CN108471497A - A kind of ship target real-time detection method based on monopod video camera - Google Patents
A kind of ship target real-time detection method based on monopod video camera Download PDFInfo
- Publication number
- CN108471497A CN108471497A CN201810173195.8A CN201810173195A CN108471497A CN 108471497 A CN108471497 A CN 108471497A CN 201810173195 A CN201810173195 A CN 201810173195A CN 108471497 A CN108471497 A CN 108471497A
- Authority
- CN
- China
- Prior art keywords
- ship
- video camera
- monopod video
- detection
- trained
- 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 36
- 238000011897 real-time detection Methods 0.000 title claims abstract description 8
- 238000001514 detection method Methods 0.000 claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 230000002045 lasting effect Effects 0.000 claims description 3
- 238000002360 preparation method Methods 0.000 claims description 3
- 238000002054 transplantation Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 206010041349 Somnolence Diseases 0.000 description 1
- 230000007812 deficiency 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
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/64—Computer-aided capture of images, e.g. transfer from script file into camera, check of taken image quality, advice or proposal for image composition or decision on when to take image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of ship target real-time detection method based on monopod video camera, including video flowing of the intake comprising ship target, training data pretreatment, the calculating speed for determining training pattern and improving trained ship detection model, it is trained according to actual conditions model adjustment, by trained ship detection model and monopod video camera linkage step.This method is by video flowing being trained data prediction, determining training pattern and improving the calculating speed of trained ship detection model, adjusting training model, finally trained ship detection model and monopod video camera are linked, finally the ship object width being presented in display main bit stream is kept to account for one third of picture width or so, and be shown centered on, realize the purpose that monopod video camera persistently tracks.Design is scientific and reasonable for the method for the present invention, is skillfully constructed, is easily achieved, and effectively increases accuracy rate, and arithmetic speed is fast, the value with wide popularization and application.
Description
Technical field
The invention belongs to technical field of video monitoring, especially a kind of ship target based on monopod video camera side of detection in real time
Method.
Background technology
Development based on deep learning in recent years, in industrial embedded product field, due to the computing capability of computing element
Enhancing, product obtain the development advanced by leaps and bounds in image recognition and two big algorithm direction of speech recognition.
Artificial intelligence application is generated on product intellectual product task be desirable to product can be to biographies such as video cameras
The data that sensor obtains understand and the rational thinking possessed by the mankind in a manner of make decision, to help the mankind to complete
The decision-making work of various uninteresting repetitions.The existing equipment using traditional images processing method mostly only moves up remote sea
Moving-target identifies, and there is the difficulty of mobile target classification, especially camera lens further after complex background under big object ship detection deposit
In difficulty.
And existing training pattern is effective to big target detection but to distal end Small object, especially haze weather in recent years
Ship shape object cannot detect.The high deep learning method training pattern of detection discrimination leads to single frames due to computationally intensive mostly
Processing speed is slow, and live video stream detection is carried out in front end embedded platform it is difficult to apply.Cannot be real-time, it is unable to zoom identification,
Cannot apply headend equipment ensure user data privacy the shortcomings of, be all unsuitable for Sea Surface Ship observation identification alarm and
Track demand.
Invention content
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of, and the ship target based on monopod video camera is real-time
Detection method, in the guarantee camera lens object no matter radar can be provided whether haze weather orientation object ship object it is real-time
Detecting and tracking method.
