CN103077501A - Road target marking method and device - Google Patents

Abstract

The invention discloses a road target marking method and device. The method comprises the steps of obtaining image data I (x, y) of a fog-containing image of a traffic scene; conducting defogging processing to the image data I (x, y) of the fog-containing image to obtain defogged image data J (x, y); detecting a road target in the defogged image data J (x, y); and marking the detected road target. By adopting the technical scheme, drivers can accurately know the situations of roads ahead and the driving safety is improved.

Description

The labeling method of road target and device

Technical field

The present invention relates to the computer image processing technology field, in particular to a kind of labeling method and device of road target.

Background technology

In greasy weather gas situation, because the visibility of scene reduces, the feature such as target contrast and color is attenuated in the image, causes the life outdoor videos system to work, and therefore need to eliminate haze to the impact of scene image in video image.The image mist elimination is the important content of computer vision field research always.

In the greasy weather situation, the edge fog of the target such as road, vehicle, cross-color, profile information are lost in the image, and this is very disadvantageous for human pilot.

Summary of the invention

The present invention aims to provide a kind of labeling method and device of road target, can't differentiate the problem of road target with human pilot under the traffic scene in the greasy weather situation in the solution prior art.

To achieve these goals, according to an aspect of the present invention, provide a kind of labeling method of road target, having comprised: the view data I (x, y) that contains the mist image that obtains traffic scene; The above-mentioned picture number I (x, y) that contains the mist image is carried out mist elimination process, obtain the view data J (x, y) behind the mist elimination; Detect the road target among the view data J (x, y) behind the above-mentioned mist elimination; The road target that mark detects.

Further, the above-mentioned view data I (x, y) that contains the mist image is carried out mist elimination process, comprising: utilization is helped Dow process secretly the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination processing.

Further, utilization is helped Dow process secretly the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination processing, comprising: determine the above-mentioned dark primary I that contains the view data I (x, y) of mist image ^Drak(x, y); According to above-mentioned dark primary I ^Drak(x, y) determines the above-mentioned sky brightness A that contains the view data of mist image; According to above-mentioned sky brightness A and above-mentioned dark primary I ^Drak(x, y) determines transmissivity t (x, y); According to above-mentioned transmissivity t (x, y) and above-mentioned sky brightness A the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination and process, obtain the view data J (x, y) behind the above-mentioned mist elimination.

Further, determine in such a way above-mentioned transmissivity t (x, y):

$t(x,y)=\frac{A\left(1\right)+A\left(2\right)+A\left(3\right)}{3}-\mathrm{\ω}{I}^{\mathrm{dark}}(x,y),$ Wherein, A (1), A (2), A (3) are respectively the value of above-mentioned sky brightness A on three passages, and ω ∈ (0,1) is used for regulating the mist elimination degree for regulating parameter.

Further, carrying out in such a way above-mentioned mist elimination processes:

$J^c(x,y)=\frac{I(x,y)-A^c}{\max (\frac{t(x,y)}{A^c},t_0)}+A^c,$ C ∈ r, and g, b}, wherein, t ₀Be the default lower limit of t (x, y).

Further, utilize and to help Dow process secretly the above-mentioned view data I (x, y) that contains the mist image is carried out after above-mentioned mist elimination processes, also comprise: to the view data J (x behind the above-mentioned mist elimination, y) carry out the stretch processing of Linear Comparison degree, view data J'(x, y after obtaining stretching); Utilize the Brovey conversion to recover the original color view data B (x, y) of the view data J ' (x, y) after the above-mentioned stretching.

According to another aspect of the present invention, provide a kind of labelling apparatus of road target, having comprised: acquisition module, for the view data I (x, y) that contains the mist image that obtains traffic scene; The mist elimination module is used for that the above-mentioned picture number I (x, y) that contains the mist image is carried out mist elimination and processes, and obtains the view data J (x, y) behind the mist elimination; Detection module is for detection of the road target among the view data J (x, y) behind the above-mentioned mist elimination; Mark module is used for the road target that mark detects.

Further, above-mentioned mist elimination module be used for to be utilized and to be helped Dow process secretly and the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination process.

Further, above-mentioned mist elimination module comprises: the first determining unit is used for determining the above-mentioned dark primary I that contains the view data I (x, y) of mist image ^Drak(x, y); The second determining unit is used for according to above-mentioned dark primary I ^Drak(x, y) determines the above-mentioned sky brightness A that contains the view data of mist image; The 3rd determining unit is used for according to above-mentioned sky brightness A and above-mentioned dark primary I ^Drak(x, y) determines transmissivity t (x, y); The mist elimination unit is used for according to above-mentioned transmissivity t (x, y) and above-mentioned sky brightness A the above-mentioned view data I (x, y) that contains the mist image being carried out above-mentioned mist elimination processing, obtains the view data J (x, y) behind the above-mentioned mist elimination.

Further, above-mentioned mist elimination module also comprises: draw unit is used for the view data J (x, y) behind the above-mentioned mist elimination being carried out the stretch processing of Linear Comparison degree, the view data J ' (x, y) after obtaining stretching; Recovery unit is used for utilizing the Brovey conversion to recover the original color view data B (x, y) of the view data J ' (x, y) after the above-mentioned stretching.

Use technical scheme of the present invention, the mist image that contains to traffic scene carries out the mist elimination processing, detect road target in the image after mist elimination is processed, and in image, mark the road target that detects, so that human pilot also can be recognized the situation of road ahead accurately in the greasy weather, improved the security that traffic is driven.

Description of drawings

The Figure of description that consists of the application's a part is used to provide a further understanding of the present invention, and illustrative examples of the present invention and explanation thereof are used for explaining the present invention, do not consist of improper restriction of the present invention.In the accompanying drawings:

Fig. 1 is the process flow diagram according to the labeling method of the road target of the embodiment of the invention;

Fig. 2 is the schematic flow sheet according to the labeling method of the road target of embodiment of the invention preferred embodiment;

Fig. 3 is the process flow diagram according to the preferred image defogging method capable of the embodiment of the invention;

Fig. 4 is the process flow diagram according to the preferred Approach for road detection of the embodiment of the invention;

Fig. 5 is the schematic flow sheet according to the preferred vehicle checking method of the embodiment of the invention; And

Fig. 6 is the structured flowchart according to the labelling apparatus of the road target of the embodiment of the invention.

Embodiment

Need to prove, in the situation that do not conflict, embodiment and the feature among the embodiment among the application can make up mutually.Describe below with reference to the accompanying drawings and in conjunction with the embodiments the present invention in detail.

Can't differentiate the problem of road target for human pilot under the traffic scene in the greasy weather situation in the correlation technique, the embodiment of the invention provides a kind of tagging scheme of road target.For the traffic Driving Scene, in the greasy weather situation, input is contained the mist image carry out the mist elimination pre-service, road target in the image behind the detection mist elimination also marks in image, the image that will be marked with road target is again presented to human pilot, so that human pilot is unlikely to leave the track because of dense fog and collide vehicle.The below is described in detail the scheme of the embodiment of the invention.

According to the embodiment of the invention, provide a kind of labeling method of road target.

Fig. 1 is the process flow diagram according to the labeling method of the road target of the embodiment of the invention, and as shown in Figure 1, the method comprising the steps of S102 is to step S108.

Step S102 obtains the view data I (x, y) that contains the mist image of traffic scene.

Step S104 carries out mist elimination to the above-mentioned picture number I (x, y) that contains the mist image and processes, and obtains the view data J (x, y) behind the mist elimination.

Step S106 detects the road target among the view data J (x, y) behind the above-mentioned mist elimination.

The road target that step S108, mark detect.

Use the technical scheme of the embodiment of the invention, the mist image that contains to traffic scene carries out the mist elimination processing, detect road target in the image after mist elimination is processed, and in image, mark the road target that detects, so that human pilot also can be recognized the situation of road ahead accurately in the greasy weather, improved the security that traffic is driven.

In the embodiment of the invention, above-mentioned road target comprises: barrier in road and/or the road.Barrier in the road comprises that vehicle, pedestrian etc. affect the object that human pilot is driven.

The below is described the preferred implementation of above-mentioned each step of the invention process.

(1) step S104

In embodiments of the present invention, can adopt various ways that the above-mentioned view data I (x, y) that contains the mist image is carried out mist elimination and process, the embodiment of the invention is not construed as limiting this.In a preferred implementation of the embodiment of the invention, can utilize and help Dow process secretly and the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination process.

Further, can according to the above-mentioned dark primary that contains the view data I (x, y) of mist image, the above-mentioned view data I (x, y) that contains the mist image be carried out above-mentioned mist elimination process.In embodiments of the present invention, can determine the above-mentioned dark primary I that contains the view data I (x, y) of mist image ^Drak(x, y) is according to above-mentioned dark primary I ^Drak(x, y) determines to contain the sky brightness A of the view data of mist image, and according to above-mentioned sky brightness A and above-mentioned dark primary I ^Drak(x, y) determines transmissivity t (x, y), then according to above-mentioned transmissivity t (x, y) and above-mentioned sky brightness A the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination processing, obtain the view data J (x, y) behind the above-mentioned mist elimination.

In an embodiment of the embodiment of the invention, can determine in such a way to contain the view data I (x of mist image, y) dark primary: to containing the view data I (x of mist image, y) three passages of RGB carry out mini-value filtering, and the masterplate size can be set to N * N (N ∈ { 3,5,7..}), in three Color Channels, use the minimum operation computing, obtain the dark primary I of I (x, y) ^Drak(x, y) is: $I^{\mathrm{drak}}(x,y)=\underset{c\∈\{r,g,b\}}{\min }\left(\underset{(x^{\′},y^{\′})\∈\mathrm{\Ω}(x,y)}{\min }\left(I^c(x^{\′},y^{\′})\right)\right).$

Further, can choose dark primary I ^Drak0.1% of the brightness maximum pixel in (x, y), in the middle of above pixel, the pixel that contains the middle intensity maximum of view data I (x, y) of mist image is chosen to be atmosphere light A.Certainly, be not limited to definite method of above-mentioned sky brightness in the embodiment of the invention, according to actual needs, can adopt definite method of other sky brightnesses.

In an embodiment of the embodiment of the invention, can determine in such a way above-mentioned transmissivity t (x, y):

$t(x,y)=\frac{A\left(1\right)+A\left(2\right)+A\left(3\right)}{3}-\mathrm{\ω}{I}^{\mathrm{dark}}(x,y),$ Wherein, A (1), A (2), A (3) are respectively the value of above-mentioned sky brightness A on three passages, and ω ∈ (0,1) is used for regulating the mist elimination degree for regulating parameter.

In the embodiment of the invention, ω is used for regulating the mist elimination degree to the view data I (x, y) that contains the mist image, in specific implementation process, the value of ω can be set according to actual needs.If need complete mist elimination, then can be set to 1 by ω, to realize complete mist elimination.

Further, in an embodiment of the embodiment of the invention, can carry out in such a way above-mentioned mist elimination and process:

$J^c(x,y)=\frac{I(x,y)-A^c}{\max (\frac{t(x,y)}{A^c},t_0)}+A^c,$ C ∈ r, and g, b}, wherein, t ₀Be the default lower limit of t (x, y).

In above-mentioned formula, the RGB triple channel is carried out respectively mist elimination process, with the image J (x, y) after the synthetic recovery of three channel image of RGB that obtain.For fear of

Uncertain type occurs, and has set a lower limit t can for t (x, y) ₀, in actual applications, t ₀A representative value be 0.1.

Further, in an embodiment of the embodiment of the invention, utilization is helped Dow process secretly the above-mentioned view data I (x, y) that contains the mist image is carried out after above-mentioned mist elimination processes, can also be to the view data J (x behind the above-mentioned mist elimination, y) carry out the stretch processing of Linear Comparison degree, view data J'(x, y after obtaining stretching), and utilize view data J'(x after the Brovey conversion recovers above-mentioned stretching, y) original color view data B (x, y).

In an embodiment of the embodiment of the invention, can be in such a way three passages of the view data J (x, y) behind the above-mentioned mist elimination be carried out respectively the Linear Comparison degree and stretch:

$J^{\′c}(x,y)=\frac{J^c(x,y)-J_{\min }^c}{J_{\max }^c-J_{\min }^c}(g_{\max }-g_{\min })+g_{\min },c\∈(r,g,b);$

Wherein, g _MaxAnd g _MinBe respectively maximal value and the minimum value of passage g,

With

Be respectively maximal value and the minimum value of J (x, y) each passage in above-mentioned three passages.

Further, in an embodiment of the embodiment of the invention, can recover in such a way the original color view data B (x, y) of the view data J ' (x, y) after the above-mentioned stretching:

$B(x,y,i)=J^{\′}(x,y,i)\×\frac{3\×J(x,y,i)}{J(x,y,1)+J(x,y,2)+J(x,y,3)},i=\mathrm{1,2,3}.$

(2) step S106-step S108

In embodiments of the present invention, can adopt different detection methods to detect road target, the below describes the embodiment that the embodiment of the invention detects the vehicle in road and the detection road.

Can adopt based on the road in the Approach for road detection detected image of color images edge detection, the color space conversion of the view data after the input picture mist elimination processed is the LAB color space, extract the image road edge by the information fusion to colour difference, and extract the image border by the Hough conversion and obtain road edge information, the mark road information of engaging in this profession in image then.

Can adopt HOG﹠amp; The SVM method detects front vehicles.Load the good HOG﹠amp that is used for the identification vehicle of off-line training; The svm classifier device calculates its HOG vector for each the normal size subwindow of the view data behind the mist elimination, uses above-mentioned sorter that the HOG feature of these windows is classified, and judges whether it belongs to vehicle, and mark vehicle location in original image.

After marking the road target that detects, the view data of the road target that detects can be arranged by the display device show tags, so that human pilot can carry out driver behavior according to the road target of mark in the image.

The below is described an instantiation of the embodiment of the invention.Fig. 2 is the schematic flow sheet according to the labeling method of the road target of embodiment of the invention preferred embodiment, and as shown in Figure 2, the method comprises the following aspects:

1, the image mist elimination

Fig. 3 is the process flow diagram according to the preferred image defogging method capable of the embodiment of the invention, and as shown in Figure 3, S302 is to step S310 for the method step step.

Step S302 carries out mini-value filtering to the view data I (x, y) that contains the mist image and obtains dark primary I ^Drak(x, y) is referred to as I.

Three passages of RGB to the view data I (x, y) that contains the mist image carry out mini-value filtering, masterplate be of a size of N * N (N ∈ 3,5,7..}), in three Color Channels, use the minimum operation computing, obtain the dark primary I of I (x, y) ^Drak(x, y) is: $I^{\mathrm{drak}}(x,y)=\underset{c\∈\{r,g,b\}}{\min }\left(\underset{(x^{\′},y^{\′})\∈\mathrm{\Ω}(x,y)}{\min }\left(I^c(x^{\′},y^{\′})\right)\right).$

Step S304 determines sky brightness A and transmissivity t (x, y) according to the dark primary that obtains, and transmissivity t (x, y) is referred to as t.

Choose dark primary I ^Drak0.1% of the brightness maximum pixel in (x, y), in the middle of above pixel, the pixel that contains the middle intensity maximum of view data I (x, y) of mist image is chosen to be atmosphere light A.

According to the dark primary I that obtains ^Drak(x, y) estimates transmissivity t (x, y):

$t(x,y)=\frac{A\left(1\right)+A\left(2\right)+A\left(3\right)}{3}-\mathrm{\ω}{I}^{\mathrm{dark}}(x,y);$

Wherein, ω ∈ (0,1) is used for regulating the mist elimination degree for regulating parameter.ω is larger, and mist elimination is more thorough.

Step S306 recovers original image J (x, y) according to sky brightness A and transmissivity t (x, y), referred to as J from the view data I (x, y) that contains the mist image.

Can determine restored image behind the mist elimination according to transmissivity t (x, y) and sky brightness A, formula is:

$J^c(x,y)=\frac{I(x,y)-A^c}{\max (\frac{t(x,y)}{A^c},t_0)}+A^c,c\∈\{r,g,b\};$

In above-mentioned formula, the RGB triple channel is restored respectively, with the image J (x, y) after the synthetic recovery of three channel image of RGB that obtain.For fear of

Uncertain type occurs, and has set a lower limit t in the formula t (x, y) ₀, t ₀A representative value be 0.1.

Step S308 obtains J ' (x, y) to J (x, y) linear stretch, referred to as J '.

Can carry out respectively the Linear Comparison degree to the triple channel of J (x, y) according to following formula stretches:

$J^{\′c}(x,y)=\frac{J^c(x,y)-J_{\min }^c}{J_{\max }^c-J_{\min }^c}(g_{\max }-g_{\min })+g_{\min },$ Wherein, g _MaxAnd g _MinBe respectively maximal value and the minimum value of passage g,

With

Be respectively maximal value and the minimum value of J (x, y) each passage in above-mentioned three passages.

Step S310 adopts Brovey transfer pair J ' (x, y) to recover original image color B (x, y), referred to as B.

Can carry out Recovery processing according to following formula:

$B(x,y,i)=J^{\′}(x,y,i)\×\frac{3\×J(x,y,i)}{J(x,y,1)+J(x,y,2)+J(x,y,3)},i=\mathrm{1,2,3};$

2, Road Detection

In embodiments of the present invention, adopt based on the road in the Approach for road detection detected image of color images edge detection.Fig. 4 is the process flow diagram according to the preferred Approach for road detection of the embodiment of the invention, and as shown in Figure 4, the method comprises that mainly step S402 is to step S408.

Step S402, the color space conversion of the B (x, y) that step S310 is obtained is the LAB color space.

Step S404 adopts Sobel operator Edge detected, extracts the image road edge by the information fusion to colour difference.

Step S406, the edge image of binaryzation.

Step S408 extracts the image border by the Hough conversion and obtains road edge information.

After step S408, can in former figure, mark the road information that detects.

3, front vehicles detects

In example of the present invention, adopt HOG﹠amp; The SVM method detects front vehicles.Fig. 5 is the schematic flow sheet according to the preferred vehicle checking method of the embodiment of the invention, and as shown in Figure 5, the method comprises the following aspects.

1) loads the good HOG﹠amp that is used for the identification vehicle of off-line training; The svm classifier device.

2) each the normal size subwindow for input picture B (x, y) calculates its HOG vector, uses sorter that the HOG feature of these windows is classified, and judges whether it belongs to vehicle.

3) in original image, mark vehicle location.

Use the technical scheme of this example, for the greasy weather low visibility, drive adventurous problem, carry out the mist elimination pre-service with helping the mist image that contains of algorithm to traffic scene secretly, and the result further processed strengthen its contrast, recover simultaneously the original image color, Road Detection on this basis and vehicle detection have further guaranteed the security of driving in the greasy weather.

According to the embodiment of the invention, also provide a kind of labelling apparatus of road target, in order to the said method of realizing that the embodiment of the invention provides.

Fig. 6 is the structured flowchart according to the labelling apparatus of the road target of the embodiment of the invention, and as shown in Figure 6, this device mainly comprises: acquisition module 10, mist elimination module 20, detection module 30 and mark module 40.Wherein, acquisition module 10 is for the view data I (x, y) that contains the mist image that obtains traffic scene; Mist elimination module 20 is connected with acquisition module 10, is used for that the above-mentioned picture number I (x, y) that contains the mist image is carried out mist elimination and processes, and obtains the view data J (x, y) behind the mist elimination; Detection module 30 is connected with mist elimination module 20, for detection of the road target among the view data J (x, y) behind the above-mentioned mist elimination; Mark module 40 is connected with detection module 30, is used for the road target that mark detects.

Use the technical scheme of the embodiment of the invention, the mist image that contains to traffic scene carries out the mist elimination processing, detect road target in the image after mist elimination is processed, and in image, mark the road target that detects, so that human pilot also can be recognized the situation of road ahead accurately in the greasy weather, improved the security that traffic is driven.

In the embodiment of the invention, above-mentioned road target comprises: barrier in road and/or the road.Barrier in the road comprises that vehicle, pedestrian etc. affect the object that human pilot is driven.

The below is described the preferred implementation of above-mentioned modules of the invention process.

(1) the mist elimination module 20

In embodiments of the present invention, mist elimination module 20 can adopt various ways that the above-mentioned view data I (x, y) that contains the mist image is carried out mist elimination and process, and the embodiment of the invention is not construed as limiting this.In a preferred implementation of the embodiment of the invention, mist elimination module 20 can be utilized and help Dow process secretly and the above-mentioned view data I (x, y) that contains the mist image is carried out above-mentioned mist elimination process.

Further, can according to the above-mentioned dark primary that contains the view data I (x, y) of mist image, the above-mentioned view data I (x, y) that contains the mist image be carried out above-mentioned mist elimination process.

In embodiments of the present invention, above-mentioned mist elimination module 20 can comprise: the first determining unit is used for determining the above-mentioned dark primary I that contains the view data I (x, y) of mist image ^Drak(x, y); The second determining unit is used for according to above-mentioned dark primary I ^Drak(x, y) determines the above-mentioned sky brightness A that contains the view data of mist image; The 3rd determining unit is used for according to above-mentioned sky brightness A and above-mentioned dark primary I ^Drak(x, y) determines transmissivity t (x, y); The mist elimination unit is used for according to above-mentioned transmissivity t (x, y) and above-mentioned sky brightness A the above-mentioned view data I (x, y) that contains the mist image being carried out above-mentioned mist elimination processing, obtains the view data J (x, y) behind the above-mentioned mist elimination.

In an embodiment of the embodiment of the invention, the first determining unit can determine to contain the view data I (x of mist image in such a way, y) dark primary: to containing the view data I (x of mist image, y) three passages of RGB carry out mini-value filtering, and the masterplate size can be set to N * N (N ∈ { 3,5,7..}), in three Color Channels, use the minimum operation computing, obtain the dark primary I of I (x, y) ^Drak(x, y) is: $I^{\mathrm{drak}}(x,y)=\underset{c\∈\{r,g,b\}}{\min }\left(\underset{(x^{\′},y^{\′})\∈\mathrm{\Ω}(x,y)}{\min }\left(I^c(x^{\′},y^{\′})\right)\right).$

Further, the second determining unit can be chosen dark primary I ^Drak0.1% of the brightness maximum pixel in (x, y), in the middle of above pixel, the pixel that contains the middle intensity maximum of view data I (x, y) of mist image is chosen to be atmosphere light A.Certainly, be not limited to definite method of above-mentioned sky brightness in the embodiment of the invention, according to actual needs, can adopt definite method of other sky brightnesses.

In an embodiment of the embodiment of the invention, the 3rd determining unit can be determined above-mentioned transmissivity t (x, y) in such a way:

$t(x,y)=\frac{A\left(1\right)+A\left(2\right)+A\left(3\right)}{3}-\mathrm{\ω}{I}^{\mathrm{dark}}(x,y);$

Wherein, A (1), A (2), A (3) are respectively the value of above-mentioned sky brightness A on three passages, and ω ∈ (0,1) is used for regulating the mist elimination degree for regulating parameter.

In the embodiment of the invention, ω is used for regulating the mist elimination degree to the view data I (x, y) that contains the mist image, in specific implementation process, the value of ω can be set according to actual needs.If need complete mist elimination, then can be set to 1 by ω, to realize complete mist elimination.

Further, in an embodiment of the embodiment of the invention, the mist elimination unit can carry out above-mentioned mist elimination in such a way to be processed:

$J^c(x,y)=\frac{I(x,y)-A^c}{\max (\frac{t(x,y)}{A^c},t_0)}+A^c,c\∈\{r,g,b\},$ Wherein, t ₀Be the default lower limit of t (x, y).

In above-mentioned formula, the RGB triple channel is carried out respectively mist elimination process, with the image J (x, y) after the synthetic recovery of three channel image of RGB that obtain.For fear of

Uncertain type occurs, and has set a lower limit t can for t (x, y) ₀, in actual applications, t ₀A representative value be 0.1.

Further, in an embodiment of the embodiment of the invention, utilization is helped Dow process secretly to the above-mentioned view data I (x that contains the mist image, y) carry out after the above-mentioned mist elimination processing, the processing that can also carry out to the view data J (x, y) behind the above-mentioned mist elimination stretch processing of Linear Comparison degree and recover former figure color.Therefore, above-mentioned mist elimination module 20 can also comprise: draw unit is used for the view data J (x, y) behind the above-mentioned mist elimination being carried out the stretch processing of Linear Comparison degree, the view data J ' (x, y) after obtaining stretching; Recovery unit is used for utilizing the Brovey conversion to recover the original color view data B (x, y) of the view data J ' (x, y) after the above-mentioned stretching.

In an embodiment of the embodiment of the invention, draw unit can be in such a way carried out respectively the Linear Comparison degree to three passages of the view data J (x, y) behind the above-mentioned mist elimination and is stretched:

$J^{\′c}(x,y)=\frac{J^c(x,y)-J_{\min }^c}{J_{\max }^c-J_{\min }^c}(g_{\max }-g_{\min })+g_{\min },c\∈(r,g,b);$

Wherein, g _MaxAnd g _MinBe respectively maximal value and the minimum value of passage g,

With

Be respectively maximal value and the minimum value of J (x, y) each passage in above-mentioned three passages.

Further, in an embodiment of the embodiment of the invention, recovery unit is to recover in such a way view data J'(x, the y after the above-mentioned stretching) original color view data B (x, y):

$B(x,y,i)=J^{\′}(x,y,i)\×\frac{3\×J(x,y,i)}{J(x,y,1)+J(x,y,2)+J(x,y,3)},i=\mathrm{1,2,3}.$

(2) detection module 30 and mark module 40

In embodiments of the present invention, detection module 30 can adopt different detection methods to detect road target, and the below describes the embodiment that the embodiment of the invention detects the vehicle in road and the detection road.

Detection module 30 can adopt based on the road in the Approach for road detection detected image of color images edge detection, the color space conversion of the view data after detection module 30 is processed the input picture mist elimination is the LAB color space, extract the image road edge by the information fusion to colour difference, and extract the image border by the Hough conversion and obtain road edge information, mark module 40 mark in image road information of engaging in this profession then.

In addition, detection module 30 can adopt HOG﹠amp; The SVM method detects front vehicles.Detection module 30 loads the good HOG﹠amp that is used for the identification vehicle of off-line training; The svm classifier device calculates its HOG vector for each the normal size subwindow of the view data behind the mist elimination, uses above-mentioned sorter that the HOG feature of these windows is classified, and judges whether it belongs to vehicle, and mark module 40 marks vehicle location in original image.

After marking the road target that detects, the view data of the road target that detects can be arranged by the display device show tags, so that human pilot can carry out driver behavior according to the road target of mark in the image.Therefore, said apparatus can also comprise: display module, and for the view data that the road target that detects is arranged by the display device show tags.

Through the above description of the embodiments, those skilled in the art can be well understood to the present invention and can realize by the mode that software adds essential hardware (comprising camera, processor etc.).Based on such understanding, the part that the technical scheme of the embodiment of the invention contributes to prior art in essence in other words can embody with the form of entity apparatus (comprising software section), and this device can be carried out the described method of some part of each embodiment of the present invention or embodiment.

The above is the preferred embodiments of the present invention only, is not limited to the present invention, and for a person skilled in the art, the present invention can have various modifications and variations.Within the spirit and principles in the present invention all, any modification of doing, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. the labeling method of a road target is characterized in that, comprising:

Obtain the view data I (x, y) that contains the mist image of traffic scene;

The described picture number I (x, y) that contains the mist image is carried out mist elimination process, obtain the view data J (x, y) behind the mist elimination;

Detect the road target among the view data J (x, y) behind the described mist elimination;

The road target that mark detects.

2. method according to claim 1 is characterized in that, the described view data I (x, y) that contains the mist image is carried out mist elimination process, and comprising:

Utilization is helped Dow process secretly the described view data I (x, y) that contains the mist image is carried out described mist elimination processing.

3. method according to claim 2 is characterized in that, utilization is helped Dow process secretly the described view data I (x, y) that contains the mist image is carried out described mist elimination processing, comprising:

Determine the described dark primary I that contains the view data I (x, y) of mist image ^Drak(x, y);

According to described dark primary I ^Drak(x, y) determines the described sky brightness A that contains the view data of mist image;

According to described sky brightness A and described dark primary I ^Drak(x, y) determines transmissivity t (x, y);

According to described transmissivity t (x, y) and described sky brightness A the described view data I (x, y) that contains the mist image is carried out described mist elimination and process, obtain the view data J (x, y) behind the described mist elimination.

4. method according to claim 3 is characterized in that, determines in such a way described transmissivity t (x, y):

$t(x,y)=\frac{A\left(1\right)+A\left(2\right)+A\left(3\right)}{3}-\mathrm{\ω}{I}^{\mathrm{dark}}(x,y),$ Wherein, A (1), A (2), A (3) are respectively the value of described sky brightness A on three passages, and ω ∈ (0,1) is used for regulating the mist elimination degree for regulating parameter.

5. method according to claim 3 is characterized in that, carries out in such a way described mist elimination and processes:

$J^c(x,y)=\frac{I(x,y)-A^c}{\max (\frac{t(x,y)}{A^c},t_0)}+A^c,c\∈\{r,g,b\},$ Wherein, t ₀Be the default lower limit of t (x, y).

6. each described method in 5 according to claim 3 is characterized in that, utilizes to help Dow process secretly the described view data I (x, y) that contains the mist image is carried out also comprising after described mist elimination processes:

View data J (x, y) behind the described mist elimination is carried out the stretch processing of Linear Comparison degree, the view data J ' (x, y) after obtaining stretching;

Utilize the Brovey conversion to recover the original color view data B (x, y) of the view data J ' (x, y) after the described stretching.

7. the labelling apparatus of a road target is characterized in that, comprising:

Acquisition module is for the view data I (x, y) that contains the mist image that obtains traffic scene;

The mist elimination module is used for that the described picture number I (x, y) that contains the mist image is carried out mist elimination and processes, and obtains the view data J (x, y) behind the mist elimination;

Detection module is for detection of the road target among the view data J (x, y) behind the described mist elimination;

Mark module is used for the road target that mark detects.

8. device according to claim 7 is characterized in that, described mist elimination module is helped Dow process secretly for utilization the described view data I (x, y) that contains the mist image is carried out described mist elimination processing.

9. device according to claim 8 is characterized in that, described mist elimination module comprises:

The first determining unit is used for determining the described dark primary I that contains the view data I (x, y) of mist image ^Drak(x, y);

The second determining unit is used for according to described dark primary I ^Drak(x, y) determines the described sky brightness A that contains the view data of mist image;

The 3rd determining unit is used for according to described sky brightness A and described dark primary I ^Drak(x, y) determines transmissivity t (x, y);

The mist elimination unit is used for according to described transmissivity t (x, y) and described sky brightness A the described view data I (x, y) that contains the mist image being carried out described mist elimination processing, obtains the view data J (x, y) behind the described mist elimination.

10. device according to claim 9 is characterized in that, described mist elimination module also comprises:

Draw unit is used for the view data J (x, y) behind the described mist elimination being carried out the stretch processing of Linear Comparison degree, view data J'(x, y after obtaining stretching);

Recovery unit is used for utilizing the Brovey conversion to recover the original color view data B (x, y) of the view data J ' (x, y) after the described stretching.

Images