CN109993168A - Intelligent polling method - Google Patents
Intelligent polling method Download PDFInfo
- Publication number
- CN109993168A CN109993168A CN201910279871.4A CN201910279871A CN109993168A CN 109993168 A CN109993168 A CN 109993168A CN 201910279871 A CN201910279871 A CN 201910279871A CN 109993168 A CN109993168 A CN 109993168A
- Authority
- CN
- China
- Prior art keywords
- frame
- candidate
- candidate frame
- image
- score
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of intelligent polling method, comprising: acquisition inspection image;Processing is carried out to the inspection image and obtains multiple candidate frames on the inspection image, obtains each candidate frame information;Adaptive threshold function table curve is obtained, the adaptive thresholding value function includes multiple variables;According to the adaptive threshold function curve and each candidate frame information, the image in satisfactory candidate frame is acquired using secondary method for inspecting.Adaptive threshold of the present invention by setting including multiple parameters, and secondary inspection is carried out to inspection image according to adaptive threshold, false detection rate can be reduced while reducing omission factor.
Description
Technical field
The present invention relates to monitoring fields, and in particular to a kind of intelligent polling method.
Background technique
With the development of image processing techniques, the performance of monitoring device is also greatly improved, such as present intelligent video camera head
Target in some scene or certain width picture can be detected and be identified, by interested target object and be lost interest in
Non-targeted object separate, to improve routing inspection efficiency and accuracy.Existing intelligent video camera head identifies figure during inspection
When as target, by associated picture algorithm can produce it is multiple may include target frame, while to the frame of each frame
Information is labeled, for example is marked to frame score (frame score indicates that image belongs to the probability of a certain type in frame)
Note, if certain frame score be more than preset threshold, retain the frame and include by the frame image information collecting back into
Row processing, there are a deficiencies for such acquisition mode: since preset threshold pertains only to a parameter, if preset threshold is arranged
It is too much, then cause omission factor high, if preset threshold is arranged too low, false detection rate is high.
Summary of the invention
In view of this, the present invention provides a kind of intelligent polling method, it include the adaptive thresholding of multiple parameters by setting
Value, and secondary inspection is carried out to inspection image according to adaptive threshold, false detection rate can be reduced while reducing omission factor.This
Scheme is realized by following technological means:
Intelligent polling method, comprising:
Acquire inspection image;
Processing is carried out to the inspection image and obtains multiple candidate frames on the inspection image, obtains each candidate
Frame information;
Adaptive threshold function table curve is obtained, the adaptive thresholding value function includes multiple variables;
According to the adaptive threshold function curve and each candidate frame information, met using the acquisition of secondary method for inspecting
It is required that candidate frame in image.
Further, the inspection image handle and in institute according to Region Proposal Network algorithm
It states and obtains multiple candidate frames on inspection image.
Further, the candidate frame information includes candidate frame size, candidate frame score and candidate bezel locations
Coordinate.
Further, the variable of the adaptive thresholding value function includes frame size and frame score.
Further, the construction process of the adaptive threshold function curve includes:
Acquire standard picture;
Processing is carried out to the standard picture and obtains multiple spare frames on the standard picture, is obtained each spare
The frame size and frame score of frame;
The image in each spare frame is acquired, judges whether the image in each spare frame meets the requirements, if symbol
It closes and requires, record the frame size and frame score of corresponding spare frame;
According to the frame size and frame score of all satisfactory spare frames, adaptive thresholding value function is calculated
And construct adaptive threshold function curve.
Further, the abscissa of the adaptive threshold function curve and ordinate respectively indicate frame size and frame
Score.
Further, described according to the adaptive threshold function curve and each candidate frame information, it is patrolled using secondary
Detecting method acquires the step for image in satisfactory candidate frame including an inspection step and secondary inspection step, institute
Stating an inspection step includes:
Obtain the frame size and frame score of each candidate frame on inspection image;
In the coordinate system of the adaptive threshold function curve, judge that the frame size of each candidate frame and frame obtain
Divide whether the coordinate points constituted are located above the adaptive threshold function curve, if it is, corresponding candidate frame is to meet
It is required that candidate frame, if it is not, then corresponding candidate frame is candidate frame undetermined;
Acquire the satisfactory image wait adopt in candidate frame.
