CN104573655A - Blind sidewalk direction detection method based on video - Google Patents

Abstract

The invention provides a blind sidewalk direction detection method based on a video. The blind sidewalk direction detection method comprises the steps that blind sideway video images are collected; the current frame of color image is transformed into a gray image; fuzzy processing in the horizontal direction is conducted on the gray image through a mean filter; image enhancement is conducted on the gray image obtained after the fuzzy processing is conducted; a binary value perpendicular edge feature map of the gray image obtained after the image enhancement is conducted is obtained; a perpendicular edge connected region of a blind sideway is obtained; the inclined angle of the blind sideway is obtained; the angle amount needing to be corrected in the marching direction is obtained through calculation; the angle in the marching direction of the current frame is updated; a user is prompted through voice with the direction of the blind sideway when marching continually. According to the blind sidewalk direction detection method, the video image processing technology is adopted, the direction of the blind sideway is judged based on the optimal edge angle on the traditional significance, and the method has the advantages of being high in detection accuracy, high in robustness, small in influence of the environment, high in speed, low in cost and the like.

Description

A kind of sidewalk for visually impaired people direction detection method based on video

Technical field

The present invention relates to technical field of image processing, specifically a kind of sidewalk for visually impaired people direction detection method based on video.

Background technology

Sidewalk for visually impaired people is the accessible infrastructure of one in city, is intended to for visually impaired person provides walking along the street convenience and safety.Along with the progress of society, nearly all road is all furnished with sidewalk for visually impaired people.But, sidewalk for visually impaired people does not but play its due effect, trace it to its cause, existing blind-guiding method is as guide-stick for blind people and seeing-eye dog etc., there is larger deficiency, sidewalk for visually impaired people directional information or guide cost is too high or inconvenient operation etc. accurately maybe can not be provided, real trip cannot be provided convenient for vast visually impaired person.

At present, for the detection of sidewalk for visually impaired people directional information, conventional technical method has following several:

(1) based on ultrasound wave or ultrared detection method, as Chinese utility model patent CN203935390U discloses a kind of electronics guide-stick for blind people.These class methods can detect that working direction has clear, but for the concrete trend of sidewalk for visually impaired people, or judged by traditional angle sheave, therefore sidewalk for visually impaired people directional information accurately can not be provided.

(2) based on the detection method of depth survey, as Chinese invention patent application CN104055657A discloses a kind of blind-guide brick based on Kinect and its implementation.These class methods, by three dimensional depth measurement mechanism, obtain the depth information of working direction, and so as to determining whether barrier, the shortcoming of these class methods is that equipment is complicated, and be not suitable for long-time use, can not provide sidewalk for visually impaired people directional information accurately, cost is relatively high simultaneously.

(3) based on the detection method of Computer Vision, as Chinese invention patent application CN103908365A discloses a kind of electronic travel aid device.These class methods are by canny boundary operator and hough line detection algorithm, obtain the angle information on Liang Tiao border, sidewalk for visually impaired people, these class methods are for the sidewalk for visually impaired people of standard, effect is better, but in actual environment, the many factors of interference sidewalk for visually impaired people, as surrounding background exists stronger marginal interference, sidewalk for visually impaired people part edge disappearance etc., these class methods often lost efficacy.

Summary of the invention

The object of the present invention is to provide a kind of sidewalk for visually impaired people direction detection method based on video, Corpus--based Method technical know-how, obtain optimum sidewalk for visually impaired people edge angle information, improve the accuracy of testing result and the robustness for various interference environment.

Technical scheme of the present invention is:

Based on a sidewalk for visually impaired people direction detection method for video, comprise the following steps:

(1) sidewalk for visually impaired people video image is gathered;

(2) current frame color image is transformed to gray level image;

(3) mean filter is utilized to carry out horizontal direction Fuzzy Processing to gray level image;

(4) image enhaucament is carried out to the gray level image after Fuzzy Processing;

(5) the two-value vertical edge characteristic pattern of the gray level image after image enhaucament is obtained;

(6) the vertical edge connected region of sidewalk for visually impaired people is obtained;

(7) angle of inclination, sidewalk for visually impaired people is obtained;

(8) angular metric that working direction needs to revise is calculated;

(9) the working direction angle of present frame is upgraded;

(10) direction, sidewalk for visually impaired people when voice message user moves on.

The described sidewalk for visually impaired people direction detection method based on video, in step (2), is describedly transformed to gray level image by coloured image, realizes especially by following formula:

f(x，y)＝0.299r(x，y)+0.587g(x，y)+0.114b(x，y)

Wherein, f (x, y) pixel (x in the rear gray level image of conversion is represented, y) gray-scale value at place, r (x, y), g (x, y), b (x, y) the three-channel value of red, green, blue at pixel (x, y) place in coloured image is represented respectively.

The described sidewalk for visually impaired people direction detection method based on video, in step (3), the rectangle convolution kernel that described mean filter adopts is:

$K=\frac{1}{3*h}\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ \·& \·& \·\\ \·& \·& \·\\ \·& \·& \·\\ 1& 1& 1\\ 1& 1& 1\end{array}\right]$

Wherein, K represents rectangle convolution kernel, and h represents the height of rectangle convolution kernel.

The described sidewalk for visually impaired people direction detection method based on video, in step (4), describedly carries out image enhaucament to the gray level image after Fuzzy Processing, realizes especially by following formula:

$g(x,y)=255*{\left(\frac{f(x,y)}{255}\right)}^{\mathrm{\γ}}$

Wherein, g (x, y) represents the gray-scale value at pixel (x, y) place in image after image enhaucament, and f (x, y) represents the gray-scale value at pixel (x, y) place in image before image enhaucament, and γ represents gamma filter factor.

The described sidewalk for visually impaired people direction detection method based on video, in step (5), the two-value vertical edge characteristic pattern of the gray level image after described acquisition image enhaucament, comprises the following steps:

(51) utilize the sobel edge detection operator improved, obtain the vertical edge characteristic pattern of the gray level image after image enhaucament, the sobel edge detection operator of described improvement adopts following formula:

$K=\begin{array}{}\left[\begin{array}{ccccc}0& -1& 0& 1& 0\\ -1& -2& 0& 2& 1\\ 0& -1& 0& 1& 0\end{array}\right]\end{array}$

Wherein, K represents the sobel edge detection operator of improvement;

(52) based on local binarization algorithm, adopt following formula, obtain two-value vertical edge characteristic pattern:

$g(x,y)=\left\{\begin{array}{cc}255& f(x,y)\&GreaterEqual;T\\ 0& f(x,y)<T\end{array}\right.$

$T=\frac{1}{M*N}\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x_m,y_n)+C$

Wherein, g (x, y) represents the gray-scale value at pixel (x, y) place in two-value vertical edge characteristic pattern, and f (x, y) represents the gray-scale value at pixel (x, y) place in vertical edge characteristic pattern, f (x _m, y _n) represent pixel (x in M*N neighborhood centered by pixel (x, y) in vertical edge characteristic pattern _m, y _n) gray-scale value at place, M, N represent width and the height of neighborhood respectively, and C represents side-play amount.

The described sidewalk for visually impaired people direction detection method based on video, in step (6), the vertical edge connected region of described acquisition sidewalk for visually impaired people, comprises the following steps:

(61) utilize convolution mask, remove the isolated point in two-value vertical edge characteristic pattern by medium filtering, described convolution mask adopts following formula:

$K=\left[\begin{array}{ccccc}0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0\\ 1& 1& 1& 1& 1\\ 0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0\end{array}\right]$

Wherein, K represents convolution mask;

(62) the Clutter edge connected region in two-value vertical edge characteristic pattern is removed;

(63) utilize structural element template, n closing operation of mathematical morphology is carried out to the two-value vertical edge characteristic pattern through above-mentioned process;

(64) Bock Altitude is greater than the connected region of predetermined threshold value.

The described sidewalk for visually impaired people direction detection method based on video, in step (7), angle of inclination, described acquisition sidewalk for visually impaired people, comprises the following steps:

(71) based on principle of least square method, adopt following formula to carry out fitting a straight line to the data of each connected region foreground target point respectively, calculate the angle of inclination of each connected region, and result of calculation is corrected:

$\mathrm{\&alpha;}\left[j\right]=\arctan \left(\frac{\mathrm{N\&Sigma;}{x}_i{y}_i-\mathrm{\&Sigma;}{x}_i\mathrm{\&Sigma;}{y}_i}{\mathrm{N\&Sigma;}{x_i}^2-{\left(\mathrm{\&Sigma;}{x}_i\right)}^2}\right)$

${\mathrm{\&alpha;}}^{\&prime;}\left[j\right]=\left\{\begin{array}{c}\mathrm{\&alpha;}\left[j\right]+\mathrm{\&pi;},\mathrm{\&alpha;}\left[j\right]<0\\ \mathrm{\&alpha;}\left[j\right],\mathrm{\&alpha;}\left[j\right]\&GreaterEqual;0\end{array}\right.$

Wherein, α [j] represents the angle of inclination of the connected region calculated, x _i, y _irepresent horizontal ordinate, the ordinate of foreground target point in connected region, N represents the number of foreground target point in connected region; α ' [j] represents the angle of inclination of the connected region after correcting;

(72) following formula is adopted to add up the angle of inclination histogram of connected region:

$\left\{\begin{array}{c}\mathrm{hist}\left[\mathrm{bin}\right]=\mathrm{hist}\left[\mathrm{bin}\right]+1\\ \mathrm{bin}=\mathrm{floor}\left(\frac{{\mathrm{\&alpha;}}^{\&prime;}\left[j\right]*30}{\mathrm{\&pi;}}\right)\end{array}\right.$

Wherein, hist [bin] represents the angle of inclination histogram of connected region, function representation is chosen and is not more than maximum integer, the span of bin is 0 ~ 29;

(73) maximum histogram dimension bi is obtained;

(74) mean value at all angles of inclination fallen in hist [bi] is calculated be angle of inclination, sidewalk for visually impaired people.

The described sidewalk for visually impaired people direction detection method based on video, in step (8), described in calculate working direction need revise angular metric, realize especially by following formula:

Wherein, d_ α represents the angular metric that working direction needs are revised, γ represents the working direction angle of previous frame, represent the angle of inclination, sidewalk for visually impaired people of present frame, d_ α > 0 represents that needing angle correction d_ α, d_ α < 0 left to represent needs angle correction to the right | d_ α | and, d_ α=0 represents and does not need to carry out working direction correction.

The described sidewalk for visually impaired people direction detection method based on video, in step (9), the working direction angle of described renewal present frame, realizes especially by following formula:

${\mathrm{\&gamma;}}^{\&prime;}=\mathrm{\&gamma;}*0.4+\stackrel{\&OverBar;}{\mathrm{\&alpha;}}*0.6$

Wherein, the working direction angle of γ ' expression present frame, γ represents the working direction angle of previous frame, represent the angle of inclination, sidewalk for visually impaired people of present frame.

The present invention adopts video image processing technology, and the optimal edge angle in Corpus--based Method meaning judges direction, sidewalk for visually impaired people, has that accuracy in detection is high, strong robustness, little a, feature such as speed is fast, cost is low affected by environment.

Accompanying drawing explanation

Fig. 1 is method flow diagram of the present invention;

Fig. 2 is sidewalk for visually impaired people gray level image;

Fig. 3 is Fuzzy Processing and image enhaucament Overlay figure;

Fig. 4 is two-value vertical edge characteristic pattern;

Fig. 5 removes isolated point design sketch;

Fig. 6 removes Clutter edge connected region design sketch;

Fig. 7 is morphology operations design sketch;

Fig. 8 is the vertical edge connected region design sketch of sidewalk for visually impaired people.

Embodiment

Below, the present invention is further illustrated with specific embodiment by reference to the accompanying drawings.

In order to the superiority of the method for the invention provides is described, the illustration adopted in the present embodiment is typical strong jamming sidewalk for visually impaired people image, and as shown in Figure 2, in this figure, sidewalk for visually impaired people does not have obvious color distortion, and surrounding background exists the marginal interference in same tilt direction simultaneously.

As shown in Figure 1, a kind of sidewalk for visually impaired people direction detection method based on video, comprises the following steps:

S1, collection sidewalk for visually impaired people video image, require that sidewalk for visually impaired people is positioned at the horizontal center position of image as far as possible.

S2, according to formula [1], coloured image is transformed to gray level image, and effect is as Fig. 2:

Formula [1]:

f(x，y)＝0.299r(x，y)+0.587g(x，y)+0.114b(x，y)

Wherein, f (x, y) is pixel (x in the rear gray level image of conversion, y) gray-scale value at place, r (x, y), g (x, y), b (x, y) is the three-channel value of red, green, blue at pixel (x, y) place in coloured image.

S3, use mean filter carry out horizontal direction Fuzzy Processing to gray level image.Because the rectangle protruding block region of inside, sidewalk for visually impaired people and outer edge zone all have obvious vertical direction edge feature, the present invention uses rectangle convolution kernel as shown in formula [2], complete the horizontal direction Fuzzy Processing of image, retain the extrorse interference of simultaneously removing horizontal direction edge of Vertical Square:

Formula [2]:

$K=\frac{1}{3*h}\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;\\ 1& 1& 1\\ 1& 1& 1\end{array}\right]$

Wherein, h is the height of convolution kernel, and value is larger, and horizontal direction edge is fuzzyyer.

S4, carry out image enhaucament to the gray level image after Fuzzy Processing, according to following formula [3], based on gamma filtering theory correcting image, strengthen the intensity of image border, sidewalk for visually impaired people, effect is as Fig. 3:

Formula [3]:

$g(x,y)=255*{\left(\frac{f(x,y)}{255}\right)}^{\mathrm{\&gamma;}}$

Wherein, g (x, y) is the gray-scale value at pixel (x, y) place in image after image enhaucament, f (x, y) be the gray-scale value at pixel (x, y) place in image before image enhaucament, γ is gamma filter factor, as γ < 1, significantly can strengthen the contrast of gray-scale value lower region, as γ > 1, significantly can strengthen the contrast of gray-scale value upper zone.

S5, obtain through Fuzzy Processing, strengthen after the two-value vertical edge characteristic pattern of gray level image, effect is as Fig. 4, and concrete steps are as follows:

S51, based on the sobel edge detection operator improved, as shown in formula [4], obtain vertical edge characteristic pattern:

Formula [4]:

$K=\begin{array}{}\left[\begin{array}{ccccc}0& -1& 0& 1& 0\\ -1& -2& 0& 2& 1\\ 0& -1& 0& 1& 0\end{array}\right]\end{array}$

S52, based on local binarization algorithm, obtain two-value vertical edge characteristic pattern; Because the edge strength of edge image zones of different is different, use overall Binarization methods can cause the disappearance of part edge, so, according to formula [5] and formula [6], complete the local binarization of vertical edge characteristic pattern:

Formula [5]:

$g(x,y)=\left\{\begin{array}{cc}255& f(x,y)\&GreaterEqual;T\\ 0& f(x,y)<T\end{array}\right.$

Formula [6]:

$T=\frac{1}{M*N}\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x_m,y_n)+C$

Wherein, g (x, y) is the gray-scale value at pixel (x, y) place in two-value vertical edge characteristic pattern, and f (x, y) is the gray-scale value at pixel (x, y) place in vertical edge characteristic pattern, f (x _m, y _n) be pixel (x in M*N neighborhood centered by pixel (x, y) in vertical edge characteristic pattern _m, y _n) gray-scale value at place, M, N are width and the height of neighborhood, and C is side-play amount, are generally positive constants.

The vertical edge connected region of S6, acquisition sidewalk for visually impaired people, concrete steps are as follows:

S61, according to convolution mask, as shown in formula [7], remove isolated point by medium filtering, fill less gap, effect is as Fig. 5:

Formula [7]:

$K=\left[\begin{array}{ccccc}0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0\\ 1& 1& 1& 1& 1\\ 0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0\end{array}\right]$

S62, remove Clutter edge connected region, comprise the less connected region of area and be highly less than the connected region of 3 times of width, effect is as Fig. 6;

S63, according to structural element template, as shown in formula [8], carry out n closing operation of mathematical morphology, connect the connected region of vertical direction close together, the image that the value of n gathers for distinct device is slightly different, value 20 in the present embodiment, and effect is as Fig. 7:

Formula [8]:

$K=\left[\begin{array}{ccc}0& 1& 0\\ 0& 1& 0\\ 1& 1& 1\\ 0& 1& 0\\ 0& 1& 0\end{array}\right]$

S64, Bock Altitude are greater than the connected region of h, and in the present embodiment, the value of h is 0.4 times of picture altitude, and effect is as Fig. 8:

S7, acquisition angle of inclination, sidewalk for visually impaired people, concrete steps are as follows:

S71, based on principle of least square method, as shown in formula [9], utilize the foreground target point data of each connected region, carry out fitting a straight line respectively, calculate the tilt angle alpha [j] of current connected region, owing to the scope at angle of inclination, sidewalk for visually impaired people being limited to [0, π], so again according to formula [10], the tilt angle alpha [j] of the connected region calculated is corrected:

Formula [9]:

$\mathrm{\&alpha;}\left[j\right]=\arctan \left(\frac{\mathrm{N\&Sigma;}{x}_i{y}_i-\mathrm{\&Sigma;}{x}_i\mathrm{\&Sigma;}{y}_i}{\mathrm{N\&Sigma;}{x_i}^2-{\left(\mathrm{\&Sigma;}{x}_i\right)}^2}\right)$

Formula [10]:

${\mathrm{\&alpha;}}^{\&prime;}\left[j\right]=\left\{\begin{array}{c}\mathrm{\&alpha;}\left[j\right]+\mathrm{\&pi;},\mathrm{\&alpha;}\left[j\right]<0\\ \mathrm{\&alpha;}\left[j\right],\mathrm{\&alpha;}\left[j\right]\&GreaterEqual;0\end{array}\right.$

Wherein, x _i, y _irepresent horizontal ordinate, the ordinate of foreground target point in current connected region, N represents the number of foreground target point in current connected region; α ' [j] represents the angle of inclination after the correction of each connected region.

S72, as shown in formula [11], statistics connected region angle of inclination histogram hist [bin]:

Formula [11]:

$\left\{\begin{array}{c}\mathrm{hist}\left[\mathrm{bin}\right]=\mathrm{hist}\left[\mathrm{bin}\right]+1\\ \mathrm{bin}=\mathrm{floor}\left(\frac{{\mathrm{\&alpha;}}^{\&prime;}\left[j\right]*30}{\mathrm{\&pi;}}\right)\end{array}\right.$

Wherein, floor (x) function representation chooses the maximum integer being not more than x, and the span of bin is 0 ~ 29.

S73, obtain maximum histogram dimension bi;

S74, calculate the mean value of all tilt angle alpha ' [j] fallen in hist [bi] namely be angle of inclination, sidewalk for visually impaired people.

S8, based on formula [12], calculate working direction need revise angular metric d_ α:

Formula [12]:

Wherein, γ is the working direction angle of previous frame, and d_ α > 0 represents that needing angle correction d_ α, d_ α < 0 left to represent needs angle correction to the right | d_ α | and, d_ α=0 represents and does not need to carry out working direction correction.

S9, according to formula [13], upgrade the working direction angle γ of present frame ^':

Formula [13]:

${\mathrm{\&gamma;}}^{\&prime;}=\mathrm{\&gamma;}*0.4+\stackrel{\&OverBar;}{\mathrm{\&alpha;}}*0.6$

S10, voice broadcast, the direction, sidewalk for visually impaired people when reminding user to move on.

The above embodiment is only be described the preferred embodiment of the present invention; not scope of the present invention is limited; under not departing from the present invention and designing the prerequisite of spirit; the various distortion that those of ordinary skill in the art make technical scheme of the present invention and improvement, all should fall in protection domain that claims of the present invention determine.