The scheme of the invention is be achieved：
A kind of ship target real-time detection method based on monopod video camera, includes the following steps：
Step 1: including the video flowing of ship target by monopod video camera intake
The video stream packets of intake contain main bit stream and auxiliary code stream：
(1) main bit stream of rtsp video flowings for clearly indicating real-time video in the display；
(2) the auxiliary code stream of rtsp video flowings is handled as the input video stream of ship detection, and carries out target inspection
It surveys；
Step 2: training data pre-processes
The source of training data picture includes two parts：
A part of image credit is extracted in by the picture for containing ship target from main bit stream, that is, after obtaining rtsp video flowings
Resolution processes are compressed and changed to picture using the method batch of image procossing, ensure that key message is not lost；
Another part picture then comes from the ship target in extraction common data sets such as Pascal VOC, arranges needs
Part, this part are mainly used for improving the generalization of ship detection, the recognition detection for ship target after furthering to zoom；
The above-mentioned picture classification handled well is arranged, ensure distal end ship type Small object in training set, various types of STOWAGE PLAN piece and
Possible error detection picture accounts for the certain ratio of training data；
Step 3: determining training pattern and improving the calculating speed of trained ship detection model
The darknet frame structures in linux system are applied in selection, facilitate the later stage using c/c++ language embedded
It transplanting in equipment and is docked with monopod video camera, using the principle of yolo trains ship detection model；
Image is RGB triple channel images in auxiliary code stream in the step 1；
The method for improving detection speed is, using one type objects of training ship target, and using the volume base core of multiple 1*1 sizes
The principle of the convolution kernel of a n*n is substituted, meanwhile, auxiliary code stream input image size is changed by image processing algorithm, and ensure
Key message is not lost, and calculative pixel number is reduced, and the method by reducing calculation amount improves calculating speed；
Step 4: being trained model adjustment according to actual conditions
Training data strictly screens, and meets error detection factor that specific environment ship detection and this environment may introduce not
It introduces；
Step 5: trained ship detection model and monopod video camera are linked
By trained ship detection model transplantations to monopod video camera, the problem of in order to improve accuracy, statistics is simultaneously
It filters out most believable coordinate and monopod video camera is issued in size preparation.
Moreover, in the step 5, the method for screening most credible coordinate and size is：
Present frame records coordinate (continuous record 20-25 frame of the ship target in entire picture at this time if there are object ship
Target's center's point coordinates), often get 20 groups of data, data processings just carried out to this 20 groups of data, will most believable coordinate
Point information is sent to monopod video camera by serial ports, and monopod video camera uses 3D algorithms, believes according to the most credible coordinate received
The rotation of the length and width dimensions information of breath and ship target and zoom, monopod video camera, which remains, at this time is presented in display main bit stream
Ship object width account for one third of picture width or so, and keep placed in the middle, and lasting tracking.
The advantages and positive effects of the present invention are：
1, this method is by being trained data prediction to video flowing, determining training pattern and improving the inspection of training ship target
The calculating speed of model, adjusting training model are surveyed, finally by trained ship detection model and monopod video camera
Linkage finally keeps the ship object width being presented in display main bit stream to account for one third of picture width or so, and placed in the middle
It has been shown that, realizes the purpose that monopod video camera persistently tracks.
2, the method for the present invention improves working efficiency and the work comfort sense of staff, make staff can with having time and
Energy does more and its relevant excelsior action of profession, for example personnel in charge of the case faster can more timely obtain
Clue helps each post staff preferably to realize that the industry with craftsman's spirit is promoted.
3, design is scientific and reasonable, is skillfully constructed, is easily achieved for the method for the present invention, effectively increases accuracy rate, and operation
Speed is fast, the value with wide popularization and application.
Description of the drawings
Fig. 1 is the flow chart of the present invention；
Fig. 2 is the specific workflow figure of the present invention；
Fig. 3 is the calculating speed block diagram that training ship detection model is improved in Fig. 1；
Fig. 4 is the design sketch of the method for the present invention intake image.
Specific implementation mode
The invention will be further described below in conjunction with the accompanying drawings and by specific embodiment.
A kind of ship target real-time detection method based on monopod video camera, as shown in Figure 1, including the following steps：
Step 1: including the video flowing of ship target by monopod video camera intake
Monopod video camera shooting positioned at ground is likely to occur the video of ship type object at a distance, and can after the zoom that furthers
The video of spot target can be gone out, at this time the usually shared pixel ratio very little of ship target in the camera lens of monopod video camera, sometimes
Due to haze weather, the not high problem of color contrast is also will present out, can only about show the picture effect of shape.(referring to
Fig. 4, left side are the auxiliary code stream video sectional drawings of monopod video camera intake, and right side is auxiliary code stream detection identification effect of this method to intake
Fruit shows)
The video stream packets of intake contain main bit stream and auxiliary code stream：
1, the main bit stream of rtsp video flowings for clearly indicating real-time video in the display；
2, the auxiliary code stream of rtsp video flowings is handled as the input video stream of ship detection, and carries out target inspection
It surveys；
Step 2: training data pre-processes
The source of training data picture includes two parts：
A part of image credit is extracted in by the picture for containing ship target from main bit stream, that is, after obtaining rtsp video flowings
The processing such as resolution ratio are compressed and changed to picture using the method batch of image procossing, ensure that key message is not lost；
Another part picture then comes from the ship target in extraction common data sets such as Pascal VOC, arranges needs
Part, this part are mainly used for improving the generalization of ship detection, the recognition detection for ship target after furthering to zoom.
The above-mentioned picture classification handled well is arranged, ensures distal end ship type Small object, variety classes STOWAGE PLAN piece in training set
The certain ratio of training data is accounted for possible error detection picture (bird of such as water surface, beacon, the building etc. on opposite bank).
Step 3: determining training pattern and improving the calculating speed of trained ship detection model
The darknet frame structures in linux system are applied in selection, it is therefore an objective to the later stage be facilitated to exist using c/c++ language
It transplanting on embedded device and is docked with monopod video camera, using the principle of yolo trains ship detection model.
Image is RGB triple channel images in auxiliary code stream in the step 1, if wishing just to need video bag each second in real time
Containing images more than 20 frames, and being usually detected processing is carried out on the headend equipment of low operational capability, therefore for ship
The single frames processing speed of target detection will be as quickly as possible.
To improve detection speed (as shown in Figure 3), using one type objects of training ship target, network structure is relatively easy, and
Calculation amount is reduced using the principle of the convolution kernel of one n*n of volume base nuclear subsitution of multiple 1*1 sizes, meanwhile, pass through image procossing
Algorithm changes auxiliary code stream input image size, and ensures that key message is not lost, and calculative pixel number is reduced, by subtracting
The method of few calculation amount, improves calculating speed.
Step 4: being trained model adjustment according to actual conditions
Since trained network is relatively easy, adds new feature entrance and be very easy to, but subtracting for error detection feature
It is not easy to less, it is therefore desirable to which the stringent screening for noticing training data will meet specific environment ship detection and this environment can
The error detection factor that can be introduced does not introduce.
Step 5: trained ship detection model and monopod video camera are linked
By trained ship detection model transplantations to monopod video camera, each frame picture may be implemented in monopod video camera
The arithmetic speed all detected, but the case where might have error detection or detection object scale error occur.
In order to solve the problems, such as accuracy, using coordinate information issue monopod video camera to monopod video camera act in the presence of
Between the characteristics of being spaced, do not go to send each frame detection information, and be changed to send for each second it is primary, to collect the institute within this second
The coordinate information and dimension information for all detection frames collected, count and filter out most believable coordinate and cloud is issued in size preparation
Platform video camera.
Monopod video camera acts incoherent problem in order to prevent at this time, can according to the direction of motion of ship target and speed
Letter coordinate transformation is to predict credible coordinate and ship target size and be sent to monopod video camera by serial ports, monopod video camera according to
The information action received is realized from motion tracking and zoom.
