CN117372719A - Target detection method based on screening - Google Patents
Target detection method based on screening Download PDFInfo
- Publication number
- CN117372719A CN117372719A CN202311669123.XA CN202311669123A CN117372719A CN 117372719 A CN117372719 A CN 117372719A CN 202311669123 A CN202311669123 A CN 202311669123A CN 117372719 A CN117372719 A CN 117372719A
- Authority
- CN
- China
- Prior art keywords
- similarity
- screening
- connected domain
- target sample
- target
- 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
- 238000012216 screening Methods 0.000 title claims abstract description 46
- 238000001514 detection method Methods 0.000 title claims abstract description 32
- 230000004927 fusion Effects 0.000 claims abstract description 21
- 238000000034 method Methods 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000004891 communication Methods 0.000 claims description 5
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses a target detection method based on screening, and belongs to the technical field of computer vision. A screening-based target detection method comprising: detecting a connected domain in an image to be detected; calculating the color similarity of the connected domain and the target sample; calculating HOG similarity between the connected domain and the target sample; calculating fusion similarity according to the color similarity and the HOG similarity; and screening the connected domain according to the fusion similarity to obtain a target detection result. The invention reduces false detection through a screening mechanism and improves the accuracy of small target detection.
Description
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a screening-based target detection method.
Background
Target detection is an important research direction in the field of computer vision, and is widely applied to the fields of image analysis, video monitoring, automatic driving and the like. However, the conventional target detection method often has problems of low precision and recall rate when processing small targets, which is mainly caused by the characteristics of low signal-to-noise ratio, low contrast, abundant background interference and the like of the small targets. Therefore, the development of a high-efficiency and accurate small target detection method has important significance.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a screening-based target detection method.
The aim of the invention is realized by the following technical scheme: a screening-based target detection method comprising:
detecting a connected domain in an image to be detected;
calculating the color similarity of the connected domain and the target sample;
calculating HOG similarity between the connected domain and the target sample;
calculating fusion similarity according to the color similarity and the HOG similarity;
and screening the connected domain according to the fusion similarity to obtain a target detection result.
Further, detecting the connected domain in the image to be detected includes:
extracting connected domains in an image to be detected;
and screening the connected domain according to a preset rule.
Further, screening the connected domain according to a preset rule includes:
screening out connected domains with widths and heights meeting a screening formula, wherein the screening formula is as follows:
in the method, in the process of the invention,for the width of the target sample, +.>High, high for the target sample>For the set screening coefficient, +.>Is the width of the communicating region->Is the height of the connected domain.
Further, calculating the color similarity of the connected domain and the target sample includes:
respectively counting color histograms of the connected domain and the target sample;
and calculating the color similarity of the connected domain and the target sample based on the color histogram, wherein the calculation formula of the color similarity is as follows:
in the method, in the process of the invention,for color similarity, ++>Is->Personal communication domain(s)>For the target sample, ++>Represents->The->The value of the individual bin, ">Sample of object->The value in the individual bin; the pixel values of 0-255 are divided into +.>And a histogram bin, each bin representing a value normalized by the L1 norm for the number of pixels having pixel values between 0-15, 16-31, …, 240-255.
Further, calculating the HOG similarity between the connected domain and the target sample includes:
respectively extracting HOG characteristics of the connected domain and the target sample;
and calculating the HOG similarity of the connected domain and the target sample, wherein the HOG similarity has a calculation formula as follows:
in the method, in the process of the invention,for HOG similarity, ++>Is->Candidate connected domain->For the target sample, ++>Represents->The->The value of the individual bin, ">Sample of object->The value of the individual bin; dividing 0-179 degree into +.>The bins are values normalized to the histogram with the L1 norm, 0-29, 30-59, 60-89, …,150-179, respectively.
Further, the calculation formula of the fusion similarity is as follows:
in the method, in the process of the invention,for color similarity, ++>For HOG similarity, ++>And->And s is fusion similarity, which is a weight scale factor.
Further, screening the connected domain according to the fusion similarity to obtain a target detection result, including:
and determining the connected domain with the fusion similarity being greater than or equal to a similarity threshold as the detected target.
The beneficial effects of the invention are as follows:
(1) According to the invention, the targets with the non-conforming aspect ratio and the relatively larger targets are eliminated through the screening mechanism, the false detection of the small targets is reduced, the accuracy of the detection of the small targets is improved, the operand of the subsequent steps is reduced, and the performance of the algorithm is improved;
(2) The method adopts a mode of fusing the color similarity and the HOG similarity, is insensitive to the rotation and illumination influence of the object, and improves the robustness of the detection result;
(3) Because the small target occupies a smaller area, the required operation amount is relatively smaller when the color similarity and the HOG similarity are calculated, and therefore, the invention has higher performance in the small target detection task.
