CN107992875B - A kind of well-marked target detection method based on image bandpass filtering - Google Patents
A kind of well-marked target detection method based on image bandpass filtering Download PDFInfo
- Publication number
- CN107992875B CN107992875B CN201711422362.XA CN201711422362A CN107992875B CN 107992875 B CN107992875 B CN 107992875B CN 201711422362 A CN201711422362 A CN 201711422362A CN 107992875 B CN107992875 B CN 107992875B
- Authority
- CN
- China
- Prior art keywords
- image
- bandpass filtering
- result
- point
- carried out
- 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.)
- Active
Links
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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of well-marked target detection methods based on image bandpass filtering, are pre-processed to original image;By image zoom to being sized, and floating-point conversion is carried out, obtains floating-point image;Multiple bandpass filtering is carried out using iir filter to floating-point image, bandpass filtering is carried out using IIR digital filter, bandpass filter is obtained by two low-pass filter difference, further bandpass filtering is carried out to filter result again after carrying out a bandpass filtering to floating-point image, obtains bandpass filtering result;Bandpass filtering result is split, Morphological scale-space is carried out to segmentation result, cancelling noise influences;Result after Morphological scale-space is clustered, cluster result is screened according to target priori, in Screening Treatment, is screened according to known characteristic, and obtain final testing result.The present invention realizes that simply, operation efficiency is high, can effectively detect target.
Description
Technical field
The present invention relates to a kind of well-marked target detection methods based on image bandpass filtering, are suitable for high-definition image complicated field
Well-marked target detection field under scape.
Background technology
The mankind can quickly select a part for image to carry out depth before carrying out high-level vision processing analysis visual scene
Analysis, to reduce the complexity of overall calculation, the mechanism that this selected section key area is handled is that vision is aobvious
Work property.The data volume of processing can be greatly reduced by this method, have great significance for subsequent analysis.
Vision significance includes top-down and two kinds of mechanism from bottom to top.It is to be shown by what image data drove from bottom to top
Work property, i.e. attraction of the image itself to human eye attention;Top-down system be then by purpose drive to image section region
Concern.Usually in Digital Image Processing, mode from bottom to top is more paid close attention to, that is, considers image itself to degree of concern
It influences.
Target detection technique is Digital Image Processing and one important technology of artificial intelligence field, is dedicated to from complexity
Candidate target is partitioned into background, the processing such as further to be tracked, identified.Target detection seeks to fixed in the background
Position goes out position and the size of interesting target.For target motion conditions, it is big with quiet target detection two that moving-target detection can be divided into
Class.
The common method of moving-target detection has:Inter-frame difference, background difference and motion segmentation etc..Inter-frame difference utilizes interframe
Variation realize detection, common method has two frames or three-frame difference method.Background subtraction is using Background or passes through model weight
Structure background obtains motion target area with current frame difference, and common method is mixed Gauss model background modeling method.Movement point
Segmentation method mainly uses light stream to extract motion vector, and is split.
Target detection technique can be divided mainly into two classes in static image:Spy is searched in the picture using feature and sorter network
It sets the goal, and possible interesting target in image is found using the notable method of vision.
It is common to search for specific objective method in the picture using feature and sorter network and be divided into two classes:Traditional classifier side
Method and deep learning frame method.Include commonly in conventional method:Target identification based on HOG features and SVM classifier, base
In Haar features and the target identification of Adaboost graders.Using the target identification of deep learning frame method, including
The Faster RCNN algorithms that R.Girshick is proposed, the SSD algorithms etc. of the propositions such as W.Liu.
In recent years, the notable method of numerous visions is suggested, for detecting to the significant target of human eye in image, such as early stage
ITTI models, the spectrum residual error method etc. that Hou Xiaodi is proposed.Conspicuousness method does not differentiate between target type, only considers target visually
Significance degree, recall rate is high, but it has more flase drop accordingly.
But above-mentioned existing method there are the shortcomings that be mainly reflected in:
(1) for static object detection method, the characteristics of due to grader, the target of particular category can be only identified,
When target is not in identification range, target can not be partitioned into.And due to the limitation of characteristic mass and grader generalization ability,
Detection capability is poor, when there are when difference, possibly can not being detected between target shape and training sample, omission factor and false alarm rate compared with
It is high.Deep learning frame method in static object detection, possesses compared with high detection rate, but detection speed is slower, it is difficult in reality
It is used in the real-time application on border.
(2) notable figure that common vision significance algorithm obtains is ineffective, and the time of complicated conspicuousness method disappears
It consumes huge, can not be applied in real time.
Invention content
Present invention solves the technical problem that being:What deficiency for overcoming the prior art, for image object in complex scene
Detection, a kind of method detected based on conspicuousness is provided, improves the speed of target detection, realizes simple, operation efficiency is high,
Target can effectively be detected and be easy to realize on a hardware platform.
Technical solution of the invention is:A kind of well-marked target detection method based on image bandpass filtering, step is such as
Under:
(1) original image is pre-processed, by the original image zoom to being sized, and carries out floating-point conversion,
Obtain floating-point image;
(2) bandpass filtering twice is carried out using iir filter to floating-point image;
(3) bandpass filtering result is split, obtains segmentation result;
(4) Morphological scale-space is carried out to the segmentation result, cancelling noise point influences;
(5) result after Morphological scale-space is clustered, obtains cluster result;
(6) cluster result is screened according to target priori, in Screening Treatment, is carried out according to known characteristic
Screening, and obtain final testing result.
In the step (1), if described image is coloured image, gray processing is first carried out, obtains gray level image, then will figure
As zoom is to being sized, and floating-point conversion is carried out, obtains floating-point image.
In the step (2), bandpass filtering is carried out using IIR digital filter, bandpass filter is by two low-pass filtering
Device difference obtains, and digital IIR low-pass filters are made of positive and negative filtering operation twice, and successively to horizontal direction and vertical direction
It carries out, the forward filtering recurrence formula of low-pass filter is as follows:
x′n=(1-a) × x 'n-1+a×xn
X in formulanIt is nth point grey scale pixel value, x 'nIt is the forward filtering of nth point as a result, a is filtering parameter;
Inverse filtering recurrence formula is as follows:
x″n=(1-a) x "n+1+a×x′n
Second of bandpass filtering is carried out to filter result again after carrying out first time bandpass filtering to floating-point image, obtains band logical
Filter result.
The value of first time bandpass filtering parameter a is respectively 0.6 and 0.2, and the value of second of bandpass filtering parameter a is respectively
0.3 and 0.1, it can be adjusted according to practical application scene.
In the step (3) binary segmentation is carried out using fixed threshold.
The fixed threshold is 1.0, can be adjusted according to practical application scene.
The step (5), is clustered, progressive scanning picture using breadth first algorithm, by be not classified and segmentation
As a result cluster point is labeled as using 4 neighborhood region-growing methods progress breadth first search cluster as seed point for 1 point
Classify a little, and continued to scan on, has finally obtained cluster result.
In the step (6), it is known that characteristic include:Target sizes, target length-width ratio, the known characteristic can basis
Practical application scene is adjusted.
The advantages of the present invention over the prior art are that:
(1) present invention can effectively detect well-marked target all in image, with existing grader learning algorithm phase
Than can adapt in a variety of different targets.Algorithm missing inspection proposed by the present invention is few, can effectively be detected to target.
(2) iir filter that the present invention uses realizes very high effect, only carries out 4 traversals to image and filter can be completed
Wave calculates, and the relatively used conspicuousness algorithm having, computational efficiency is obviously improved, and processing water in real time can be reached on hardware system
It is flat.
(3) algorithm that the present invention uses is realized simple, can quickly be developed, it is only necessary to slightly adjusting parameter, i.e. a's
Value can adjust on demand, and threshold value can adjust on demand, and characteristic can adjust on demand, you can be suitable for different scenes.
Description of the drawings
Fig. 1 is a kind of flow diagram of the well-marked target detection method based on image bandpass filtering of the present invention;
Fig. 2 is the original image of input;
Fig. 3 is the binary saliency figure after the segmentation that step (3) obtains;
Fig. 4 is the output result finally detected.
Specific implementation mode
The following describes the present invention in detail with reference to the accompanying drawings and embodiments.
A set of target detection software, input image resolution 1280x720, RGB color image.
(1) image is pre-processed.Gray processing is carried out to colour, obtains gray level image, it is down-sampled to 320x180, and
Gray level image is subjected to floating-point conversion.Gray level image is as shown in Figure 2.
(2) multiple bandpass filtering is carried out using iir filter to floating-point image.The present invention using IIR digital filter into
Row bandpass filtering.The bandpass filter is obtained by two low-pass filter difference.
Digital IIR low-pass filters are made of positive and negative filtering operation twice, and successively to horizontal direction and vertical direction into
Row, forward filtering recurrence formula are as follows:
x′n=(1-a) × x 'n-1+a×xn
X in formulanIt is nth point grey scale pixel value, x 'nIt is the forward filtering of nth point as a result, a is filtering parameter.
Inverse filtering recurrence formula is as follows:
x″n=(1-a) x "n+1+a×x′n
X in formula "nIt is the output of nth point horizontal filtering result, filtering parameter a is identical as forward filtering.
After carrying out horizontal direction low-pass filtering operation line by line, vertical direction low-pass filtering operation is carried out by column.
Two filters have different filtering parameters, keep the cutoff frequency of two filter low pass filtering different, are dividing
It is other poor to making after image filtering, obtain bandpass filtering result.
FBP=|b×(FH-FL)|
In formula, FHIt is off the higher low-pass filter filter result of frequency, FLIt is off the lower low-pass filter of frequency
Filter result is higher here to refer to opposite FLIt is higher, it is relatively low here to refer to opposite FLIt is higher, FBPIt is that bandpass filter is defeated
Go out, b is amplification factor.
In the present invention, a bandpass filtering is carried out to original image, then bandpass filtering is further carried out to filter result
Obtain final vision significance figure.
During first time bandpass filtering, the parameter a of two low-pass filters is respectively 0.6 and 0.2, and amplification factor b is
50, carry out a bandpass filtering again to result after obtaining result, filtering parameter is respectively 0.3 and 0.1, and amplification factor b is 1, is obtained
It is distributed to conspicuousness, parameter can be adjusted according to actual demand.
(3) bandpass filtering result is split, carries out binary segmentation using fixed threshold, obtains segmentation result:
T is threshold value in formula, and threshold value is decided to be 1.0 by experiment and experience, can be adjusted according to actual demand.
(4) opening operation is used to operate the image used, the influence of noise smaller for filtering out scale.Filtered knot
Fruit is as shown in Figure 3.
(5) segmentation result is clustered, is clustered using breadth first algorithm.Progressive scanning picture will not divided
Class and segmentation result be 1 point be used as seed point, using 4 neighborhood region-growing methods progress breadth first search cluster, will gather
Class point is labeled as having classified a little, and continues to scan on, and obtains the centre coordinate and outer rim of each target.
(6) each target that step (5) obtains is screened according to target priori.In Screening Treatment, according to
The characteristic known, such as:Target sizes, target length-width ratio etc. are screened, and export final testing result.The embodiment of the present invention
In, use following priori:
(1) target size is more than 5x5;
(2) target size is less than 60x60;
(3) when target length is more than 20, length-width ratio should be less than 4.
Screening analysis is carried out to result according to this three, ineligible target is screened out, obtains final detection knot
Fruit, prior information can be adjusted according to actual demand.
Detection block is drawn on original image after detection block length and width are expanded 4 times, obtained image is as shown in Figure 4.
Two cars in figure are to the significant target of human eye simultaneously it is also desirable that the target searched, is effectively split
And Overlapping display frame.In figure 3, most of flase drop target is all screened out in subsequent screening process, is only had in Fig. 4
4 flase drop objects.The present invention handles frame per second up to 30 frames/second on DSP architecture, reaches real-time processing requirement.
Above example is provided just for the sake of the description purpose of the present invention, and is not intended to limit the scope of the present invention.This
The range of invention is defined by the following claims.It does not depart from spirit and principles of the present invention and the various equivalent replacements made and repaiies
Change, should all cover within the scope of the present invention.
Claims (7)
1. a kind of well-marked target detection method based on image bandpass filtering, which is characterized in that include the following steps:
(1) original image is pre-processed, by the original image zoom to being sized, and carries out floating-point conversion, obtains
Floating-point image;
(2) bandpass filtering twice is carried out using iir filter to floating-point image;
(3) bandpass filtering result is split, obtains segmentation result;
(4) Morphological scale-space is carried out to the segmentation result, cancelling noise point influences;
(5) result after Morphological scale-space is clustered, obtains cluster result;
(6) cluster result is screened according to target priori, in Screening Treatment, is sieved according to known characteristic
Choosing, and obtain final testing result;
In the step (2), bandpass filtering is carried out using IIR digital filter, bandpass filter is poor by two low-pass filters
Get, digital IIR low-pass filters are made of positive and negative filtering operation twice, and successively to horizontal direction and vertical direction into
Row, the forward filtering recurrence formula of low-pass filter are as follows:
x'n=(1-a) × x'n-1+a×xn
X in formulanIt is nth point grey scale pixel value, x'nIt is the forward filtering of nth point as a result, a is filtering parameter;
Inverse filtering recurrence formula is as follows:
x”n=(1-a) x "n+1+a×x'n
Second of bandpass filtering is carried out to filter result again after carrying out first time bandpass filtering to floating-point image, obtains bandpass filtering
As a result.
2. the well-marked target detection method according to claim 1 based on image bandpass filtering, it is characterised in that:The step
Suddenly in (1), if described image is coloured image, gray processing is first carried out, obtains gray level image, then by image zoom to setting ruler
It is very little, and floating-point conversion is carried out, obtain floating-point image.
3. the well-marked target detection method according to claim 1 based on image bandpass filtering, it is characterised in that:For the first time
The value of bandpass filtering parameter a is respectively 0.6 and 0.2, and the value of second of bandpass filtering parameter a is respectively 0.3 and 0.1.
4. the well-marked target detection method according to claim 1 based on image bandpass filtering, it is characterised in that:The step
Suddenly in (3) binary segmentation is carried out using fixed threshold.
5. the well-marked target detection method according to claim 4 based on image bandpass filtering, it is characterised in that:It is described solid
It is 1.0 to determine threshold value.
6. the well-marked target detection method according to claim 1 based on image bandpass filtering, it is characterised in that:The step
Suddenly (5), are clustered using breadth first algorithm, progressive scanning picture, using be not classified and segmentation result be 1 point as
Seed point carries out breadth first search cluster using 4 neighborhood region-growing methods, cluster point is labeled as having classified a little, and continue
Scanning, finally obtains cluster result.
7. the well-marked target detection method according to claim 1 based on image bandpass filtering, it is characterised in that:The step
Suddenly in (6), it is known that characteristic include:Target sizes, target length-width ratio, the known characteristic can according to practical application scene into
Row adjustment.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711422362.XA CN107992875B (en) | 2017-12-25 | 2017-12-25 | A kind of well-marked target detection method based on image bandpass filtering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711422362.XA CN107992875B (en) | 2017-12-25 | 2017-12-25 | A kind of well-marked target detection method based on image bandpass filtering |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107992875A CN107992875A (en) | 2018-05-04 |
CN107992875B true CN107992875B (en) | 2018-10-26 |
Family
ID=62041860
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711422362.XA Active CN107992875B (en) | 2017-12-25 | 2017-12-25 | A kind of well-marked target detection method based on image bandpass filtering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107992875B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110910421B (en) * | 2019-11-11 | 2022-11-11 | 西北工业大学 | Weak and small moving object detection method based on block characterization and variable neighborhood clustering |
CN111145156A (en) * | 2019-12-27 | 2020-05-12 | 创新奇智(南京)科技有限公司 | Rapid screw surface defect detection method |
CN111784630A (en) * | 2020-05-18 | 2020-10-16 | 广州信瑞医疗技术有限公司 | Method and device for segmenting components of pathological image |
CN111784698A (en) * | 2020-07-02 | 2020-10-16 | 广州信瑞医疗技术有限公司 | Image self-adaptive segmentation method and device, electronic equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103714537A (en) * | 2013-12-19 | 2014-04-09 | 武汉理工大学 | Image saliency detection method |
CN104217438A (en) * | 2014-09-19 | 2014-12-17 | 西安电子科技大学 | Image significance detection method based on semi-supervision |
CN104992183A (en) * | 2015-06-25 | 2015-10-21 | 中国计量学院 | Method for automatic detection of substantial object in natural scene |
CN105975911A (en) * | 2016-04-28 | 2016-09-28 | 大连民族大学 | Energy perception motion significance target detection algorithm based on filter |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104103082A (en) * | 2014-06-06 | 2014-10-15 | 华南理工大学 | Image saliency detection method based on region description and priori knowledge |
CN105574534B (en) * | 2015-12-17 | 2019-03-26 | 西安电子科技大学 | Conspicuousness object detection method based on sparse subspace clustering and low-rank representation |
CN106447699B (en) * | 2016-10-14 | 2019-07-19 | 中国科学院自动化研究所 | High iron catenary object detecting and tracking method based on Kalman filtering |
CN107229917B (en) * | 2017-05-31 | 2019-10-15 | 北京师范大学 | A kind of several remote sensing image general character well-marked target detection methods based on iteration cluster |
-
2017
- 2017-12-25 CN CN201711422362.XA patent/CN107992875B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103714537A (en) * | 2013-12-19 | 2014-04-09 | 武汉理工大学 | Image saliency detection method |
CN104217438A (en) * | 2014-09-19 | 2014-12-17 | 西安电子科技大学 | Image significance detection method based on semi-supervision |
CN104992183A (en) * | 2015-06-25 | 2015-10-21 | 中国计量学院 | Method for automatic detection of substantial object in natural scene |
CN105975911A (en) * | 2016-04-28 | 2016-09-28 | 大连民族大学 | Energy perception motion significance target detection algorithm based on filter |
Non-Patent Citations (1)
Title |
---|
SDSP: A novel saliency detection method by combining simple priors;LinZhang 等;《2013 IEEE International Conference on Image Processing》;20130918;第171-175页 * |
Also Published As
Publication number | Publication date |
---|---|
CN107992875A (en) | 2018-05-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107992875B (en) | A kind of well-marked target detection method based on image bandpass filtering | |
Bautista et al. | Convolutional neural network for vehicle detection in low resolution traffic videos | |
WO2022099598A1 (en) | Video dynamic target detection method based on relative statistical features of image pixels | |
CN110929593B (en) | Real-time significance pedestrian detection method based on detail discrimination | |
CN110688987A (en) | Pedestrian position detection and tracking method and system | |
CN107315990B (en) | Pedestrian detection algorithm based on XCS-LBP characteristics | |
CN109255326B (en) | Traffic scene smoke intelligent detection method based on multi-dimensional information feature fusion | |
CN105405138B (en) | Waterborne target tracking based on conspicuousness detection | |
CN110807384A (en) | Small target detection method and system under low visibility | |
CN109190456B (en) | Multi-feature fusion overlook pedestrian detection method based on aggregated channel features and gray level co-occurrence matrix | |
Tuominen et al. | Cloud detection and movement estimation based on sky camera images using neural networks and the Lucas-Kanade method | |
CN114627269A (en) | Virtual reality security protection monitoring platform based on degree of depth learning target detection | |
CN108073940A (en) | A kind of method of 3D object instance object detections in unstructured moving grids | |
Hassan et al. | A comparative study on hsv-based and deep learning-based object detection algorithms for pedestrian traffic light signal recognition | |
WO2014066218A2 (en) | Cast recognition method and device, and urine analyzer | |
CN111027564A (en) | Low-illumination imaging license plate recognition method and device based on deep learning integration | |
Patro | Design and implementation of novel image segmentation and BLOB detection algorithm for real-time video surveillance using DaVinci processor | |
Shi et al. | Moving cast shadow detection in video based on new chromatic criteria and statistical modeling | |
CN116311212B (en) | Ship number identification method and device based on high-speed camera and in motion state | |
CN109241932A (en) | A kind of thermal infrared human motion recognition method based on movement variogram phase property | |
Meshram et al. | Convolution Neural Network based Hand Gesture Recognition System | |
US20170053399A1 (en) | Method and apparatus for processing block to be processed of urine sediment image | |
Gooda et al. | Automatic detection of road cracks using EfficientNet with residual U-net-based segmentation and YOLOv5-based detection | |
CN114463379A (en) | Dynamic capturing method and device for video key points | |
Shi et al. | License plate detection based on convolutional neural network and visual feature |
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 |