In above-mentioned steps five, the method for the most credible coordinate of the screening and size is：
Present frame records coordinate (continuous record 20-25 frame of the ship target in entire picture at this time if there are object ship
Target's center's point coordinates), often get 20 groups of data, data processings just carried out to this 20 groups of data, will most believable coordinate
Point information is sent to monopod video camera by serial ports, and monopod video camera uses 3D algorithms, believes according to the most credible coordinate received
The rotation of the length and width dimensions information of breath and ship target and zoom, monopod video camera, which remains, at this time is presented in display main bit stream
Ship object width account for one third of picture width or so, and keep placed in the middle, and lasting tracking.
Artificial intelligence application on observation tracing task of the sea to ship, is exactly being made up work by this method using equipment
Personnel generate sleepy sense due to staring at similar picture for a long time, or require staff to be found from one section of playing back videos and want to search
What period the object of rope appears at, and similar needs the mankind to think deeply judgement but in very not humane work.
It is emphasized that embodiment of the present invention is illustrative, without being restrictive, therefore packet of the present invention
Include the embodiment being not limited to described in specific implementation mode, it is every by those skilled in the art according to the technique and scheme of the present invention
The other embodiment obtained, also belongs to the scope of protection of the invention.
Claims (2)
1. a kind of ship target real-time detection method based on monopod video camera, it is characterised in that：Include the following steps：
Step 1: including the video flowing of ship target by monopod video camera intake
The video stream packets of intake contain main bit stream and auxiliary code stream：
(1) main bit stream of rtsp video flowings for clearly indicating real-time video in the display；
(2) the auxiliary code stream of rtsp video flowings is handled as the input video stream of ship detection, and carries out target detection；
Step 2: training data pre-processes
The source of training data picture includes two parts：
A part of image credit is extracted in by the picture for containing ship target from main bit stream, that is, is used after obtaining rtsp video flowings
The method batch of image procossing compresses picture and is changed resolution processes, ensures that key message is not lost；
Another part picture then comes from the ship target in extraction common data sets such as Pascal VOC, arranges the portion of needs
Point, this part is mainly used for improving the generalization of ship detection, the recognition detection for ship target after furthering to zoom；
The above-mentioned picture classification handled well is arranged, ensures distal end ship type Small object in training set, various types of STOWAGE PLAN piece and possibility
Error detection picture account for the certain ratio of training data；
Step 3: determining training pattern and improving the calculating speed of trained ship detection model
The darknet frame structures in linux system are applied in selection, facilitate the later stage using c/c++ language in embedded device
On transplanting and docked with monopod video camera, train ship detection model using the principle of yolo；
Image is RGB triple channel images in auxiliary code stream in the step 1；
The method for improving detection speed is, using one type objects of training ship target, and using the volume base nuclear subsitution of multiple 1*1 sizes
The principle of the convolution kernel of one n*n, meanwhile, auxiliary code stream input image size is changed by image processing algorithm, and ensure key
Information is not lost, and calculative pixel number is reduced, and the method by reducing calculation amount improves calculating speed；
Step 4: being trained model adjustment according to actual conditions
Training data strictly screens, and meets the error detection factor that specific environment ship detection and this environment may introduce and does not draw
Enter；
Step 5: trained ship detection model and monopod video camera are linked
By trained ship detection model transplantations to monopod video camera, the problem of in order to improve accuracy, counts and screen
Go out most believable coordinate and monopod video camera is issued in size preparation.
2. the ship target real-time detection method according to claim 1 based on monopod video camera, it is characterised in that：The step
In rapid five, the method for screening most credible coordinate and size is：
Present frame records the coordinate (mesh of continuous record 20-25 frame of the ship target in entire picture at this time if there are object ship
Mark center point coordinate), 20 groups of data are often got, data processing just is carried out to this 20 groups of data, it will most believable coordinate points letter
Breath is sent to monopod video camera by serial ports, monopod video camera using 3D algorithms, according to the most credible coordinate information received and
The length and width dimensions information of ship target rotates and zoom, and monopod video camera remains the ship being presented in display main bit stream at this time
Object width accounts for one third of picture width or so, and keeps placed in the middle, and lasting tracking.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810173195.8A CN108471497A (en) | 2018-03-02 | 2018-03-02 | A kind of ship target real-time detection method based on monopod video camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810173195.8A CN108471497A (en) | 2018-03-02 | 2018-03-02 | A kind of ship target real-time detection method based on monopod video camera |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108471497A true CN108471497A (en) | 2018-08-31 |
Family
ID=63264084
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810173195.8A Pending CN108471497A (en) | 2018-03-02 | 2018-03-02 | A kind of ship target real-time detection method based on monopod video camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108471497A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934088A (en) * | 2019-01-10 | 2019-06-25 | 海南大学 | Sea ship discrimination method based on deep learning |
| CN110278417A (en) * | 2019-07-25 | 2019-09-24 | 上海莫吉娜智能信息科技有限公司 | Monitoring device method for rapidly positioning and system based on millimetre-wave radar |
| CN110376593A (en) * | 2019-08-05 | 2019-10-25 | 上海埃威航空电子有限公司 | A kind of target apperception method and device based on laser radar |
| CN110427030A (en) * | 2019-08-08 | 2019-11-08 | 上海大学 | A kind of unmanned boat based on Tiny-YOLOship algorithm of target detection independently docks recovery method |
| CN111476163A (en) * | 2020-04-07 | 2020-07-31 | 浙江大华技术股份有限公司 | High-altitude parabolic monitoring method and device and computer storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1186923C (en) * | 2003-04-03 | 2005-01-26 | 上海交通大学 | Abnormal object automatic finding and tracking video camera system |
| CN102231798A (en) * | 2011-06-24 | 2011-11-02 | 天津市亚安科技电子有限公司 | Method for controlling PTZ (Pan/Tilt/Zoom) camera to zoom automatically and system thereof |
| CN104217215A (en) * | 2014-08-28 | 2014-12-17 | 哈尔滨工程大学 | Classification and identification method for foggy water surface image and clear water surface image |
| CN205179256U (en) * | 2015-12-02 | 2016-04-20 | 北京视酷伟业科技股份有限公司 | Automatic follow tracks of video linked system of boats and ships |
| CN106372590A (en) * | 2016-08-29 | 2017-02-01 | 江苏科技大学 | Sea surface ship intelligent tracking system and method based on machine vision |
| KR101737430B1 (en) * | 2015-11-06 | 2017-05-19 | 주식회사 지오멕스소프트 | A method of detecting objects in the image with moving background |
| CN107563387A (en) * | 2017-09-14 | 2018-01-09 | 成都掌中全景信息技术有限公司 | Frame method is selected in a kind of image object detection based on Recognition with Recurrent Neural Network |
-
2018
- 2018-03-02 CN CN201810173195.8A patent/CN108471497A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1186923C (en) * | 2003-04-03 | 2005-01-26 | 上海交通大学 | Abnormal object automatic finding and tracking video camera system |
| CN102231798A (en) * | 2011-06-24 | 2011-11-02 | 天津市亚安科技电子有限公司 | Method for controlling PTZ (Pan/Tilt/Zoom) camera to zoom automatically and system thereof |
| CN104217215A (en) * | 2014-08-28 | 2014-12-17 | 哈尔滨工程大学 | Classification and identification method for foggy water surface image and clear water surface image |
| KR101737430B1 (en) * | 2015-11-06 | 2017-05-19 | 주식회사 지오멕스소프트 | A method of detecting objects in the image with moving background |
| CN205179256U (en) * | 2015-12-02 | 2016-04-20 | 北京视酷伟业科技股份有限公司 | Automatic follow tracks of video linked system of boats and ships |
| CN106372590A (en) * | 2016-08-29 | 2017-02-01 | 江苏科技大学 | Sea surface ship intelligent tracking system and method based on machine vision |
| CN107563387A (en) * | 2017-09-14 | 2018-01-09 | 成都掌中全景信息技术有限公司 | Frame method is selected in a kind of image object detection based on Recognition with Recurrent Neural Network |
Non-Patent Citations (1)
| Title |
|---|
| 黄洁，姜志国等: "基于卷积神经网络的遥感图像舰船目标检测", 《北京航空航天大学学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934088A (en) * | 2019-01-10 | 2019-06-25 | 海南大学 | Sea ship discrimination method based on deep learning |
| CN110278417A (en) * | 2019-07-25 | 2019-09-24 | 上海莫吉娜智能信息科技有限公司 | Monitoring device method for rapidly positioning and system based on millimetre-wave radar |
| CN110376593A (en) * | 2019-08-05 | 2019-10-25 | 上海埃威航空电子有限公司 | A kind of target apperception method and device based on laser radar |
| CN110376593B (en) * | 2019-08-05 | 2021-05-04 | 上海埃威航空电子有限公司 | Target sensing method and device based on laser radar |
| CN110427030A (en) * | 2019-08-08 | 2019-11-08 | 上海大学 | A kind of unmanned boat based on Tiny-YOLOship algorithm of target detection independently docks recovery method |
| CN111476163A (en) * | 2020-04-07 | 2020-07-31 | 浙江大华技术股份有限公司 | High-altitude parabolic monitoring method and device and computer storage medium |
| CN111476163B (en) * | 2020-04-07 | 2022-02-18 | 浙江大华技术股份有限公司 | High-altitude parabolic monitoring method and device and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106096577B (en) | A kind of target tracking method in camera distribution map | |
| CN108471497A (en) | A kind of ship target real-time detection method based on monopod video camera | |
| CN103164706B (en) | Object counting method and device based on video signal analysis | |
| CN103413444B (en) | A kind of traffic flow based on unmanned plane HD video is investigated method | |
| CN104978567B (en) | Vehicle checking method based on scene classification | |
| CN108154110B (en) | Intensive people flow statistical method based on deep learning people head detection | |
| EP2956891B1 (en) | Segmenting objects in multimedia data | |
| CN109145708B (en) | Pedestrian flow statistical method based on RGB and D information fusion | |
| CN108334847A (en) | A kind of face identification method based on deep learning under real scene | |
| CN108197604A (en) | Fast face positioning and tracing method based on embedded device | |
| CN113762009B (en) | Crowd counting method based on multi-scale feature fusion and double-attention mechanism | |
| WO2020114116A1 (en) | Pedestrian detection method based on dense crowds, and storage medium and processor | |
| CN105160297A (en) | Masked man event automatic detection method based on skin color characteristics | |
| CN109063549A (en) | High-resolution based on deep neural network is taken photo by plane video moving object detection method | |
| CN104156978A (en) | Multi-target dynamic following method based on balloon platform | |
| CN109558790B (en) | Pedestrian target detection method, device and system | |
| Zhang et al. | I-MMCCN: Improved MMCCN for RGB-T crowd counting of drone images | |
| CN112464893A (en) | Congestion degree classification method in complex environment | |
| CN110866453B (en) | Real-time crowd steady state identification method and device based on convolutional neural network | |
| Wu et al. | Vehicle Classification and Counting System Using YOLO Object Detection Technology. | |
| Chen et al. | Real-time garbage object detection with data augmentation and feature fusion using SUAV low-altitude remote sensing images | |
| CN116311071A (en) | Substation perimeter foreign matter identification method and system integrating frame difference and CA | |
| CN112329550A (en) | Weak supervision learning-based disaster-stricken building rapid positioning evaluation method and device | |
| CN112258552A (en) | Pedestrian multi-target tracking method under community monitoring scene | |
| CN115880643A (en) | Social distance monitoring method and device based on target detection algorithm |
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 | ||
| AD01 | Patent right deemed abandoned | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20240419 |