Drawings
FIG. 1 is a flow chart of a method for detecting an object in the present invention.
Detailed Description
The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by a person skilled in the art without any inventive effort, are intended to be within the scope of the present invention, based on the embodiments of the present invention.
Referring to fig. 1, the present invention provides a screening-based target detection method:
as shown in fig. 1, a screening-based target detection method includes S100 to S500.
S100, detecting connected domains in the image to be detected.
In some embodiments, connected domains in the image to be detected are detected, including S110 and S120.
S110, extracting connected domains in the image to be detected.
Specifically, an input image to be detected is obtained, and then the connected domain of the image to be detected is extracted.
S120, screening the connected domain according to a preset rule.
In some embodiments, screening the connected domain according to a preset rule includes: screening all the extracted connected domains according to the following formula, screening out connected domains with the width and the height meeting the screening formula, and eliminating connected domains which do not meet the screening formula:
in the method, in the process of the invention,for the width of the target sample, +.>High, high for the target sample>For the set screening coefficient, in this embodimentThe value of (2) is 0.2, in other embodiments +.>Is a value of->Is the width of the communicating region->Is the height of the connected domain.
S200, calculating the color similarity of the connected domain and the target sample.
In some embodiments, calculating the color similarity of the connected domain to the target sample includes: and respectively counting color histograms of the connected domain and the target sample obtained by screening, counting the number of pixels according to 1 bin of every 17 pixels, totaling n=15bins, normalizing the color histograms by using an L1 norm, and calculating the color similarity of the connected domain and the target sample. The calculation formula of the color similarity is as follows:
in the method, in the process of the invention,for color similarity, ++>Is->Personal communication domain(s)>For the target sample, ++>Represents->The->The value of the individual bin, ">Sample of object->The value in the individual bin; the pixel values of 0-255 are divided into +.>And a histogram bin, each bin representing a value normalized by the L1 norm for the number of pixels having pixel values between 0-15, 16-31, …, 240-255.
S300, calculating the HOG similarity of the connected domain and the target sample.
In some embodiments, calculating how similar the connected domain is to the target sample comprises: respectively extracting HOG characteristics of the connected domain and the target sample obtained by screening; the histograms were normalized with L1 norms and HOG similarity between connected domain and target samples was calculated by dividing 0-180 degrees into n=6 bins, 0, 30, 60..150, respectively. The calculation formula of the HOG similarity is as follows:
in the method, in the process of the invention,for HOG similarity, ++>Is->Candidate connected domain->For the target sample, ++>Represents->The->The value of the individual bin, ">Sample of object->The value of the individual bin; dividing 0-179 degree into +.>The bins are values normalized to the histogram with the L1 norm, 0-29, 30-59, 60-89, …,150-179, respectively.
S400, calculating fusion similarity according to the color similarity and the HOG similarity.
In some embodiments, the fused similarity calculation formula is:
in the method, in the process of the invention,for color similarity, ++>For HOG similarity, ++>And->For the set weight scale factor, s is the fusion similarity,/->Is->Personal communication domain(s)>Is the target sample. In this embodiment +.>The value of (2) is 0.4,/o>The value of (2) is 0.6, and in other embodiments may be set according to the implementation condition>And->Is a value of (2).
S500, screening the connected domain according to the fusion similarity to obtain a target detection result.
In some embodiments, screening the connected domain according to the fusion similarity to obtain a target detection result includes: and determining the connected domain with the fusion similarity being greater than or equal to a similarity threshold as the detected target.
For example, a similarity thresholdSet to 0.25, the connected domain is screened according to the following formula:
that is, the connected domain with the fusion similarity larger than or equal to the similarity threshold value is reserved, the connected domain with the fusion similarity smaller than the similarity threshold value is excluded, the reserved connected domain is the detected target, and the position coordinate of the reserved connected domain is the position coordinate of the target.
The numbers of the steps in this embodiment do not indicate that the steps must be performed in the order of the numbers, for example, S200 and S300 may be performed simultaneously.
The foregoing is merely a preferred embodiment of the invention, and it is to be understood that the invention is not limited to the form disclosed herein but is not to be construed as excluding other embodiments, but is capable of numerous other combinations, modifications and environments and is capable of modifications within the scope of the inventive concept, either as taught or as a matter of routine skill or knowledge in the relevant art. And that modifications and variations which do not depart from the spirit and scope of the invention are intended to be within the scope of the appended claims.
