CN109766780A - A kind of ship smog emission on-line checking and method for tracing based on deep learning - Google Patents
A kind of ship smog emission on-line checking and method for tracing based on deep learning Download PDFInfo
- Publication number
- CN109766780A CN109766780A CN201811561334.0A CN201811561334A CN109766780A CN 109766780 A CN109766780 A CN 109766780A CN 201811561334 A CN201811561334 A CN 201811561334A CN 109766780 A CN109766780 A CN 109766780A
- Authority
- CN
- China
- Prior art keywords
- smog
- model
- training
- detection
- ship
- 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
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 73
- 238000012549 training Methods 0.000 claims abstract description 48
- 230000008569 process Effects 0.000 claims abstract description 12
- 239000000779 smoke Substances 0.000 claims abstract description 11
- 238000004321 preservation Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 claims description 2
- 230000003252 repetitive effect Effects 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 description 8
- 239000000523 sample Substances 0.000 description 8
- 238000013527 convolutional neural network Methods 0.000 description 5
- 239000003643 water by type Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000003915 air pollution Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000003921 oil Substances 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- NINIDFKCEFEMDL-UHFFFAOYSA-N Sulfur Chemical compound [S] NINIDFKCEFEMDL-UHFFFAOYSA-N 0.000 description 1
- 239000000443 aerosol Substances 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 239000002283 diesel fuel Substances 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 239000000295 fuel oil Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 239000000700 radioactive tracer Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000011593 sulfur Substances 0.000 description 1
- 229910052717 sulfur Inorganic materials 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Abstract
The invention discloses a kind of ship smog emission on-line checking and method for tracing based on deep learning, mainly include two parts of Smoke Detection and tracking, the ship smog video data in the Changjiang river Wuhan basin and PORT OF SHANZHEN acquisition is separately converted to the standard data set format for training detection model and tracing model first.Secondly it uses TensorFlow to construct detection network and tracking network respectively and the hyper parameters such as learning rate, batch sample number, training the number of iterations is set, then load training dataset and start to train, training is detected after the completion and tracing model.It recently enters a video clip or Online Video flows in detection model, when detecting in some coordinates regional in a certain frame there are when smog, stops detection process and detect that the frame of smog is set as being input to that tracing model is tracked the initial position message of process and template image is tracked with this coordinate information and for the first time.Compared with prior art, the beneficial effects of the invention are as follows low cost and Gao Shixiao.
Description
Technical field
The invention belongs to intelligent green technical field of transportation, it is related to a kind of ship smog emission on-line checking and tracking side
A kind of method, and in particular to ship smog vision-based detection and method for tracing based on deep learning.
Technical background
With the continuous growth of economic fast development and volume of trade, China's cargo transport on water amount and the traffic of a port connect
Continuous many years are sure to occupy the first in the world.Busy water transport and sea-freight also brings air pollution while bringing economic benefit to increase,
But waters is generally off-site from downtown area, bring air pollution is easy to be ignored by people.In fact, in Jing-jin-ji region, length
Triangle, Pearl River Delta and coastal riparian area, ship smog emission have become one of important sources of atmosphere pollution.Ship is used
Fuel be mainly residual oil or heavy oil, they belong to diesel oil, and sulfur content is 100 to 3500 times of automobile oil, therefore in navigation channel
It inside can be appreciated that the ship with " black tail ".
In recent years, the improvement of each bound pair atmosphere pollution of Chinese society has higher attention rate, and the Chinese government is also ecology
Environmental protection rises to the height of national strategy.On December 15th, 2015, in Department of Transportation, the People's Republic of China (PRC) the 25th time
" People's Republic of China (PRC) prevents and treats Pollution From Ships inland waters environmental regulations " is passed through in meeting, the appearance of this regulation makes
Problem of environmental pollution caused by ship discharge smog brings effective improvement.One good regulation is no doubt important, but section
It learns efficient supervision and is just important guarantee.
The method of ship smog emission detection at present is broadly divided into following several:
(1) full artificial vision's observation method;
So-called full artificial vision's detection, i.e., not against any software systems and hardware device, according only to water bodies management portion
Scene experience when staff in door is to the cognition of the visual characteristic of smog appearance and ship discharge smog judges waters
In ship whether have discharge smog behavior.From the angle analysis of technical method, this method is not related to technological difficulties, only
Need to select the Observation personnel for having certain scientific literacy.And from the angle analysis of economic cost, although this method is not
Purchase system and hardware device resources are needed, but a large amount of cost of human resources can be paid.In addition, the disadvantage of this method maximum
It is excessively to rely on the standard of artificial experience, because everyone experience of life and human-subject test have this limitation of difference,
So the case where missing inspection or erroneous detection, happens occasionally.
(2) sensing equipment probe method;
So-called sensing equipment detection, i.e., by the side that the smoke sensor device of monitoring water area laying certain scale is detected
Formula.Sensitivity of this method under normal weather environment is higher, however if encountering the weather of high wind, it just will appear detection letter
Number unstable problem.In addition to this, this method is affected by smokescope, when the concentration of smog reaches a threshold value
When, the phenomenon that sensor can just generate response concurrent and go out alarm signal, and this inflexible detection feature easilys lead to missing inspection
Occur.Compared with full artificial vision's observation method, this method saves the expense of a large amount of human cost really, but if monitoring
Waters range it is larger, then a large amount of sensing equipment resource can be expended, so sensing equipment probe method is more suitable for emphasis waters
Small range detect scene.
(3) Machine Vision Detection method;
Machine Vision Detection method is a kind of detection method towards video image, is specifically calculated using computer vision
Method carries out the technological means of smog emission detection to the ship occurred in monitor video.Such method and full artificial vision's observation method
Compared to more efficient, the influence degree compared with sensing equipment probe method by weather conditions and smokescope is relatively small.Although
This method is more relatively reliable than first two method, but there are also very big rooms for promotion for accuracy.Therefore in some industrial systems,
Have to then send the video image marked to monitoring personnel confirms using by video monitoring automatic early-warning
Working mechanism.Generally speaking, the key point of such method is how accurately to extract the feature of smog.Because of the feature of smog
Type is relative complex, so if the method for taking manual construction feature, so spending in the time cost in Feature Engineering stage
Relatively high, furthermore again because manual construction feature is limited to the limitation of artificial experience, it is difficult to extract arrive reflection smog
More essential further feature information, and then influence the effect of Smoke Detection.Currently, carrying out fire hazard aerosol fog detection using such method
Application case it is more, mainly based on forest fire, and apply ship smog emission detect also in jejune starting
Stage.
In conclusion the method for ship smog emission detection at present still excessively relies on the branch of manual work and hardware device
It holds, under the intelligentized epoch tide of application, intelligent green traffic becomes important developing direction and industrial application requirement.It is deep
Degree study is that one of the important technical of development intelligent use is adopted to free people from many and diverse Feature Engineering
It more allows model to be automatically performed the work of latent structure and feature selecting with the method for deep learning, Feature Engineering can be saved
Huge investment.The concept of deep learning is derived from the research of artificial neural network, and the extensive use in vision is mainly benefited
In the rapid development of convolutional neural networks.In order to improve the efficiency of ship Smoke Detection, the powerful characterology of deep learning is utilized
Habit ability, and provide and the deep learning model of monitoring ship smog emission task is suitble to be necessary, show with important
Sincere justice and learning value.
Summary of the invention
In order to overcome the shortcomings of above-mentioned prior art, the present invention provides a kind of ship smog based on deep learning
Discharge on-line checking and method for tracing.
