CN105931220B - Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy - Google Patents
Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy Download PDFInfo
- Publication number
- CN105931220B CN105931220B CN201610227754.XA CN201610227754A CN105931220B CN 105931220 B CN105931220 B CN 105931220B CN 201610227754 A CN201610227754 A CN 201610227754A CN 105931220 B CN105931220 B CN 105931220B
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- road area
- seed point
- detected
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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]
-
- 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
- G06V10/60—Extraction of image or video features relating to illumination properties, e.g. using a reflectance or lighting model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30192—Weather; Meteorology
Abstract
The present invention relates to the traffic haze visibility detecting methods based on dark channel prior Yu minimum image entropy.In image characteristics extraction module, image to be detected I is handled by dark channel prior, obtains the rough estimate evaluation of atmospheric transmissivity, smooth micronization processes are carried out to transmissivity rough estimate evaluation using Steerable filter edge-smoothing operator, obtain the depth information of each pixel;In road area extraction module, using region growing methods extract the road area in I, it includes that initial seed point is arranged, setting target growth region, the minimum value for calculating neighboring gradation difference, judges whether object pixel is road area, update seed point that the region, which increases,;In visibility estimation module, the minimum image entropy in the region is calculated, extinction coefficient optimal value is obtained, haze visibility size is effectively estimated out.The advantage of the present invention is addition target interest region extraction step during seeking image entropy, reduces the calculation amount of algorithm, improves arithmetic speed and robustness.
Description
Technical field
The present invention relates to the enhancing of image and recovery techniques, and in particular to one kind is based on dark channel prior and minimum image entropy
Traffic haze visibility detecting method and detecting system.
Background technology
Atmospheric visibility reflects a regional air pollution degree, can also influence human health and traffic safety.It is special
Not being the appearance of haze can make visibility become very poor, and can lead to many problems:People's go off daily will become very not
Convenient, driver needs higher attention and faster response ability, in order to avoid cause traffic accident.To sum up, visibility detects
Become and its necessary and meaningful.
It is the classical technology of the haze visibility detection based on video image to seek image inflection method, according to human eye vision spy
Property, the corresponding position of visibility is just that eye-observation object is visible with invisible separation, still can pass through
The position of the inflection point of image brightness properties is found to acquire visibility value.For seeking the classical haze visibility of image inflection method
Detection technique, the video camera to acquiring image are accurately demarcated, and need to know the focal length of video camera, inclination angle and vertically
The parameters such as height, therefore be applicable in different scenes and be restricted;And the result of visibility detection is marked by whether Accurate Calibration is influenced
It is more sensitive to noise to determine process.
It is the new haze visibility based on video image relative to image inflection method, dark channel prior theory is asked
Detection technique.Since most of characteristics of image substantially conforms to this feature:In the regional area of the non-sky of the overwhelming majority of image
In, certain some pixel, which always has at least one Color Channel, has very low value;So dark channel prior theory can be adapted for
The road of any scene only fails to sky areas.Since road area is target area, therefore the theory is not limited by scene condition
System, it is applied widely, it is widely used in the fields such as intelligent transportation, image defogging, image enhancement at present, and show very
Good application prospect.But countless facts have proved relies only on the technology there are many defects in image processing fields such as defoggings.Such as public affairs
The number of opening be CN104063853A, entitled " a kind of raising traffic video image clarity method based on dark technology " hair
Bright patent is to carry out defogging processing to traffic video image using dark channel prior theory, it is a kind of image enchancing method, but
The problem is that the image transmission rate figure obtained by dark channel prior theory is there are apparent halo effect and blocking artifact,
Transmittance figure is accurate not enough, can only be used as a thick value, still have at a distance from larger with the precision of expectation.
Invention content
The technical problem to be solved in the present invention is to provide a kind of road video image traffic haze visibility detecting methods, accurate
Really estimation atmospheric extinction coefficient effectively accurately estimates the difficulty of atmospheric visibility under the conditions of solution haze on daytime.
In order to solve the above technical problems, the present invention proposes a kind of traffic haze based on dark channel prior Yu minimum image entropy
Visibility detecting method, the technical scheme comprises the following steps:
1) under the conditions of haze on daytime, the video sequence of road monitoring is acquired, selects a wherein frame as to be detected
Image I (x);
2) atmospheric transmissivity is calculated
J (x) is the recovery image obtained by above-mentioned image to be detected I (x), and A is sky brightness, and t (x) is image transmission
Rate, then foggy image formed model be:
I (x)=J (x) t (x)+A (1-t (x))
Then minimum operation twice is asked to above formula, obtains following formula:
Wherein JcAnd IcIndicate that each channel of image, ω (x) indicate a window centered on pixel x,
For arbitrary input picture J (x), at least one Color Channel has very low value in image, i.e.,:
Atmospheric transmissivity is estimated as follows:
Wherein, sky brightness A is the mean value size of 0.1% brightest pixel point in dark channel diagram;
3) transmissivity is optimized and revised
For as the rectangular window w centered on number kkIf t (x) is after Steerable filterImage is oriented to using to be detected
Image I (x), filtering output imageMeet following relationship with image I (x) is oriented to:
Wherein, (ak,bk) it is rectangular window wkInvariable linear coefficient,
Using following cost function come weigh t (x) withDifference, i.e.,:
Wherein, ε is the parameter of regularization parameter and a control smoothness, then finds out picture using linear regression
Rectangular window w centered on number kkCorresponding linear coefficient (ak,bk), i.e.,:
Wherein,And μkRectangular window w in image I is indicated respectivelykVariance and mean value,Indicate t in rectangular window wkIn it is equal
Value, covk(I, t) indicates to be oriented to image I and transmittance figure t in rectangular window wkIn cross-correlation function,
Since pixel i can be included by multiple rectangular windows, different rectangular windows can obtain differentThen it is taken
The mean value of all probable values:
Wherein, | w | indicate the number of pixel in rectangular window,And
4) depth information is calculated
For road image, for any one lane line R (i) in image to be detected I (x), (i=1,2 ... n), can obtain
To corresponding extinction coefficient β (i):
Wherein, L is the national standard length scale of lane line, t1、t2Respectively lane line both ends are corresponded in image to be detected
Then transmissivity after position optimization takes mean value if N indicates the number of lane line in image to be detected to extinction coefficient:
For any one in image to be detected as number x, the depth information of corresponding position is calculated by following formula
It arrives, i.e.,:
5) road area extracts
The road area in image to be detected I (x) is extracted using region growing methods, takes road area lowermost end in I (x)
One-row pixels are that the initial seed point increased handles line by line as several points if the lastrow pixel of seed is target area, uses area
The step of domain growing method extraction road area, is as follows:
(5.1) setting initial seed point is ps, current seed point p=ps;
(5.2) three pixels adjacent above target area selection with pixel seed point, calculate seed point and this three
The gray scale difference of pixel(α takes -1,0,1 to indicate upper left side adjacent with seed point, surface and upper right side pixel respectively);
(5.3) minimum value of adjacent 3 pixel grey scales difference is calculated
(5.4) judge whether target area picture number point p (u, v) and seed point mean value mean (p) meets following relationship, i.e.,:
Wherein, ρ < 1, nrIt is target area picture number point p (u, v) and initial seed point psThe line number being separated by, if satisfied, then will
Road area is added as several points in this, traverses target area all pixels;If condition is not satisfied for all pixels of the row, region
Increase and terminates;
(5.5) current seed point and target area are updated, being put as new seed point as several for road area is added by new,
Using the new lastrow pixel that road area picture number is added as new target area, return to step 5.2 after update;
6) visibility is estimated
The depth information that the road area and step 4 obtained based on step 5 is obtained, Mead law is wished according to section
(Koschmieder's Law), which can be calculated, restores image J (x), i.e.,:
Wherein, the value range of β isThe value of τ is right between 0.01~0.02The topography entropy H for calculating the road area for restoring to be extracted in image J (x) is:
H=-E [log p (y)]=- Σ p (y) log p (y)
Wherein, stochastic variable y indicates gray value, and p (y) is the probability density function about y, then finds topography's entropy
Corresponding value is extinction coefficient optimal value β when minimum*, visibility size is finally calculated according to visibility etection theory:
Further, preferably, the size of the window ω (x) in step 2 is 15 × 15.
Further, the present invention also provides the traffic haze visibility inspections based on above-mentioned dark channel prior and minimum image entropy
The detecting system of survey method comprising image capture module, image characteristics extraction module, road area extraction module and visibility
Estimation module, the respective function of above-mentioned module and its correlation are:
1) image capture module:Under the conditions of haze on daytime, original input picture acquires institute through monitoring camera electronic equipment
, referred to as image to be detected I;
2) image characteristics extraction module:Image to be detected I is handled by dark channel prior, obtains the thick of atmospheric transmissivity
Estimated value;Smooth micronization processes are carried out to transmissivity rough estimate evaluation using Steerable filter edge-smoothing operator, are obtained finer
Transmissivity;In conjunction with the lane line information of transmissivity and original road image after optimization, each pixel of image Scene is obtained
The depth information of point;
3) road area extraction module:The road area in image to be detected I, the area are extracted using region growing methods
It includes that initial seed point is arranged, setting target growth region, the minimum value for calculating neighboring gradation difference, judges object pixel that domain, which increases,
Whether it is road area, update seed point;
4) visibility estimation module:After extracting road area, the minimum image entropy in the region is calculated, delustring system is obtained
Number optimal value, is then effectively estimated out haze visibility size.
Theoretical compared to traditional dark channel prior, technical scheme of the present invention is filtered using dark channel prior theory and guiding
Wave is combined, and estimates the atmospheric transmissivity of image to be detected;It is based on dark channel prior theory and lane detection basis again, proposes
Completely new image depth information computational methods;Using region growing algorithm, extraction target interest region, and it is based on minimum image entropy
Theory, the method for proposing new calculating atmospheric extinction coefficient.Advantageous effect is the present invention after obtaining transmittance figure, to transmission
Rate figure carries out Steerable filter processing, and further fine optimization transmittance figure can effectively reduce follow-up visibility estimating step
Detection error.Moreover, using Steerable filter for edge-smoothing operator in fine atmospheric transmissivity, and seeking image entropy process
Middle addition target interest region extraction step, reduces the calculation amount of algorithm, improves arithmetic speed.The method of the present invention can make
The error for obtaining visibility detection meets China Meteorological professional standard, and robustness is good.
Description of the drawings
Fig. 1 is the system schematic of the present invention.
Fig. 2 is the relative error schematic diagram of the visibility testing result and practical reference value of the present invention.
Specific implementation mode
The specific implementation mode of the present invention is described in detail below in conjunction with the accompanying drawings.
Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy, including image capture module,
Image characteristics extraction module, road area extraction module and visibility detection module.As shown in Figure 1, described image acquisition module,
Under the conditions of haze on daytime, original input picture acquires gained through monitoring camera electronic equipment, and a selection frame therein, which is used as, to be waited for
Detection image.
Described image characteristic extracting module handles image to be detected I by dark channel prior, obtains atmospheric transmissivity
Rough estimate evaluation;Smooth micronization processes are carried out to transmissivity rough estimate evaluation using Steerable filter edge-smoothing operator, are obtained finer
Transmissivity;In conjunction with the lane line information of transmissivity and original road image after optimization, each picture of image Scene is obtained
The depth information of vegetarian refreshments.Detailed step is described below.
Involved relevant parameter, used in present invention verification, the protection domain of invention includes but not limited to this.
1. calculating atmospheric transmissivity
If I (x) is image to be detected, J (x) is the recovery image obtained by I (x), and A is sky brightness, and t (x) is image
Transmissivity wishes the mathematical model that Mead law defines by section, and following equations are that foggy image forms model:
I (x)=J (x) t (x)+A (1-t (x)) (1)
Then minimum operation twice is asked to formula (1), obtains following formula:
Wherein JcAnd IcIndicate each channel of image, ω (x) indicates a window centered on pixel x, in of the invention
The size of window takes 15 × 15.