Further, the secondary inspection step includes:
Focal length and the direction for adjusting image capture device, according to the bezel locations coordinate of the candidate frame undetermined, acquisition
The topography of the inspection image;
Processing is carried out to the topography and obtains one or more secondary candidate frames in the topography, is obtained
Take the frame size and frame score of each secondary candidate frame;
According to the frame size and frame score of the adaptive threshold function curve and each secondary candidate frame, acquisition
Image in satisfactory secondary candidate frame.
Further, it is described according to the adaptive threshold function curve and the frame size of each secondary candidate frame and
Frame score, the step for acquiring the image in satisfactory secondary candidate frame include:
Obtain default frame size and default frame score;
In the coordinate system of the adaptive threshold function curve, frame size and the side of each secondary candidate frame are judged
Whether the coordinate points that frame score is constituted are located above the adaptive threshold function curve, if it is, the secondary candidate frame
For satisfactory secondary candidate frame, if it is not, then the secondary candidate frame is secondary candidate frame undetermined;
Judge whether the frame size of the secondary candidate frame undetermined is greater than the default frame size, if it is,
The secondary candidate frame undetermined is undesirable secondary candidate frame, is directly given up, if it is not, then carrying out in next step;
Judge whether the frame score of the secondary candidate frame undetermined is greater than the default frame score, if it is,
The secondary candidate frame undetermined is satisfactory secondary candidate frame;
Acquire the image in satisfactory secondary candidate frame.
Further, execution is described handle to the inspection image and obtains multiple candidates on the inspection image
Frame when the step for obtaining each candidate frame information, if adjacent have overlapping wait adopt candidate frame, gives up frame score
Low candidate frame.
Adaptive threshold of the present invention by setting including multiple parameters, and inspection image is carried out according to adaptive threshold
Secondary inspection can reduce false detection rate while reducing omission factor.
Detailed description of the invention
Fig. 1 is a kind of intelligent polling method flow chart shown according to an exemplary embodiment.
Specific embodiment
It is with reference to the accompanying drawing and specific real in order to make those skilled in the art more fully understand technical solution of the present invention
Applying example, the present invention is described in further detail.
Embodiment
As shown in Figure 1, the present embodiment provides a kind of intelligent polling methods, comprising:
Step S1: acquisition inspection image;
Step S2: carrying out processing to the inspection image and multiple candidate frames are obtained on the inspection image, obtains
Each candidate's frame information;
Step S3: obtaining adaptive threshold function table curve, and the adaptive thresholding value function includes multiple variables;
Step S4: according to the adaptive threshold function curve and each candidate frame information, using secondary method for inspecting
Acquire the image in satisfactory candidate frame.
In this implementation, the intelligent polling method can be applied to intelligent camera, the adaptive threshold letter in step S3
Number curve can be constructed in advance and be stored in intelligent camera, in inspection can directly from the memory of intelligent camera or
It is transferred in memory block, to improve routing inspection efficiency, since adaptive thresholding value function includes multiple variables, compares unitary variant
It says, the present embodiment property taken into account is more preferable, and there is no the problems that omission factor is high or false detection rate is high, can be well in omission factor and mistake
It is balanced between inspection rate, simultaneously as missing inspection can reduced by further ensuring the present embodiment by the way of secondary inspection
False detection rate is reduced while rate.
Preferably, in step S2, can according to Region Proposal Network algorithm to the inspection image into
Row handles and obtains multiple candidate frames on the inspection image, wherein candidate frame information includes candidate frame size, waits
Select frame score and candidate bezel locations coordinate.In image procossing, obtained candidate frame is limited, raw in candidate frame
Cheng Shi, target is bigger or target is closer, and easier detection comes, thus big target or close target there is no or be difficult to deposit
In missing inspection, when screening satisfactory candidate frame, if candidate frame size is very big but score is very low, illustrates this time
Selecting the image in frame not is the target image of our needs, can be given up.
Preferably, the variable of the adaptive thresholding value function includes frame size and frame score.
Preferably, the construction process of adaptive threshold function curve described in step S3 includes:
Step P1: acquisition standard picture;
Step P2: the standard picture handle and obtains multiple spare frames on the standard picture, is obtained
The frame size and frame score of each spare frame;
Step P3: the image in each spare frame of acquisition judges whether the image in each spare frame meets the requirements,
If met the requirements, the frame size and frame score of corresponding spare frame are recorded;
Step P4: it according to the frame size and frame score of all satisfactory spare frames, is calculated adaptive
Threshold function table simultaneously constructs adaptive threshold function curve.
Preferably, the abscissa and ordinate of the adaptive threshold function curve respectively indicate frame size and frame
Score.
It is preferably, step S4, i.e., described according to the adaptive threshold function curve and each candidate frame information, it adopts
The step for acquiring the image in satisfactory candidate frame with secondary method for inspecting includes an inspection step and secondary patrols
Step is examined, an inspection step includes:
Step S411: the frame size and frame score of each candidate frame on inspection image are obtained;
Step S412: in the coordinate system of the adaptive threshold function curve, judge the frame ruler of each candidate frame
Whether the coordinate points that very little and frame score is constituted are located above the adaptive threshold function curve, if it is, corresponding candidate
Frame is satisfactory candidate frame, if it is not, then corresponding candidate frame is candidate frame undetermined;
Step S413: the satisfactory image wait adopt in candidate frame of acquisition.
Preferably, the secondary inspection step includes:
Step S421: adjusting focal length and the direction of image capture device, according to the bezel locations of the candidate frame undetermined
Coordinate acquires the topography of the inspection image;
Step S422: carrying out processing to the topography and one or more secondary times are obtained in the topography
Frame is selected, the frame size and frame score of each secondary candidate frame are obtained;
Step S423: according to the frame size and frame of the adaptive threshold function curve and each secondary candidate frame
Score acquires the image in satisfactory secondary candidate frame.
It is preferably, step S423, i.e., described according to the adaptive threshold function curve and each secondary candidate frame
Frame size and frame score, the step for acquiring the image in satisfactory secondary candidate frame includes:
Step S4231: default frame size and default frame score are obtained;
Step S4232: in the coordinate system of the adaptive threshold function curve, judge the side of each secondary candidate frame
Whether the coordinate points that frame size and frame score are constituted are located above the adaptive threshold function curve, if it is, this two
Secondary candidate's frame is satisfactory secondary candidate frame, if it is not, then the secondary candidate frame is secondary candidate side undetermined
Frame;
Step S4233: judging whether the frame size of the secondary candidate frame undetermined is greater than the default frame size,
If it is, the secondary candidate frame undetermined is undesirable secondary candidate frame, directly give up, if it is not, then into
Row is in next step;
Step S4234: judging whether the frame score of the secondary candidate frame undetermined is greater than the default frame score,
If it is, the secondary candidate frame undetermined is satisfactory secondary candidate frame;
Step S4235: the image in the satisfactory secondary candidate frame of acquisition.
Preferably, execute step S2, i.e., it is described that processing is carried out to the inspection image and is obtained on the inspection image
To multiple candidate frames, when the step for obtaining each candidate frame information, if adjacent have overlapping wait adopt candidate frame, give up
Abandon the low candidate frame of frame score.
The above is only the preferred embodiment of the present invention, it is noted that above-mentioned preferred embodiment is not construed as pair
Limitation of the invention, protection scope of the present invention should be defined by the scope defined by the claims..For the art
For those of ordinary skill, without departing from the spirit and scope of the present invention, several improvements and modifications can also be made, these change
It also should be regarded as protection scope of the present invention into retouching.
Claims (10)
1. intelligent polling method characterized by comprising
Acquire inspection image;
Processing is carried out to the inspection image and obtains multiple candidate frames on the inspection image, obtains each candidate frame
Information;
Adaptive threshold function table curve is obtained, the adaptive thresholding value function includes multiple variables;
According to the adaptive threshold function curve and each candidate frame information, met the requirements using the acquisition of secondary method for inspecting
Candidate frame in image.
2. intelligent polling method according to claim 1, which is characterized in that according to Region Proposal Network
Algorithm carries out processing to the inspection image and obtains multiple candidate frames on the inspection image.
3. intelligent polling method according to claim 1, which is characterized in that candidate's frame information includes candidate frame
Size, candidate frame score and candidate bezel locations coordinate.
4. intelligent polling method according to claim 3, which is characterized in that the variable of the adaptive thresholding value function includes
Frame size and frame score.
5. intelligent polling method according to claim 4, which is characterized in that the construction of the adaptive threshold function curve
Process includes:
Acquire standard picture;
Processing is carried out to the standard picture and obtains multiple spare frames on the standard picture, obtains each spare frame
Frame size and frame score;
The image in each spare frame is acquired, judges whether the image in each spare frame meets the requirements, if conformed to
It asks, records the frame size and frame score of corresponding spare frame;
According to the frame size and frame score of all satisfactory spare frames, adaptive thresholding value function and structure is calculated
Make adaptive threshold function curve.
6. intelligent polling method according to claim 4, which is characterized in that the horizontal seat of the adaptive threshold function curve
Mark and ordinate respectively indicate frame size and frame score.
7. intelligent polling method according to claim 6, which is characterized in that described bent according to the adaptive thresholding value function
Line and each candidate frame information, the step for image in satisfactory candidate frame is acquired using secondary method for inspecting packet
An inspection step and secondary inspection step are included, an inspection step includes:
Obtain the frame size and frame score of each candidate frame on inspection image;
In the coordinate system of the adaptive threshold function curve, the frame size and frame score structure of each candidate frame are judged
At coordinate points whether be located above the adaptive threshold function curve, if it is, corresponding candidate frame is to meet the requirements
Candidate frame, if it is not, then corresponding candidate frame is candidate frame undetermined;
Acquire the satisfactory image wait adopt in candidate frame.
8. intelligent polling method according to claim 7, which is characterized in that the secondary inspection step includes:
Focal length and the direction for adjusting image capture device, according to the bezel locations coordinate of the candidate frame undetermined, described in acquisition
The topography of inspection image;
Processing is carried out to the topography and obtains one or more secondary candidate frames in the topography, is obtained every
The frame size and frame score of a secondary candidate frame;
Met according to the frame size and frame score, acquisition of the adaptive threshold function curve and each secondary candidate frame
It is required that secondary candidate frame in image.
9. intelligent polling method according to claim 8, which is characterized in that described bent according to the adaptive thresholding value function
The frame size and frame score of line and each secondary candidate frame, acquire image in satisfactory secondary candidate frame this
One step includes:
Obtain default frame size and default frame score;
In the coordinate system of the adaptive threshold function curve, judge that the frame size of each secondary candidate frame and frame obtain
Divide whether the coordinate points constituted are located above the adaptive threshold function curve, if it is, the secondary candidate frame is to accord with
Desired secondary candidate frame is closed, if it is not, then the secondary candidate frame is secondary candidate frame undetermined;
Judge whether the frame size of the secondary candidate frame undetermined is greater than the default frame size, if it is, should be to
Fixed secondary candidate frame is undesirable secondary candidate frame, is directly given up, if it is not, then carrying out in next step;
Judge whether the frame score of the secondary candidate frame undetermined is greater than the default frame score, if it is, should be to
Fixed secondary candidate frame is satisfactory secondary candidate frame;
Acquire the image in satisfactory secondary candidate frame.
10. intelligent polling method according to claim 1, which is characterized in that execute described to inspection image progress
When the step for handling and obtain on the inspection image multiple candidate frames, obtaining each candidate frame information, if phase
Neighbour has overlapping wait adopt candidate frame, then gives up the low candidate frame of frame score.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910279871.4A CN109993168B (en) | 2019-04-09 | 2019-04-09 | Intelligent inspection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910279871.4A CN109993168B (en) | 2019-04-09 | 2019-04-09 | Intelligent inspection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109993168A true CN109993168A (en) | 2019-07-09 |
CN109993168B CN109993168B (en) | 2021-07-16 |
Family
ID=67132590
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910279871.4A Active CN109993168B (en) | 2019-04-09 | 2019-04-09 | Intelligent inspection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109993168B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106504233A (en) * | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
