output matrix of the first hidden layer is the result of the input data after the weight vector and the bias treatment, and the output of the second hidden layerThe matrix is the result of further processing of the first hidden layer output.
Will be from encoderCalculating +.>Calculating an output weight matrix of the self-encoder>Wherein->The number of layers for the self-encoder; repeating at each hidden layer until the number of layers of the self-encoder is reached +.>。
Calculating an output matrix of each hidden node based on the input data and the output weight matrix of the self-encoder。
And taking the output matrix of the last hidden layer as the output matrix of the hidden layer. Output matrix of last hidden layerIs used as an hidden layer output matrix for the entire network. This output matrix contains deep features representative of the input data.
Example IV
A water flow rate identification system based on image identification, comprising:
unmanned aerial vehicle and flight control system thereof; the unmanned aerial vehicle is used for obtaining real-time video data of rivers. And is equipped with a flight control system to ensure that the unmanned aerial vehicle can fly stably and reach a designated water area for shooting.
The high-definition video shooting device is arranged on the unmanned plane; and the high-definition video is used for capturing water flow. High definition video capture devices are critical to obtaining sufficiently clear, detailed images of water flow because the quality of the image directly affects the accuracy of subsequent image processing and flow rate identification.
The wireless communication assembly is arranged on the unmanned aerial vehicle, and the high-definition video shooting device is communicated with the main control computer through the wireless communication assembly; a wireless communication component installed on the unmanned aerial vehicle allows the high-definition video shooting device to transmit captured video data to the main control computer in real time. The wireless transmission mechanism enables data processing to be performed while the unmanned aerial vehicle flies, and improves the working efficiency of the system.
The main control computer comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the water flow rate identification method based on image identification when executing the computer program.
The main control computer is the core of the system and comprises a memory, a processor and a preloaded computer program. This computer program contains the previously described image recognition based water flow rate recognition method. When the processor executes this program, it will process the video data received from the drone, perform steps such as NSST processing, CNN feature extraction, LBP-TOP algorithm processing, etc., and finally predict the flow rate of the water flow through the constructed kernel stochastic weight neural network model.
The memory may be used to store software programs and modules, and the processor executes various functional applications of the terminal and data processing by running the software programs and modules stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an execution program required for at least one function, and the like.
The storage data area may store data created according to the use of the terminal, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.
A computer readable storage medium storing a computer program which when executed by a processor implements a water flow rate identification method based on image identification as above.
Computer readable media may include computer storage media and communication media without loss of generality. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instruction data structures, program modules or other data. Computer storage media includes RAM, ROM, EPROM, EEPROM, flash memory or other solid state memory technology, CD-ROM, DVD or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices. Of course, those skilled in the art will recognize that computer storage media are not limited to the ones described above. The above-described system memory and mass storage devices may be collectively referred to as memory.
A computer program product comprising computer programs/instructions which when executed by a processor implement a water flow rate identification method based on image identification as above.
The computer program product comprises a computer program or set of instructions for performing specific tasks or implementing specific functions. These programs or instructions are designed to be executed by a processor to implement a series of predefined steps or operations. The program product may be stored on various forms of computer storage media, such as memory, hard disk, solid state drive, optical disk, or other forms of digital storage devices. Either in the form of compiled binary code or in the form of scripts or bytecodes that can be executed by an interpreter. The program product enables the processor to process data in a specific order and manner through well-designed algorithms and logic instructions to perform various functions such as data analysis, user interaction, device control, etc.
In the description of the present specification, reference to the terms "one embodiment/manner," "some embodiments/manner," "example," "specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment/manner or example is included in at least one embodiment/manner or example of the present application. In this specification, the schematic representations of the above terms are not necessarily for the same embodiment/manner or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments/modes or examples. Furthermore, the various embodiments/modes or examples described in this specification and the features of the various embodiments/modes or examples can be combined and combined by persons skilled in the art without contradiction.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present application, the meaning of "plurality" is at least two, such as two, three, etc., unless explicitly defined otherwise.
It will be appreciated by persons skilled in the art that the above embodiments are provided for clarity of illustration only and are not intended to limit the scope of the invention. Other variations or modifications of the above-described invention will be apparent to those of skill in the art, and are still within the scope of the invention.
Claims (10)
1. A water flow rate identification method based on image identification, comprising:
acquiring video samples of different water qualities at different flow rates, and taking frame images in the video samples as sample pictures;
NSST processing is carried out on the sample picture, and the sample picture is divided into a training sample and a test sample;
performing feature extraction on the training sample and the test sample through a CNN neural network to obtain image features;
processing the training sample and the test sample by using an LBP-TOP algorithm, and extracting time-frequency dynamic characteristics;
carrying out feature fusion on the image features and the time-frequency dynamic features to obtain fusion features;
constructing a nuclear random weight neural network model, and training through fusion characteristics to obtain a water flow velocity identification model;
and acquiring a water flow real-time video of the water area to be identified, processing the water flow real-time video to acquire fusion characteristics for identification, and inputting the fusion characteristics into a water flow velocity identification model to acquire the current water flow velocity of the water area.
2. The method for identifying the flow rate of water based on image identification according to claim 1, wherein the method for performing NSST processing on the sample picture comprises the following steps:
performing multi-scale decomposition on the sample picture through a non-downsampled pyramid filter bank to obtainLow frequency image and->A layer high frequency subband image;
high frequency subband images by shear wave filtersStage multidirectional decomposition to obtain->A plurality of multidirectional subbands; the size of the multidirectional subband is the same as the size of the sample picture.
3. The method for identifying the flow velocity of water based on image identification according to claim 2, wherein the method for extracting the time-frequency dynamic characteristics comprises the following steps:
orthogonalization segmentation of sequences of video into spatial and temporal relationshipsPlane, & gt>Plane, & gt>A plane;
for a pair ofPlane, & gt>Plane, & gt>The plane is respectively provided with a radius and a field point;
determining a central pixel, obtaining LBP values of the central pixel on three planes, generating a characteristic vector of the central pixel on each plane, and normalizing the characteristic vector;
the within-class divergence matrix is calculated and,wherein->For category->Mean vector of>For category->Feature vector set of>Is the total number of categories;
the inter-class divergence matrix is calculated and,wherein->For the total mean vector of all feature vectors, +.>For category->Is a feature vector number of (a);
solving eigenvalues of inter-class divergence matrix and intra-class divergence matrixAnd feature vector->，/>Selecting the best discrimination feature vector corresponding to the largest feature value;
and projecting the original feature vector to the optimal discrimination feature vector to obtain the time-frequency dynamic feature.
4. The method for recognizing a flow rate of water based on image recognition according to claim 2, wherein the method for obtaining the recognition model of the flow rate of water comprises:
constructing a random weight neural network model and defining a loss function;
calculating hidden layer output matrix by using double hidden layer self-coding random weight neural network；
Obtaining a weight matrixWherein->For the target value matrix +.>Is a unitary matrix->Is a punishment parameter;
selecting RBF kernel function as kernel function of kernel random weight neural network modelAnd introducing a kernel matrix into the random weight neural network model>Constructing a kernel random weight neural network model, wherein +_>Input +_for hidden layer pair>Output of->Input +_for hidden layer pair>An output of (2);
updating a target value matrix based on a kernel function and a kernel matrixAnd weight matrix->；
Determining hidden layer outputAnd updated weight matrix +.>Obtaining the output of the kernel random weight neural network model。
5. The method for identifying a water flow rate based on image recognition according to claim 4, wherein the loss function is:wherein->For training error +.>Is Lagrangian multiplier +.>For training sample total number>For outputting the number of categories +.>Input +_for hidden layer pair>Output of->For samples in the target value matrix->In category->Is (are) true tags->For sample->In category->Training error of->Is->Weight value of category corresponding in weight matrix, < ->For sample->In category->Lagrangian multiplier of (c).
6. The method for identifying water flow rate based on image recognition according to claim 4, wherein the calculating method of the hidden layer output matrix comprises the following steps:
the first hidden layer and the second hidden layer of the random weight neural network model are replaced by constructing a double hidden layer self-coding random weight neural network;
randomly generating an input weight vector for a first hidden layer nodeAnd bias->The method comprises the steps of carrying out a first treatment on the surface of the Randomly generating an input weight vector for a second hidden layer node>And bias->；
Input device、/>、/>Calculating an output matrix of the first hidden layer;
through the output matrix of the first hidden layer,、/>Calculating a second hidden layer output matrix +.>；
Will be from encoderCalculating +.>Calculating an output weight matrix of the self-encoder>Wherein->The number of layers for the self-encoder;
calculating an output matrix for each hidden node；
And taking the output matrix of the last hidden layer as the output matrix of the hidden layer.
7. The method for identifying a water flow rate based on image recognition according to claim 4, wherein the target value matrix is updated based on a kernel functionThe formula of (2) is: />，/>Wherein->，/>，/>Width parameters that are kernel functions;
the method for updating the weight matrix comprises the steps of updating the target value matrixSubstituted into->Obtaining an updated weight matrix +.>。
8. A water flow rate identification system based on image identification, comprising:
unmanned aerial vehicle and flight control system thereof;
the high-definition video shooting device is installed on the unmanned aerial vehicle;
the wireless communication assembly is arranged on the unmanned aerial vehicle, and the high-definition video shooting device is communicated with the main control computer through the wireless communication assembly;
the main control computer comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the water flow rate identification method based on image identification according to any one of claims 1-7 when executing the computer program.
9. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements a method for identifying a water flow rate based on image identification according to any one of claims 1 to 7.
10. A computer program product comprising computer programs/instructions which, when executed by a processor, implement a method of identifying water flow rate based on image recognition as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410225742.8A CN117809230A (en) | 2024-02-29 | 2024-02-29 | Water flow velocity identification method based on image identification and related products |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410225742.8A CN117809230A (en) | 2024-02-29 | 2024-02-29 | Water flow velocity identification method based on image identification and related products |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117809230A true CN117809230A (en) | 2024-04-02 |
Family
ID=90434904
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410225742.8A Pending CN117809230A (en) | 2024-02-29 | 2024-02-29 | Water flow velocity identification method based on image identification and related products |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117809230A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106156734A (en) * | 2016-06-28 | 2016-11-23 | 浙江工业大学 | A kind of current speed-measuring method based on convolutional neural networks image recognition |
| CN109784206A (en) * | 2018-12-25 | 2019-05-21 | 杭州电子科技大学 | A kind of gait recognition method based on the learning machine that transfinites |
| CN110287975A (en) * | 2019-06-28 | 2019-09-27 | 福州大学 | Flotation dosing abnormity detection method based on NSST morphological characteristics and depth KELM |
| CN110969086A (en) * | 2019-10-31 | 2020-04-07 | 福州大学 | Handwritten image recognition method based on multi-scale CNN (CNN) features and quantum flora optimization KELM |
| US20210090270A1 (en) * | 2018-11-16 | 2021-03-25 | Central South University | Method and system for detecting flow velocity of high-temperature molten fluid |
| CN113869234A (en) * | 2021-09-29 | 2021-12-31 | 中国平安财产保险股份有限公司 | Facial expression recognition method, device, equipment and storage medium |
| CN116503378A (en) * | 2023-05-16 | 2023-07-28 | 福州大学 | Concentrate grade prediction method for bimodal CNN secondary transfer learning |
-
2024
- 2024-02-29 CN CN202410225742.8A patent/CN117809230A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106156734A (en) * | 2016-06-28 | 2016-11-23 | 浙江工业大学 | A kind of current speed-measuring method based on convolutional neural networks image recognition |
| US20210090270A1 (en) * | 2018-11-16 | 2021-03-25 | Central South University | Method and system for detecting flow velocity of high-temperature molten fluid |
| CN109784206A (en) * | 2018-12-25 | 2019-05-21 | 杭州电子科技大学 | A kind of gait recognition method based on the learning machine that transfinites |
| CN110287975A (en) * | 2019-06-28 | 2019-09-27 | 福州大学 | Flotation dosing abnormity detection method based on NSST morphological characteristics and depth KELM |
| CN110969086A (en) * | 2019-10-31 | 2020-04-07 | 福州大学 | Handwritten image recognition method based on multi-scale CNN (CNN) features and quantum flora optimization KELM |
| CN113869234A (en) * | 2021-09-29 | 2021-12-31 | 中国平安财产保险股份有限公司 | Facial expression recognition method, device, equipment and storage medium |
| CN116503378A (en) * | 2023-05-16 | 2023-07-28 | 福州大学 | Concentrate grade prediction method for bimodal CNN secondary transfer learning |
Non-Patent Citations (6)
| Title |
|---|
| GUANG-BIN HUANG: "An Insight into Extreme Learning Machines: Random Neurons, Random Features and Kernels", 《COGNITIVE COMPUTATION》, vol. 06, no. 03, 3 April 2014 (2014-04-03), pages 376 * |
| XIAOPING JIANG等: "Research on Multi-Scale Feature Extraction and Working Condition Classification Algorithm of Lead-Zinc Ore Flotation Foam", 《APPLIED SCIENCES》, vol. 13, no. 06, 22 March 2023 (2023-03-22), pages 1 - 16 * |
| 廖一鹏等: "结合双模多尺度CNN特征及自适应深度 KELM的浮选工况识别", 《光学精密工程》, vol. 28, no. 18, 14 August 2020 (2020-08-14), pages 1785 - 1798 * |
| 张延良等: "基于Fisher 特征选择的微表情识别", 《计算机应用与软件》, vol. 37, no. 09, 10 September 2020 (2020-09-10), pages 68 - 74 * |
| 蒋小平等: "基于多尺度卷积神经网络和LBP算法的浮选工况识别", 《矿业科学学报》, vol. 08, no. 02, 26 October 2022 (2022-10-26), pages 202 - 212 * |
| 黄河等: "视觉测流技术在中小河流洪水流量监测中的应用研究", 《长江技术经济》, vol. 07, no. 04, 15 August 2023 (2023-08-15), pages 112 - 117 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110135267B (en) | Large-scene SAR image fine target detection method | |
| Springenberg et al. | Improving deep neural networks with probabilistic maxout units | |
| CN111507378A (en) | Method and apparatus for training image processing model | |
| CN114202672A (en) | Small target detection method based on attention mechanism | |
| Pally et al. | Application of image processing and convolutional neural networks for flood image classification and semantic segmentation | |
| CN114022759A (en) | Airspace finite pixel target detection system and method fusing neural network space-time characteristics | |
| CN114676777A (en) | Self-supervision learning fine-grained image classification method based on twin network | |
| Demertzis et al. | Geo-AI to aid disaster response by memory-augmented deep reservoir computing | |
| CN111797970A (en) | Method and apparatus for training neural network | |
| Alsanad et al. | Real-time fuel truck detection algorithm based on deep convolutional neural network | |
| Deepthi et al. | Detection and classification of objects in satellite images using custom CNN | |
| CN113989574B (en) | Image interpretation method, image interpretation device, electronic device, and storage medium | |
| Deepan et al. | Remote sensing image scene classification using dilated convolutional neural networks | |
| CN114332075A (en) | Rapid structural defect identification and classification method based on lightweight deep learning model | |
| CN111914949B (en) | Zero sample learning model training method and device based on reinforcement learning | |
| CN116665114A (en) | Multi-mode-based remote sensing scene identification method, system and medium | |
| CN115965968A (en) | Small sample target detection and identification method based on knowledge guidance | |
| Deepan et al. | Comparative analysis of scene classification methods for remotely sensed images using various convolutional neural network | |
| Jiang et al. | A robust end-to-end deep learning framework for detecting Martian landforms with arbitrary orientations | |
| CN117809230A (en) | Water flow velocity identification method based on image identification and related products | |
| Wang et al. | Sonar image detection based on multi-scale multi-column convolution neural networks | |
| Royer et al. | Localizing grouped instances for efficient detection in low-resource scenarios | |
| CN117690164B (en) | Airport bird identification and driving method and system based on edge calculation | |
| US20230196541A1 (en) | Defect detection using neural networks based on biological connectivity | |
| CN116030347B (en) | High-resolution remote sensing image building extraction method based on attention network |
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 |