According to the mathematical definition of dark, for arbitrary input picture J (x), at least one Color Channel tool in image
There is very low value, i.e.,:
Atmospheric transmissivity is estimated as follows:
Wherein, sky brightness A is the mean value size of 0.1% brightest pixel point in dark channel diagram.
2. transmissivity is optimized and revised
Steerable filter is the smoothing operator of an Edge preservation, for as the rectangular window w centered on number kkIt is interior, if t (x) is passed through
It is after Steerable filterIt is oriented to image and uses image to be detected I (x), filtering output imageIt is about guiding image I (x)
Linear model is filtered, following relationship is met:
Wherein, (ak,bk) it is rectangular window wkInvariable linear coefficient.In order to find and the guiding of t (x) difference minimums
It is filteredUsing following cost function come weigh t (x) withDifference, i.e.,:
Wherein, ε is the parameter of regularization parameter and a control smoothness.Then picture is found out using linear regression
Rectangular window w centered on number kkCorresponding linear coefficient (ak,bk), i.e.,:
Wherein,And μkRectangular window w in image I is indicated respectivelykVariance and mean value,Indicate t in rectangular window wkIn it is equal
Value, covk(I, t) indicates to be oriented to image I and transmittance figure t in rectangular window wkIn cross-correlation function.
Since pixel i can be included by multiple rectangular windows, different rectangular windows can obtain differentThen it is taken
The mean value of all probable values:
Wherein, | w | indicate the number of pixel in rectangular window,
3. calculating depth information
The present invention is directed road image, there is national standard at the middle lane line length of highway and interval, profit
Extinction coefficient initial value is found out with the Given information.For any one lane line R (i) in image to be detected I (x) (i=1,2 ...
N), corresponding extinction coefficient β (i) can be obtained:
Wherein, L is the national standard length scale of lane line, t1、t2Respectively lane line both ends are corresponded in image to be detected
Transmissivity after position optimization.If N indicates the number of lane line in image to be detected, mean value then is taken to extinction coefficient:
For any one in image to be detected as number x, the depth information of corresponding position is calculated by following formula
It arrives, i.e.,:
The road area extraction module extracts the road area in image to be detected I using region growing methods, takes I
Middle road area lowermost end one-row pixels are the initial seed point increased.If the lastrow pixel of seed is target area, by
Row is handled as several points, and it is as follows to increase the step of extracting road area with region:
(1) setting initial seed point is ps, current seed point p=ps;
(2) three pixels adjacent above target area selection with pixel seed point, calculate seed point and this three pictures
The gray scale difference of element(α takes -1,0,1 to indicate upper left side adjacent with seed point, surface and upper right side pixel respectively);
(3) minimum value of adjacent 3 pixel grey scales difference is calculated
(4) judge whether target area picture number point p (u, v) and seed point mean value mean (p) meets following relationship, i.e.,:
Wherein, ρ < 1, nrIt is target area picture number point p (u, v) and initial seed point psThe line number being separated by.If satisfied, then will
Road area is added as several points in this, traverses target area all pixels;If condition is not satisfied for all pixels of the row, region
Increase and terminates;
(5) current seed point and target area are updated.Being put as new seed point as several for road area is added by new, it will
The new lastrow pixel that road area picture number is added has updated return to step (2) as new target area;
Heretofore described visibility estimation module calculates the minimum image entropy in the region after extracting road area,
Extinction coefficient optimal value is obtained, haze visibility size is then effectively estimated out.It is specifically described as follows:Extracting road
Behind interest region and acquisition depth information, wishes Mead law according to section and can calculate and restore image J (x), i.e.,:
Wherein, the value range of β isThe value of τ is between 0.01~0.02.It is rightCalculate the topography entropy H for restoring that road interest region is extracted in image J (x):
H=-E [log p (y)]=- ∑ p (y) log p (y) (16)
Wherein, stochastic variable y indicates grey scale pixel value, and p (y) is the probability density function about y.Then image entropy is found
Corresponding value is extinction coefficient optimal value β when minimum*, visibility V is finally calculated according to visibility etection theory:
Fig. 2 is the relative error schematic diagram of the visibility testing result and practical reference value of the present invention.Wherein, subgraph (a),
(b), (c) is respectively that visibility is less than 100m (reference value 83m), and visibility is between 100~150m (reference value 143m), visibility
Between 150~200m (reference value 200m), 3 width subgraphs are distinguished using the relative error of the method for the present invention measured value and reference value
For:7.83%, 6.78%, 6.84%.Allow to examine when visibility is less than 2000m according to the professional standard that China Meteorological Administration issues
It is 10% to survey error.It can thus be seen that the method for the present invention can make the error that visibility detects meet China Meteorological industry
Standard, robustness are good.
