CN110458147A - The liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing - Google Patents
The liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing Download PDFInfo
- Publication number
- CN110458147A CN110458147A CN201910788600.1A CN201910788600A CN110458147A CN 110458147 A CN110458147 A CN 110458147A CN 201910788600 A CN201910788600 A CN 201910788600A CN 110458147 A CN110458147 A CN 110458147A
- Authority
- CN
- China
- Prior art keywords
- image
- artificial intelligence
- pollutants
- disposal
- filter
- 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.)
- Granted
Links
- 239000003344 environmental pollutant Substances 0.000 title claims abstract description 57
- 231100000719 pollutant Toxicity 0.000 title claims abstract description 57
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 34
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012544 monitoring process Methods 0.000 title claims abstract description 26
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 27
- 238000004422 calculation algorithm Methods 0.000 claims description 21
- 239000000126 substance Substances 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 18
- 239000007788 liquid Substances 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 14
- 241000195493 Cryptophyta Species 0.000 claims description 10
- 238000013527 convolutional neural network Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000007620 mathematical function Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000011109 contamination Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 230000000996 additive effect Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 239000011248 coating agent Substances 0.000 claims description 3
- 238000000576 coating method Methods 0.000 claims description 3
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000013500 data storage Methods 0.000 claims description 3
- 230000007547 defect Effects 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims 1
- 239000007789 gas Substances 0.000 description 21
- RAHZWNYVWXNFOC-UHFFFAOYSA-N Sulphur dioxide Chemical compound O=S=O RAHZWNYVWXNFOC-UHFFFAOYSA-N 0.000 description 12
- JCXJVPUVTGWSNB-UHFFFAOYSA-N nitrogen dioxide Inorganic materials O=[N]=O JCXJVPUVTGWSNB-UHFFFAOYSA-N 0.000 description 9
- MGWGWNFMUOTEHG-UHFFFAOYSA-N 4-(3,5-dimethylphenyl)-1,3-thiazol-2-amine Chemical compound CC1=CC(C)=CC(C=2N=C(N)SC=2)=C1 MGWGWNFMUOTEHG-UHFFFAOYSA-N 0.000 description 7
- 238000001514 detection method Methods 0.000 description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 5
- 238000009434 installation Methods 0.000 description 4
- 238000003973 irrigation Methods 0.000 description 4
- 230000002262 irrigation Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000000151 deposition Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000003595 mist Substances 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000009738 saturating Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 241000195628 Chlorophyta Species 0.000 description 1
- 101150042248 Mgmt gene Proteins 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 241000199919 Phaeophyceae Species 0.000 description 1
- 241000206572 Rhodophyta Species 0.000 description 1
- 244000269722 Thea sinensis Species 0.000 description 1
- 239000002253 acid Substances 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000010828 animal waste Substances 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000003337 fertilizer Substances 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013618 particulate matter Substances 0.000 description 1
- 239000000575 pesticide Substances 0.000 description 1
- 239000000049 pigment Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 239000013049 sediment Substances 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 231100000331 toxic Toxicity 0.000 description 1
- 230000002588 toxic effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
- 238000003911 water pollution Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
Abstract
The invention discloses the liquefied gas disposal of pollutants monitoring methods of the non-Application inductor of artificial intelligence cloud computing, industrial digital camera including being configured with microprocessing unit, the industry digital camera is used to capture the image that disposal of pollutants position may occur, the industry digital camera chooses to install in fixed platform, on track mobile platform or flying platform, and industrial digital camera microprocessing unit and global position system are established into connection, system position is recognized when capturing image in order to industrial digital camera, the location information of image that capture may the image of disposal of pollutants and capture by microprocessing unit occur, cloud calculation server is sent to by network, it is analyzed again by the artificial intelligence machine learning software in Cloud Server.Whether having determined that having occurred as a result, will send information on terminal device in real time through internet, terminal device can also read result through internet to artificial intelligence calculator from cloud server.
Description
Technical field
The present invention relates to electronic information technical fields, and in particular to a kind of liquid of the non-Application inductor of artificial intelligence cloud computing
Body other polluting emissions monitoring method.
Background technique
Now, the frequent exhaust emission liquid of people or gas are discharged into private or the public irrigation canals and ditches, water small stream, river, lake
Pool, chimney, open spaces etc., and lead to economic loss and serious harm is caused to environment.For detecting or monitoring this
Usually there are two types of common methods for the prior art of (such as gas and liquid disposal of pollutants) a bit.First method: sensor is penetrated
In hardware and software setting come the chemicals or material that detect, such as the physical gas or liquid of disposal of pollutants.This can be transported
It is completed with various sensors, these sensors can be used to detect with liquid or gas chemicals itself or detection sensor
Some physical parameters variation, such as resistance, capacitor or light reflection are reflected with the presence or absence of disposal of pollutants.Second method:
Through the hardware and software setting in sensor to detect the ginseng of the environment as caused by (such as disposal of pollutants of gas or liquid)
Several some variations.These parameters can be audio sound level, the variation of the vibration of physical resonant vibration machinery or gas or fluid pressure
Variation.
In order to identify gas or liquid disposal of pollutants position, sensor is necessarily mounted at and actually occurs a little relatively
Close distance, for example be mounted on gas or liquid and pollute the position that may be discharged.But this must be along discharge irrigation canals and ditches, water small stream, river
The different location of stream, lake, chimney, open spaces installs big quantity sensor.Or sensor be can along discharge sites week
It is trapped among on platform and moves, so that the range for belonging to the short distance from disposal of pollutants position can be reached.When disposal of pollutants occurs, inspection
The position for measuring the nearest sensor of disposal of pollutants can be used to determine the position of disposal of pollutants.Therefore, big quantity sensor is installed
To identify that the position of disposal of pollutants is very expensive.When being mounted in outdoor environment, from substance, such as it is dirty
The gas or liquid for contaminating discharge, can rapidly be diffused into open-air atmosphere, in addition the concentration of disposal of pollutants substance can reduce rapidly,
Detect that the substance of this low concentration is more difficult than environment indoors.Therefore, in outdoor environment, more sensors to be installed
Or the more expensive sensor sensitive to low concentration, cost can be higher.The position of some discharges, it may be possible to privately owned venue location,
Some may be remote or very inconvenient arrival region, and installation sensor may be not to be allowed to or very expensive.
The step of these sensor hardwares and software are to detect this disposal of pollutants is installed in fact it is very labor intensive,
In addition operator inevitably will appear incorrect operation, it is that these detection sensor systems obtain the result is that will not be correct
's.Due to installing and operating these sensors all unusual labor intensives, in practical situations, the sensor of installation is up to not
It is as a result exactly that detection process is invalid and can not reliably identify practical (such as gas or liquid pollution to enough quantity
Discharge) position.In general, if alarm is sounded, artificial patrol party will be sent to know in the limited situation of number of sensors
The position not occurred.But the people search team of this type, which will lead to, handles such time extra delay.
Further, since the extra cost of real time communication is high, disposal of pollutants testing result generally can not immediately with all management
What the member of team, supervision government official or rescue group shared.Sometimes, local alarm system is only used, government official is supervised
Possibly it can not be notified disposal of pollutants accident in time with the rescue group far from spot, such communication delay will be led again
Cause to handle such time extra delay again.If disposal of pollutants testing result can send all Executive Teams in real time
Member and government official, then entire society will also greatly improve the confidence and satisfaction that handle this kind of disposal of pollutants influence.
Summary of the invention
The object of the present invention is to provide the liquefied gas disposal of pollutants of the non-Application inductor of artificial intelligence cloud computing monitoring
Method avoids and is mounted with a large amount of sensor hardware and software, to detect, eliminates the people search team peace of this type
Row, and reduce the extra delay for handling such accident.Reducing will test result and is communicated to all Mgmt. Team members, supervise
The cost of government official and the rescue group from a distant place, while more shortening the time for sharing result.And these testing results
It can be checked with any mobile device.
In addition, gaseous contamination, which usually has special color relevant to the chemical substance in pollutant, to be discharged by coal combustion
Sulfur dioxide (SO2) and particle when generating acid mist degree in conjunction with the water in mist, form industrial smog, this smog pollution material
Usually look like filemot.Nitrogen dioxide (NO2) is another common pollutant, it carries a kind of toxic dark red orange
Color gas.Liquid and three main sources of water pollution are: agricultural, including chemical fertilizer, animal wastes and other wastes, pesticide etc..
Algae and suspended sediment particle are the rotten organic substances of very common particulate matter, they have common yellow " tea
Color ".Algae carries red, green and brown algae, because wherein containing different types of pigment.Method of the invention will make
The presence of pollutant is identified with these color attributes.
To achieve the above object, the technical solution adopted by the present invention is that:
The liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing, which is characterized in that including
Industrial digital camera configured with microprocessing unit, the industry digital camera are used to capture the image that position may occur, such as
Private or the public irrigation canals and ditches, water small stream, river, lake, chimney, open spaces etc., the industry digital camera choose to install fixed flat
On platform, track mobile platform or flying platform, especially when requiring to be to cover long range and large area in outdoor environment
Monitoring, and industrial digital camera microprocessing unit and global position system are established into connection, in order to the capture of industrial digital camera
Recognize system position when image, the image of disposal of pollutants may be occurred for capture and the image that captures by microprocessing unit
Location information is sent to cloud calculation server by network, then by the artificial intelligence machine learning software in Cloud Server into
Row analysis.
Captured image can be transmitted to microprocessing unit from industrial digital camera, and microprocessing unit again sends these images
Wire/radio network communicator goes to local router/demodulator through wire/radio network local network, and local router/
Image is sent cloud calculation server through internet by demodulator, or due to safety concerns, these image files can also be saturating
It crosses wire/radio network and sends local computing server, all cloud calculates and all completes in local computing device, cloud servo
Image transmission is removed artificial intelligence calculator by device, and in useful data storage beyond the clouds reservoir.
Artificial intelligence machine study specifically prediction modeling, concern is primarily with the errors of minimum model, or enable most
Accurately prediction be possibly realized, intelligence machine study need to pass through a trained step, comprising largely there is no when position
Image, and have location drawing picture when generation, these images should include the image in monitoring region, it needs to include specific gas
And liquid pollutant, the particular chemicals in discharge-channel, such as sulfur dioxide, nitrogen dioxide, deposit and Organic Algae.
In training, training image data set is come from, 75% is extracted and is used as training sample, remaining 25% is used as test specimens
This is used to assessment result precision.When complete this enough image training after, intelligence machine study can according to the new image of input,
Predict what new possibility occurred, training step can be continued in the future to collect more data, such model prediction
The accuracy of defect will also be continuously improved, after giveing training to artificial intelligence machine, when input is just in the new of monitoring area
When image, image can be analyzed and inform whether the region being monitored actually has occurred, when an accident occurs, be
System can send a warning automatically by internet to each side.
Image procossing is the first step for handling new images, in the image new from input, it is possible to specify how to be directed to specific dirt
Dye object (such as sulfur dioxide, nitrogen dioxide and Organic Algae) is monitored, and increases the sensibility to the chemicals, application image
Processing software algorithm enhances specificity to match the pollutant type being look for.When depositing for desired identification particular chemicals
When, if that chemical substance is that band is red, such as when content of nitrogen dioxide is low, image processing algorithm can be used
Enhancing the red in image, the mathematical algorithm of image procossing can pass through i) addition or ii) subtraction solves.Addition algorithm is such as
Under:
Color image is rgb pixels RGB,
Let,
R=img (::, 1) it is red pixel value,
G=img (::, 2) it is green pixel values,
B=img (::, 3) it is blue pixel value,
It includes: r+g+b=gray component, cyan=g+b, magenta=r+b, yellow=red that mathematical algorithm, which can be extended to,
Color+green.
Image procossing addition mathematical function is as follows:
Let,
Extra_red=max (r-max (b, g), 0),
New red pixel value function Fr (r, g, b, k)=r_mod=min (r+k*extra_red, 1)
%k is the scale of additional amount of red.
R_mod is applied to red pixel value and is up to expected additive effect,
Other than addition, mathematical algorithm can be solved by subtraction, for example, when wanting to position another Organic Algae
When substance, that Organic Algae is band green, when pollutant concentration is low, needs to enhance the green in image, passes through subtraction
Non-green component can be reduced.The amount mathematical function for subtracting additional green components is as follows,
Let,
Extra_green=max (g-max (r, b), 0),
New red pixel value function Fr (r, g, b, k)=r_mod=max (r-k*extra_green, 0),
New blue pixel value function Fb (r, g, b, k)=b_mod=max (b-k*extra_green, 0),
R_mod is applied to red pixel value, b_mod is applied to blue pixel value, be up to it is expected subtract effect,
By image procossing, the algorithm that method carries out image contamination discharge identification application is that artificial intelligence machine learns wherein
Branch's convolutional neural networks, convolutional neural networks operation method includes the following:
It is formed altogether by 6 layers of convolutional layer/ReLU/ is down-sampled.
(1) convolution: image is inputted convolutional calculation, then convolution is executed to image, has the matrix of pixel value defeated image
Enter, each image size is 1024*1024 pixel after cutting, and the reading of input matrix is since the upper left corner of image, next
A lesser matrix, referred to as filter are selected there, and then filter generates convolution, i.e., moves with the x of input picture and y-axis
It is dynamic；The task of filter is by its value multiplied by original pixel value, and all these multiplication are all to be added, and finally obtains a number
Word；Since filter only reads image in the upper left corner, it further moves right 1 or N unit, then executes similar behaviour again
Make；After filter passes through all positions, a new matrix is obtained, new matrix size is less than input matrix；First layer
The long * wide of the size of filter is 32*32, and depth 16, step number 2, when crossing the border, with 0 filling, the size of second layer filter is
Long * wide is 32*32, and depth 16, step number 2, with 0 filling, the long * wide of the size of the filter of third layer is 16*16, and depth is
32, step number 1, with 0 filling, the long * wide of size of the 4th layer of filter is 12*12, depth 32, step number 1, with 0 filling,
The long * wide of the size of layer 5 filter is 4*4, depth 64, step number 1, with 0 filling, the long * wide of the size of layer 6 filter
3*3, depth 128, step number 1, with 0 filling.
(2) ReLU is activated: after each convolution algorithm, into linear R eLU active coating, ReLU activation being applied to
Matrix, ReLU represent rectification linear unit, are used for nonlinear operation, it is exported as f (x)=max (0, x) with this equation formula,
The purpose of ReLU be introduced in calculating it is non-linear, because need calculate study be non-negative value, this step result can be one group
Characteristic pattern；
(3) down-sampled: by these characteristic patterns be supplied to it is down-sampled collect calculating, down-sampled also referred to as sub-sampling or under adopt
Sample, it reduces the dimension of matrix, but remains important information, this step executes that data maximums are down-sampled collects calculatings, this
The element with maximum value in interception ReLU activation characteristic pattern is calculated, and is applied to all elements；First four layers down-sampled, maximum drop
It adopts to set to drop and adopts window as (PoolingSize) 4*4, sliding step 4, rear two layers of precision maximum drop adopts setting drop and adopts window
Size 2*2, sliding step 2.
(4) constantly repeat: calculation method is established by 6 layers altogether, if it is desired, reduction adjustment can be increased, convolution, ReLU activation
With down-sampled process, until obtained characteristic pattern shows satisfactory key parameter；
(5) it flattens and is fully connected layer: after repeating enough numbers, characteristic pattern being flattened, the matrix of characteristic pattern, will be turned
It is changed to vector, and is transmitted to be formed and be fully connected layer, finally, output tool softmax activation primitive is fully connected layer, is made
It is returned with SoftMax, the result for propagating neural network forward becomes a probability distribution.Softmax is known as normalizing
Exponential function, expression formula are as follows:
f:
Z1 indicates that node belongs to the 1st class, and zk indicates that node belongs to kth class.
(6) result: activation primitive is applied after being fully connected layer, result is classified as no disposal of pollutants, and different
The disposal of pollutants of type.
Whether artificial intelligence calculator is having occurred as a result, will send information to end in real time through internet having determined that
In end equipment, terminal device can also read result through internet from cloud server.The terminal device can be movement
Phone, tablet computer.
The invention has the advantages that: the present invention to provide the liquefied gas disposal of pollutants of the non-Application inductor of artificial intelligence cloud computing
Monitoring method, solves at present that traditional way installs a large amount of sensor hardware and software, to detect, especially when requiring
It is the place that covering may be discharged with large area over long distances in outdoor environment, detects these disposals of pollutants and need a large amount of sensing
Device and manpower go to install so that detection is effective, and entire sensor detecting system cost is very expensive.When being installed in outdoor environment,
Substance in discharge can rapidly diffuse into open area such as the gas or liquid of disposal of pollutants, and concentration can also decline rapidly,
Detection in environment indoors even than being more difficult.Therefore, in outdoor environment, the more expensive sensing sensitive to low concentration to be installed
Device or the greater number of sensor of installation, cost are therefore higher.The position of some discharges, it may be possible to privately owned venue location, it is some
It may be remote or very inconvenient arrival region, installation sensor may be not to be allowed to or very expensive.
The artificial intelligent cloud of the present invention calculates the liquefied gas disposal of pollutants monitoring method of non-Application inductor, can identify reality
Border, such as gas or liquid disposal of pollutants position, in long range or large area possibility discharge sites, once alarm is sent out
Out, also artificial patrol party is no longer needed to go to determining position.The people search team that system eliminates this type arranges, and reduces
The extra delay of the such accident of processing.
In order to explain the structural features and functions of the invention more clearly, come with reference to the accompanying drawing with specific embodiment to its into
Row is described in detail.
Detailed description of the invention
Fig. 1 show present system structural schematic diagram；
Fig. 2 show convolutional neural networks operation workflow figure of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Refering to attached drawing 1 to Fig. 2, the present embodiment provides the pollutions of the liquefied gas of the non-Application inductor of artificial intelligence cloud computing to arrange
Monitoring method is put, the industrial digital camera including being configured with microprocessing unit, the industry digital camera is used to capture possibility
The image of disposal of pollutants position, such as private or the public irrigation canals and ditches, water small stream, river, lake, chimney, open spaces etc., the work
Industry digital camera chooses to install on fixed platform, track mobile platform or flying platform, it is made to allow to image these regions of capture
The position of more preferable view and will be industrial especially when requiring to be to cover the monitoring of long range and large area in outdoor environment
Digital camera microprocessing unit and global position system (such as Beidou satellite navigation) establish connection, in order to which industrial digital camera is caught
System position, the image that capture may the image of disposal of pollutants and capture by microprocessing unit occur are recognized when obtaining image
Location information, cloud calculation server is sent to by network, then by the artificial intelligence machine learning software in Cloud Server
It is analyzed.
Captured image can be transmitted to microprocessing unit from industrial digital camera, and microprocessing unit again sends these images
Wire/radio network communicator goes to local router/demodulator through wire/radio network local network, and local router/
Image is sent cloud calculation server through internet by demodulator, or due to safety concerns, these image files can also be saturating
It crosses wire/radio network and sends local computing server, all cloud calculates and all completes in local computing device, cloud servo
Image transmission is removed artificial intelligence calculator by device, and in useful data storage beyond the clouds reservoir.
Artificial intelligence machine study specifically prediction modeling, concern is primarily with the errors of minimum model, or enable most
Accurately prediction be possibly realized, intelligence machine study need to pass through a trained step, comprising largely there is no when position
Image, and have location drawing picture when generation, these images should include the image in monitoring region, it needs to include specific gas
And liquid pollutant, the particular chemicals in discharge-channel, such as sulfur dioxide, nitrogen dioxide, deposit and Organic Algae.
In training, training image data set is come from, 75% is extracted and is used as training sample, remaining 25% is used as test specimens
This is used to assessment result precision.When complete this enough image training after, intelligence machine study can according to the new image of input,
Predict what new possibility occurred, training step can be continued in the future to collect more data, such model prediction
The accuracy of defect will also be continuously improved, after giveing training to artificial intelligence machine, when input is just in the new of monitoring area
When image, image can be analyzed and inform whether the region being monitored actually has occurred, when an accident occurs, be
System can send a warning automatically by internet to each side.
Image procossing is the first step for handling new images, in the image new from input, it is possible to specify how to be directed to specific dirt
Dye object (such as sulfur dioxide, nitrogen dioxide and Organic Algae) is monitored, and increases the sensibility to the chemicals, application image
Processing software algorithm enhances specificity to match the pollutant type being look for.When depositing for desired identification particular chemicals
When, such as when content of nitrogen dioxide is low, image processing algorithm can be used to enhance the red in image, at image
The mathematical algorithm of reason can pass through i) addition or ii) subtraction solves.Addition algorithm is as follows:
Color image is rgb pixels RGB,
Let,
R=img (::, 1) it is red pixel value,
G=img (::, 2) it is green pixel values,
B=img (::, 3) it is blue pixel value,
It includes: r+g+b=gray component, cyan=g+b, magenta=r+b, yellow=red that mathematical algorithm, which can be extended to,
Color+green.
Image procossing addition mathematical function is as follows:
Let,
Extra_red=max (r-max (b, g), 0),
New red pixel value function Fr (r, g, b, k)=r_mod=min (r+k*extra_red, 1)
%k is the scale of additional amount of red.
R_mod is applied to red pixel value and is up to expected additive effect,
Other than addition, mathematical algorithm can be solved by subtraction, for example, when wanting to position another Organic Algae
It when substance, needs to enhance the green in image, non-green component can be reduced by subtraction.Subtract the amount number of additional green components
Mathematic(al) function is as follows,
Let,
Extra_green=max (g-max (r, b), 0),
New red pixel value function Fr (r, g, b, k)=r_mod=max (r-k*extra_green, 0),
New blue pixel value function Fb (r, g, b, k)=b_mod=max (b-k*extra_green, 0),
R_mod is applied to red pixel value, b_mod is applied to blue pixel value, be up to it is expected subtract effect,
By image procossing, the algorithm that method carries out image contamination discharge identification application is that artificial intelligence machine learns wherein
Branch's convolutional neural networks, convolutional neural networks operation method includes the following:
It is formed altogether by 6 layers of convolutional layer/ReLU/ is down-sampled.
(1) convolution: image is inputted convolutional calculation, then convolution is executed to image, has the matrix of pixel value defeated image
Enter, each image size is 1024*1024 pixel after cutting, and the reading of input matrix is since the upper left corner of image, next
A lesser matrix, referred to as filter are selected there, and then filter generates convolution, i.e., moves with the x of input picture and y-axis
It is dynamic；The task of filter is by its value multiplied by original pixel value, and all these multiplication are all to be added, and finally obtains a number
Word；Since filter only reads image in the upper left corner, it further moves right 1 or N unit, then executes similar behaviour again
Make；After filter passes through all positions, a new matrix is obtained, new matrix size is less than input matrix；First layer
The long * wide of the size of filter is 32*32, and depth 16, step number 2, when crossing the border, with 0 filling, the size of second layer filter is
Long * wide is 32*32, and depth 16, step number 2, with 0 filling, the long * wide of the size of the filter of third layer is 16*16, and depth is
32, step number 1, with 0 filling, the long * wide of size of the 4th layer of filter is 12*12, depth 32, step number 1, with 0 filling,
The long * wide of the size of layer 5 filter is 4*4, depth 64, step number 1, with 0 filling, the long * wide of the size of layer 6 filter
3*3, depth 128, step number 1, with 0 filling.
(2) ReLU is activated: after each convolution algorithm, into linear R eLU active coating, ReLU activation being applied to
Matrix, ReLU represent rectification linear unit, are used for nonlinear operation, it is exported as f (x)=max (0, x) with this equation formula,
The purpose of ReLU be introduced in calculating it is non-linear, because need calculate study be non-negative value, this step result can be one group
Characteristic pattern；
(3) down-sampled: by these characteristic patterns be supplied to it is down-sampled collect calculating, down-sampled also referred to as sub-sampling or under adopt
Sample, it reduces the dimension of matrix, but remains important information, this step executes that data maximums are down-sampled collects calculatings, this
The element with maximum value in interception ReLU activation characteristic pattern is calculated, and is applied to all elements；First four layers down-sampled, maximum drop
It adopts to set to drop and adopts window as (PoolingSize) 4*4, sliding step 4, rear two layers of precision maximum drop adopts setting drop and adopts window
Size 2*2, sliding step 2.
(4) constantly repeat: calculation method is established by 6 layers altogether, if it is desired, reduction adjustment can be increased, convolution, ReLU activation
With down-sampled process, until obtained characteristic pattern shows satisfactory key parameter；
(5) it flattens and is fully connected layer: after repeating enough numbers, characteristic pattern being flattened, the matrix of characteristic pattern, will be turned
It is changed to vector, and is transmitted to be formed and be fully connected layer, finally, output tool softmax activation primitive is fully connected layer, is made
It is returned with SoftMax, the result for propagating neural network forward becomes a probability distribution.Softmax is known as normalizing
Exponential function, expression formula are as follows:
f:
Z1 indicates that node belongs to the 1st class, and zk indicates that node belongs to kth class.
(6) result: activation primitive is applied after being fully connected layer, result is classified as no disposal of pollutants, and different
The disposal of pollutants of type.
Whether artificial intelligence calculator is having occurred as a result, will send information to end in real time through internet having determined that
In end equipment (such as mobile phone or tablet computer), terminal device can also read knot through internet from cloud server
Fruit.
The above described is only a preferred embodiment of the present invention, being not intended to limit the present invention in any form.Appoint
What those skilled in the art, without departing from the scope of the technical proposal of the invention, all using the side of the disclosure above
Method and technology contents make many possible changes and modifications to technical solution of the present invention, or are revised as the equivalent reality of equivalent variations
Apply example.Therefore anything that does not depart from the technical scheme of the invention, equivalent change made by shape according to the present invention, construction and principle
Change, should all be covered by protection scope of the present invention.
Claims (7)
1. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing, which is characterized in that including matching
It is equipped with the industrial digital camera of microprocessing unit, which is used to capture the image that position may occur, the industry
Digital camera chooses to install on fixed platform, track mobile platform or flying platform, especially when require being covered in outdoor environment
Lid long range and large area monitoring, and industrial digital camera microprocessing unit and global position system are established into connection, with
System position is recognized when capturing image convenient for industrial digital camera, the figure of disposal of pollutants may be occurred for capture by microprocessing unit
The location information of picture and the image captured, is sent to cloud calculation server by network, then by the people in Cloud Server
Work intelligence machine learning software is analyzed.
2. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing according to claim 1,
It is characterized in that, captured image can be transmitted to microprocessing unit from industrial digital camera, microprocessing unit is again by these images
Wire/radio network communicator is sent, goes to local router/demodulator, local routing through wire/radio network local network
Image is sent cloud calculation server through internet by device/demodulator, or due to safety concerns, these image files can also be with
Local computing server is sent through wire/radio network, all cloud calculates and all completes in local computing device, and cloud is watched
It takes device and artificial intelligence calculator is removed in image transmission, and in useful data storage beyond the clouds reservoir.
3. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing according to claim 1,
It is characterized in that, artificial intelligence machine study specifically prediction modeling, concern is primarily with minimize the error of model, or order
Most accurately prediction be possibly realized, intelligence machine study need to pass through a trained step, comprising largely there is no when position
Image is set, these images should include the image in monitoring region, it needs to include specific gas and liquid pollutant；In training
In, training image data set is come from, 75% is extracted and is used as training sample, remaining 25% is used as test sample to be used to assessment result
Precision；After completing the training of this enough image, intelligence machine study can predict new possibility hair according to the new image of input
Raw, training step can be continued in the future to collect more data, the accuracy of such model prediction defect
It will be continuously improved, after being giveed training to artificial intelligence machine, when inputting the new images just in monitoring area, can analyze
Image simultaneously informs whether the region being monitored actually has occurred, and when an accident occurs, system can be automatically by mutual
It networks and sends a warning to each side.
4. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing according to claim 1,
It is characterized in that, image procossing is the first step for handling new images, in the image new from input, it is possible to specify how for specific
Pollutant is monitored, increase to the sensibility of the chemicals, and application image processing software algorithm enhances specificity to match
The pollutant type being look for, when want identification particular chemicals there are when pollutant concentration is low, can be with
Enhance the red in image using image processing algorithm, the mathematical algorithm of image procossing can pass through i) addition or ii) subtraction
It solves, addition algorithm is as follows:
Color image is rgb pixels RGB,
Let,
R=img (::, 1) it is red pixel value,
G=img (::, 2) it is green pixel values,
B=img (::, 3) it is blue pixel value,
Mathematical algorithm can be include: r+g+b=gray component, cyan=g+b, magenta=r+b, yellow=red+green,
Image procossing addition mathematical function is as follows:
Let,
Extra_red=max (r-max (b, g), 0),
New red pixel value function Fr (r, g, b, k)=r_mod=min (r+k*extra_red, 1)
%k is the scale of additional amount of red,
R_mod is applied to red pixel value and is up to expected additive effect；
Other than addition, mathematical algorithm can be solved by subtraction, when wanting to position another Organic Algae substance, be needed
Enhance the green in image, non-green component can be reduced by subtraction, subtracts the amount mathematical function of additional green components such as
Under,
Let,
Extra_green=max (g-max (r, b), 0),
New red pixel value function Fr (r, g, b, k)=r_mod=max (r-k*extra_green, 0),
New blue pixel value function Fb (r, g, b, k)=b_mod=max (b-k*extra_green, 0),
R_mod is applied to red pixel value, b_mod is applied to blue pixel value, is up to and expected subtracts effect.
5. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing according to claim 4,
It is characterized in that, the algorithm that method carries out image contamination discharge identification application is artificial intelligence machine study by image procossing
One of branch's convolutional neural networks, convolutional neural networks operation method include the following:
(1) convolution: image is inputted convolutional calculation, then convolution is executed to image, image is had to the Input matrix of pixel value, is cut out
Cutting rear each image size is 1024*1024 pixel, and the reading of input matrix is since the upper left corner of image, next there
A lesser matrix, referred to as filter are selected, then filter generates convolution, i.e., mobile with the x of input picture and y-axis；It crosses
The task of filter is by its value multiplied by original pixel value, and all these multiplication are all to be added, and finally obtains a number；Due to
Filter only reads image in the upper left corner, it further moves right 1 or N unit, then executes similar operation again；It has served as
After filter passes through all positions, a new matrix is obtained, new matrix size is less than input matrix；First layer filter
The long * wide of size is 32*32, and depth 16, step number 2, when crossing the border, with 0 filling, the size of second layer filter is that long * wide is
32*32, depth 16, step number 2, with 0 filling, the long * wide of the size of the filter of third layer is 16*16, depth 32, step number
It is 1, with 0 filling, the long * wide of size of the 4th layer of filter is 12*12, depth 32, step number 1, with 0 filling, layer 5
The long * wide of the size of filter is 4*4, and depth 64, step number 1, with 0 filling, the long * wide 3*3 of the size of layer 6 filter is deep
Degree is 128, step number 1, with 0 filling；
(2) ReLU is activated: after each convolution algorithm, into linear R eLU active coating, ReLU activation is applied to matrix,
ReLU represents rectification linear unit, is used for nonlinear operation, it is exported as f (x)=max (0, x), ReLU with this equation formula
Purpose be introduced in calculating it is non-linear, because need calculate study be non-negative value, this step result can be one group of feature
Figure；
(3) down-sampled: by these characteristic patterns be supplied to it is down-sampled collect calculating, down-sampled also referred to as sub-sampling or down-sampling, it
The dimension of matrix is reduced, but remains important information, this step executes that data maximums are down-sampled collects calculatings, the calculating section
The element with maximum value in ReLU activation characteristic pattern is taken, and is applied to all elements；First four layers down-sampled, and maximum drop adopts setting
It is (Pooling Size) 4*4 that drop, which adopts window, and sliding step 4, rear two layers of precision maximum drop, which is adopted, sets the size that drop adopts window
2*2, sliding step 2；
(4) constantly repeat: calculation method is established by 6 layers altogether, is adjusted if it is desired, can increase reduction, convolution, ReLU activation and drop
The process of sampling, until obtained characteristic pattern shows satisfactory key parameter；
(5) it flattens and is fully connected layer: after repeating enough numbers, characteristic pattern being flattened, the matrix of characteristic pattern is converted into
Vector, and transmitted to be formed and be fully connected layer, finally, output tool softmax activation primitive is fully connected layer, use
SoftMax is returned, and the result for propagating neural network forward becomes a probability distribution, and Softmax, which is known as normalizing, to be referred to
Number function, expression formula are as follows:
f:
Z1 indicates that node belongs to the 1st class, and zk indicates that node belongs to kth class；
(6) result: activation primitive is applied after being fully connected layer, and result is classified as no disposal of pollutants and different type
Disposal of pollutants.
6. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing according to claim 1,
It is characterized in that, whether artificial intelligence calculator is having determined that having occurred as a result, in real time sending information through internet
Onto terminal device, terminal device can also read result through internet from cloud server.
7. the liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing according to claim 6,
It is characterized in that, the terminal device can be mobile phone, tablet computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910788600.1A CN110458147B (en) | 2019-08-26 | 2019-08-26 | Liquid gas pollution emission monitoring method of artificial intelligent cloud computing non-application sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910788600.1A CN110458147B (en) | 2019-08-26 | 2019-08-26 | Liquid gas pollution emission monitoring method of artificial intelligent cloud computing non-application sensor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110458147A true CN110458147A (en) | 2019-11-15 |
| CN110458147B CN110458147B (en) | 2024-01-09 |
Family
ID=68488994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910788600.1A Active CN110458147B (en) | 2019-08-26 | 2019-08-26 | Liquid gas pollution emission monitoring method of artificial intelligent cloud computing non-application sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110458147B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113311119A (en) * | 2021-07-28 | 2021-08-27 | 深圳市图元科技有限公司 | Gas source tracking method, device and system |
| CN114646126A (en) * | 2020-12-21 | 2022-06-21 | 研能科技股份有限公司 | Indoor gas pollution filtering method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203265A (en) * | 2016-06-28 | 2016-12-07 | 江苏大学 | A kind of Construction Fugitive Dust Pollution based on unmanned plane collection image is derived from dynamic monitoring and coverage prognoses system and method |
| CN107798381A (en) * | 2017-11-13 | 2018-03-13 | 河海大学 | A kind of image-recognizing method based on convolutional neural networks |
| CN108038450A (en) * | 2017-12-14 | 2018-05-15 | 海安常州大学高新技术研发中心 | Marine pollution object detecting method based on unmanned plane and image recognition |
| CN108931970A (en) * | 2018-08-09 | 2018-12-04 | 湖南湘牛环保实业有限公司 | A kind of industrial park wisdom environment supervisory systems based on cloud platform technology |
| KR101933856B1 (en) * | 2017-07-03 | 2018-12-31 | (주)시정 | System for image processing using the convolutional neural network and method using the same |
| CN110414478A (en) * | 2019-08-08 | 2019-11-05 | 东莞德福得精密五金制品有限公司 | The contingency liquefied gas leak supervision method of the non-Application inductor of artificial intelligence cloud computing |
-
2019
- 2019-08-26 CN CN201910788600.1A patent/CN110458147B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106203265A (en) * | 2016-06-28 | 2016-12-07 | 江苏大学 | A kind of Construction Fugitive Dust Pollution based on unmanned plane collection image is derived from dynamic monitoring and coverage prognoses system and method |
| KR101933856B1 (en) * | 2017-07-03 | 2018-12-31 | (주)시정 | System for image processing using the convolutional neural network and method using the same |
| CN107798381A (en) * | 2017-11-13 | 2018-03-13 | 河海大学 | A kind of image-recognizing method based on convolutional neural networks |
| CN108038450A (en) * | 2017-12-14 | 2018-05-15 | 海安常州大学高新技术研发中心 | Marine pollution object detecting method based on unmanned plane and image recognition |
| CN108931970A (en) * | 2018-08-09 | 2018-12-04 | 湖南湘牛环保实业有限公司 | A kind of industrial park wisdom environment supervisory systems based on cloud platform technology |
| CN110414478A (en) * | 2019-08-08 | 2019-11-05 | 东莞德福得精密五金制品有限公司 | The contingency liquefied gas leak supervision method of the non-Application inductor of artificial intelligence cloud computing |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114646126A (en) * | 2020-12-21 | 2022-06-21 | 研能科技股份有限公司 | Indoor gas pollution filtering method |
| CN113311119A (en) * | 2021-07-28 | 2021-08-27 | 深圳市图元科技有限公司 | Gas source tracking method, device and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110458147B (en) | 2024-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414478A (en) | The contingency liquefied gas leak supervision method of the non-Application inductor of artificial intelligence cloud computing | |
| Biber | The problem of environmental monitoring | |
| CN108956484B (en) | Method and device for integrally tracking pollution source | |
| Bell et al. | Evaluation of next generation emission measurement technologies under repeatable test protocols | |
| CN110458147A (en) | The liquefied gas disposal of pollutants monitoring method of the non-Application inductor of artificial intelligence cloud computing | |
| CN110320163A (en) | Online water quality monitoring method, system and the storage medium of unmanned plane | |
| Zhang et al. | Novel method for industrial sewage outfall detection: Water pollution monitoring based on web crawler and remote sensing interpretation techniques | |
| CN116361737B (en) | Lake abnormity dynamic monitoring method and device, electronic equipment and storage medium | |
| Højberg et al. | Use of models to support the monitoring requirements in the water framework directive | |
| CN115436573A (en) | Intelligent monitoring method and device for atmospheric pollution source | |
| Imen et al. | Developing a model-based drinking water decision support system featuring remote sensing and fast learning techniques | |
| CN116148188A (en) | Air-space-ground integrated lake water quality tracing method, system, equipment and storage medium | |
| Jehn et al. | Incremental model breakdown to assess the multi-hypotheses problem | |
| CN116659568A (en) | On-spot environmental monitoring system of boats and ships application workshop | |
| CN110988286B (en) | Intelligent water resource long-term detection system | |
| CN115665690A (en) | River buffer zone soil restoration feedback system and restoration method | |
| Walton et al. | The need for environmental monitoring in Antarctica: baselines, environmental impact assessments, accidents and footprints | |
| Shahriar et al. | Spatial-temporal prediction of algal bloom | |
| Halabisky et al. | The wetland intrinsic potential tool: mapping wetland intrinsic potential through machine learning of multi-scale remote sensing proxies of wetland indicators | |
| CN113466135A (en) | GEE-based seawater quality online inversion monitoring method | |
| Brando et al. | Reef Rescue Marine Monitoring Program: Using Remote Sensing for GBR Wide Water Quality: Final Report for 2008/09 Activities | |
| Halabisky et al. | The Wetland Intrinsic Potential tool: Mapping wetland intrinsic potential through machine learning of multi-scale remote sensing proxies of wetland indicators | |
| CN112710623A (en) | Method and equipment for remotely sensing and monitoring diffusion range and concentration of toxic and harmful gas | |
| Jang et al. | Spatiotemporal algal bloom prediction of geum river, Korea using the deep learning models in company with the EFDC model | |
| KR20210088133A (en) | analyzing method for integrated environmental pollution |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |