CN103473753A - Target detection method based on multi-scale wavelet threshold denoising - Google Patents
Target detection method based on multi-scale wavelet threshold denoising Download PDFInfo
- Publication number
- CN103473753A CN103473753A CN2013103916208A CN201310391620A CN103473753A CN 103473753 A CN103473753 A CN 103473753A CN 2013103916208 A CN2013103916208 A CN 2013103916208A CN 201310391620 A CN201310391620 A CN 201310391620A CN 103473753 A CN103473753 A CN 103473753A
- Authority
- CN
- China
- Prior art keywords
- image
- images
- denoising
- wavelet
- frame
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 claims abstract description 36
- 238000012544 monitoring process Methods 0.000 claims abstract description 22
- 238000012545 processing Methods 0.000 claims abstract description 20
- 230000009466 transformation Effects 0.000 claims description 8
- 238000011410 subtraction method Methods 0.000 claims description 7
- 230000002159 abnormal effect Effects 0.000 abstract description 5
- 230000000694 effects Effects 0.000 abstract description 4
- 230000003287 optical effect Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000007306 turnover Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Closed-Circuit Television Systems (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a target detection method based on multi-scale wavelet threshold denoising, and belongs to the field of target detection. The target detection method comprises the steps: firstly performing grayscale processing on collected colorful images, then performing space multi-scale wavelet threshold denoising processing on the images to obtain denoised original grayscale images; then performing the target detection on an image at a certain moment by using a combined model of a background different method and a frame difference method, to obtain denoised three-frame-difference images; finally, determining whether switching to a next group of monitoring images or giving an alarm and switching to the next group of monitoring images according to the obtained three-frame-difference images. According to the target detection method, the negative influence of noise signals can be eliminated as far as possible by adopting a wavelet threshold denoising method to denoise noise-containing images, so that individual abnormal data in samples can not generate large influence on the final results; then the target detection can be performed by constructing the combined detection model of the background difference method and the frame difference method, and the moving detection effect can be enhanced.
Description
Technical field
The present invention relates to a kind of object detection method based on the multi-scale wavelet threshold denoising, belong to object detection field.
Background technology
Target detection is an important subject in vision field, is to carry out the basis that target classification, tracking and behavior are understood, in numerous areas extensive application such as intelligent video monitoring, robot navigation and Medical Image Processing.And the existence of noise has destroyed the correlativity of image at aspects such as structures, be unfavorable for the extraction of feature.But up to now, on the basis of detecting at the reference all types of target, research finds that detection model in the past do not consider the impact on testing result of the noise that produces in image acquisition process and abnormal data, as the shake of the fluctuation of some object in the unexpected variation of illumination, real background image, video camera, moving object turnover scene on the impact of former scene etc.
Detection algorithm for target has: background subtraction method, centroid tracking method, template matching method, frame differential method, Mean shift method, particle filter (PF) algorithm, optical flow method, pipeline filter method etc.Wherein relatively more commonly used is optical flow method, background subtraction method and frame difference method.Optical flow method utilizes the time domain of the pixel intensity in image sequence to change and correlativity is determined " motion " of location of pixels separately, can not need to know in advance any information of scene; The frame-to-frame differences method is more responsive to the motion of object, is subject to the impact of light less, but it is not accurate enough to detect target; The background subtraction method can be partitioned into moving object quickly and accurately, but easily is subject to the variable effect of environment and illumination.
On the basis of reference all types of target detection method, research finds that detection model in the past do not consider the impact on testing result of the noise that produces in image acquisition process and abnormal data, and the existence of noise has destroyed the correlativity of image at aspects such as structures, be unfavorable for the extraction of feature.
Summary of the invention
The invention provides a kind of object detection method based on the multi-scale wavelet threshold denoising, for overcoming prior art, not consider the impact on testing result of the noise that produces in image acquisition process and abnormal data.
Technical scheme of the present invention is: a kind of object detection method based on the multi-scale wavelet threshold denoising, at first the coloured image collected is carried out to the gray scale processing, again image is carried out to the processing of Multi scale wavelet threshold denoising, obtain the original-gray image after denoising; Then adopt the built-up pattern of background subtraction method and frame difference method to look like to carry out target detection to a certain time chart, obtain three frame difference images; Last three frame difference images according to obtaining determine that being switched to next organizes monitoring image or warning and be switched to next group monitoring image.
The concrete steps of the object detection method of described multi-scale wavelet threshold denoising are as follows:
A, at first the coloured image collected is carried out to the gray scale processing, then it is carried out to wavelet transformation, obtain one group of wavelet coefficient
w j,
k ;
B, wavelet coefficient is carried out to the threshold denoising processing, by the wavelet coefficient of signals and associated noises and threshold value coefficient
compare: for being greater than
point be punctured into the difference of this point and threshold value, for being less than
point be punctured into this point value and threshold value and:
(1)
In formula:
,
for the mean variance of noise,
nit is the pixel number of image;
yardstick after meaning to remap
jupper position
kthe wavelet coefficient at place,
k=0,1 ... 2
j -1;
C, the point to wavelet coefficient in ± 20% scope, carry out reviewing of Multi scale, further judges that this point is signal or noise: noise in this way, and this wavelet coefficient is zero; Signal, do not do operation to this coefficient in this way;
D, the wavelet coefficient after denoising is carried out to wavelet inverse transformation, obtain the original-gray image after denoising;
E, by the current frame image that obtains respectively with denoising after original-gray image carry out the background subtraction computing, obtain the background subtraction image
b i (
x,
y):
In formula:
a i (
x,
y) be current frame image;
a t (
x,
y) be the background image of continuous three frames;
F, make respectively frame difference method according to three groups of background subtraction images obtaining and process, with the
ithe frame picture deducts
i-1 frame picture, obtain bianry image
d 1(
x,
y), then use
i+ 1 frame picture deducts
ithe frame picture, obtain bianry image
d 2(
x,
y):
(3)
In formula:
tfor the gray threshold of setting;
G, two groups of bianry images that will obtain carry out and computing, obtain three frame difference images
d(
x,
y):
Three frame difference images that H, basis obtain are judged:
If
d(
x,
y) be 0, be switched to next group monitoring image;
If
d(
x,
y) be 1, report to the police and be switched to next group monitoring image.
Described yardstick is 3, gray threshold
tbe 20.
Principle of work of the present invention is:
At first the coloured image collected is carried out to the gray scale processing, then image is carried out to the processing of Multi scale wavelet threshold denoising, obtain the original-gray image after denoising.
And then by the current frame image that obtains respectively with denoising after original image carry out the background subtraction computing, obtain the background subtraction image
b i (
x,
y); Three groups of background subtraction images that obtain are done respectively to frame difference method again and process, with the
ithe frame picture deducts
i-1 frame picture, obtain bianry image
d 1(
x,
y) (if the gray-scale value of bianry image is greater than threshold value, thinks that this point is foreground point, otherwise be background dot), then use
i+ 1 frame picture deducts
ithe frame picture, obtain bianry image
d 2(
x,
y), finally use
d 1(
x,
y) and
d 2(
x,
y) carry out AND operation, just can obtain three frame difference images
d(
x,
y):
According to three frame difference images that obtain
d(
x,
y) judged:
If
d(
x,
y) be 0, be switched to next group monitoring image;
If
d(
x,
y) be 1, report to the police and be switched to next group monitoring image.
The invention has the beneficial effects as follows: at first by the wavelet threshold denoising method, noisy image is carried out to denoising, eliminated as much as possible the negative effect of noise signal, make the indivedual abnormal datas in sample not have too much influence to net result; Then the combine detection model by the poor method of structural setting and frame difference method is detected target, has improved the effect of motion detection.
The accompanying drawing explanation
The FB(flow block) that Fig. 1 is multi-scale wavelet threshold denoising in the present invention;
Fig. 2 is target detection process flow diagram in the present invention;
The gray level image that Fig. 3 is the original background image that in the present invention, embodiment 1 gathers;
Fig. 4 is the original-gray image after embodiment 1 denoising in the present invention;
Fig. 5 is the gradation of image image that in the present invention, a certain moment of embodiment 1 gathers;
Three frame difference images that Fig. 6 is embodiment 1 in the present invention.
Embodiment
Below in conjunction with drawings and Examples, the invention will be further described, but content of the present invention is not limited to described scope.
Embodiment 1: as shown in Fig. 1-6, a kind of object detection method based on the multi-scale wavelet threshold denoising, at first the coloured image collected is carried out to the gray scale processing and obtain gray level image as shown in Figure 3, again image is carried out to the processing of Multi scale wavelet threshold denoising, obtain original-gray image after denoising as shown in Figure 4; Then adopt the built-up pattern of background subtraction method and frame difference method to look like to carry out target detection (a certain time chart picture as shown in Figure 5) to a certain time chart, obtain three frame difference images as shown in Figure 6; Last three frame difference images according to obtaining determine that being switched to next organizes monitoring image or warning and be switched to next group monitoring image:
At first the coloured image collected is carried out to gray scale and processes and to obtain gray level image as shown in Figure 3, then to its carry out wavelet transformation (yardstick is 3,
k=0,1 ... 2
3-1), obtain one group of wavelet coefficient
w j,
k ; According to formula (1), wavelet coefficient is carried out to the threshold denoising processing, by the wavelet coefficient of signals and associated noises and threshold value coefficient
compare, be greater than
point be punctured into the difference of this point and threshold value, be less than
be punctured into this point value and threshold value and; The point in ± 20% scope to wavelet coefficient, carry out reviewing of Multi scale again, further judges that this point is signal or noise: noise in this way, and this wavelet coefficient is zero; Signal, do not do operation to this coefficient in this way; Then the wavelet coefficient after denoising is carried out to wavelet inverse transformation, obtain original-gray image after denoising as shown in Figure 4;
Then according to formula (2) by the current frame image that obtains respectively with denoising after original-gray image carry out the background subtraction computing, obtain the background subtraction image
b i (
x,
y); Three groups of background subtraction images that will obtain according to formula (3) and formula (4) are made respectively frame difference method and are processed, with the
ithe frame picture deducts
i-1 frame picture, obtain bianry image
d 1(
x,
y) (if the gray-scale value of bianry image is greater than threshold value, thinks that this point is foreground point, otherwise be background dot, take herein
tbe 20), then use
i+ 1 frame picture deducts
ithe frame picture, obtain bianry image
d 2(
x,
y); Two groups of bianry images that will obtain according to formula (5) carry out and computing, obtain three frame difference images
d(
x,
y) as shown in Figure 6;
Last three frame difference images according to obtaining
d(
x,
y) judged: if
d(
x,
y) be 0, be switched to next group monitoring image; If
d(
x,
y) be 1, report to the police and be switched to next group monitoring image.Reported to the police and be switched to next group monitoring image according to three frame difference images as shown in Figure 6 that obtain (being the target detection result).
Embodiment 2: as shown in Fig. 1-6, a kind of object detection method based on the multi-scale wavelet threshold denoising, at first the coloured image collected is carried out to the gray scale processing, then image is carried out to the processing of Multi scale wavelet threshold denoising, obtain the original-gray image after denoising; Then adopt the built-up pattern of background subtraction method and frame difference method to look like to carry out target detection to a certain time chart, obtain three frame difference images; Last three frame difference images according to obtaining determine that being switched to next organizes monitoring image or warning and be switched to next group monitoring image.
The concrete steps of the object detection method of described multi-scale wavelet threshold denoising are as follows:
A, at first the coloured image collected is carried out to the gray scale processing, then it is carried out to wavelet transformation, obtain one group of wavelet coefficient
w j,
k ;
B, wavelet coefficient is carried out to the threshold denoising processing, by the wavelet coefficient of signals and associated noises and threshold value coefficient
compare: for being greater than
point be punctured into the difference of this point and threshold value, for being less than
point be punctured into this point value and threshold value and:
In formula:
,
for the mean variance of noise,
nit is the pixel number of image;
yardstick after meaning to remap
jupper position
kthe wavelet coefficient at place,
k=0,1 ... 2
j -1;
C, the point to wavelet coefficient in ± 20% scope, carry out reviewing of Multi scale, further judges that this point is signal or noise: noise in this way, and this wavelet coefficient is zero; Signal, do not do operation to this coefficient in this way;
D, the wavelet coefficient after denoising is carried out to wavelet inverse transformation, obtain the original-gray image after denoising;
E, by the current frame image that obtains respectively with denoising after original-gray image carry out the background subtraction computing, obtain the background subtraction image
b i (
x,
y):
In formula:
a i (
x,
y) be current frame image;
a t (
x,
y) be the background image of continuous three frames;
F, make respectively frame difference method according to three groups of background subtraction images obtaining and process, with the
ithe frame picture deducts
i-1 frame picture, obtain bianry image
d 1(
x,
y), then use
i+ 1 frame picture deducts
ithe frame picture, obtain bianry image
d 2(
x,
y):
In formula:
tfor the gray threshold of setting;
G, two groups of bianry images that will obtain carry out and computing, obtain three frame difference images
d(
x,
y):
Three frame difference images that H, basis obtain are judged:
If
d(
x,
y) be 0, be switched to next group monitoring image;
If
d(
x,
y) be 1, report to the police and be switched to next group monitoring image.
Described yardstick is 3, gray threshold
tbe 20.
The above is explained in detail the specific embodiment of the present invention by reference to the accompanying drawings, but the present invention is not limited to above-mentioned embodiment, in the ken possessed those of ordinary skills, can also under the prerequisite that does not break away from aim of the present invention, make various variations.
Claims (3)
1. the object detection method based on the multi-scale wavelet threshold denoising is characterized in that: at first the coloured image collected is carried out to the gray scale processing, then image is carried out to the processing of Multi scale wavelet threshold denoising, obtain the original-gray image after denoising; Then adopt the built-up pattern of background subtraction method and frame difference method to look like to carry out target detection to a certain time chart, obtain three frame difference images; Last three frame difference images according to obtaining determine that being switched to next organizes monitoring image or warning and be switched to next group monitoring image.
2. the object detection method based on the multi-scale wavelet threshold denoising according to claim 1, it is characterized in that: the concrete steps of the object detection method of described multi-scale wavelet threshold denoising are as follows:
A, at first the coloured image collected is carried out to the gray scale processing, then it is carried out to wavelet transformation, obtain one group of wavelet coefficient
w j,
k ;
B, wavelet coefficient is carried out to the threshold denoising processing, by the wavelet coefficient of signals and associated noises and threshold value coefficient
compare: for being greater than
point be punctured into the difference of this point and threshold value, for being less than
point be punctured into this point value and threshold value and:
In formula:
,
for the mean variance of noise,
nit is the pixel number of image;
yardstick after meaning to remap
jupper position
kthe wavelet coefficient at place,
k=0,1 ... 2
j -1;
C, the point to wavelet coefficient in ± 20% scope, carry out reviewing of Multi scale, further judges that this point is signal or noise: noise in this way, and this wavelet coefficient is zero; Signal, do not do operation to this coefficient in this way;
D, the wavelet coefficient after denoising is carried out to wavelet inverse transformation, obtain the original-gray image after denoising;
E, by the current frame image that obtains respectively with denoising after original-gray image carry out the background subtraction computing, obtain the background subtraction image
b i (
x,
y):
In formula:
a i (
x,
y) be current frame image;
a t (
x,
y) be the background image of continuous three frames;
F, make respectively frame difference method according to three groups of background subtraction images obtaining and process, with the
ithe frame picture deducts
i-1 frame picture, obtain bianry image
d 1(
x,
y), then use
i+ 1 frame picture deducts
ithe frame picture, obtain bianry image
d 2(
x,
y):
In formula:
tfor the gray threshold of setting;
G, two groups of bianry images that will obtain carry out and computing, obtain three frame difference images
d(
x,
y):
Three frame difference images that H, basis obtain are judged:
If
d(
x,
y) be 0, be switched to next group monitoring image;
If
d(
x,
y) be 1, report to the police and be switched to next group monitoring image.
3. the object detection method based on the multi-scale wavelet threshold denoising according to claim 1 and 2, it is characterized in that: described yardstick is 3, gray threshold
tbe 20.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013103916208A CN103473753A (en) | 2013-09-02 | 2013-09-02 | Target detection method based on multi-scale wavelet threshold denoising |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013103916208A CN103473753A (en) | 2013-09-02 | 2013-09-02 | Target detection method based on multi-scale wavelet threshold denoising |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103473753A true CN103473753A (en) | 2013-12-25 |
Family
ID=49798587
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013103916208A Pending CN103473753A (en) | 2013-09-02 | 2013-09-02 | Target detection method based on multi-scale wavelet threshold denoising |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103473753A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103886290A (en) * | 2014-03-13 | 2014-06-25 | 西安交通大学 | Coal flow video detection method under complex background |
CN109460705A (en) * | 2018-09-26 | 2019-03-12 | 北京工业大学 | Oil pipeline monitoring method based on machine vision |
CN110097132A (en) * | 2019-05-07 | 2019-08-06 | 电子科技大学 | A method of identification digital photos and shooting camera |
CN110458144A (en) * | 2019-08-21 | 2019-11-15 | 杭州品茗安控信息技术股份有限公司 | Object area intrusion detection method, system, device and readable storage medium storing program for executing |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009116686A (en) * | 2007-11-07 | 2009-05-28 | Nippon Telegr & Teleph Corp <Ntt> | Imaging target detection apparatus and method |
CN101715070A (en) * | 2009-11-16 | 2010-05-26 | 南京邮电大学 | Method for automatically updating background in specifically monitored video |
CN101883209A (en) * | 2010-05-31 | 2010-11-10 | 中山大学 | Method by integrating background model and three-frame difference to detect video background |
-
2013
- 2013-09-02 CN CN2013103916208A patent/CN103473753A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009116686A (en) * | 2007-11-07 | 2009-05-28 | Nippon Telegr & Teleph Corp <Ntt> | Imaging target detection apparatus and method |
CN101715070A (en) * | 2009-11-16 | 2010-05-26 | 南京邮电大学 | Method for automatically updating background in specifically monitored video |
CN101883209A (en) * | 2010-05-31 | 2010-11-10 | 中山大学 | Method by integrating background model and three-frame difference to detect video background |
Non-Patent Citations (6)
Title |
---|
BHAVIK R.BAKSHI: "MULTISCALE ANALYSIS AND MODELING USING WAVELETS", 《JOURNAL OF CHEMOMETRICS》 * |
匡海鹏等: "基于中值预滤波的航空图像小波去噪算法研究", 《应用光学》 * |
唐俐等: "运动目标检测的三帧差分和背景消减研究", 《科技信息》 * |
李泽峰等: "基于改进背景差法的运动目标检测", 《仪表技术》 * |
王孝艳等: "运动目标检测的三帧差法算法研究", 《沈阳理工大学学报》 * |
莫林等: "一种基于背景减除与三帧差分的运动目标检测算法", 《微计算机信息》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103886290A (en) * | 2014-03-13 | 2014-06-25 | 西安交通大学 | Coal flow video detection method under complex background |
CN109460705A (en) * | 2018-09-26 | 2019-03-12 | 北京工业大学 | Oil pipeline monitoring method based on machine vision |
CN110097132A (en) * | 2019-05-07 | 2019-08-06 | 电子科技大学 | A method of identification digital photos and shooting camera |
CN110458144A (en) * | 2019-08-21 | 2019-11-15 | 杭州品茗安控信息技术股份有限公司 | Object area intrusion detection method, system, device and readable storage medium storing program for executing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104392468B (en) | Based on the moving target detecting method for improving visual background extraction | |
CN106327520B (en) | Moving target detection method and system | |
CN102307274B (en) | Motion detection method based on edge detection and frame difference | |
CN108596169B (en) | Block signal conversion and target detection method and device based on video stream image | |
CN103093198B (en) | A kind of crowd density monitoring method and device | |
CN104700430A (en) | Method for detecting movement of airborne displays | |
CN106548488B (en) | A kind of foreground detection method based on background model and inter-frame difference | |
CN108022249B (en) | Automatic extraction method for target region of interest of remote sensing video satellite moving vehicle | |
CN105374049B (en) | Multi-corner point tracking method and device based on sparse optical flow method | |
CN105760846A (en) | Object detection and location method and system based on depth data | |
CN105741319B (en) | Improvement visual background extracting method based on blindly more new strategy and foreground model | |
CN110555868A (en) | method for detecting small moving target under complex ground background | |
Lian et al. | A novel method on moving-objects detection based on background subtraction and three frames differencing | |
KR20130053980A (en) | Obstacle detection method using image data fusion and apparatus | |
Zhao et al. | Principal curvature for infrared small target detection | |
CN110728668A (en) | Airspace high-pass filter for maintaining small target form | |
CN109166137A (en) | For shake Moving Object in Video Sequences detection algorithm | |
CN103473753A (en) | Target detection method based on multi-scale wavelet threshold denoising | |
CN108876807B (en) | Real-time satellite-borne satellite image moving object detection tracking method | |
CN112435278B (en) | Visual SLAM method and device based on dynamic target detection | |
CN108010050B (en) | Foreground detection method based on adaptive background updating and selective background updating | |
CN103400395A (en) | Light stream tracking method based on HAAR feature detection | |
CN103093481B (en) | A kind of based on moving target detecting method under the static background of watershed segmentation | |
CN112613456A (en) | Small target detection method based on multi-frame differential image accumulation | |
CN116228712A (en) | Multi-scale slope disaster monitoring method, system and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20131225 |