The present invention is based on dark channel prior theories, are aided with region and increase and the image processing techniques such as minimum image entropy, proposition
A kind of new haze visibility detecting method is then effectively estimated with estimating the transmissivity and depth information of image to be detected
Go out the atmospheric visibility of image to be detected.The present invention is in image characteristics extraction module, to image to be detected I by dark elder generation
Processing is tested, the rough estimate evaluation of atmospheric transmissivity is obtained;Transmissivity rough estimate evaluation is carried out using Steerable filter edge-smoothing operator
Smooth micronization processes obtain finer transmissivity;In conjunction with the lane line information of transmissivity and original road image after optimization,
Obtain the depth information of each pixel of image Scene;In road area extraction module, extracted using region growing methods
Road area in image to be detected I, it includes setting initial seed point that the region, which increases, setting target growth region, is calculated
The minimum value of neighboring gradation difference judges whether object pixel is road area, update seed point;Estimate mould in visibility
Block calculates the minimum image entropy in the region, obtains extinction coefficient optimal value, be then effectively estimated after extracting road area
Go out haze visibility size.
Advantage of the invention is that the dark channel prior theory in the present invention can be adapted for the road of any scene, it is only right
Sky areas is failed, and since road area is target area, therefore the theory is not limited by scene condition, applied widely;Meanwhile
It uses Steerable filter for edge-smoothing operator in fine atmospheric transmissivity, and target interest is added during seeking image entropy
Region extraction step reduces the calculation amount of algorithm, improves arithmetic speed.The method of the present invention can make what visibility detected
Error meets China Meteorological professional standard, therefore robustness is good.
Claims (3)
1. the traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy, it is characterised in that including following step
Suddenly:
1) under the conditions of haze on daytime, the video sequence of road monitoring is acquired, selects a wherein frame as image to be detected I
(x);
2) atmospheric transmissivity is calculated
J (x) is the recovery image obtained by above-mentioned image to be detected I (x), and A is sky brightness, and t (x) is image transmission rate, then
Foggy image forms model:
I (x)=J (x) t (x)+A (1-t (x))
Then minimum operation twice is asked to above formula, obtains following formula:
Wherein JcAnd IcIndicate that each channel of image, ω (x) indicate a window centered on pixel x,
For arbitrary input picture J (x), at least one Color Channel has very low value in image, i.e.,:
Atmospheric transmissivity is estimated as follows:
Wherein, sky brightness A is the mean value size of 0.1% brightest pixel point in dark channel diagram;
3) transmissivity is optimized and revised
For the rectangular window w centered on pixel kkIf t (x) is after Steerable filterIt is oriented to image and uses image to be detected I
(x), filtering output imageMeet following relationship with image I (x) is oriented to:
Wherein, (ak,bk) it is rectangular window wkInvariable linear coefficient,
Using following cost function come weigh t (x) withDifference, i.e.,:
Wherein, ε is the parameter of regularization parameter and a control smoothness, and then finding out pixel k using linear regression is
The rectangular window w at centerkCorresponding linear coefficient (ak,bk), i.e.,:
Wherein,And μkRectangular window w in image I is indicated respectivelykVariance and mean value,Indicate t in rectangular window wkIn mean value,
covk(I, t) indicates to be oriented to image I and transmittance figure t in rectangular window wkIn cross-correlation function,
Since pixel i can be included by multiple rectangular windows, different rectangular windows can obtain differentThen take it all
The mean value of probable value:
Wherein, | w | indicate the number of pixel in rectangular window,And
4) depth information is calculated