Claims (7)
1. A screening-based target detection method, comprising:
detecting a connected domain in an image to be detected;
calculating the color similarity of the connected domain and the target sample;
calculating HOG similarity between the connected domain and the target sample;
calculating fusion similarity according to the color similarity and the HOG similarity;
and screening the connected domain according to the fusion similarity to obtain a target detection result.
2. The screening-based object detection method according to claim 1, wherein detecting connected domains in an image to be detected comprises:
extracting connected domains in an image to be detected;
and screening the connected domain according to a preset rule.
3. The screening-based object detection method according to claim 2, wherein screening the connected domain according to a preset rule comprises:
screening out connected domains with widths and heights meeting a screening formula, wherein the screening formula is as follows:
in the method, in the process of the invention,for the width of the target sample, +.>High, high for the target sample>For the set screening coefficient, +.>Is the width of the communicating region->Is the height of the connected domain.
4. The screening-based target detection method according to claim 1, wherein calculating the color similarity between the connected domain and the target sample comprises:
respectively counting color histograms of the connected domain and the target sample;
and calculating the color similarity of the connected domain and the target sample based on the color histogram, wherein the calculation formula of the color similarity is as follows:
in the method, in the process of the invention,for color similarity, ++>Is->Personal communication domain(s)>For the target sample, ++>Represents->The->The value of the individual bin, ">Sample of object->The value in the individual bin; the pixel values of 0-255 are divided into +.>And a histogram bin, each bin representing a value normalized by the L1 norm for the number of pixels having pixel values between 0-15, 16-31, …, 240-255.
5. The screening-based target detection method according to claim 1, wherein calculating HOG similarity between the connected domain and the target sample comprises:
respectively extracting HOG characteristics of the connected domain and the target sample;
and calculating the HOG similarity of the connected domain and the target sample, wherein the HOG similarity has a calculation formula as follows:
in the method, in the process of the invention,for HOG similarity, ++>Is->Candidate connected domain->For the target sample, ++>Represents->The->The value of the individual bin, ">Sample of object->The value of the individual bin; dividing 0-179 degree into +.>The bins are values normalized to the histogram with the L1 norm, 0-29, 30-59, 60-89, …,150-179, respectively.
6. The screening-based target detection method according to claim 1, wherein the calculation formula of the fusion similarity is:
in the method, in the process of the invention,for color similarity, ++>For HOG similarity, ++>And->S is the fusion similarity and is the weight scale factor,is->Personal communication domain(s)>Is the target sample.
7. The screening-based target detection method according to claim 1, wherein screening the connected domain according to the fusion similarity to obtain the target detection result comprises:
and determining the connected domain with the fusion similarity being greater than or equal to a similarity threshold as the detected target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311669123.XA CN117372719A (en) | 2023-12-07 | 2023-12-07 | Target detection method based on screening |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311669123.XA CN117372719A (en) | 2023-12-07 | 2023-12-07 | Target detection method based on screening |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117372719A true CN117372719A (en) | 2024-01-09 |
Family
ID=89402570
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311669123.XA Pending CN117372719A (en) | 2023-12-07 | 2023-12-07 | Target detection method based on screening |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117372719A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080187220A1 (en) * | 2006-12-04 | 2008-08-07 | Lockheed Martin Corporation | Device and method for fast computation of region based image features |
| CN101408942A (en) * | 2008-04-17 | 2009-04-15 | 浙江师范大学 | Method for locating license plate under a complicated background |
| CN102226907A (en) * | 2011-05-24 | 2011-10-26 | 武汉嘉业恒科技有限公司 | License plate positioning method and apparatus based on multiple characteristics |
| CN102663377A (en) * | 2012-03-15 | 2012-09-12 | 华中科技大学 | Character recognition method based on template matching |
| CN102799653A (en) * | 2012-06-29 | 2012-11-28 | 中国科学院自动化研究所 | Logo detection method based on spatial connected domain prepositioning |
| KR20150041428A (en) * | 2013-10-08 | 2015-04-16 | 재단법인대구경북과학기술원 | Method for object detection and apparatus thereof |
| JP2016081271A (en) * | 2014-10-16 | 2016-05-16 | セコム株式会社 | Object detection device |
| CN108765468A (en) * | 2018-05-30 | 2018-11-06 | 中国石油大学(华东) | A kind of method for tracking target and device of feature based fusion |
| CN112819094A (en) * | 2021-02-25 | 2021-05-18 | 北京时代民芯科技有限公司 | Target detection and identification method based on structural similarity measurement |
| CN114373079A (en) * | 2022-01-10 | 2022-04-19 | 东南大学 | Rapid and accurate ground penetrating radar target detection method |
| CN114529584A (en) * | 2022-02-21 | 2022-05-24 | 沈阳理工大学 | Single-target vehicle tracking method based on unmanned aerial vehicle aerial photography |
-
2023
- 2023-12-07 CN CN202311669123.XA patent/CN117372719A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080187220A1 (en) * | 2006-12-04 | 2008-08-07 | Lockheed Martin Corporation | Device and method for fast computation of region based image features |
| CN101408942A (en) * | 2008-04-17 | 2009-04-15 | 浙江师范大学 | Method for locating license plate under a complicated background |
| CN102226907A (en) * | 2011-05-24 | 2011-10-26 | 武汉嘉业恒科技有限公司 | License plate positioning method and apparatus based on multiple characteristics |
| CN102663377A (en) * | 2012-03-15 | 2012-09-12 | 华中科技大学 | Character recognition method based on template matching |
| CN102799653A (en) * | 2012-06-29 | 2012-11-28 | 中国科学院自动化研究所 | Logo detection method based on spatial connected domain prepositioning |
| KR20150041428A (en) * | 2013-10-08 | 2015-04-16 | 재단법인대구경북과학기술원 | Method for object detection and apparatus thereof |
| JP2016081271A (en) * | 2014-10-16 | 2016-05-16 | セコム株式会社 | Object detection device |
| CN108765468A (en) * | 2018-05-30 | 2018-11-06 | 中国石油大学(华东) | A kind of method for tracking target and device of feature based fusion |
| CN112819094A (en) * | 2021-02-25 | 2021-05-18 | 北京时代民芯科技有限公司 | Target detection and identification method based on structural similarity measurement |
| CN114373079A (en) * | 2022-01-10 | 2022-04-19 | 东南大学 | Rapid and accurate ground penetrating radar target detection method |
| CN114529584A (en) * | 2022-02-21 | 2022-05-24 | 沈阳理工大学 | Single-target vehicle tracking method based on unmanned aerial vehicle aerial photography |
Non-Patent Citations (1)
| Title |
|---|
| 叶俊杰等: "基于HOG-SVM分类器的电力屏柜锁孔快速检测方法", 《计算机应用与软件》, vol. 40, no. 01, 12 January 2023 (2023-01-12), pages 134 - 138 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115351598A (en) | Numerical control machine tool bearing detection method | |
| US4542526A (en) | Character recognition apparatus | |
| CN106529559A (en) | Pointer-type circular multi-dashboard real-time reading identification method | |
| CN117351008B (en) | Smart phone panel surface defect detection method | |
| CN107301412B (en) | Pointer instrument reading method based on image processing | |
| CN106643965B (en) | Method for accurately identifying liquid level by utilizing template matching | |
| CN116091504B (en) | Connecting pipe connector quality detection method based on image processing | |
| CN110084830B (en) | Video moving object detection and tracking method | |
| CN115018850A (en) | Method for detecting burrs of punched hole of precise electronic part based on image processing | |
| CN112308854A (en) | Automatic detection method and system for chip surface flaws and electronic equipment | |
| CN112419260A (en) | PCB character area defect detection method | |
| CN116309577B (en) | Intelligent detection method and system for high-strength conveyor belt materials | |
| CN115115637A (en) | Cloth defect detection method based on image pyramid thought | |
| CN113313109A (en) | Semi-quantitative analysis method of fluorescence immunochromatographic test paper | |
| CN109975196B (en) | Reticulocyte detection method and system | |
| CN107729863B (en) | Human finger vein recognition method | |
| CN117372719A (en) | Target detection method based on screening | |
| CN110930393A (en) | Chip material pipe counting method, device and system based on machine vision | |
| CN1739111A (en) | Particle extraction for automatic flow microscope | |
| CN115661194A (en) | Moving object extraction method, system, electronic device and medium | |
| CN114255540A (en) | Method, device, equipment and storage medium for identifying stained paper money | |
| CN112015960A (en) | Clustering method of vehicle-mounted radar measurement data, storage medium and electronic device | |
| CN110428430B (en) | Linear positioning fitting method in region and application | |
| CN114067123A (en) | Instrument automatic identification system, identification method and computer readable storage medium | |
| CN113705672A (en) | Threshold value selection method, system and device for image target detection and storage medium |
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 |