CN106845502A (en) * | 2017-01-23 | 2017-06-13 | 东南大学 | It is a kind of to visualize guidance method for the Wearable servicing unit of overhaul of the equipments and overhaul of the equipments |
CN107369162A (en) * | 2017-07-21 | 2017-11-21 | 华北电力大学(保定) | A kind of generation method and system of insulator candidate target region |
CN109145872A (en) * | 2018-09-20 | 2019-01-04 | 北京遥感设备研究所 | A kind of SAR image Ship Target Detection method merged based on CFAR with Fast-RCNN |
EP3432271A1 (en) * | 2017-07-20 | 2019-01-23 | Tata Consultancy Services Limited | Systems and methods for detecting grasp poses for handling target objects |
US10198671B1 (en) * | 2016-11-10 | 2019-02-05 | Snap Inc. | Dense captioning with joint interference and visual context |
CN109344766A (en) * | 2018-09-29 | 2019-02-15 | 南京理工大学 | Slide block type breaker recognition methods based on crusing robot |
-
2019
- 2019-04-09 CN CN201910279871.4A patent/CN109993168B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106504233A (en) * | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
US10198671B1 (en) * | 2016-11-10 | 2019-02-05 | Snap Inc. | Dense captioning with joint interference and visual context |
CN106845502A (en) * | 2017-01-23 | 2017-06-13 | 东南大学 | It is a kind of to visualize guidance method for the Wearable servicing unit of overhaul of the equipments and overhaul of the equipments |
EP3432271A1 (en) * | 2017-07-20 | 2019-01-23 | Tata Consultancy Services Limited | Systems and methods for detecting grasp poses for handling target objects |
CN107369162A (en) * | 2017-07-21 | 2017-11-21 | 华北电力大学(保定) | A kind of generation method and system of insulator candidate target region |
CN109145872A (en) * | 2018-09-20 | 2019-01-04 | 北京遥感设备研究所 | A kind of SAR image Ship Target Detection method merged based on CFAR with Fast-RCNN |
CN109344766A (en) * | 2018-09-29 | 2019-02-15 | 南京理工大学 | Slide block type breaker recognition methods based on crusing robot |
Non-Patent Citations (4)
Title |
---|
HERMAN HAVERKORT等: "Locality and Bounding-Box Quality of Two-Dimensional Space-Filling Curves", 《ARXIV》 * |
KHAN IZHAR ALI: "基于特征学习的视频行人检测", 《万方数据库》 * |
SHAOQING REN等: "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks", 《ARXIV》 * |
张红颖等: "结合特征在线选择与协方差矩阵的压缩跟踪算法", 《光学精密工程》 * |
Also Published As
Publication number | Publication date |
---|---|
CN109993168B (en) | 2021-07-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102103753B (en) | Use method and the terminal of real time camera estimation detect and track Moving Objects | |
CN110334635A (en) | Main body method for tracing, device, electronic equipment and computer readable storage medium | |
CN101299239B (en) | Method and device for acquiring character area image and character recognition system | |
JP2010045613A (en) | Image identifying method and imaging device | |
CN108549852A (en) | Pedestrian detector's Auto-learning Method under special scenes based on the enhancing of depth network | |
CN110674886B (en) | Video target detection method fusing multi-level features | |
CN110826612A (en) | Training and identifying method for deep learning | |
CN110782385A (en) | Image watermark removing method based on deep learning | |
CN109101985A (en) | It is a kind of based on adaptive neighborhood test image mismatch point to elimination method | |
JP2007110932A (en) | Automatic method for culturing and cell culture apparatus | |
CN111028263A (en) | Moving object segmentation method and system based on optical flow color clustering | |
CN111050027B (en) | Lens distortion compensation method, device, equipment and storage medium | |
CN109993168A (en) | Intelligent polling method | |
CN106204889B (en) | A kind of localization method and device of bank note fluorescence area | |
CN112465702A (en) | Synchronous self-adaptive splicing display processing method for multi-channel ultrahigh-definition video | |
CN109886963B (en) | Image processing method and system | |
CN111008944A (en) | Method and system for searching unstable pixels of infrared focal plane detector | |
KR101915538B1 (en) | System For Detecting Objects Motion Vectors And Method For Detecting Objects Motion Vectors | |
CN109409159A (en) | A kind of fuzzy two-dimensional code detection method and device | |
JP2006331266A (en) | Method and apparatus for detecting object | |
US8538142B2 (en) | Face-detection processing methods, image processing devices, and articles of manufacture | |
CN112052883B (en) | Clothes detection method, device and storage medium | |
CN104935837B (en) | For optical imagery pointing device and the image processing method of optical video equipment | |
CN112101331B (en) | Security video fusion scheduling method and device | |
CN112347975B (en) | Visual converged communication scheduling system and implementation method thereof |
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 |