For road image, for any one lane line R (i) in image to be detected I (x), (i=1,2 ... n), can obtain pair
The extinction coefficient β (i) answered:
Wherein, L is the national standard length scale of lane line, t1、t2Respectively lane line both ends are in image to be detected corresponding position
Then transmissivity after optimization takes mean value if N indicates the number of lane line in image to be detected to extinction coefficient:
For any one in image to be detected as number x, the depth information of corresponding position is calculated by following formula,
I.e.:
5) road area extracts
The road area in image to be detected I (x) is extracted using region growing methods, takes road area lowermost end a line in I (x)
Pixel is that the initial seed point increased handles line by line as several points, increased with region if the lastrow pixel of seed is target area
The step of long method extraction road area, is as follows:
(5.1) setting initial seed point is ps, current seed point p=ps;
(5.2) three pixels adjacent above target area selection with pixel seed point, calculate seed point and this three pixels
Gray scale differenceα takes -1,0,1 to indicate upper left side adjacent with seed point, surface and upper right side pixel respectively;
(5.3) minimum value of adjacent 3 pixel grey scales difference is calculated
(5.4) judge whether target area picture number point p (u, v) and seed point mean value mean (p) meets following relationship, i.e.,:
Wherein, ρ < 1, nrIt is target area picture number point p (u, v) and initial seed point psThe line number being separated by, if satisfied, then by the picture
Road area is added in several points, traverses target area all pixels;If condition is not satisfied for all pixels of the row, region increases
It terminates;
(5.5) current seed point and target area are updated, being put as new seed point as several for road area is added by new, it will be new
The lastrow pixel of addition road area picture number is as new target area, return to step 5.2 after update;
6) visibility is estimated
Based on step 5 obtain road area and step 4 obtain depth information, according to section wish Mead law can calculate it is extensive
Complex pattern J (x), i.e.,:
Wherein, the value range of β isThe value of τ is right between 0.01~0.02
The topography entropy H for calculating the road area for restoring to be extracted in image J (x) is:
H=-E [logp (y)]=- ∑ p (y) logp (y)
Wherein, stochastic variable y indicates gray value, and p (y) is the probability density function about y, and it is minimum then to find topography's entropy
When corresponding β value be extinction coefficient optimal value β*, visibility size is finally calculated according to visibility etection theory:
2. the traffic haze visibility detecting method according to claim 1 based on dark channel prior Yu minimum image entropy,
It is characterized in that the size of the window ω (x) in step 2 is 15 × 15.
3. a kind of realizing the traffic haze visibility detection side described in claim 1 based on dark channel prior Yu minimum image entropy
The detecting system of method, it is characterised in that including image capture module, image characteristics extraction module, road area extraction module and energy
Degree of opinion estimation module, the respective function of above-mentioned module and its correlation are:
1) image capture module:Under the conditions of haze on daytime, original input picture acquires gained through monitoring camera electronic equipment, claims
For image to be detected I;
2) image characteristics extraction module:Image to be detected I is handled by dark channel prior, obtains the rough estimate of atmospheric transmissivity
Value;Smooth micronization processes are carried out to transmissivity rough estimate evaluation using Steerable filter edge-smoothing operator, obtain finer transmission
Rate;In conjunction with the lane line information of transmissivity and original road image after optimization, each pixel of image Scene is obtained
Depth information;
3) road area extraction module:The road area in image to be detected I is extracted using region growing methods, the region increases
Length includes that initial seed point is arranged, setting target growth region, the minimum value for calculating neighboring gradation difference, whether judges object pixel
