CN116026528A - High waterproof safe type tri-proof light - Google Patents
High waterproof safe type tri-proof light Download PDFInfo
- Publication number
- CN116026528A CN116026528A CN202310061219.1A CN202310061219A CN116026528A CN 116026528 A CN116026528 A CN 116026528A CN 202310061219 A CN202310061219 A CN 202310061219A CN 116026528 A CN116026528 A CN 116026528A
- Authority
- CN
- China
- Prior art keywords
- matrix
- humidity
- scale
- feature
- vector
- 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
- 239000011159 matrix material Substances 0.000 claims abstract description 244
- 239000013598 vector Substances 0.000 claims description 191
- 230000008859 change Effects 0.000 claims description 148
- 238000000605 extraction Methods 0.000 claims description 72
- 238000013527 convolutional neural network Methods 0.000 claims description 25
- 230000007246 mechanism Effects 0.000 claims description 22
- 230000004927 fusion Effects 0.000 claims description 9
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 7
- 238000012544 monitoring process Methods 0.000 claims description 6
- 230000001427 coherent effect Effects 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 3
- 238000009412 basement excavation Methods 0.000 abstract description 3
- 241001465382 Physalis alkekengi Species 0.000 abstract description 2
- 238000000034 method Methods 0.000 description 29
- 230000006870 function Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 10
- 238000005457 optimization Methods 0.000 description 7
- HDDSHPAODJUKPD-UHFFFAOYSA-N fenbendazole Chemical compound C1=C2NC(NC(=O)OC)=NC2=CC=C1SC1=CC=CC=C1 HDDSHPAODJUKPD-UHFFFAOYSA-N 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 230000007547 defect Effects 0.000 description 4
- 239000000243 solution Substances 0.000 description 4
- 230000007797 corrosion Effects 0.000 description 3
- 238000005260 corrosion Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000003647 oxidation Effects 0.000 description 3
- 238000007254 oxidation reaction Methods 0.000 description 3
- 241000258971 Brachiopoda Species 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
Images
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
- Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
- Y02B20/00—Energy efficient lighting technologies, e.g. halogen lamps or gas discharge lamps
- Y02B20/40—Control techniques providing energy savings, e.g. smart controller or presence detection
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The application relates to the technical field of lighting lamps and lanterns, and more specifically relates to a high waterproof safe type tri-proof light. The high waterproof safety type tri-proof light can be used for carrying out deep feature excavation on surface images acquired in real time, and whether the waterproof characteristics of the surface of the tri-proof light in different humidity environments meet requirements or not can be evaluated by utilizing a transfer matrix between the surface image features and the humidity features of the surface of the tri-proof light, so that intelligent early warning is carried out, and the safety of the tri-proof light is improved.
Description
Technical Field
The application relates to the technical field of lighting lamps and lanterns, and more specifically relates to a high waterproof safe type tri-proof light.
Background
The three-proofing lamp is a waterproof, dustproof and anticorrosive lamp, has a protection grade of IP65 and is suitable for illumination of places such as food factories, parking lots, factories, goods yards and the like. The existing three-proofing lamp is complex in structure, the upper lamp shell and the lower lamp shell are mainly injection molded, and a T-shaped tube light source is used as a light source.
Although the tri-proof light has relatively strong waterproof capability, if the tri-proof light is placed in a humid or underwater environment for a long time, water easily enters the light tube body from the gaps due to oxidation or corrosion of the surface of the tri-proof light, so that the light tube is damaged, and the safety is insufficient.
The existing improvement scheme mostly improves the waterproof performance of the tri-proof light along the direction of material optimization or structure optimization so as to improve the safety of the tri-proof light, and the optimization also achieves good effect, but the change of the waterproof performance of the tri-proof light in the using process cannot be known, so that the tri-proof light can not be known when water leakage occurs, and the tri-proof light fails.
Therefore, an optimized highly waterproof safe type tri-proof light is desired.
Disclosure of Invention
The present application has been made in order to solve the above technical problems. The embodiment of the application provides a high waterproof safe type tri-proof light, wherein, high waterproof safe type tri-proof light can carry out the characteristic excavation of degree of depth through the surface image to the real-time collection to utilize the surface image characteristic with the transfer matrix between the humidity characteristic of tri-proof light surface evaluates whether satisfy the requirement in the humidity environment of difference tri-proof light surface, and then carries out intelligent early warning, in order to improve the security of tri-proof light.
According to one aspect of the present application, there is provided a highly waterproof safe-type tri-proof light, comprising:
the data monitoring and collecting module is used for obtaining humidity values of the surface of the three-proofing lamp at a plurality of preset time points in a preset time period collected by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp collected by a camera arranged in the three-proofing lamp;
The surface state analysis module is used for obtaining a surface feature matrix through a convolutional neural network model using a spatial attention mechanism according to the surface image of the tri-proof light;
the humidity characteristic extraction module is used for arranging humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period into humidity input vectors according to the time dimension, and obtaining multiscale humidity change characteristic vectors through the multiscale neighborhood characteristic extraction module;
the Gaussian enhancement module is used for carrying out data enhancement on the multiscale humidity change characteristic vector by using a Gaussian density chart so as to obtain a multiscale humidity change matrix;
the characteristic correction module is used for correcting the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix;
the characteristic fusion module is used for fusing the surface characteristic matrix and the multi-scale humidity change matrix to obtain a classification characteristic matrix; and the early warning module is used for enabling the classification feature matrix to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
In the above Gao Fangshui safe three-proofing lamp, the surface state analysis module is further configured to: input data are carried out on each layer of the convolutional neural network model in forward transfer of the layer: inputting the surface image of the tri-proof light into a multi-layer convolution layer of the convolution neural network model to output an initial surface state feature matrix from the last layer of the multi-layer convolution layer; inputting the initial surface state feature matrix into a spatial attention module of the convolutional neural network model to obtain a spatial attention matrix; and calculating the space attention matrix and the initial surface state feature matrix, and multiplying the space attention matrix and the initial surface state feature matrix according to the position points to obtain the surface feature matrix.
In the above Gao Fangshui safe type tri-proof light, the humidity characteristic extraction module comprises: a first scale feature extraction unit, configured to input the humidity input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first scale humidity feature vector, where the first convolution layer has a first one-dimensional convolution kernel with a first length; a second scale feature extraction unit configured to input the humidity input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector, where the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and the multi-scale cascading unit is used for cascading the first scale humidity characteristic vector and the second scale humidity characteristic vector to obtain the multi-scale humidity change characteristic vector.
In the above Gao Fangshui safe three-proofing lamp, the first scale feature extraction unit is further configured to: performing one-dimensional convolution encoding on the humidity input vector by using a first convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a first-scale humidity feature vector; wherein, the formula is:
Wherein a is the width of the first convolution kernel in the X direction, F (a) is a first convolution kernel parameter vector, G (X-a) is a local vector matrix calculated by a convolution kernel function, w is the size of the first convolution kernel, and X represents the humidity input vector.
In the above Gao Fangshui safe three-proofing lamp, the second scale feature extraction unit is further configured to: performing one-dimensional convolution encoding on the humidity input vector by using a second convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a second scale humidity feature vector; wherein, the formula is:
wherein b is the width of the second convolution kernel in the X direction, F (b) is a second convolution kernel parameter vector, G (X-b) is a local vector matrix calculated with a convolution kernel function, m is the size of the second convolution kernel, and X represents the humidity input vector.
In the above Gao Fangshui safe-guard lamp, the gaussian enhancement module comprises: a gaussian density map construction unit, configured to construct a self-gaussian density map of the multiscale humidity change feature vector, where a mean vector of the self-gaussian density map is the multiscale humidity change feature vector, and a value of each position in a covariance matrix of the gaussian density map is a variance between feature values of two corresponding positions in the multiscale humidity change feature vector; and a Gaussian discrete unit, configured to discretize a Gaussian distribution of each position of the Gaussian density map to obtain the multi-scale humidity change matrix.
In the above Gao Fangshui safe three-proofing lamp, the feature correction module is further configured to: correcting the multi-scale humidity change matrix by using a pre-classification-based class probability coherent compensation mechanism according to the following formula to obtain a corrected multi-scale humidity change matrix; wherein, the formula is:
M′=p p ·M p-1 ⊙e -p·M
wherein M is the multi-scale humidity change matrix, p is a probability value obtained by pre-classifying the multi-scale humidity change matrix M by the classifier, and by-represents multiplication by location.
In the above Gao Fangshui safe type tri-proof light, the feature fusion module is further configured to: fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix according to the following formula; wherein, the formula is:
wherein M is 1 Representing the surface feature matrix, M 2 Representing the multi-scale humidity change matrix, M representing the classification feature matrix,representing matrix multiplication.
In the above-mentioned Gao Fangshui safe type tri-proof light, the early warning module includes: the matrix unfolding unit is used for unfolding the classification characteristic matrix into a classification characteristic vector according to a row vector or a column vector; the full-connection coding unit is used for carrying out full-connection coding on the classification characteristic vectors by using a full-connection layer of the classifier so as to obtain coded classification characteristic vectors; and the classification unit is used for inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
According to another aspect of the present application, there is also provided a method of operating a highly waterproof and safe type tri-proof light, comprising:
acquiring humidity values of the surface of a three-proofing lamp at a plurality of preset time points in a preset time period acquired by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp acquired by a camera arranged in the three-proofing lamp;
the surface image of the tri-proof light is subjected to a convolutional neural network model by using a spatial attention mechanism to obtain a surface feature matrix;
arranging humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period into humidity input vectors according to a time dimension, and then obtaining multiscale humidity change feature vectors through a multiscale neighborhood feature extraction module;
data enhancement is carried out on the multiscale humidity change feature vector by using a Gaussian density chart so as to obtain a multiscale humidity change matrix;
correcting the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix;
fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix; and passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
In the above Gao Fangshui safe three-proofing lamp operation method, the step of obtaining the surface feature matrix by using a convolutional neural network model of a spatial attention mechanism from the surface image of the three-proofing lamp includes: input data are carried out on each layer of the convolutional neural network model in forward transfer of the layer: inputting the surface image of the tri-proof light into a multi-layer convolution layer of the convolution neural network model to output an initial surface state feature matrix from the last layer of the multi-layer convolution layer; inputting the initial surface state feature matrix into a spatial attention module of the convolutional neural network model to obtain a spatial attention matrix; and calculating the space attention matrix and the initial surface state feature matrix, and multiplying the space attention matrix and the initial surface state feature matrix according to the position points to obtain the surface feature matrix.
In the above Gao Fangshui safe three-proofing lamp operation method, the arranging the humidity values of the three-proofing lamp surfaces at a plurality of predetermined time points in the predetermined time period into the humidity input vector according to the time dimension, and then obtaining the multiscale humidity change feature vector through the multiscale neighborhood feature extraction module includes: inputting the humidity input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first scale humidity feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel with a first length; inputting the humidity input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and cascading the first scale humidity feature vector and the second scale humidity feature vector to obtain the multi-scale humidity change feature vector.
In the above Gao Fangshui safe three-proofing lamp operation method, the inputting the humidity input vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale humidity feature vector includes: performing one-dimensional convolution encoding on the humidity input vector by using a first convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a first-scale humidity feature vector; wherein, the formula is:
wherein a is the width of the first convolution kernel in the X direction, F (a) is a first convolution kernel parameter vector, G (X-a) is a local vector matrix calculated by a convolution kernel function, w is the size of the first convolution kernel, and X represents the humidity input vector.
In the above Gao Fangshui safe three-proofing lamp operation method, the inputting the humidity input vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector includes: performing one-dimensional convolution encoding on the humidity input vector by using a second convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a second scale humidity feature vector; wherein, the formula is:
Wherein b is the width of the second convolution kernel in the X direction, F (b) is a second convolution kernel parameter vector, G (X-b) is a local vector matrix calculated with a convolution kernel function, m is the size of the second convolution kernel, and X represents the humidity input vector.
In the above operation method of Gao Fangshui safe three-proofing lamp, the data enhancement is performed on the multiscale humidity change feature vector by using a gaussian density chart to obtain a multiscale humidity change matrix, which includes: constructing a self-Gaussian density map of the multi-scale humidity change feature vector, wherein the mean value vector of the self-Gaussian density map is the multi-scale humidity change feature vector, and the value of each position in the covariance matrix of the Gaussian density map is the variance between the feature values of two corresponding positions in the multi-scale humidity change feature vector; and discretizing the Gaussian distribution of each position of the Gaussian density map to obtain the multi-scale humidity change matrix.
In the above operation method of Gao Fangshui safe three-proofing lamp, the correcting the characteristic distribution of the multiscale humidity change matrix to obtain a corrected multiscale humidity change matrix includes: correcting the multi-scale humidity change matrix by using a pre-classification-based class probability coherent compensation mechanism according to the following formula to obtain a corrected multi-scale humidity change matrix; wherein, the formula is:
M′=p p ·M p-1 ⊙e -p·M
Wherein M is the multi-scale humidity change matrix, p is a probability value obtained by pre-classifying the multi-scale humidity change matrix M by the classifier, and by-represents multiplication by location.
In the above Gao Fangshui safe three-proofing lamp operation method, the fusing the surface feature matrix and the multiscale humidity change matrix to obtain a classification feature matrix includes: fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix according to the following formula; wherein, the formula is:
wherein M is 1 Representing the surface feature matrix, M 2 Representing the multi-scale humidity change matrix, M representing the classification feature matrix,representing matrix multiplication.
In the above Gao Fangshui safe three-proofing lamp operation method, the step of passing the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether to generate a safety early warning prompt, and the method includes: expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors; performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
Compared with the prior art, the high waterproof safety type tri-proof light that this application provided, wherein, high waterproof safety type tri-proof light can carry out the characteristic excavation of degree of depth through the surface image to the real-time collection to utilize the surface image characteristic with the transfer matrix between the humidity characteristic of tri-proof light surface evaluates whether satisfy the requirement in the humidity environment of difference tri-proof light surface, and then carries out intelligent early warning, in order to improve the security of tri-proof light.
Drawings
Fig. 1 illustrates an application scenario diagram of a highly waterproof safe-type tri-proof light according to an embodiment of the present application.
Fig. 2 illustrates a block diagram of a highly waterproof safe-type tri-proof light according to an embodiment of the present application.
Fig. 3 illustrates a system architecture diagram of a highly waterproof safe-type tri-proof light according to an embodiment of the present application.
Fig. 4 illustrates a block diagram of a humidity signature extraction module in a highly waterproof and safe tri-proof light, according to an embodiment of the present application.
Fig. 5 illustrates a block diagram of an early warning module in a highly waterproof and safe type tri-proof light according to an embodiment of the present application.
Fig. 6 illustrates a flow chart of a method of operating a highly waterproof safe-type tri-proof light in accordance with an embodiment of the present application.
Reference numerals: 100-high waterproof safety type tri-proof light, 110-data monitoring and acquisition module, 120-surface state analysis module, 130-humidity characteristic extraction module, 140-Gaussian enhancement module, 150-characteristic correction module, 160-characteristic fusion module, 170-early warning module, 131-first scale characteristic extraction unit, 132-second scale characteristic extraction unit, 133-multi-scale cascade unit, 171-matrix expansion unit, 172-full-connection coding unit and 173-classification unit.
Detailed Description
The invention is further described below with reference to the drawings and examples.
Examples
As described above, although the tri-proof light has a relatively strong waterproof ability, if the tri-proof light is placed in a humid or underwater environment for a long time, water easily enters the tube body from these gaps due to oxidation or corrosion of the surface of the tri-proof light, thereby causing damage to the tube and safety deficiency. The waterproof performance of the three-proofing lamp is improved along the direction of material optimization or structure optimization to improve the safety of the three-proofing lamp, but the change of the waterproof performance of the three-proofing lamp in the using process cannot be known, so that the three-proofing lamp can not be known at all times when water leakage occurs to the three-proofing lamp, and the three-proofing lamp fails.
Therefore, an optimized high waterproof safety type tri-proof light is expected, which can monitor the surface waterproof performance of the tri-proof light in real time and perform intelligent early warning based on the humidity characteristic of the surface of the tri-proof light so as to improve the safety of the tri-proof light.
Specifically, in the technical scheme of the application, the waterproof characteristic of the surface of the tri-proof light can be represented by the surface image of the tri-proof light collected by the camera arranged in the tri-proof light, and it is understood that if the surface defect such as a crack or scratch is observed on the surface of the tri-proof light, the waterproof performance of the tri-proof light can be considered to be weakened. Further, when the security pre-warning is performed on the tri-proof light, besides the water-proof characteristic of the surface of the tri-proof light is to be observed, the humidity condition of the surface of the tri-proof light is to be considered, and it is to be understood that if the surface humidity of the tri-proof light is low, the water-proof performance requirement on the tri-proof light is relatively low, and if the surface humidity of the tri-proof light is high, the water-proof performance requirement on the tri-proof light is relatively high.
Specifically, first, humidity values of the surface of the three-proofing lamp at a plurality of preset time points in a preset time period acquired by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp acquired by a camera arranged in the three-proofing lamp are acquired.
And then, the surface image of the tri-proof light is processed through a convolutional neural network model using a spatial attention mechanism to obtain a surface feature matrix. That is, in the technical scheme of the present application, a convolutional neural network model having a surface with excellent performance in the field of image feature extraction is used as a feature extractor to extract the surface state features of the surface of the tri-proof light, so as to represent the waterproof property of the tri-proof light. In particular, it is considered that if the surface state of the tri-proof light changes, for example, cracks, scratches, etc. occur, the cracks, scratches, etc. are perceived in the image, but because the cracks, scratches, etc. belong to shallow line features, the features are easily confused with other shallow features when extraction is performed, and are easily blurred or even ignored as the coding depth deepens in the convolutional coding process. Therefore, in the technical solution of the present application, a spatial attention mechanism is integrated in the convolutional neural network model so that the surface features for representing the surface defects are given a higher attention in the feature extraction, thereby improving the accuracy of the surface state feature extraction.
Aiming at humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period, firstly, arranging the humidity values of the surfaces of the three-proofing lamps at the preset time points in the preset time period into humidity input vectors according to a time dimension, and then obtaining multiscale humidity change feature vectors through a multiscale neighborhood feature extraction module. That is, a multi-scale neighborhood feature extraction module comprising a plurality of parallel one-dimensional convolution layers is used as a feature extractor to capture pattern features of humidity distribution over different time spans to obtain the multi-scale humidity change feature vector.
However, because the humidity value of the surface of the tri-proof light is less in data quantity at the source domain end, even if the multi-scale neighborhood feature extraction module can extract quite rich humidity features due to the specificity of the model structure, the sparse data quantity at the source domain end still causes slight impoverishment of humidity feature representation. In particular, in the technical solution of the present application, the gaussian density map concept is used to data enhance the humidity characteristic distribution.
In particular, gaussian density maps are widely used in deep learning for a priori based estimation of target posterior and can therefore be used to correct data distribution to achieve the above objective. Specifically, firstly, constructing a self-Gaussian density map of the multi-scale humidity change feature vector, wherein the mean value vector of the self-Gaussian density map is the multi-scale humidity change feature vector, and the value of each position in the covariance matrix of the Gaussian density map is the variance between the feature values of two corresponding positions in the multi-scale humidity change feature vector. And then, carrying out Gaussian discretization on the Gaussian distribution of each position in the Gaussian density map to obtain a multi-scale humidity change matrix.
In the technical scheme of the application, when the gaussian density map is used for carrying out data enhancement on the multiscale humidity change feature vector to obtain a multiscale humidity change matrix, partial local random distribution is inevitably introduced into the multiscale humidity change matrix due to the gaussian discretization carried out on the gaussian density map, so that negative influence of the local random distribution on a classification result is expected to be inhibited when the surface feature matrix and the multiscale humidity change matrix are fused to obtain the classification feature matrix.
Thus, the multiscale humidity change matrix is preferably corrected for a pre-classification based class probability coherence compensation mechanism, expressed as:
M′=p p ·M p-1 ⊙e -p·M
wherein M is the multi-scale humidity change matrix, and p is a probability value obtained by pre-classifying the multi-scale humidity change matrix M by the classifier.
In the correction of the class probability coherence compensation mechanism based on pre-classification, class probability values of the classifier are taken as multiplicative interference noise items of classification features to carry out coherence compensation of class probabilities on the classification features by considering class coherence interference generated by possible local random distribution in a weight matrix of the classifier and feature distribution, so that equivalent probability intensity representation under the condition of no interference of the multi-scale humidity change matrix M can be recovered, and the surface feature matrix and the multi-scale humidity change matrix are fused, so that the accuracy of classification results of the classification feature matrix can be improved.
And then fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix. In a specific example of the present application, the classification feature vector may be obtained by calculating a response estimate of the multi-scale humidity change matrix with respect to the surface feature matrix, that is, calculating a relative amount of a surface humidity feature of the tri-proof light with respect to a waterproof feature of the tri-proof light as a feature representation for evaluating safety of the tri-proof light. And then, the classification feature matrix is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
Based on this, this application provides a high waterproof safe type tri-proof light, and it includes: the data monitoring and collecting module is used for obtaining humidity values of the surface of the three-proofing lamp at a plurality of preset time points in a preset time period collected by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp collected by a camera arranged in the three-proofing lamp; the surface state analysis module is used for obtaining a surface feature matrix through a convolutional neural network model using a spatial attention mechanism according to the surface image of the tri-proof light; the humidity characteristic extraction module is used for arranging humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period into humidity input vectors according to the time dimension, and obtaining multiscale humidity change characteristic vectors through the multiscale neighborhood characteristic extraction module; the Gaussian enhancement module is used for carrying out data enhancement on the multiscale humidity change characteristic vector by using a Gaussian density chart so as to obtain a multiscale humidity change matrix; the characteristic correction module is used for correcting the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix; the characteristic fusion module is used for fusing the surface characteristic matrix and the multi-scale humidity change matrix to obtain a classification characteristic matrix; and the early warning module is used for enabling the classification feature matrix to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
Fig. 1 illustrates an application scenario diagram of a highly waterproof safe-type tri-proof light according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, humidity values of a surface of a tri-proof light at a plurality of predetermined time points within a predetermined period of time acquired by a humidity sensor (e.g., se) disposed on the surface of the tri-proof light (e.g., L as illustrated in fig. 1) and a surface image of the tri-proof light acquired by a camera (e.g., C as illustrated in fig. 1) disposed within the tri-proof light are acquired. Further, the humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period and the surface images of the three-proofing lamps are input into a server (for example, S as shown in fig. 1) provided with a high waterproof safety three-proofing lamp algorithm, wherein the server can process the humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period and the surface images of the three-proofing lamps based on the high waterproof safety three-proofing lamp algorithm so as to obtain a classification result for indicating whether a safety warning prompt is generated.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
Exemplary System
Fig. 2 illustrates a block diagram of a highly waterproof safe-type tri-proof light according to an embodiment of the present application. As shown in fig. 2, a highly waterproof safe-type tri-proof light 100 according to an embodiment of the present application includes: the data monitoring and collecting module 110 is configured to obtain humidity values of a surface of a tri-proof light at a plurality of predetermined time points in a predetermined time period collected by a humidity sensor disposed on the surface of the tri-proof light and surface images of the tri-proof light collected by a camera disposed in the tri-proof light; the surface state analysis module 120 is configured to obtain a surface feature matrix by using a convolutional neural network model of a spatial attention mechanism for the surface image of the tri-proof light; the humidity feature extraction module 130 is configured to arrange humidity values of the surfaces of the tri-proof lamps at a plurality of predetermined time points in the predetermined time period into humidity input vectors according to a time dimension, and then obtain multiscale humidity change feature vectors through the multiscale neighborhood feature extraction module; the gaussian enhancement module 140 is configured to perform data enhancement on the multiscale humidity change feature vector by using a gaussian density chart to obtain a multiscale humidity change matrix; the feature correction module 150 is configured to correct the feature distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix; a feature fusion module 160, configured to fuse the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix; and the early warning module 170 is configured to pass the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether to generate a safety early warning prompt.
Fig. 3 illustrates a system architecture diagram of the high waterproof safety type tri-proof light 100 according to an embodiment of the present application. As shown in fig. 3, in the system architecture of the Gao Fangshui safe-type tri-proof light 100, first, humidity values of the surface of the tri-proof light at a plurality of predetermined time points within a predetermined period of time acquired by a humidity sensor disposed on the surface of the tri-proof light and surface images of the tri-proof light acquired by a camera disposed in the tri-proof light are acquired. And then, the surface image of the tri-proof light is processed through a convolutional neural network model using a spatial attention mechanism to obtain a surface feature matrix. And then, arranging humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period into humidity input vectors according to a time dimension, and obtaining multiscale humidity change feature vectors through a multiscale neighborhood feature extraction module. And further, carrying out data enhancement on the multiscale humidity change characteristic vector by using a Gaussian density chart to obtain a multiscale humidity change matrix. And correcting the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix. And then fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix. And then, the classification feature matrix passes through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
In the above Gao Fangshui safe-type tri-proof light 100, the data monitoring and collecting module 110 is configured to obtain humidity values of a surface of the tri-proof light at a plurality of predetermined time points within a predetermined time period collected by a humidity sensor disposed on the surface of the tri-proof light and a surface image of the tri-proof light collected by a camera disposed in the tri-proof light. As described above, although the tri-proof light has a relatively strong waterproof ability, if the tri-proof light is placed in a humid or underwater environment for a long time, water easily enters the tube body from these gaps due to oxidation or corrosion of the surface of the tri-proof light, thereby causing damage to the tube and safety deficiency. The waterproof performance of the three-proofing lamp is improved along the direction of material optimization or structure optimization to improve the safety of the three-proofing lamp, but the change of the waterproof performance of the three-proofing lamp in the using process cannot be known, so that the three-proofing lamp can not be known at all times when water leakage occurs to the three-proofing lamp, and the three-proofing lamp fails.
Therefore, an optimized high waterproof safety type tri-proof light is expected, which can monitor the surface waterproof performance of the tri-proof light in real time and perform intelligent early warning based on the humidity characteristic of the surface of the tri-proof light so as to improve the safety of the tri-proof light.
Specifically, in the technical scheme of the application, the waterproof characteristic of the surface of the tri-proof light can be represented by the surface image of the tri-proof light collected by the camera arranged in the tri-proof light, and it is understood that if the surface defect such as a crack or scratch is observed on the surface of the tri-proof light, the waterproof performance of the tri-proof light can be considered to be weakened. Further, when the security pre-warning is performed on the tri-proof light, besides the water-proof characteristic of the surface of the tri-proof light is to be observed, the humidity condition of the surface of the tri-proof light is to be considered, and it is to be understood that if the surface humidity of the tri-proof light is low, the water-proof performance requirement on the tri-proof light is relatively low, and if the surface humidity of the tri-proof light is high, the water-proof performance requirement on the tri-proof light is relatively high.
Specifically, first, humidity values of the surface of the three-proofing lamp at a plurality of preset time points in a preset time period acquired by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp acquired by a camera arranged in the three-proofing lamp are acquired.
In the above Gao Fangshui safe-type tri-proof light 100, the surface state analysis module 120 is configured to obtain a surface feature matrix by using a convolutional neural network model of a spatial attention mechanism for the surface image of the tri-proof light. That is, in the technical scheme of the present application, a convolutional neural network model having a surface with excellent performance in the field of image feature extraction is used as a feature extractor to extract the surface state features of the surface of the tri-proof light, so as to represent the waterproof property of the tri-proof light. In particular, it is considered that if the surface state of the tri-proof light changes, for example, cracks, scratches, etc. occur, the cracks, scratches, etc. are perceived in the image, but because the cracks, scratches, etc. belong to shallow line features, the features are easily confused with other shallow features when extraction is performed, and are easily blurred or even ignored as the coding depth deepens in the convolutional coding process. Therefore, in the technical solution of the present application, a spatial attention mechanism is integrated in the convolutional neural network model so that the surface features for representing the surface defects are given a higher attention in the feature extraction, thereby improving the accuracy of the surface state feature extraction.
In one example, in the Gao Fangshui safe-guard lamp 100, the surface state analyzing module 120 is further configured to: input data are carried out on each layer of the convolutional neural network model in forward transfer of the layer: inputting the surface image of the tri-proof light into a multi-layer convolution layer of the convolution neural network model to output an initial surface state feature matrix from the last layer of the multi-layer convolution layer; inputting the initial surface state feature matrix into a spatial attention module of the convolutional neural network model to obtain a spatial attention matrix; and calculating the space attention matrix and the initial surface state feature matrix, and multiplying the space attention matrix and the initial surface state feature matrix according to the position points to obtain the surface feature matrix.
In the above Gao Fangshui safe-type tri-proof light 100, the humidity feature extraction module 130 is configured to arrange humidity values of the tri-proof light surface at a plurality of predetermined time points within the predetermined time period into humidity input vectors according to a time dimension, and then obtain multiscale humidity change feature vectors through the multiscale neighborhood feature extraction module. Aiming at humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period, firstly, arranging the humidity values of the surfaces of the three-proofing lamps at the preset time points in the preset time period into humidity input vectors according to a time dimension, and then obtaining multiscale humidity change feature vectors through a multiscale neighborhood feature extraction module. That is, a multi-scale neighborhood feature extraction module comprising a plurality of parallel one-dimensional convolution layers is used as a feature extractor to capture pattern features of humidity distribution over different time spans to obtain the multi-scale humidity change feature vector.
Fig. 4 illustrates a block diagram of a humidity signature extraction module in a highly waterproof and safe tri-proof light, according to an embodiment of the present application. As shown in fig. 4, in the Gao Fangshui safe-type tri-proof light 100, the humidity characteristic extraction module 130 includes: a first scale feature extraction unit 131, configured to input the humidity input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first scale humidity feature vector, where the first convolution layer has a first one-dimensional convolution kernel with a first length; a second scale feature extraction unit 132 configured to input the humidity input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector, where the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and a multi-scale cascade unit 133, configured to cascade the first scale humidity feature vector and the second scale humidity feature vector to obtain the multi-scale humidity variation feature vector.
In one example, in the above Gao Fangshui safe-type tri-proof light 100, the first scale feature extraction unit 131 is further configured to: performing one-dimensional convolution encoding on the humidity input vector by using a first convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a first-scale humidity feature vector; wherein, the formula is:
Wherein a is the width of the first convolution kernel in the X direction, F (a) is a first convolution kernel parameter vector, G (X-a) is a local vector matrix calculated by a convolution kernel function, w is the size of the first convolution kernel, and X represents the humidity input vector.
In one example, in the above Gao Fangshui safe-guard lamp 100, the second scale feature extracting unit 132 is further configured to: performing one-dimensional convolution encoding on the humidity input vector by using a second convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a second scale humidity feature vector; wherein, the formula is:
wherein b is the width of the second convolution kernel in the X direction, F (b) is a second convolution kernel parameter vector, G (X-b) is a local vector matrix calculated with a convolution kernel function, m is the size of the second convolution kernel, and X represents the humidity input vector.
In the above Gao Fangshui safe-type tri-proof light 100, the gaussian enhancement module 140 is configured to perform data enhancement on the multiscale humidity change feature vector by using a gaussian density chart to obtain a multiscale humidity change matrix. Because the data volume of the humidity value of the surface of the tri-proof light at the source domain end is less, even if the multi-scale neighborhood feature extraction module can extract quite rich humidity features due to the specificity of the model structure, the sparse data volume of the source domain end still can cause the humidity feature representation to be slightly lean. In particular, in the technical solution of the present application, the gaussian density map concept is used to data enhance the humidity characteristic distribution.
In particular, gaussian density maps are widely used in deep learning for a priori based estimation of target posterior and can therefore be used to correct data distribution to achieve the above objective. Specifically, firstly, constructing a self-Gaussian density map of the multi-scale humidity change feature vector, wherein the mean value vector of the self-Gaussian density map is the multi-scale humidity change feature vector, and the value of each position in the covariance matrix of the Gaussian density map is the variance between the feature values of two corresponding positions in the multi-scale humidity change feature vector. And then, carrying out Gaussian discretization on the Gaussian distribution of each position in the Gaussian density map to obtain a multi-scale humidity change matrix.
In one example, in the Gao Fangshui safe-guard lamp 100 described above, the gaussian enhancement module 140 includes: a gaussian density map construction unit, configured to construct a self-gaussian density map of the multiscale humidity change feature vector, where a mean vector of the self-gaussian density map is the multiscale humidity change feature vector, and a value of each position in a covariance matrix of the gaussian density map is a variance between feature values of two corresponding positions in the multiscale humidity change feature vector; and a Gaussian discrete unit, configured to discretize a Gaussian distribution of each position of the Gaussian density map to obtain the multi-scale humidity change matrix.
In the above Gao Fangshui safe-type tri-proof light 100, the characteristic correction module 150 is configured to correct the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix.
In the technical scheme of the application, when the gaussian density map is used for carrying out data enhancement on the multiscale humidity change feature vector to obtain a multiscale humidity change matrix, partial local random distribution is inevitably introduced into the multiscale humidity change matrix due to the gaussian discretization carried out on the gaussian density map, so that negative influence of the local random distribution on a classification result is expected to be inhibited when the surface feature matrix and the multiscale humidity change matrix are fused to obtain the classification feature matrix.
Thus, preferably, the multiscale humidity change matrix is corrected for a pre-classification based class probability coherence compensation mechanism. In one example, in the Gao Fangshui safe-guard lamp 100, the feature correction module 150 is further configured to: correcting the multi-scale humidity change matrix by using a pre-classification-based class probability coherent compensation mechanism according to the following formula to obtain a corrected multi-scale humidity change matrix; wherein, the formula is:
M′=p p ·M p-1 ⊙e -p·M
Wherein M is the multi-scale humidity change matrix, p is a probability value obtained by pre-classifying the multi-scale humidity change matrix M by the classifier, and by-represents multiplication by location.
In the correction of the class probability coherence compensation mechanism based on pre-classification, class probability values of the classifier are taken as multiplicative interference noise items of classification features to carry out coherence compensation of class probabilities on the classification features by considering class coherence interference generated by possible local random distribution in a weight matrix of the classifier and feature distribution, so that equivalent probability intensity representation under the condition of no interference of the multi-scale humidity change matrix M can be recovered, and the surface feature matrix and the multi-scale humidity change matrix are fused, so that the accuracy of classification results of the classification feature matrix can be improved.
In the Gao Fangshui safe-type tri-proof light 100, the feature fusion module 160 is configured to fuse the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix. In a specific example of the present application, the classification feature vector may be obtained by calculating a response estimate of the multi-scale humidity change matrix with respect to the surface feature matrix, that is, calculating a relative amount of a surface humidity feature of the tri-proof light with respect to a waterproof feature of the tri-proof light as a feature representation for evaluating safety of the tri-proof light.
In one example, in the Gao Fangshui safe-guard lamp 100, the feature fusion module 160 is further configured to: fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix according to the following formula; wherein, the formula is:
wherein M is 1 Representing the surface feature matrix, M 2 Representing the multi-scale humidity change matrix, M representing the classification feature matrix,representing matrix multiplication.
In the above Gao Fangshui safe-type tri-proof light 100, the early warning module 170 is configured to pass the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether a safety early warning prompt is generated. Therefore, the surface waterproof performance of the three-proofing lamp is monitored in real time, and intelligent early warning is performed based on the humidity characteristic of the surface of the three-proofing lamp, so that the safety of the three-proofing lamp is improved.
Fig. 5 illustrates a block diagram of an early warning module in a highly waterproof and safe type tri-proof light according to an embodiment of the present application. As shown in fig. 5, in the Gao Fangshui safe-type tri-proof light 100, the pre-warning module 170 includes: a matrix expansion unit 171 for expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors; a full-connection encoding unit 172, configured to perform full-connection encoding on the classification feature vector by using a full-connection layer of the classifier to obtain an encoded classification feature vector; and a classification unit 173, configured to input the encoded classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
In summary, the high waterproof safety type tri-proof light 100 according to the embodiment of the present application is illustrated, wherein the high waterproof safety type tri-proof light is capable of performing feature mining of depth through surface images acquired in real time, and evaluating whether waterproof characteristics of the tri-proof light surface in different humidity environments meet requirements or not by using a transfer matrix between surface image features and humidity features of the tri-proof light surface, and further performing intelligent early warning, so as to improve safety of the tri-proof light.
Exemplary method
Fig. 6 illustrates a flow chart of a method of operating a highly waterproof safe-type tri-proof light in accordance with an embodiment of the present application. As shown in fig. 6, the operation method of the high waterproof safety type tri-proof light according to the embodiment of the present application includes the steps of: s110, acquiring humidity values of the surface of the three-proofing lamp at a plurality of preset time points in a preset time period acquired by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp acquired by a camera arranged in the three-proofing lamp; s120, the surface image of the tri-proof light is processed through a convolutional neural network model using a spatial attention mechanism to obtain a surface feature matrix; s130, arranging humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period into humidity input vectors according to a time dimension, and then obtaining multiscale humidity change feature vectors through a multiscale neighborhood feature extraction module; s140, carrying out data enhancement on the multiscale humidity change feature vector by using a Gaussian density chart to obtain a multiscale humidity change matrix; s150, correcting the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix; s160, fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix; and S170, passing the classification feature matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
In one example, in the method for operating a Gao Fangshui safe-type tri-proof light, the step of obtaining the surface feature matrix from the surface image of the tri-proof light by using a convolutional neural network model of a spatial attention mechanism includes: input data are carried out on each layer of the convolutional neural network model in forward transfer of the layer: inputting the surface image of the tri-proof light into a multi-layer convolution layer of the convolution neural network model to output an initial surface state feature matrix from the last layer of the multi-layer convolution layer; inputting the initial surface state feature matrix into a spatial attention module of the convolutional neural network model to obtain a spatial attention matrix; and calculating the space attention matrix and the initial surface state feature matrix, and multiplying the space attention matrix and the initial surface state feature matrix according to the position points to obtain the surface feature matrix.
In an example, in the above-mentioned Gao Fangshui safe three-proofing lamp operation method, the arranging the humidity values of the three-proofing lamp surfaces at a plurality of predetermined time points within the predetermined time period into the humidity input vector according to the time dimension, and then obtaining the multiscale humidity change feature vector through the multiscale neighborhood feature extraction module includes: inputting the humidity input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first scale humidity feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel with a first length; inputting the humidity input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and cascading the first scale humidity feature vector and the second scale humidity feature vector to obtain the multi-scale humidity change feature vector.
In one example, in the method for operating a Gao Fangshui safe-type tri-proof light, the inputting the humidity input vector into the first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first-scale humidity feature vector includes: performing one-dimensional convolution encoding on the humidity input vector by using a first convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a first-scale humidity feature vector; wherein, the formula is:
wherein a is the width of the first convolution kernel in the X direction, F (a) is a first convolution kernel parameter vector, G (X-a) is a local vector matrix calculated by a convolution kernel function, w is the size of the first convolution kernel, and X represents the humidity input vector.
In one example, in the method for operating a Gao Fangshui safe three-protection lamp, the inputting the humidity input vector into the second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector includes: performing one-dimensional convolution encoding on the humidity input vector by using a second convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a second scale humidity feature vector; wherein, the formula is:
Wherein b is the width of the second convolution kernel in the X direction, F (b) is a second convolution kernel parameter vector, G (X-b) is a local vector matrix calculated with a convolution kernel function, m is the size of the second convolution kernel, and X represents the humidity input vector.
In one example, in the method for operating a Gao Fangshui safe-type tri-proof light, the data enhancement on the multiscale humidity change feature vector using a gaussian density map to obtain a multiscale humidity change matrix includes: constructing a self-Gaussian density map of the multi-scale humidity change feature vector, wherein the mean value vector of the self-Gaussian density map is the multi-scale humidity change feature vector, and the value of each position in the covariance matrix of the Gaussian density map is the variance between the feature values of two corresponding positions in the multi-scale humidity change feature vector; and discretizing the Gaussian distribution of each position of the Gaussian density map to obtain the multi-scale humidity change matrix.
In one example, in the method for operating a Gao Fangshui safe-type tri-proof light, the correcting the characteristic distribution of the multiscale humidity change matrix to obtain a corrected multiscale humidity change matrix includes: correcting the multi-scale humidity change matrix by using a pre-classification-based class probability coherent compensation mechanism according to the following formula to obtain a corrected multi-scale humidity change matrix; wherein, the formula is:
M′=p p ·M p-1 ⊙e -p·M
Wherein M is the multi-scale humidity change matrix, p is a probability value obtained by pre-classifying the multi-scale humidity change matrix M by the classifier, and by-represents multiplication by location.
In one example, in the method for operating a Gao Fangshui safe-type tri-proof light, the fusing the surface feature matrix and the multiscale humidity change matrix to obtain the classification feature matrix includes: fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix according to the following formula; wherein, the formula is:
wherein M is 1 Representing the surface feature matrix, M 2 Representing the multi-scale humidity change matrix, M representing the classification feature matrix,representing matrix multiplication.
In an example, in the method for operating a Gao Fangshui safe-type tri-proof light, the step of passing the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether to generate a safety pre-warning prompt, includes: expanding the classification feature matrix into classification feature vectors according to row vectors or column vectors; performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
In summary, the operation method of the high waterproof safety type tri-proof light according to the embodiment of the application is clarified, wherein the high waterproof safety type tri-proof light can perform deep feature mining on surface images acquired in real time, and whether the waterproof characteristics of the surface of the tri-proof light in different humidity environments meet requirements or not is evaluated by using a transfer matrix between the surface image features and the humidity features of the surface of the tri-proof light, so that intelligent early warning is performed, and the safety of the tri-proof light is improved.
Claims (9)
1. A highly waterproof safe type tri-proof light, which is characterized by comprising:
the data monitoring and collecting module is used for obtaining humidity values of the surface of the three-proofing lamp at a plurality of preset time points in a preset time period collected by a humidity sensor arranged on the surface of the three-proofing lamp and surface images of the three-proofing lamp collected by a camera arranged in the three-proofing lamp;
the surface state analysis module is used for obtaining a surface feature matrix through a convolutional neural network model using a spatial attention mechanism according to the surface image of the tri-proof light;
the humidity characteristic extraction module is used for arranging humidity values of the surfaces of the three-proofing lamps at a plurality of preset time points in the preset time period into humidity input vectors according to the time dimension, and obtaining multiscale humidity change characteristic vectors through the multiscale neighborhood characteristic extraction module;
The Gaussian enhancement module is used for carrying out data enhancement on the multiscale humidity change characteristic vector by using a Gaussian density chart so as to obtain a multiscale humidity change matrix;
the characteristic correction module is used for correcting the characteristic distribution of the multi-scale humidity change matrix to obtain a corrected multi-scale humidity change matrix;
the characteristic fusion module is used for fusing the surface characteristic matrix and the multi-scale humidity change matrix to obtain a classification characteristic matrix; and the early warning module is used for enabling the classification feature matrix to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether a safety early warning prompt is generated or not.
2. The highly waterproof and safe type tri-proof light of claim 1, wherein the surface state analysis module is further configured to: input data are carried out on each layer of the convolutional neural network model in forward transfer of the layer:
inputting the surface image of the tri-proof light into a multi-layer convolution layer of the convolution neural network model to output an initial surface state feature matrix from the last layer of the multi-layer convolution layer;
inputting the initial surface state feature matrix into a spatial attention module of the convolutional neural network model to obtain a spatial attention matrix; and calculating the space attention matrix and the initial surface state feature matrix by multiplying the position points to obtain the surface feature matrix.
3. The highly waterproof and safe type tri-proof light of claim 2, wherein the humidity characteristic extraction module comprises:
a first scale feature extraction unit, configured to input the humidity input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first scale humidity feature vector, where the first convolution layer has a first one-dimensional convolution kernel with a first length;
a second scale feature extraction unit configured to input the humidity input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second scale humidity feature vector, where the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; and
and the multi-scale cascading unit is used for cascading the first scale humidity characteristic vector and the second scale humidity characteristic vector to obtain the multi-scale humidity change characteristic vector.
4. The highly water-resistant safe type tri-proof light of claim 3, wherein the first scale feature extraction unit is further configured to: performing one-dimensional convolution encoding on the humidity input vector by using a first convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a first-scale humidity feature vector;
Wherein, the formula is:
wherein a is the width of the first convolution kernel in the X direction, F (a) is a first convolution kernel parameter vector, G (X-a) is a local vector matrix calculated by a convolution kernel function, w is the size of the first convolution kernel, and X represents the humidity input vector.
5. The highly waterproof and safe three-proof lamp of claim 4, wherein the second scale feature extraction unit is further configured to: performing one-dimensional convolution encoding on the humidity input vector by using a second convolution layer of the multi-scale neighborhood feature extraction module according to the following formula to obtain a second scale humidity feature vector;
wherein, the formula is:
wherein b is the width of the second convolution kernel in the X direction, F (b) is a second convolution kernel parameter vector, G (X-b) is a local vector matrix calculated with a convolution kernel function, m is the size of the second convolution kernel, and X represents the humidity input vector.
6. The highly waterproof and safe three-proof lamp of claim 5, wherein the gaussian enhancement module comprises:
a gaussian density map construction unit, configured to construct a self-gaussian density map of the multiscale humidity change feature vector, where a mean vector of the self-gaussian density map is the multiscale humidity change feature vector, and a value of each position in a covariance matrix of the gaussian density map is a variance between feature values of two corresponding positions in the multiscale humidity change feature vector; and the Gaussian discrete unit is used for discretizing the Gaussian distribution of each position of the Gaussian density map to obtain the multi-scale humidity change matrix.
7. The highly waterproof and safe type tri-proof light of claim 6, wherein the characteristic correction module is further configured to: correcting the multi-scale humidity change matrix by using a pre-classification-based class probability coherent compensation mechanism according to the following formula to obtain a corrected multi-scale humidity change matrix;
wherein, the formula is:
M′=p p ·M p-1 ⊙e -p·M
wherein M is the multi-scale humidity change matrix, p is a probability value obtained by pre-classifying the multi-scale humidity change matrix M by the classifier, and by-represents multiplication by location.
8. The highly waterproof and safe type tri-proof light of claim 7, wherein the feature fusion module is further configured to: fusing the surface feature matrix and the multi-scale humidity change matrix to obtain a classification feature matrix according to the following formula;
wherein, the formula is:
9. The highly waterproof and safe type tri-proof light of claim 8, wherein the pre-warning module comprises:
the matrix unfolding unit is used for unfolding the classification characteristic matrix into a classification characteristic vector according to a row vector or a column vector;
The full-connection coding unit is used for carrying out full-connection coding on the classification characteristic vectors by using a full-connection layer of the classifier so as to obtain coded classification characteristic vectors; and the classification unit is used for inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310061219.1A CN116026528A (en) | 2023-01-14 | 2023-01-14 | High waterproof safe type tri-proof light |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310061219.1A CN116026528A (en) | 2023-01-14 | 2023-01-14 | High waterproof safe type tri-proof light |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116026528A true CN116026528A (en) | 2023-04-28 |
Family
ID=86079298
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310061219.1A Pending CN116026528A (en) | 2023-01-14 | 2023-01-14 | High waterproof safe type tri-proof light |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116026528A (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019080488A1 (en) * | 2017-10-27 | 2019-05-02 | 东南大学 | Three-dimensional human face recognition method based on multi-scale covariance descriptor and local sensitive riemann kernel sparse classification |
US20210073637A1 (en) * | 2019-08-16 | 2021-03-11 | Leidos, Inc. | Deep Rapid Class Augmentation |
CN114475350A (en) * | 2022-02-14 | 2022-05-13 | 杭州鸽然科技有限公司 | Intelligent charging system and working method thereof |
WO2022160771A1 (en) * | 2021-01-26 | 2022-08-04 | 武汉大学 | Method for classifying hyperspectral images on basis of adaptive multi-scale feature extraction model |
CN114913401A (en) * | 2022-07-13 | 2022-08-16 | 江苏烨明光电有限公司 | Welding equipment for LED lamp core column and shell and welding quality monitoring method thereof |
CN115099684A (en) * | 2022-07-18 | 2022-09-23 | 江西中科冠物联网科技有限公司 | Enterprise safety production management system and management method thereof |
CN115150984A (en) * | 2022-07-14 | 2022-10-04 | 惠州市慧昊光电有限公司 | LED lamp strip and control method thereof |
CN115294639A (en) * | 2022-07-11 | 2022-11-04 | 惠州市慧昊光电有限公司 | Color temperature adjustable lamp strip and control method thereof |
CN115564203A (en) * | 2022-09-23 | 2023-01-03 | 杭州国辰智企科技有限公司 | Equipment real-time performance evaluation system and method based on multi-dimensional data cooperation |
-
2023
- 2023-01-14 CN CN202310061219.1A patent/CN116026528A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019080488A1 (en) * | 2017-10-27 | 2019-05-02 | 东南大学 | Three-dimensional human face recognition method based on multi-scale covariance descriptor and local sensitive riemann kernel sparse classification |
US20210073637A1 (en) * | 2019-08-16 | 2021-03-11 | Leidos, Inc. | Deep Rapid Class Augmentation |
WO2022160771A1 (en) * | 2021-01-26 | 2022-08-04 | 武汉大学 | Method for classifying hyperspectral images on basis of adaptive multi-scale feature extraction model |
CN114475350A (en) * | 2022-02-14 | 2022-05-13 | 杭州鸽然科技有限公司 | Intelligent charging system and working method thereof |
CN115294639A (en) * | 2022-07-11 | 2022-11-04 | 惠州市慧昊光电有限公司 | Color temperature adjustable lamp strip and control method thereof |
CN114913401A (en) * | 2022-07-13 | 2022-08-16 | 江苏烨明光电有限公司 | Welding equipment for LED lamp core column and shell and welding quality monitoring method thereof |
CN115150984A (en) * | 2022-07-14 | 2022-10-04 | 惠州市慧昊光电有限公司 | LED lamp strip and control method thereof |
CN115099684A (en) * | 2022-07-18 | 2022-09-23 | 江西中科冠物联网科技有限公司 | Enterprise safety production management system and management method thereof |
CN115564203A (en) * | 2022-09-23 | 2023-01-03 | 杭州国辰智企科技有限公司 | Equipment real-time performance evaluation system and method based on multi-dimensional data cooperation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6586430B2 (en) | Estimation of vehicle position | |
CN111652227B (en) | Method for detecting damage fault of bottom floor of railway wagon | |
CN112232349A (en) | Model training method, image segmentation method and device | |
Akarsu et al. | A fast and adaptive road defect detection approach using computer vision with real time implementation | |
CN101315701B (en) | Movement destination image partition method | |
CN111275627B (en) | Image snow removing algorithm based on snow model and deep learning fusion | |
CN101286239A (en) | Aerial shooting traffic video frequency vehicle rapid checking method | |
CN113052159B (en) | Image recognition method, device, equipment and computer storage medium | |
CN116485717B (en) | Concrete dam surface crack detection method based on pixel-level deep learning | |
CN112489023A (en) | Pavement crack detection method based on multiple scales and multiple layers | |
CN114119586A (en) | Intelligent detection method for aircraft skin defects based on machine vision | |
CN114821018B (en) | Infrared dim target detection method for constructing convolutional neural network by utilizing multidirectional characteristics | |
CN109815802A (en) | A kind of monitor video vehicle detection and recognition method based on convolutional neural networks | |
CN106709903A (en) | PM2.5 concentration prediction method based on image quality | |
CN109784479A (en) | Restrictor bar neural network based is anti-to pound method, system, equipment and storage medium | |
CN114359838A (en) | Cross-modal pedestrian detection method based on Gaussian cross attention network | |
CN107529647B (en) | Cloud picture cloud amount calculation method based on multilayer unsupervised sparse learning network | |
Ren et al. | Environment influences on uncertainty of object detection for automated driving systems | |
CN113222898B (en) | Double-navigation SAR image trace detection method based on multi-element statistics and deep learning | |
CN111598793A (en) | Method and system for defogging image of power transmission line and storage medium | |
CN116026528A (en) | High waterproof safe type tri-proof light | |
Hodne et al. | Detecting and suppressing marine snow for underwater visual slam | |
Sarafraz et al. | Performance assessment in solving the correspondence problem in underwater stereo imagery | |
Husain et al. | VRHAZE: The simulation of synthetic haze based on visibility range for dehazing method in single image | |
CN114550297B (en) | Pedestrian intention analysis method and 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230428 |