The technical scheme adopted by the invention is that: a kind of ship smog emission on-line checking and tracking based on deep learning
Method, which comprises the following steps:
Step 1: the ship smog video data of the ship smog video data of two different zones of acquisition is examined as training
Survey the standard data set of model criteria data set and tracing model;
Step 2: related hyper parameter, including learning rate, batch sample is arranged in building detection network and tracking network respectively
Number, training the number of iterations;
Step 3: the standard data set of training detection model standard data set and tracing model is directed respectively into detection network
It is trained with tracking network, detection model and tracing model is obtained after the completion of training;
Step 4: one video clip of input or Online Video flow in detection model, when detecting certain in a certain frame
There are when smog in a coordinates regional, stops detection process and detect that the frame of smog is set as defeated with this coordinate information and for the first time
Enter to tracing model to be tracked the initial position message and template image of process, and according to this position to the smog of subsequent frame into
Row tracking;
Step 5: saving the smog coordinate information exported in tracing process, and use each frame in rectangle mark tracking sequence
In location of smoke.
Compared with prior art, the beneficial effects of the invention are as follows low cost and Gao Shixiao.Specifically, with full artificial vision
Monitoring method is compared, and not only saves the manually-operated time, and advanced scientific method and technological means make monitoring result more
Add and has science;Compared with sensing equipment probe method, the economic cost of extensive laying hardware device has not only been saved, but also
Because in training data including the ship smog image under a variety of environment, model is in the stability in dynamic environment
It is higher;Compared with Machine Vision Detection method, the program has been selected according to the high feature of the feature complexity of ship smog in depth
It is suitble to the depth convolutional neural networks of processing image data as Feature Selection Model in study, not only considerably reduces and passing
The complexity of Feature Engineering in the Machine Vision Detection method of system, and because of its powerful independent learning ability, it can extract
The shallow-layer visual characteristic that human eye can perceive, and can be easier to extract the Deep Semantics feature of reflection smog essence.
Detailed description of the invention
Fig. 1 is the ship smog emission video database contracting that the embodiment of the present invention is acquired in the Changjiang river Wuhan basin and PORT OF SHANZHEN
Sketch map;
Fig. 2 is that the embodiment of the present invention constructs and training detects network and tracking network flow chart;
Fig. 3 is that the ship smog emission video clip progress example that the embodiment of the present invention was shot on May 30th, 2015 is said
Bright flow diagram.
Specific embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, with reference to the accompanying drawings and embodiments to this hair
It is bright to be described in further detail, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, not
For limiting the present invention.
The present embodiment is using ship's navigation video as application scenarios, using target detection and tracer technique based on deep learning
Feature extraction and prediction are carried out to the ship smog in video.It is provided by the invention a kind of based on deep see Fig. 1, Fig. 2 and Fig. 3
Spend the ship smog emission on-line checking and method for tracing of study, comprising the following steps:
Step 1: the ship smog video data of the ship smog video data of acquisition the Changjiang river Wuhan basin and PORT OF SHANZHEN
Ship smog video data, the standard data set as training detection model standard data set and tracing model;
Referring to Fig.1, the ship smog emission video database acquired for the present embodiment in the Changjiang river Wuhan basin and PORT OF SHANZHEN
Thumbnail, the ship's navigation view that the database purchase in May, 2015 shoots in the Changjiang river Wuhan basin and in June, 2018 in PORT OF SHANZHEN
There is the video clip of smog emission behavior in frequency, these video clips are both used to make detection and tracing model involved in the present invention
Training dataset, be used also as the local test video of Fig. 2 and Fig. 3 detection model input terminal.
Step 2: constructing detection network and tracking network respectively using TensorFlow, related hyper parameter is set, including learns
Habit rate, batch sample number, training the number of iterations;
Step 3: the standard data set of training detection model standard data set and tracing model is directed respectively into detection network
It is trained with tracking network, detection model and tracing model is obtained after the completion of training;
Training detection model standard data set is wherein imported detection network to be trained, detection mould is obtained after the completion of training
Type, specific implementation include following sub-step:
Step 3A.1: initial location of smoke of the building based on YOLO3 detects network model;
Step 3A.2: design a model Training strategy, i.e., the initial value of learning rate is 1e-4, and batch sample number is 16, training
The number of iterations is 100;
Step 3A.3: load training pattern is finely adjusted training, the i.e. power of load YOLO3 of training on COCO data set
Value parameter starts to be trained to carry out model initialization according to the Training strategy in step 3A.2, saves after the completion of training
The weighting parameter of model.
The standard data set of training tracing model is imported tracking network to be trained, tracking mould is obtained after the completion of training
Type, specific implementation include following sub-step:
Step 3B.1: smog tracking network model of the building based on the twin network of full convolution;
Step 3B.2: the training hyper parameter of setting tracking network model, the i.e. initial value of learning rate are 1e-2, batch sample
Number is 32, and training the number of iterations is 50;
Step 3B.3: being arranged according to the hyper parameter in step 3B.2, completes the weighting parameter of repetitive exercise and preservation model.
As shown in Fig. 2, left side block diagram is the input terminal of ship's navigation video, it is responsible for the loading of ship's navigation video and locates in advance
Reason, ship's navigation video both can choose the local video file in Fig. 1, be also possible to carry the online view of video monitoring equipment
Frequency flow data.
As shown in Fig. 2, right side block diagram is ship smog emission prediction model, it is divided into detection submodel (left side) and tracking submodule
Type (right side).In this step, first detection submodel is illustrated, tracking submodel continues in a subsequent step
It is bright.The task of detection submodel is to receive ship's navigation video data, and get ship discharge smog in video simultaneously and go out
Existing initial position message.Because smog initial position be track submodel main input data, it requires have it is higher
Accuracy in detection.In order to guarantee the accuracy of smog initial position message, using the detection framework based on YOLO3.YOLO3 is
The best mainstream detection framework of comprehensive performance at present, its inspection accuracy on common data sets VOC2007 and COCO are higher than
YOLO2 and SSD.However, training the general of obtained model on COCO because smog is different from general " rigid objects "
Change it is indifferent, so need be finely adjusted training on the fine granularity ship smog label data collection made by Fig. 1, thus
To the detection model of suitable ship smog.While the testing requirements in order to guarantee real-time, characteristic is detected retaining part YOLO3
On the basis of, the feature extraction network of YOLO3 is simplified, the convolutional layer group that convolution kernel size is 3*3 and 1*1 is dramatically reduced
It closes, finally obtains the lightweight detection model for meeting requirement of real-time.
Above-mentioned is the training stage of the model, and using the model carry out it is actually detected when do not need to each frame all into
Row detection, it is only necessary to which there are when smog when detecting a certain frame, so that it may stop the detection of subsequent frame and save present frame
Smog initial position message.The method that statistics is consecutively detected the frame number of smog can be used in a possibility that in order to reduce erroneous detection, and
Statistical data is compared with preset threshold value, if it is greater than the threshold value, then with the smog detected for the last time
Otherwise position continues the detection of subsequent frame as smog initial position message.
As shown in Fig. 2, the right side block diagram in detection submodel is tracking submodel.The task of the model is to receive detection
After the smog initial position message of submodel output, dynamic tracing is carried out to the position of smog in each frame.For in ship's navigation
In video, in the position at each moment different variations is occurring for the smog of ship discharge, so considering video counts
Dynamic variation characteristic according to middle smog and the more stable particularity of ship smog dynamic change in inland navigation craft navigation scene
On the basis of, not only need for single-frame images carry out airspace on detection, it should also be taken into account that the two field pictures of different moments it
Between sequential relationship, i.e., the position of smog is predicted by the similarity relationships between two field pictures.
The complete twin network of convolution is a kind of tracking network for learning temporal aspect between template image and image to be searched,
The extraction that convolutional neural networks carry out spatial signature information is used simultaneously for single-frame images.For tracking task, mould
Type has important influence to final tracking effect, can pass through modification network structure, adjusting training strategy and optimization data set
Mode solves.For tracing problem, tracking speed affects the effect of tracking, and influences the major reason of tracking speed
One of be exactly single-frame images convolution feature extraction network.In the original twin network of full convolution, for extracting single-frame images
Convolutional neural networks be AlexNet, compared to the lightweight convolutional neural networks of some mainstreams, it is that a parameter amount is larger
Network structure.In order to further enhance tracking speed, the lightweights convolutional Neural such as MobileNet, ShuffleNet can be used
Network substitutes AlexNet to reduce the time of propagated forward process, so that the detection speed of tracking submodel and detection submodel
It is roughly the same, guarantee the requirement for meeting real-time of entire detection and tracing model.
Step 4: one video clip of input or Online Video flow in detection model, when detecting certain in a certain frame
There are when smog in a coordinates regional, stops detection process and detect that the frame of smog is set as defeated with this coordinate information and for the first time
Enter to tracing model to be tracked the initial position message and template image of process, and according to this position to the smog of subsequent frame into
Row tracking;
The method that statistics is consecutively detected the frame number of smog can be used in a possibility that in order to reduce erroneous detection, and by statistical number
Be compared according to preset threshold value, if it is greater than the threshold value, then using the location of smoke that detects for the last time as
Otherwise smog initial position message continues the detection of subsequent frame.
Step 5: saving the smog coordinate information exported in tracing process, and use each frame in rectangle mark tracking sequence
In location of smoke.
It should be understood that the part that this specification does not elaborate belongs to the prior art.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Benefit requires to make replacement or deformation under protected ambit, fall within the scope of protection of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (5)
1. a kind of ship smog emission on-line checking and method for tracing based on deep learning, which is characterized in that including following step
It is rapid:
Step 1: the ship smog video data of two different zones of acquisition, as training detection model standard data set and tracking
The standard data set of model;
Step 2: related hyper parameter, including learning rate, batch sample number, instruction is arranged in building detection network and tracking network respectively
Practice the number of iterations;
Step 3: the standard data set of training detection model standard data set and tracing model being directed respectively into detection network and is chased after
Track network is trained, and detection model and tracing model are obtained after the completion of training;
Step 4: input one video clip or Online Video flow in detection model, when detect in a certain frame some seat
It marks in region there are when smog, stop detection process and detects that the frame of smog is set as being input to this coordinate information and for the first time
Tracing model is tracked the initial position message and template image of process, and is chased after according to smog of this position to subsequent frame
Track;
Step 5: saving the smog coordinate information exported in tracing process, and using in each frame in rectangle mark tracking sequence
Location of smoke.
2. the ship smog emission on-line checking and method for tracing according to claim 1 based on deep learning, feature
It is: in step 2, constructs detection network and tracking network respectively using TensorFlow.
3. the ship smog emission on-line checking and method for tracing according to claim 1 based on deep learning, feature
It is: training detection model standard data set is imported into detection network described in step 3 and is trained, is examined after the completion of training
Model is surveyed, specific implementation includes following sub-step:
Step 3A.1: initial location of smoke of the building based on YOLO3 detects network model;
Step 3A.2: design a model Training strategy, and the initial value that learning rate is arranged is 1e-4, and batch sample number is 16, and training changes
Generation number is 100;
Step 3A.3: load training pattern is finely adjusted training;Load the weighting parameter of the YOLO3 of training on COCO data set
To carry out model initialization, and start to be trained according to the Training strategy in step 3A.2, preservation model after the completion of training
Weighting parameter.
4. the ship smog emission on-line checking and method for tracing according to claim 1 based on deep learning, feature
It is: the standard data set of training tracing model is imported into tracking network described in step 3 and is trained, is obtained after the completion of training
Tracing model, specific implementation include following sub-step:
Step 3B.1: smog tracking network model of the building based on the twin network of full convolution;
Step 3B.2: the training hyper parameter of setting tracking network model, the initial value that learning rate is arranged is 1e-2, batch sample number
It is 32, training the number of iterations is 50;
Step 3B.3: being arranged according to the hyper parameter in step 3B.2, completes the weighting parameter of repetitive exercise and preservation model.