For road area, update seed point step;
4) visibility estimation module:After extracting road area, the minimum image entropy in the region is calculated, obtains extinction coefficient most
Haze visibility size is then effectively estimated out in the figure of merit.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610227754.XA CN105931220B (en) | 2016-04-13 | 2016-04-13 | Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610227754.XA CN105931220B (en) | 2016-04-13 | 2016-04-13 | Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105931220A CN105931220A (en) | 2016-09-07 |
CN105931220B true CN105931220B (en) | 2018-08-21 |
Family
ID=56837973
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610227754.XA Active CN105931220B (en) | 2016-04-13 | 2016-04-13 | Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105931220B (en) |
Families Citing this family (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107194924A (en) * | 2017-05-23 | 2017-09-22 | 重庆大学 | Expressway foggy-dog visibility detecting method based on dark channel prior and deep learning |
CN107451966B (en) * | 2017-07-25 | 2021-05-11 | 四川大学 | Real-time video defogging method implemented by guiding filtering through gray level image |
CN107844761B (en) * | 2017-10-25 | 2021-08-10 | 海信集团有限公司 | Traffic sign detection method and device |
CN107945174A (en) * | 2017-12-12 | 2018-04-20 | 嘉兴四维智城信息科技有限公司 | Fastlink visibility measuring method based on video |
CN108693087B (en) * | 2018-04-13 | 2020-09-01 | 中国科学院城市环境研究所 | Air quality monitoring method based on image understanding |
CN108875794B (en) * | 2018-05-25 | 2020-12-04 | 中国人民解放军国防科技大学 | Image visibility detection method based on transfer learning |
CN108898562B (en) * | 2018-06-22 | 2022-04-12 | 大连海事大学 | Mobile equipment image defogging method based on deep learning |
CN110322431B (en) * | 2019-05-23 | 2021-04-13 | 深圳大学 | Haze image quality evaluation method and system, storage medium and electronic equipment |
CN110503634B (en) * | 2019-07-29 | 2022-07-15 | 西安理工大学 | Visibility measuring method based on automatic image identification |
CN111091601B (en) * | 2019-12-17 | 2023-06-23 | 香港中文大学深圳研究院 | PM2.5 index estimation method for real-time daytime outdoor mobile phone image |
CN111274911B (en) * | 2020-01-17 | 2020-12-01 | 河海大学 | Dense fog monitoring method based on wireless microwave attenuation characteristic transfer learning |
CN112288736B (en) * | 2020-11-09 | 2022-11-15 | 浙江大学 | Visibility estimation method based on images |
CN112365467B (en) * | 2020-11-11 | 2022-07-19 | 武汉长江通信智联技术有限公司 | Foggy image visibility estimation method based on single image depth estimation |
CN113379619B (en) * | 2021-05-12 | 2022-02-01 | 电子科技大学 | Integrated processing method for defogging imaging, visibility extraction and depth of field estimation |
CN113570554A (en) * | 2021-07-06 | 2021-10-29 | 浙江工业大学 | Single image visibility detection method based on scene depth |
CN113588507B (en) * | 2021-07-28 | 2022-07-05 | 沭阳新辰公路仪器有限公司 | Automatic haze weather detection and warning system and device for road traffic |
CN114202542B (en) * | 2022-02-18 | 2022-04-19 | 象辑科技(武汉)股份有限公司 | Visibility inversion method and device, computer equipment and storage medium |
CN114627382B (en) * | 2022-05-11 | 2022-07-22 | 南京信息工程大学 | Expressway fog visibility detection method combined with geometric prior of lane lines |
CN116664448B (en) * | 2023-07-24 | 2023-10-03 | 南京邮电大学 | Medium-high visibility calculation method and system based on image defogging |
CN117036445B (en) * | 2023-10-08 | 2023-12-26 | 埃洛克航空科技(北京)有限公司 | Depth estimation method, device, computer equipment and medium based on random distribution |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101382497A (en) * | 2008-10-06 | 2009-03-11 | 南京大学 | Visibility detecting method based on monitoring video of traffic condition |
US9177363B1 (en) * | 2014-09-02 | 2015-11-03 | National Taipei University Of Technology | Method and image processing apparatus for image visibility restoration |
CN105261018A (en) * | 2015-10-14 | 2016-01-20 | 山东交通学院 | Visibility detection method based on optical model and dark primary color priori theory |
CN105374013A (en) * | 2014-08-15 | 2016-03-02 | 台北科技大学 | Method and image processing apparatus for image visibility restoration on the base of dual dark channel prior |
CN105424655A (en) * | 2015-11-04 | 2016-03-23 | 北京交通大学 | Visibility detection method based on video images |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9305339B2 (en) * | 2014-07-01 | 2016-04-05 | Adobe Systems Incorporated | Multi-feature image haze removal |
US9361670B2 (en) * | 2014-09-04 | 2016-06-07 | National Taipei University Of Technology | Method and system for image haze removal based on hybrid dark channel prior |
-
2016
- 2016-04-13 CN CN201610227754.XA patent/CN105931220B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101382497A (en) * | 2008-10-06 | 2009-03-11 | 南京大学 | Visibility detecting method based on monitoring video of traffic condition |
CN105374013A (en) * | 2014-08-15 | 2016-03-02 | 台北科技大学 | Method and image processing apparatus for image visibility restoration on the base of dual dark channel prior |
US9177363B1 (en) * | 2014-09-02 | 2015-11-03 | National Taipei University Of Technology | Method and image processing apparatus for image visibility restoration |
CN105261018A (en) * | 2015-10-14 | 2016-01-20 | 山东交通学院 | Visibility detection method based on optical model and dark primary color priori theory |
CN105424655A (en) * | 2015-11-04 | 2016-03-23 | 北京交通大学 | Visibility detection method based on video images |
Non-Patent Citations (3)
Title |
---|
Image Based Fog Detection and Visibility Estimation for Driving Assistance Systems;Mihai Negru 等;《ICCP 2013》;20130907;第163-168页 * |
基于路况视频的气象能见度检测方法与系统设计;安明伟 等;《仪器仪表学报》;20100515;第31卷(第5期);第1148-1153页 * |
基于车道线检测与图像拐点的道路能见度估计;宋洪军 等;《计算机应用》;20121201;第32卷(第12期);第3397-3403页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105931220A (en) | 2016-09-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105931220B (en) | Traffic haze visibility detecting method based on dark channel prior Yu minimum image entropy | |
CN111209810B (en) | Boundary frame segmentation supervision deep neural network architecture for accurately detecting pedestrians in real time through visible light and infrared images | |
CN104408460B (en) | A kind of lane detection and tracking detection method | |
CN105488454B (en) | Front vehicles detection and ranging based on monocular vision | |
CN105261020B (en) | A kind of express lane line detecting method | |
CN104392468B (en) | Based on the moving target detecting method for improving visual background extraction | |
CN105512623B (en) | Based on multisensor travelling in fog day vision enhancement and visibility early warning system and method | |
CN111079556A (en) | Multi-temporal unmanned aerial vehicle video image change area detection and classification method | |
CN105404847B (en) | A kind of residue real-time detection method | |
CN105279772B (en) | A kind of trackability method of discrimination of infrared sequence image | |
Hautière et al. | Enhanced fog detection and free-space segmentation for car navigation | |
CN102831591B (en) | Gaussian filter-based real-time defogging method for single image | |
CN103735269B (en) | A kind of height measurement method followed the tracks of based on video multi-target | |
US20150086077A1 (en) | System and method of alerting a driver that visual perception of pedestrian may be difficult | |
CN108280450A (en) | A kind of express highway pavement detection method based on lane line | |
CN105279372A (en) | Building height computing method and apparatus | |
CN104318266B (en) | A kind of image intelligent analyzes and processes method for early warning | |
TW201120814A (en) | Method for determining if an input image is a foggy image, method for determining a foggy level of an input image and cleaning method for foggy images | |
JP2007235950A (en) | Method and device for detecting vehicle position | |
CN107301624A (en) | The convolutional neural networks defogging algorithm pre-processed based on region division and thick fog | |
CN108171695A (en) | A kind of express highway pavement detection method based on image procossing | |
WO2023155483A1 (en) | Vehicle type identification method, device, and system | |
CN102231844B (en) | Video image fusion performance evaluation method based on structure similarity and human vision | |
Halmaoui et al. | Contrast restoration of road images taken in foggy weather | |
CN109214331B (en) | Traffic haze visibility detection method based on image frequency spectrum |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |