CN102779349B - Foggy day detecting method based on image color spatial feature - Google Patents

Abstract

The invention provides a foggy day detecting method based on an image color spatial feature. The method comprises the following steps of first step, obtaining a background picture through a video image or a single image; step 2, performing color space conversion for the background picture, and extracting the color space feature: firstly performing color space conversion for the background picture, converting from an RGB (red-green-blue) color space into an HSV (hue, saturation, value) color space, and extracting the features of various components of the HSV according to the information included image; and step 3, dividing the weather information included image into a non-foggy day, a little foggy day and a dense foggy day according to the identification conditions: at first, if the identification condition 1 is met, the weather is the dense foggy day, otherwise the identification is continued; secondly, if the identification condition 2 is met, the weather is the non-foggy day, otherwise, the identification is continued; and finally, if the identification condition 3 is met, the weather is the little foggy day, otherwise, the weather is the dense foggy day. The method provided by the invention is suitable for being used for monitoring the foggy day in expressways, particularly pre-warning the emergency of agglomerate fog in partial road segments, thus, the safe road driving can be ensured.

Description

A kind of foggy day detection method based on color of image space characteristics

Technical field

The present invention relates to digital video image process and traffic weather detection field, is a kind of foggy day detection method based on color of image HSV feature.

Background technology

Visibility is one of key element of road weather conditions system.At traffic high speed networking, the grasp of a road visibility is related to the major issue of road user safety of life and property, reporting for work of the pernicious traffic hazard caused due to low visibility is of common occurrence, in low visibility situation, the reaction velocity of the mankind declined, especially all the more so in high speed traveling process., fast, weather conditions information just can provide and warn in advance accurately, thus management and road users choose safe road occupation route and drive speed to help high-speed transit road management department to carry out rationally effectively super expressway.

The current observation for dense fog has mainly come by artificial visually examine, special installation monitoring station, mainly by installing advanced optical devices, by receiving and measuring the intensity of scattered beam, accurately measure the air visibility on current highway, judged whether that dense fog occurs.Because the length of the expense of apparatus expensive and super expressway is comparatively large, layout research station intensive on a highway, catches the group's mist haunted everywhere, obvious too expensive.

For the detection technique of the traditional greasy weather visibility based on image, need to carry out on-site proving to video camera, wherein not only need to know the various parameter in the inside of video camera, and need the position angle etc. knowing installation, and needing corresponding corresponding scene, computing is complicated and cost is also higher.

Foggy day detection method becomes the importance of institute's primary study that high-speed transit controls relevant departments in recent years, is putting before this, researchs and proposes a kind of simple, convenient, detection method fast.

Summary of the invention

The present invention is a kind of foggy day detection method based on color of image space characteristics, first by video image or single image background extraction picture, then color space conversion (from RGB to HSV) is carried out to background picture, and then feature extraction is carried out to HSV, utilize the feature extracted to carry out qualitative detection to weather, be divided into non-greasy weather gas, little greasy weather gas, foggy weather.

The present invention adopts following technical scheme:

Based on a foggy day detection method for color of image HSV feature, implement according to following steps:

Based on a foggy day detection method for color of image space characteristics, it is characterized in that, implement according to following steps:

Step 1: initialization, reads in road traffic image or video,

Step 2: if read in piece image in step 1, then read in picture and be background picture I ₁; If what read in step 1 is video, be then handled as follows, background extraction picture I ₁:

The N frame picture read in from video camera, N=30 ~ 40, statistics comes across the number of times of same color gray-scale value on the same pixel of 30 ~ 40 frame pictures, and with the highest color gray-scale value picture I as a setting of occurrence number ₁the color gray-scale value of respective pixel point, mathematic(al) representation is:

$P(i,j,n)=\left\{\begin{array}{cc}P(i,j,n)+1& I_k(i,j,m)=n\\ P(i,j,n)& I_k(i,j,m)\≠n\end{array}\right.k=\mathrm{1,2},...,N$

Background(i,j)＝Max(P(i,j,n))n＝0,1,2,...,255

In formula, P (i, j, n) represents the number of times that pixel (i, j) place color gray level n occurs, initial value is 0, I _k(i, j, m) represents that in kth two field picture, pixel (i, j) place gray level is m, m=0,1,2, ..., 255, Background (i, j) for background picture is at the color gray-scale value of (i, j), Max (P (i, j, n)) be P (i, j, n) color gray-scale value represented when the number of times occurred is maximum, carries out traversing operation to every bit in image, obtains background picture I ₁,

Step 3: to background picture I ₁carry out color space conversion, from RGB color space conversion to hsv color space:

Each pixel form and aspect component H (i, j) are obtained by following formula:

$H(i,j)=\left\{\begin{array}{cc}\mathrm{\θ}& B(i,j)\≤G(i,j)\\ 360-\mathrm{\θ}& B(i,j)>G(i,j)\end{array}\right.$

Herein

$\mathrm{\θ}=\arccos \{\frac{\frac{1}{2}\left\{\right[R(i,j)-G(i,j)]+[R(i,j)-B(i,j)\left]\right\}}{{\{{[R(i,j)-G(i,j)]}^2+[R(i,j)-G(i,.j)\left]\right[G(i,j)-B(i,j)\left]\right\}}^{1/2}},$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

Saturation degree component S (i, j) is as follows:

$S(i,j)=1-\frac{3}{[R(i,j)+G(i,j)+B(i,j)]}\left[\min (R(i,j),G(i,j),B(i,j))\right]$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

Chrominance component V (i, j) is as follows:

$V(i,j)=\frac{1}{3}[R(i,j)+G(i,j)+B(i,j)],$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

To the conversion that each pixel of Target Photo is carried out as above, obtain background picture I ₁hsv color space,

Step 4: to background picture I ₁the color space of HSV carries out feature extraction, specific as follows:

4.1) form and aspect component H feature AveH: statistics background picture I ₁in the summation SumH of each pixel H component value, H component value be not 0 pixel add up to M ₁, then H component characterization AveH=SumH/M ₁;

4.2) saturation degree component S feature AveS: statistics background picture I ₁in the summation SumS of each pixel S component value, S component value be not 0 pixel add up to M ₂, then S component characterization AveS=SumS/M ₂;

4.3) chrominance component V feature AveV: statistics background picture I ₁in the summation SumV of each pixel V component value, V component value be not 0 pixel add up to M ₃, then V component characterization AveV=SumV/M ₃;

Step 5: according to the background picture I obtained in step 4 ₁the feature of HSV, according to decision condition 1, mark off the foggy weather with specific characteristic, decision condition 1 is expressed as follows:

AveH＜125&&AveS＜0.200&&AveV＜0.4800

Meet above-mentioned condition, then weather is dense fog, otherwise carries out next step simultaneously,

Step 6: to when being judged to be no in step 5, according to decision condition 2, mark off non-greasy weather weather, decision condition 2 expression formula is as follows:

(if AveS > 0.1050 & & AveH < 200)

Or (Aves > 0.200 & & AveH > 120),

Or 0.0700 < AveS < 0.1050 & & 120 < AveH < 200,

Meet any one condition above-mentioned, be then the non-greasy weather, no person carries out next step,

Step 7: to when being judged to be no in step 6, according to decision condition 3, mark off little greasy weather gas, decision condition 3 is expressed as follows:

AveS＜0.0700&&AveH＜100&&AveV＜0.500,

Meet above-mentioned condition, then weather is little mist simultaneously, otherwise is dense fog.

Compared with prior art, the invention has the advantages that:

1, the present invention utilizes color of image spatial information, for weather condition in road traffic, can mark off non-greasy weather gas, little greasy weather gas and foggy weather more exactly, to realize reporting to the police to the greasy weather and road carries out safety management;

2, the hsv color spatial information that comprises according to image self of the present invention, utilize the feature of its each component under different weather qualitatively to mark off different weather situation, algorithm is simpler, parameter adjustment is convenient, travelling speed is very fast, avoid and complicated on-site proving link is carried out to video camera, realize the detection of pin-point accuracy.

3, compared with the cost of traditional visibility detector costliness, cost of the present invention is lower, and make the present invention can realize large-area covering in actual field application, real realization is detected in real time road conditions, managed and early warning, ensures traffic safety

Accompanying drawing illustrates:

Fig. 1 is the particular flow sheet that whole system detects.

Embodiment:

In a particular embodiment, will by reference to the accompanying drawings, know the detailed process of the foggy day detection method intactly described based on color of image HSV feature.

Concrete steps are as follows:

Based on a foggy day detection method for color of image space characteristics, implement according to following steps:

Step 1: initialization, reads in road traffic image or video,

Step 2: if read in piece image in step 1, then read in picture and be background picture I ₁; If what read in step 1 is video, be then handled as follows, background extraction picture I ₁:

The N frame picture read in from video camera, N=30 ~ 40, statistics comes across the number of times of same color gray-scale value on the same pixel of 30 ~ 40 frame pictures, and with the highest color gray-scale value picture I as a setting of occurrence number ₁the color gray-scale value of respective pixel point, mathematic(al) representation is:

$P(i,j,n)=\left\{\begin{array}{cc}P(i,j,n)+1& I_k(i,j,m)=n\\ P(i,j,n)& I_k(i,j,m)\&NotEqual;n\end{array}\right.k=\mathrm{1,2},...,N$

Background(i,j)＝Max(P(i,j,n))n＝0,1,2,...,255

In formula, P (i, j, n) represents the number of times that pixel (i, j) place color gray level n occurs, initial value is 0, I _k(i, j, m) represents that in kth two field picture, pixel (i, j) place gray level is m, m=0,1,2, ..., 255, Background (i, j) for background picture is at the color gray-scale value of (i, j), Max (P (i, j, n)) be P (i, j, n) color gray-scale value represented when the number of times occurred is maximum, carries out traversing operation to every bit in image, obtains background picture I ₁,

Step 3: to background picture I ₁carry out color space conversion, from RGB color space conversion to hsv color space:

Each pixel form and aspect component H (i, j) are obtained by following formula:

$H(i,j)=\left\{\begin{array}{cc}\mathrm{\&theta;}& B(i,j)\&le;G(i,j)\\ 360-\mathrm{\&theta;}& B(i,j)>G(i,j)\end{array}\right.$

Herein

$\mathrm{\&theta;}=\arccos \{\frac{\frac{1}{2}\left\{\right[R(i,j)-G(i,j)]+[R(i,j)-B(i,j)\left]\right\}}{{\{{[R(i,j)-G(i,j)]}^2+[R(i,j)-G(i,.j)\left]\right[G(i,j)-B(i,j)\left]\right\}}^{1/2}},$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

Saturation degree component S (i, j) is as follows:

$S(i,j)=1-\frac{3}{[R(i,j)+G(i,j)+B(i,j)]}\left[\min (R(i,j),G(i,j),B(i,j))\right]$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

Chrominance component V (i, j) is as follows:

$V(i,j)=\frac{1}{3}[R(i,j)+G(i,j)+B(i,j)],$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

To the conversion that each pixel of Target Photo is carried out as above, obtain background picture I ₁hsv color space,

Step 4: to background picture I ₁the color space of HSV carries out feature extraction, specific as follows:

4.1) form and aspect component H feature AveH: statistics background picture I ₁in the summation SumH of each pixel H component value, H component value be not 0 pixel add up to M ₁, then H component characterization AveH=SumH/M ₁;

4.2) saturation degree component S feature AveS: statistics background picture I ₁in the summation SumS of each pixel S component value, S component value be not 0 pixel add up to M ₂, then S component characterization AveS=SumS/M ₂;

4.3) chrominance component V feature AveV: statistics background picture I ₁in the summation SumV of each pixel V component value, V component value be not 0 pixel add up to M ₃, then V component characterization AveV=SumV/M ₃;

Step 5: according to the background picture I obtained in step 4 ₁the feature of HSV, according to decision condition 1, mark off the foggy weather with specific characteristic, decision condition 1 is expressed as follows:

AveH＜125&&AveS＜0.200&&AveV＜0.4800

Meet above-mentioned condition, then weather is dense fog, otherwise carries out next step simultaneously,

Step 6: to when being judged to be no in step 5, according to decision condition 2, mark off non-greasy weather weather, decision condition 2 expression formula is as follows:

(if AveS > 0.1050 & & AveH < 200)

Or (Aves > 0.200 & & AveH > 120),

Or 0.0700 < AveS < 0.1050 & & 120 < AveH < 200,

Meet any one condition above-mentioned, be then the non-greasy weather, no person carries out next step,

Step 7: to when being judged to be no in step 6, according to decision condition 3, mark off little greasy weather gas, decision condition 3 is expressed as follows:

AveS＜0.0700&&AveH＜100&&AveV＜0.500,

Meet above-mentioned condition, then weather is little mist simultaneously, otherwise is dense fog.

Claims (1)

1. based on a foggy day detection method for color of image space characteristics, it is characterized in that, implement according to following steps:

Step 1: initialization, reads in road traffic image or video,

Step 2: if read in piece image in step 1, then read in picture and be background picture I ₁; If what read in step 1 is video, be then handled as follows, background extraction picture I ₁:

From video camera, read in N frame picture, N=30 ~ 40, statistics comes across the number of times of same color gray-scale value on the same pixel of 30 ~ 40 frame pictures, and with the highest color gray-scale value picture I as a setting of occurrence number ₁the color gray-scale value of respective pixel point, mathematic(al) representation is:

$P(i,j,n)=\left\{\begin{array}{cc}P(i,j,n)+1& I_k(i,j,m)=n\\ P(i,j,n)& I_k(i,j,m)\&NotEqual;n\end{array}\right.k=\mathrm{1,2},...,N$

Background(i,j)＝Max(P(i,j,n)) n＝0,1,2,...,255

In formula, P (i, j, n) represents the number of times that pixel (i, j) place color gray level n occurs, initial value is 0, I _k(i, j, m) represents that in kth two field picture, pixel (i, j) place gray level is m, m=0,1,2, ..., 255, Background (i, j) for background picture is at the color gray-scale value of (i, j), Max (P (i, j, n)) be P (i, j, n) color gray-scale value represented when the number of times occurred is maximum, carries out traversing operation to every bit in image, obtains background picture I ₁,

Step 3: to background picture I ₁carry out color space conversion, from RGB color space conversion to hsv color space:

Each pixel form and aspect component H (i, j) are obtained by following formula:

$H(i,j)=\left\{\begin{array}{cc}\mathrm{\&theta;}& B(i,j)\&le;G(i,j)\\ 360-\mathrm{\&theta;}& B(i,j)>G(i,j)\end{array}\right.$

Herein

$\mathrm{\&theta;}=\arccos \{\frac{\frac{1}{2}\left\{\right[R(i,j)-G(i,j)]+[R(i,j)-B(i,j)\left]\right\}}{{\{{[R(i,j)-G(i,j)]}^2+[R(i,j)-G(i,j)\left]\right[G(i,j)-B(i,j)\left]\right\}}^{1/2}},$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

Saturation degree component S (i, j) is as follows:

$S(i,j)=1-\frac{3}{[R(i,j)+G(i,j)+B(i,j)]}\left[\min (R(i,j),G(i,j),B(i,j))\right]$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

Chrominance component V (i, j) is as follows:

$V(i,j)=\frac{1}{3}[R(i,j)+G(i,j)+B(i,j)],$

Wherein, the red component intensity at pixel (i, j) place is R (i, j), effective value between 0 to 255, pixel (i, j) the green component intensity at place is G (i, j), effective value between 0 to 255, pixel (i, j) the blue component intensity at place is B (i, j), effective value is between 0 to 255

To the conversion that each pixel of Target Photo is carried out as above, obtain background picture I ₁hsv color space,

Step 4: to background picture I ₁hsv color space carry out feature extraction, specific as follows:

4.1) form and aspect component H feature AveH: statistics background picture I ₁in the summation SumH of each pixel H component value, H component value be not 0 pixel add up to M ₁, then form and aspect component H feature AveH=SumH/M ₁;

4.2) saturation degree component S feature AveS: statistics background picture I ₁in the summation SumS of each pixel S component value, S component value be not 0 pixel add up to M ₂, then saturation degree component S feature AveS=SumS/M ₂;

4.3) chrominance component V feature AveV: statistics background picture I ₁in the summation SumV of each pixel V component value, V component value be not 0 pixel add up to M ₃, then chrominance component V feature AveV=SumV/M ₃;

Step 5: according to the background picture I obtained in step 4 ₁hSV feature, according to decision condition 1, mark off the foggy weather with specific characteristic, decision condition 1 is expressed as follows:

AveH＜125&&AveS＜0.200&&AveV＜0.4800

Meet above-mentioned condition, then weather is dense fog, otherwise carries out next step simultaneously,

Step 6: to when being judged to be no in step 5, according to decision condition 2, mark off non-greasy weather weather, decision condition 2 expression formula is as follows:

(if AveS > 0.1050 & & AveH < 200),

Or (AveS > 0.200 & & AveH > 120),

Or 0.0700 < AveS < 0.1050 & & 120 < AveH < 200,

Meet any one condition above-mentioned, be then the non-greasy weather, no person carries out next step,

Step 7: to when being judged to be no in step 6, according to decision condition 3, mark off little greasy weather gas, decision condition 3 is expressed as follows:

AveS＜0.0700&&AveH＜100&&AveV＜0.500,

Meet above-mentioned condition, then weather is little mist simultaneously, otherwise is dense fog.