5. the ship smog emission on-line checking according to any one of claims 1-4 based on deep learning and tracking side
Method, it is characterised in that: in step 4, a possibility that in order to reduce erroneous detection, the side of the frame number of smog is consecutively detected using statistics
Method, and statistical data is compared with preset threshold value, if it is greater than the threshold value, then to detect for the last time
Otherwise location of smoke continues the detection of subsequent frame as smog initial position message.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811561334.0A CN109766780A (en) | 2018-12-20 | 2018-12-20 | A kind of ship smog emission on-line checking and method for tracing based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811561334.0A CN109766780A (en) | 2018-12-20 | 2018-12-20 | A kind of ship smog emission on-line checking and method for tracing based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109766780A true CN109766780A (en) | 2019-05-17 |
Family
ID=66451998
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811561334.0A Pending CN109766780A (en) | 2018-12-20 | 2018-12-20 | A kind of ship smog emission on-line checking and method for tracing based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109766780A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110309729A (en) * | 2019-06-12 | 2019-10-08 | 武汉科技大学 | Tracking and re-detection method based on anomaly peak detection and twin network |
CN111914935A (en) * | 2020-08-03 | 2020-11-10 | 哈尔滨工程大学 | Ship image target detection method based on deep learning |
CN113592842A (en) * | 2021-08-09 | 2021-11-02 | 南方医科大学南方医院 | Sample serum quality identification method and identification device based on deep learning |
Citations (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001273499A (en) * | 2000-03-27 | 2001-10-05 | Nagoya Electric Works Co Ltd | Method and device for detecting moving object |
CN102496275A (en) * | 2011-11-25 | 2012-06-13 | 大连海创高科信息技术有限公司 | Method for detecting overload of coach or not |
CN102610104A (en) * | 2012-03-16 | 2012-07-25 | 南京航空航天大学 | Onboard front vehicle detection method |
CN104318258A (en) * | 2014-09-29 | 2015-01-28 | 南京邮电大学 | Time domain fuzzy and kalman filter-based lane detection method |
CN106251861A (en) * | 2016-08-05 | 2016-12-21 | 重庆大学 | A kind of abnormal sound in public places detection method based on scene modeling |
CN107025443A (en) * | 2017-04-06 | 2017-08-08 | 江南大学 | Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks |
WO2017166098A1 (en) * | 2016-03-30 | 2017-10-05 | Xiaogang Wang | A method and a system for detecting an object in a video |
CN107491715A (en) * | 2016-06-13 | 2017-12-19 | 北京文安智能技术股份有限公司 | A kind of subway carriage passenger flow statistical method, apparatus and system based on video analysis |
CN107918765A (en) * | 2017-11-17 | 2018-04-17 | 中国矿业大学 | A kind of Moving target detection and tracing system and its method |
CN107977414A (en) * | 2017-11-22 | 2018-05-01 | 西安财经学院 | Image Style Transfer method and its system based on deep learning |
CN107992826A (en) * | 2017-12-01 | 2018-05-04 | 广州优亿信息科技有限公司 | A kind of people stream detecting method based on the twin network of depth |
CN108537269A (en) * | 2018-04-04 | 2018-09-14 | 中山大学 | A kind of the object detection deep learning method and its system of weak interactive mode |
CN108564122A (en) * | 2018-04-10 | 2018-09-21 | 北京工业大学 | The integrated training method of image description model based on convolution-cycle hybrid network |
CN108664933A (en) * | 2018-05-11 | 2018-10-16 | 中国科学院遥感与数字地球研究所 | The training method and its sorting technique of a kind of convolutional neural networks for SAR image ship classification, ship classification model |
CN108764264A (en) * | 2018-03-16 | 2018-11-06 | 深圳中兴网信科技有限公司 | Smog detection method, smoke detection system and computer installation |
CN108764142A (en) * | 2018-05-25 | 2018-11-06 | 北京工业大学 | Unmanned plane image forest Smoke Detection based on 3DCNN and sorting technique |
CN108921875A (en) * | 2018-07-09 | 2018-11-30 | 哈尔滨工业大学(深圳) | A kind of real-time traffic flow detection and method for tracing based on data of taking photo by plane |
CN108921099A (en) * | 2018-07-03 | 2018-11-30 | 常州大学 | Moving ship object detection method in a kind of navigation channel based on deep learning |
CN108961235A (en) * | 2018-06-29 | 2018-12-07 | 山东大学 | A kind of disordered insulator recognition methods based on YOLOv3 network and particle filter algorithm |
CN108985192A (en) * | 2018-06-29 | 2018-12-11 | 东南大学 | A kind of video smoke recognition methods based on multitask depth convolutional neural networks |
CN109003303A (en) * | 2018-06-15 | 2018-12-14 | 四川长虹电器股份有限公司 | Apparatus control method and device based on voice and space object identification and positioning |
CN109034863A (en) * | 2018-06-08 | 2018-12-18 | 浙江新再灵科技股份有限公司 | The method and apparatus for launching advertising expenditure are determined based on vertical ladder demographics |
-
2018
- 2018-12-20 CN CN201811561334.0A patent/CN109766780A/en active Pending
Patent Citations (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001273499A (en) * | 2000-03-27 | 2001-10-05 | Nagoya Electric Works Co Ltd | Method and device for detecting moving object |
CN102496275A (en) * | 2011-11-25 | 2012-06-13 | 大连海创高科信息技术有限公司 | Method for detecting overload of coach or not |
CN102610104A (en) * | 2012-03-16 | 2012-07-25 | 南京航空航天大学 | Onboard front vehicle detection method |
CN104318258A (en) * | 2014-09-29 | 2015-01-28 | 南京邮电大学 | Time domain fuzzy and kalman filter-based lane detection method |
WO2017166098A1 (en) * | 2016-03-30 | 2017-10-05 | Xiaogang Wang | A method and a system for detecting an object in a video |
CN107491715A (en) * | 2016-06-13 | 2017-12-19 | 北京文安智能技术股份有限公司 | A kind of subway carriage passenger flow statistical method, apparatus and system based on video analysis |
CN106251861A (en) * | 2016-08-05 | 2016-12-21 | 重庆大学 | A kind of abnormal sound in public places detection method based on scene modeling |
CN107025443A (en) * | 2017-04-06 | 2017-08-08 | 江南大学 | Stockyard smoke monitoring and on-time model update method based on depth convolutional neural networks |
CN107918765A (en) * | 2017-11-17 | 2018-04-17 | 中国矿业大学 | A kind of Moving target detection and tracing system and its method |
CN107977414A (en) * | 2017-11-22 | 2018-05-01 | 西安财经学院 | Image Style Transfer method and its system based on deep learning |
CN107992826A (en) * | 2017-12-01 | 2018-05-04 | 广州优亿信息科技有限公司 | A kind of people stream detecting method based on the twin network of depth |
CN108764264A (en) * | 2018-03-16 | 2018-11-06 | 深圳中兴网信科技有限公司 | Smog detection method, smoke detection system and computer installation |
CN108537269A (en) * | 2018-04-04 | 2018-09-14 | 中山大学 | A kind of the object detection deep learning method and its system of weak interactive mode |
CN108564122A (en) * | 2018-04-10 | 2018-09-21 | 北京工业大学 | The integrated training method of image description model based on convolution-cycle hybrid network |
CN108664933A (en) * | 2018-05-11 | 2018-10-16 | 中国科学院遥感与数字地球研究所 | The training method and its sorting technique of a kind of convolutional neural networks for SAR image ship classification, ship classification model |
CN108764142A (en) * | 2018-05-25 | 2018-11-06 | 北京工业大学 | Unmanned plane image forest Smoke Detection based on 3DCNN and sorting technique |
CN109034863A (en) * | 2018-06-08 | 2018-12-18 | 浙江新再灵科技股份有限公司 | The method and apparatus for launching advertising expenditure are determined based on vertical ladder demographics |
CN109003303A (en) * | 2018-06-15 | 2018-12-14 | 四川长虹电器股份有限公司 | Apparatus control method and device based on voice and space object identification and positioning |
CN108961235A (en) * | 2018-06-29 | 2018-12-07 | 山东大学 | A kind of disordered insulator recognition methods based on YOLOv3 network and particle filter algorithm |
CN108985192A (en) * | 2018-06-29 | 2018-12-11 | 东南大学 | A kind of video smoke recognition methods based on multitask depth convolutional neural networks |
CN108921099A (en) * | 2018-07-03 | 2018-11-30 | 常州大学 | Moving ship object detection method in a kind of navigation channel based on deep learning |
CN108921875A (en) * | 2018-07-09 | 2018-11-30 | 哈尔滨工业大学(深圳) | A kind of real-time traffic flow detection and method for tracing based on data of taking photo by plane |
Non-Patent Citations (2)
Title |
---|
史璐璐 等: "基于 Tiny Darknet 全卷积孪生网络的目标跟踪", 《南京邮电大学学报 (自然科学版)》, vol. 38, no. 4, 31 August 2018 (2018-08-31), pages 89 - 95 * |
陈慧岩: "《智能车辆理论与应用》", 北京理工大学出版社, pages: 74 - 77 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110309729A (en) * | 2019-06-12 | 2019-10-08 | 武汉科技大学 | Tracking and re-detection method based on anomaly peak detection and twin network |
CN111914935A (en) * | 2020-08-03 | 2020-11-10 | 哈尔滨工程大学 | Ship image target detection method based on deep learning |
CN111914935B (en) * | 2020-08-03 | 2022-07-15 | 哈尔滨工程大学 | Ship image target detection method based on deep learning |
CN113592842A (en) * | 2021-08-09 | 2021-11-02 | 南方医科大学南方医院 | Sample serum quality identification method and identification device based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20210319561A1 (en) | Image segmentation method and system for pavement disease based on deep learning | |
Wang et al. | RENet: Rectangular convolution pyramid and edge enhancement network for salient object detection of pavement cracks | |
CN110188720A (en) | A kind of object detection method and system based on convolutional neural networks | |
CN111091072A (en) | YOLOv 3-based flame and dense smoke detection method | |
CN114202696A (en) | SAR target detection method and device based on context vision and storage medium | |
CN110135296A (en) | Airfield runway FOD detection method based on convolutional neural networks | |
CN108171112A (en) | Vehicle identification and tracking based on convolutional neural networks | |
Bo et al. | Particle pollution estimation from images using convolutional neural network and weather features | |
CN110363104A (en) | A kind of detection method of diesel oil black smoke vehicle | |
CN107491752A (en) | Ship board character recognition method, device in a kind of natural scene based on deep learning | |
CN111047565A (en) | Method, storage medium and equipment for forest cloud image segmentation | |
CN109766780A (en) | A kind of ship smog emission on-line checking and method for tracing based on deep learning | |
CN109711407A (en) | A kind of method and relevant apparatus of Car license recognition | |
CN113553979B (en) | Safety clothing detection method and system based on improved YOLO V5 | |
CN116206223A (en) | Fire detection method and system based on unmanned aerial vehicle edge calculation | |
Sun et al. | Image recognition technology in texture identification of marine sediment sonar image | |
Wu et al. | Traffic sign detection based on SSD combined with receptive field module and path aggregation network | |
CN111062950A (en) | Method, storage medium and equipment for multi-class forest scene image segmentation | |
CN110059544B (en) | Pedestrian detection method and system based on road scene | |
CN117079095A (en) | Deep learning-based high-altitude parabolic detection method, system, medium and equipment | |
CN111222477A (en) | Vision-based method and device for detecting two hands leaving steering wheel | |
CN115100249B (en) | Intelligent factory monitoring system based on target tracking algorithm | |
CN110751077A (en) | Optical remote sensing picture ship detection method based on component matching and distance constraint | |
CN110516559A (en) | Suitable for precisely monitor method for tracking target and device, computer equipment | |
Liu et al. | Mob-YOLO: A Lightweight UAV Object Detection Method |
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: 20190517 |