CN114648738A - Image identification system and method based on Internet of things and edge calculation - Google Patents
Image identification system and method based on Internet of things and edge calculation Download PDFInfo
- Publication number
- CN114648738A CN114648738A CN202210149994.8A CN202210149994A CN114648738A CN 114648738 A CN114648738 A CN 114648738A CN 202210149994 A CN202210149994 A CN 202210149994A CN 114648738 A CN114648738 A CN 114648738A
- Authority
- CN
- China
- Prior art keywords
- unit
- image
- special work
- type operation
- work type
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of image recognition, in particular to an image recognition system and method based on the Internet of things and edge calculation, wherein the method comprises the following steps: the method comprises the steps that an image acquisition unit is used for acquiring images of special work type operation vehicles in a monitoring area, the images of the special work type operation vehicles are preprocessed, the preprocessed images of the special work type operation vehicles are transmitted to an edge computing node, and computing parameters of the images of the special work type operation vehicles are obtained after edge computing, wherein the image acquisition unit and the edge computing node are provided with a plurality of images which correspond to each other; and finishing training of the image recognition network model through calculating parameters, applying the training to each edge computing node after the training of the image recognition network model is finished, evaluating the recognition result of the image of the special work type operation vehicle by each edge computing node through a cooperation evaluation method, and voting according to the evaluation result to judge whether the image is the special work type operation vehicle.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to an image recognition system and method based on the Internet of things and edge calculation.
Background
Image recognition, which refers to a technique for processing, analyzing and understanding images by a computer to recognize various different patterns of objects and objects, is a practical application of applying a deep learning algorithm. In recent years, with the rapid development of national road traffic, a plurality of disadvantages are generated, wherein more importantly, the management and control of special work vehicles in the past are not strict enough, the special work vehicles often appear in non-designated places, the special work vehicles are large in size, heavy and dangerous, and when the vehicles run on a bridge road, because the support weight of the bridge road has a limit, once the support weight of the bridge road exceeds a set value, a safety accident that the bridge collapses is very easy to occur, and personnel injuries and deaths are caused.
However, in terms of the present, different work vehicles of different specific types are usually parked in different monitoring and management areas, a large number of cameras are required for monitoring, with the increase of monitoring cameras and the dispersion of monitoring areas, image data uploaded to an image recognition server increases greatly, so that resources of the image recognition server are occupied greatly, and a network congestion occurs.
Disclosure of Invention
The invention aims to provide an image recognition system and method based on the internet of things and edge calculation so as to solve the problems.
In order to achieve the above object, the embodiments of the present application provide the following technical solutions:
in one aspect, an embodiment of the present application provides an image recognition system based on internet of things and edge calculation, including:
an image acquisition unit: the system is used for acquiring images of the special work type operation vehicle, wherein the special work type operation vehicle is positioned in a monitoring area;
a pretreatment unit: the image preprocessing module is used for preprocessing the image of the special work type operation vehicle;
a signal transmission unit: the image preprocessing module is used for carrying out signal transmission on the image of the special work type operation vehicle after preprocessing;
a feature extraction unit: the system is used for extracting the features of the image of the special work type operation vehicle after the preprocessing is finished, wherein the feature extraction is specifically carried out by edge calculation;
an image recognition unit: the system is used for identifying the image characteristics of the special work type operation vehicle;
the image acquisition unit is connected with the preprocessing unit, the preprocessing unit is connected with the signal transmission unit, the signal transmission unit is connected with the feature extraction unit, and the feature extraction unit is connected with the image identification unit.
Preferably, the image acquisition unit may be specifically applied as a binocular camera, and the binocular camera is provided in plurality and is respectively deployed around the monitoring area.
Preferably, the pretreatment unit includes: the device comprises a denoising unit, a gray level stretching unit, a histogram equalization unit, a brightness adjusting unit, a transformation unit and a binarization unit;
the denoising unit: the noise elimination device is used for carrying out noise elimination on the image of the special work vehicle through a noise elimination algorithm, wherein the noise comprises Gaussian noise, speckle noise and salt and pepper noise;
the calculation formula of the denoising algorithm is as follows:
wherein the content of the first and second substances,
in order to complete the de-noised image of the special work type operation vehicle, P (X, Y) is the image of the special work type operation vehicle polluted by noise, (X, Y) is a certain pixel point of the image of the special work type operation vehicle polluted by noise, (X, Y) is a certain pixel point of the image of the special work type operation vehicle polluted by noise, a X b is a template in a 4 neighborhood or 8 neighborhood form, and Q is the area of the template;
the gray scale stretching unit: the gray scale stretching device is used for carrying out gray scale stretching on the image of the special work type operation vehicle;
the histogram equalization unit: the system is used for carrying out equalization processing on pixel values and gray levels of images of the special work type operation vehicle;
the brightness adjusting unit: the system is used for adjusting the brightness of the image of the special work type operation vehicle;
the transformation unit: the system is used for converting the images of the special work type operation vehicle from color images to gray images;
the binarization unit is used for: the method is used for carrying out binarization processing on the image of the special work vehicle, wherein the binarization processing adopts a binarization algorithm, and a calculation formula is as follows:
wherein (x, y) is pixel, f (x, y) is original attribute of pixel, G (x, y) is attribute after pixel change, sigma is threshold point, G0And G1And respectively completing the binaryzation colors of the image to be detected.
Preferably, the signal transmission unit is applied to the internet of things, and performs signal transmission on the image of the special work type operation vehicle after the preprocessing through the internet of things.
Preferably, the feature extraction unit is applied to an edge calculation node unit, the edge calculation node unit corresponds to the image acquisition unit and is used for performing edge calculation on the image of the special work type work vehicle acquired by the corresponding image acquisition unit to complete feature extraction on the image of the special work type work vehicle, and the edge calculation node unit comprises a node information acquisition subunit, a node analysis subunit and a node calculation unit;
the node information acquisition subunit: the system comprises a node unit, a node unit and a node unit, wherein the node unit is used for acquiring state information of an edge computing node unit;
the node analysis subunit: the system is used for receiving and analyzing the image of the special work type operation vehicle and controlling the node computing unit to compute;
the node calculation unit: the node calculation unit is used for calculating the image of the special work type operation vehicle, wherein an image optimization algorithm and an image feature extraction algorithm are preset in the node calculation unit.
Preferably, the image recognition unit comprises an acquisition unit and a recognition unit;
the acquisition unit: the training set is used for acquiring the calculation parameters of the node calculation unit and generating a training set corresponding to the edge calculation node unit;
the identification unit: the method is used for completing the identification of the image of the special work type operation vehicle through the image identification network model.
Preferably, the analysis unit comprises a model construction unit, a model optimization unit and a model training unit;
the model building unit: the method is used for constructing an image recognition network model, wherein the image recognition network model adopts a convolution network model, a loss function of the convolution network model adopts a cross entropy loss function, and a calculation formula is as follows:
where s is the sample, n is the total number of samples, H is the predicted output,
is the actual output;
the model optimization unit: the method is used for optimizing the image recognition network model, wherein a Dropout algorithm is preset in a model optimization unit, and a calculation formula of the Dropout algorithm is as follows:
L＝r×f(ωm+n)
rj～Bernoulli(p)
where f is the activation function, m is the input, r is the binary mask matrix, rjThe probability vector is a value of 0 or 1;
and the model training unit is used for finishing the training of the image recognition network model through a training set.
In one aspect, an embodiment of the present application provides an image identification method based on an internet of things and edge calculation, including the following:
the method comprises the steps that an image acquisition unit is used for acquiring images of special work type operation vehicles in a monitoring area, the images of the special work type operation vehicles are preprocessed, the preprocessed images of the special work type operation vehicles are transmitted to an edge computing node, and computing parameters of the images of the special work type operation vehicles are obtained after edge computing, wherein the preprocessing comprises denoising, gray level stretching, histogram equalization, brightness adjustment, color-to-gray conversion and binarization, the image acquisition unit is provided with a plurality of image acquisition units, and the edge computing node is provided with a plurality of edge computing nodes and corresponds to each image acquisition unit;
acquiring calculation parameters of images of the special work vehicle, forming a training set, constructing an image recognition network model, training the image recognition network model through the training set to obtain the trained image recognition network model, wherein the image recognition network model is applied to each edge calculation node after the training of the image recognition network model is completed, each edge calculation node evaluates the recognition result of the images of the special work vehicle through a collaborative evaluation method, and votes according to the evaluation result to judge whether the images are the special work vehicle, the image recognition network model is optimized by adopting a Dropout algorithm, and the collaborative evaluation method specifically comprises the following steps:
the total number of edge calculation nodes is set to S, the determination result is Puo, the value of Puo is 0 or 1, and the evaluation result is that P is the maximum value
Wherein o is 1 or 2.
In one aspect, an embodiment of the present application provides an electronic device, including:
a memory for storing a computer program;
a processor, configured to execute the computer program, and implement the steps of the image identification method based on the internet of things and edge calculation.
Finally, an embodiment of the present application provides a computer storage medium, where a computer program is stored on the readable storage medium, and when the computer program is executed by a processor, the steps of the image recognition method based on the internet of things and edge calculation are implemented as described above.
The invention has the beneficial effects that:
according to the invention, as the special work vehicle is large in size and shape, a plurality of cameras are required to be deployed to collect the special work vehicle, and collected data can be transmitted to edge nodes corresponding to the cameras through the Internet of things to be calculated, so that network congestion is greatly reduced. The efficiency of image recognition on the special work type operation vehicle is improved;
moreover, in the invention, the images acquired by the cameras are inevitably influenced by reasons such as noise, brightness and the like in the acquisition process, so that the images are unclear, and therefore, the images of the special work type operation vehicle are preprocessed in the modes of denoising, gray level stretching, histogram equalization, brightness adjustment, color-to-gray conversion, binarization and the like, the efficiency of subsequently recognizing the images of the special work type operation vehicle is greatly improved, and the failure of recognition caused by the fuzzy images of the special work type operation vehicle is further avoided;
in addition, the shapes and sizes of different special work vehicle types are different, and the same type of special work vehicle is usually positioned in the same monitoring area, so that a self-attention mechanism can be inserted into the convolution network model, and the special work vehicle corresponding to the area is mainly used for judging whether the image of the special work vehicle exists in the image of the special work vehicle or not, so that the image identification precision of the special work vehicle is improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a schematic diagram of an image recognition system based on internet of things and edge calculation according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a Dropout algorithm according to an embodiment of the present invention;
fig. 3 is a schematic diagram of an image identification method based on internet of things and edge calculation in the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers or letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined or explained in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.
Example 1
As shown in fig. 1, the embodiment provides an image recognition system based on internet of things and edge calculation, and the system includes:
an image acquisition unit: the system is used for acquiring images of the special work type operation vehicle, wherein the special work type operation vehicle is positioned in a monitoring area;
a pretreatment unit: the image preprocessing module is used for preprocessing the image of the special work type operation vehicle;
a signal transmission unit: the image preprocessing module is used for carrying out signal transmission on the image of the special work type operation vehicle after preprocessing;
a feature extraction unit: the system is used for extracting the features of the image of the special work type operation vehicle after the preprocessing is finished, wherein the feature extraction is specifically carried out by edge calculation;
an image recognition unit: the system is used for identifying the image characteristics of the special work type operation vehicle;
the image acquisition unit is connected with the preprocessing unit, the preprocessing unit is connected with the signal transmission unit, the signal transmission unit is connected with the feature extraction unit, and the feature extraction unit is connected with the image identification unit.
In a specific implementation, the image capturing unit may be specifically applied as a binocular camera, and the binocular cameras are provided in plurality and are respectively deployed around the monitoring area.
In a specific implementation, the pre-processing unit comprises: the device comprises a denoising unit, a gray level stretching unit, a histogram equalization unit, a brightness adjusting unit, a transformation unit and a binarization unit;
the denoising unit: the noise elimination device is used for carrying out noise elimination on the image of the special work vehicle through a noise elimination algorithm, wherein the noise comprises Gaussian noise, speckle noise and salt and pepper noise;
specifically, gaussian noise is a type of noise whose probability density function follows gaussian distribution, and is random noise, and is characterized in that no matter what the histogram statistical distribution graph of the original image is, how large the probability density of noise occurrence is, the statistical distribution graph of the noise intensity follows normal distribution, the image has both strong noise pixels with large pollution amplitude and weak noise pixels with small pollution amplitude, the noise mainly comes from electronic circuit noise and sensor noise caused by low illuminance or high temperature, and the probability density function is as follows:
where z is the value of an image pixel, μ is the expectation for z, and σ is the standard deviation for z.
Specifically, the speckle noise not only reduces the image quality of the image, but also seriously affects the automatic segmentation, classification, target detection and other quantitative thematic information extraction of the image.
Specifically, the salt and pepper noise is two types of noise, namely salt noise and pepper noise. Wherein the salt noise is represented as white and belongs to high-gray noise; the pepper noise is black and belongs to low-gray noise. And typically both of these noises appear simultaneously, i.e., as black and white dots in the image. Salt and pepper noise is the bright and dark point noise between black and white generated by an image sensor, a transmission channel, decoding processing and the like, the noise is often caused by image cutting, and the most common algorithm for removing pulse interference and salt and pepper noise is median filtering.
The calculation formula of the denoising algorithm is as follows:
wherein the content of the first and second substances,
for the image of the special work kind operation vehicle with noise removed, P (X, Y) is the image of the special work kind operation vehicle polluted by noise, (X, Y) is a certain pixel point of the image of the special work kind operation vehicle with noise removed, (X, Y) is a certain pixel point of the image of the special work kind operation vehicle polluted by noise, and ab is a template in a 4-neighborhood or 8-neighborhood form, and Q is the area of the template;
the gray scale stretching unit: the gray scale stretching device is used for carrying out gray scale stretching on the image of the special work type operation vehicle;
specifically, the gray stretching adopts piecewise linear transformation, and the specific transformation function expression is as follows:
wherein f is1、G1、f2、G2The gray stretching is used to expand a designated gray range for coordinates of two control points on a gray stretching transformation function graph to improve image quality.
The histogram equalization unit: the system is used for carrying out equalization processing on pixel values and gray levels of images of the special work type operation vehicle;
histogram equalization is the redistribution of pixel values of an image over gray levels such that an input image is converted to an output image having approximately the same number of pixel points per gray level, with the transformation formula:
wherein N isiThe number of pixels is i.
The brightness adjusting unit: the system is used for adjusting the brightness of the image of the special work type operation vehicle;
the brightness adjustment of the image is to increase or decrease each pixel value of the image by a certain amount to obtain a new gray value, thereby realizing the change of the whole image, and the mathematical expression is as follows:
Y(x)＝hx+d
wherein when k is 1, the brightness of the image is adjusted by changing the value of d,
the transformation unit: the system is used for converting the color image into the gray image of the special work type operation vehicle;
the binarization unit is used for: is used for carrying out binarization processing on the image of the special work vehicle, wherein, the binarization processing adopts a binarization algorithm,
the image binarization is to set the gray scale of a point on the image to 0 or 255, i.e. the whole image shows obvious black and white effect. When the image tends to have a black-and-white effect or the image has substantially only two colors of the background and the target image, a threshold point can be determined as a boundary point of the two colors in 256 brightness levels, and the image is subjected to binarization processing.
The image binarization can eliminate blur, reduce complexity and reduce image color level, after the image binarization, the gray level of the image is two, the layering sense and the reality sense of the image are damaged, but the required image information is highlighted, the method plays an important role in subsequent image identification, and the calculation formula is as follows:
wherein (x, y) is pixel, f (x, y) is original attribute of pixel, G (x, y) is attribute after pixel change, sigma is threshold point, G0And G1And respectively completing the binaryzation colors of the image to be detected.
After the image of the special work type operation vehicle is preprocessed, the image of the preprocessed special work type operation vehicle can be subjected to definition evaluation through a Tenengrad function, and if the image of the special work type operation vehicle reaches a set value, subsequent transmission work is executed.
In specific implementation, the signal transmission unit is applied to the Internet of things, and the images of the special work type operation vehicle after pretreatment are subjected to signal transmission through the Internet of things.
The definition of the internet of things is as follows: any article is connected with the Internet according to an agreed protocol through information sensing equipment such as radio frequency identification, an infrared sensor, a global positioning system and the like to carry out information exchange and communication, so that a network for intelligent identification, positioning and management is realized.
In the embodiment of the document, because the size and the shape of the special work vehicle are large, a plurality of cameras need to be deployed to collect the special work vehicle, and collected data can be transmitted to edge nodes corresponding to the cameras through the internet of things to be calculated, so that network congestion is greatly reduced. The efficiency of follow-up image recognition to special work kind operation car is improved.
In addition, in the invention, the images acquired by the cameras are inevitably influenced by noise, brightness and other reasons in the acquisition process, so that the images are unclear, and therefore, the images of the special work type operation vehicle are preprocessed in the modes of denoising, gray scale stretching, histogram equalization, brightness adjustment, color-to-gray conversion, binarization and the like, the efficiency of subsequently recognizing the images of the special work type operation vehicle is greatly improved, and the recognition failure caused by the fuzzy images of the special work type operation vehicle is further avoided.
It is conceivable that a safety accident may be caused when a special work vehicle is driven to a road by a lawless person or a worker out of a monitoring area due to a failure in identification.
In specific implementation, the feature extraction unit is applied to an edge calculation node unit, the edge calculation node unit corresponds to the image acquisition unit and is used for performing edge calculation on the image of the special work type work vehicle acquired by the corresponding image acquisition unit to complete feature extraction on the image of the special work type work vehicle, and the edge calculation node unit comprises a node information acquisition subunit, a node analysis subunit and a node calculation unit;
the calculation tasks among different edge nodes are mutually independent, more scales and richer features can be obtained by increasing the number of the edge nodes, the accuracy of image identification is improved, and the amount of image data transmitted in a network can be properly reduced if the number of the nodes is reduced, so that the transmission pressure of the data is relieved.
The node information acquisition subunit: the system comprises a node unit, a node unit and a node unit, wherein the node unit is used for acquiring state information of an edge computing node unit;
the node analysis subunit: the system is used for receiving and analyzing the image of the special work type operation vehicle and controlling the node computing unit to compute;
the node calculation unit: the node calculation unit is used for calculating the image of the special work type operation vehicle, wherein an image optimization algorithm and an image feature extraction algorithm are preset in the node calculation unit.
In a specific implementation, the image recognition unit comprises an acquisition unit and a recognition unit;
the acquisition unit: the training set is used for acquiring the calculation parameters of the node calculation unit and generating a training set corresponding to the edge calculation node unit;
the identification unit: the method is used for completing the identification of the image of the special work type operation vehicle through the image identification network model.
In a specific implementation, the analysis unit comprises a model construction unit, a model optimization unit and a model training unit;
the model building unit: the method is used for constructing an image recognition network model, wherein the image recognition network model adopts a convolution network model, a loss function of the convolution network model adopts a cross entropy loss function, and a calculation formula is as follows:
where s is the sample, n is the total number of samples, H is the predicted output,
is the actual output;
the model optimization unit: the method is used for optimizing the image recognition network model, wherein a Dropout algorithm is preset in the model optimization unit, as shown in fig. 2, the Dropout algorithm is a comparison graph between neurons adopting the Dropout algorithm, and the Dropout algorithm is a method for randomly removing part of neurons of a hidden layer and simultaneously removing all corresponding inputs and outputs in a training process. However, the neurons are only temporarily removed, and in the next training, the neurons with a fixed proportion are randomly removed from all the neurons, and in the training stage, the neural network becomes thin after dropout, so that the phenomenon that the neural network structure is too complex to cause overfitting can be avoided, and meanwhile, in the prediction process, the effect of taking the average value of the prediction result of the neural network after processing with smaller parameters is equivalent to the effect of taking the average value of the prediction result of the neural network after processing with smaller parameters, so that the prediction result is more accurate, and a specific calculation formula is as follows:
L＝r×f(ωm+n)
rj～Bernoulli(p)
where f is the activation function, m is the input, r is the binary mask matrix, rjThe probability vector is a value of 0 or 1;
and the model training unit is used for finishing training the image recognition network model through a training set.
In the invention, the related image recognition network model is specifically a convolution network model, wherein because the forms and sizes of different special work vehicle types are different, and the same type of special work vehicle type is usually located in the same monitoring area, a self-attention mechanism (which is to obtain the global geometric characteristics of an image and extract valuable information in the image by calculating the relationship between any two pixel points in the image) can be inserted into the convolution network model, and the special work vehicle type corresponding to the area is mainly determined whether the image of the special work vehicle type exists in the image so as to improve the accuracy of image recognition of the special work vehicle type.
Example 2
As shown in fig. 3, the embodiment provides an image recognition method based on internet of things and edge calculation, and the method includes:
the method comprises the steps that an image acquisition unit is used for acquiring images of special work type operation vehicles in a monitoring area, the images of the special work type operation vehicles are preprocessed, the preprocessed images of the special work type operation vehicles are transmitted to an edge computing node, and computing parameters of the images of the special work type operation vehicles are obtained after edge computing, wherein the preprocessing comprises denoising, gray level stretching, histogram equalization, brightness adjustment, color-to-gray conversion and binarization, the image acquisition unit is provided with a plurality of image acquisition units, and the edge computing node is provided with a plurality of edge computing nodes and corresponds to each image acquisition unit;
the method comprises the steps of obtaining calculation parameters of images of the special work type operation vehicle, forming a training set, constructing an image recognition network model, training the image recognition network model through the training set, and obtaining the trained image recognition network model, wherein after the training of the image recognition network model is completed, the image recognition network model is applied to each edge computing node, each edge computing node evaluates recognition results of the images of the special work type operation vehicle through a collaborative evaluation method, and judges whether the images of the special work type operation vehicle are the special work type operation vehicle or not according to evaluation result voting, the image recognition network model is optimized through a Dropout algorithm, and the collaborative evaluation method specifically comprises the following steps:
the total number of edge calculation nodes is set to S, the determination result is Puo, the value of Puo is 0 or 1, and the evaluation result is P-max
Wherein o is 1 or 2.
It should be noted that, regarding the method in the above embodiments, the specific system modules or units required to perform the corresponding operations in the method have been described in detail in the embodiments related to the system, and will not be described in detail here.
Example 3
Corresponding to the above method embodiment, the embodiment of the present disclosure further provides an electronic device, and the image identification method based on the internet of things and the edge calculation, which is described below and described above, may be referred to in correspondence with each other.
The processor is used for controlling the overall operation of the electronic equipment so as to complete all or part of the steps of the image identification method based on the internet of things and the edge calculation. The memory is used to store various types of data to support operation at the electronic device, and the data may include, for example, instructions for any application or method operating on the electronic device, as well as application-related data.
In this embodiment, a computer readable storage medium including program instructions is further provided, and the program instructions when executed by a processor implement the steps of the image recognition method based on the internet of things and edge calculation. For example, the computer readable storage medium may be the memory including program instructions executable by the processor of the electronic device to perform an image recognition method based on internet of things and edge calculation as described above.
The readable storage medium may be various readable storage media capable of storing program codes, such as a U disk, a removable hard disk, a read-only memory, a random access memory, a magnetic disk, or an optical disk.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. An image identification method based on the Internet of things and edge calculation is characterized by comprising the following steps:
the method comprises the steps that an image acquisition unit is used for acquiring images of special work type operation vehicles in a monitoring area, the images of the special work type operation vehicles are preprocessed, the preprocessed images of the special work type operation vehicles are transmitted to an edge computing node, and computing parameters of the images of the special work type operation vehicles are obtained after edge computing, wherein the preprocessing comprises denoising, gray level stretching, histogram equalization, brightness adjustment, color-to-gray conversion and binarization, the image acquisition unit is provided with a plurality of image acquisition units, and the edge computing node is provided with a plurality of edge computing nodes and corresponds to each image acquisition unit;
acquiring calculation parameters of images of the special work vehicle, forming a training set, constructing an image recognition network model, training the image recognition network model through the training set to obtain the trained image recognition network model, wherein the image recognition network model is applied to each edge calculation node after the training of the image recognition network model is completed, each edge calculation node evaluates the recognition result of the images of the special work vehicle through a collaborative evaluation method, and votes according to the evaluation result to judge whether the images are the special work vehicle, the image recognition network model is optimized by adopting a Dropout algorithm, and the collaborative evaluation method specifically comprises the following steps:
the total number of edge calculation nodes is set to be S, the judgment result is Puo, the value of Puo is 0 or 1, and the evaluation result is
Wherein o is 1 or 2.
2. An image recognition system based on internet of things and edge calculation is characterized by comprising:
an image acquisition unit: the system is used for acquiring images of the special work type operation vehicle, wherein the special work type operation vehicle is positioned in a monitoring area;
a pretreatment unit: the image preprocessing module is used for preprocessing the image of the special work type operation vehicle;
a signal transmission unit: the image preprocessing module is used for carrying out signal transmission on the image of the special work type operation vehicle after preprocessing;
a feature extraction unit: the system is used for extracting the features of the image of the special work type operation vehicle after the pretreatment, wherein the feature extraction is specifically carried out by edge calculation;
an image recognition unit: the system is used for identifying the image characteristics of the special work type operation vehicle;
the image acquisition unit is connected with the preprocessing unit, the preprocessing unit is connected with the signal transmission unit, the signal transmission unit is connected with the feature extraction unit, and the feature extraction unit is connected with the image identification unit.
3. The image recognition system based on the internet of things and the edge calculation as claimed in claim 1, wherein the image acquisition unit is specifically applicable to a binocular camera, and a plurality of binocular cameras are arranged and respectively deployed around the monitored area.
4. The image recognition system based on the internet of things and edge calculation as claimed in claim 1, wherein the preprocessing unit comprises: the device comprises a denoising unit, a gray scale stretching unit, a histogram equalization unit, a brightness adjusting unit, a transformation unit and a binarization unit;
the denoising unit: the noise elimination device is used for carrying out noise elimination on the image of the special work vehicle through a noise elimination algorithm, wherein the noise comprises Gaussian noise, speckle noise and salt and pepper noise;
the calculation formula of the denoising algorithm is as follows:
wherein, the first and the second end of the pipe are connected with each other,
in order to finish the image of the special work type operation vehicle which is denoised, P (X, Y) is the image of the special work type operation vehicle which is polluted by noise, (X, Y) is a certain pixel point of the image of the special work type operation vehicle which is denoised, (X, Y) is a certain pixel point of the image of the special work type operation vehicle which is polluted by noise, a X b is a template in a 4 neighborhood or 8 neighborhood form, and Q is the area of the template;
the gray scale stretching unit: the gray scale stretching device is used for carrying out gray scale stretching on the image of the special work type operation vehicle;
the histogram equalization unit: the system is used for carrying out equalization processing on pixel values and gray levels of images of the special work type operation vehicle;
the brightness adjusting unit: the system is used for adjusting the brightness of the image of the special work type operation vehicle;
the transformation unit: the system is used for converting the images of the special work type operation vehicle from color images to gray images;
the binarization unit is used for: the method is used for carrying out binarization processing on the image of the special work vehicle, wherein the binarization processing adopts a binarization algorithm, and a calculation formula is as follows:
wherein (x, y) is pixel, f (x, y) is original attribute of pixel, G (x, y) is attribute after pixel change, sigma is threshold point, G0And G1And respectively completing the binaryzation colors of the images to be detected.
5. The image recognition system based on the internet of things and the edge calculation as claimed in claim 1, wherein the signal transmission unit is applied to the internet of things and performs signal transmission on the image of the special work vehicle subjected to preprocessing through the internet of things.
6. The image recognition system based on the internet of things and edge calculation as claimed in claim 1, wherein the feature extraction unit is applied to an edge calculation node unit, the edge calculation node unit corresponds to the image acquisition unit and is used for performing edge calculation on the image of the special work type work vehicle acquired by the corresponding image acquisition unit to complete feature extraction on the image of the special work type work vehicle, and the edge calculation node unit comprises a node information acquisition subunit, a node analysis subunit and a node calculation unit;
the node information acquisition subunit: the system comprises a node unit, a node unit and a node unit, wherein the node unit is used for acquiring state information of an edge computing node unit;
the node analysis subunit: the system is used for receiving and analyzing the image of the special work type operation vehicle and controlling the node computing unit to compute;
the node calculation unit: the node calculation unit is used for calculating the image of the special work type operation vehicle, wherein an image optimization algorithm and an image feature extraction algorithm are preset in the node calculation unit.
7. The image recognition system based on the internet of things and the edge calculation as claimed in claim 6, wherein the image recognition unit comprises an acquisition unit and a recognition unit;
the acquisition unit: the training set is used for acquiring the calculation parameters of the node calculation unit and generating a training set corresponding to the edge calculation node unit;
the identification unit: the method is used for completing the identification of the image of the special work type operation vehicle through the image identification network model.
8. The image recognition system based on the internet of things and the edge calculation as claimed in claim 7, wherein the analysis unit comprises a model construction unit, a model optimization unit and a model training unit;
the model building unit: the method is used for constructing an image recognition network model, wherein the image recognition network model adopts a convolution network model, a loss function of the convolution network model adopts a cross entropy loss function, and a calculation formula is as follows:
where s is the sample, n is the total number of samples, H is the predicted output,
is the actual output;
the model optimization unit: the method is used for optimizing the image recognition network model, wherein a Dropout algorithm is preset in a model optimization unit, and a calculation formula of the Dropout algorithm is as follows:
L＝r×f(ωm+n)
rj～Bernoulli(p)
where f is the activation function, m is the input, r is the binary mask matrix, rjThe probability vector is a value of 0 or 1;
and the model training unit is used for finishing the training of the image recognition network model through a training set.
9. An electronic device, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the image recognition method based on internet of things and edge calculation as claimed in claim 1 when executing the computer program.
10. A computer storage medium, characterized in that the readable storage medium has stored thereon a computer program which, when being executed by a processor, implements the steps of a method for image recognition based on internet of things and edge calculation according to claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210149994.8A CN114648738A (en) | 2022-02-10 | 2022-02-10 | Image identification system and method based on Internet of things and edge calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210149994.8A CN114648738A (en) | 2022-02-10 | 2022-02-10 | Image identification system and method based on Internet of things and edge calculation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114648738A true CN114648738A (en) | 2022-06-21 |
Family
ID=81993850
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210149994.8A Pending CN114648738A (en) | 2022-02-10 | 2022-02-10 | Image identification system and method based on Internet of things and edge calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114648738A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116634638A (en) * | 2023-05-16 | 2023-08-22 | 珠海光通智装科技有限公司 | Light control strategy generation method, light control method and related device |
-
2022
- 2022-02-10 CN CN202210149994.8A patent/CN114648738A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116634638A (en) * | 2023-05-16 | 2023-08-22 | 珠海光通智装科技有限公司 | Light control strategy generation method, light control method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110033002B (en) | License plate detection method based on multitask cascade convolution neural network | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN109255326B (en) | Traffic scene smoke intelligent detection method based on multi-dimensional information feature fusion | |
| CN111896540B (en) | Water quality on-line monitoring system based on block chain | |
| CN112307919B (en) | Improved YOLOv 3-based digital information area identification method in document image | |
| CN110659649A (en) | Image processing and character recognition algorithm based on near infrared light imaging | |
| CN111209858A (en) | Real-time license plate detection method based on deep convolutional neural network | |
| CN109598200B (en) | Intelligent image identification system and method for molten iron tank number | |
| CN113256624A (en) | Continuous casting round billet defect detection method and device, electronic equipment and readable storage medium | |
| CN115311507A (en) | Building board classification method based on data processing | |
| CN116012291A (en) | Industrial part image defect detection method and system, electronic equipment and storage medium | |
| Surekha et al. | AUTOMATIC LICENSE PLATE RECOGNITION USING IMAGE PROCESSING AND NEURAL NETWORK. | |
| CN114648738A (en) | Image identification system and method based on Internet of things and edge calculation | |
| CN116935293B (en) | Automatic processing method and system for remote video exploration and damage assessment | |
| CN111160372B (en) | Large target identification method based on high-speed convolutional neural network | |
| CN110287752B (en) | Lattice code detection method and device | |
| CN110349119B (en) | Pavement disease detection method and device based on edge detection neural network | |
| CN116778460A (en) | Fatigue driving identification method based on image identification | |
| CN114758139B (en) | Method for detecting accumulated water in foundation pit | |
| CN115620121A (en) | Photoelectric target high-precision detection method based on digital twinning | |
| CN114373147A (en) | Detection method for low-texture video license plate | |
| CN111753842B (en) | Method and device for detecting text region of bill | |
| CN114694090A (en) | Campus abnormal behavior detection method based on improved PBAS algorithm and YOLOv5 | |
| Sathya et al. | Vehicle license plate recognition (vlpr) | |
| Jian et al. | Research on characters segmentation and characters recognition in intelligent license plate recognition system |
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 |