Claims (9)

1., based on a sidewalk for visually impaired people direction detection method for video, it is characterized in that, comprise the following steps:

(1) sidewalk for visually impaired people video image is gathered;

(2) current frame color image is transformed to gray level image;

(3) mean filter is utilized to carry out horizontal direction Fuzzy Processing to gray level image;

(4) image enhaucament is carried out to the gray level image after Fuzzy Processing;

(5) the two-value vertical edge characteristic pattern of the gray level image after image enhaucament is obtained;

(6) the vertical edge connected region of sidewalk for visually impaired people is obtained;

(7) angle of inclination, sidewalk for visually impaired people is obtained;

(8) angular metric that working direction needs to revise is calculated;

(9) the working direction angle of present frame is upgraded;

(10) direction, sidewalk for visually impaired people when voice message user moves on.

2. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (2), described coloured image is transformed to gray level image, realizes especially by following formula:

f(x，y)＝0.299r(x，y)+0.587g(x，y)+0.114b(x，y)

Wherein, f (x, y) pixel (x in the rear gray level image of conversion is represented, y) gray-scale value at place, r (x, y), g (x, y), b (x, y) the three-channel value of red, green, blue at pixel (x, y) place in coloured image is represented respectively.

3. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (3), the rectangle convolution kernel that described mean filter adopts is:

$K=\frac{1}{3*h}\left[\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ .& .& .\\ .& .& .\\ .& .& .\\ 1& 1& 1\\ 1& 1& 1\end{array}\right]$

Wherein, K represents rectangle convolution kernel, and h represents the height of rectangle convolution kernel.

4. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (4), describedly carries out image enhaucament to the gray level image after Fuzzy Processing, realizes especially by following formula:

$g(x,y)=255*{\left(\frac{f(x,y)}{255}\right)}^{\mathrm{\&gamma;}}$

Wherein, g (x, y) represents the gray-scale value at pixel (x, y) place in image after image enhaucament, and f (x, y) represents the gray-scale value at pixel (x, y) place in image before image enhaucament, and γ represents gamma filter factor.

5. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (5), the two-value vertical edge characteristic pattern of the gray level image after described acquisition image enhaucament, comprises the following steps:

(51) utilize the sobel edge detection operator improved, obtain the vertical edge characteristic pattern of the gray level image after image enhaucament, the sobel edge detection operator of described improvement adopts following formula:

$K=\left[\begin{array}{ccccc}0& -1& 0& 1& 0\\ -1& -2& 0& 2& 1\\ 0& -1& 0& 1& 0\end{array}\right]$

Wherein, K represents the sobel edge detection operator of improvement;

(52) based on local binarization algorithm, adopt following formula, obtain two-value vertical edge characteristic pattern:

$g(x,y)=\left\{\begin{array}{cc}255& f(x,y)\&GreaterEqual;T\\ 0& f(x,y)<T\end{array}\right.$

$T=\frac{1}{M*N}\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}f(x_m,y_n)+C$

Wherein, g (x, y) represents the gray-scale value at pixel (x, y) place in two-value vertical edge characteristic pattern, and f (x, y) represents the gray-scale value at pixel (x, y) place in vertical edge characteristic pattern, f (x _m, y _n) represent pixel (x in M*N neighborhood centered by pixel (x, y) in vertical edge characteristic pattern _m, y _n) gray-scale value at place, M, N represent width and the height of neighborhood respectively, and C represents side-play amount.

6. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (6), the vertical edge connected region of described acquisition sidewalk for visually impaired people, comprises the following steps:

(61) utilize convolution mask, remove the isolated point in two-value vertical edge characteristic pattern by medium filtering, described convolution mask adopts following formula:

$K=\left[\begin{array}{ccccc}0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0\\ 1& 1& 1& 1& 1\\ 0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0\end{array}\right]$

Wherein, K represents convolution mask;

(62) the Clutter edge connected region in two-value vertical edge characteristic pattern is removed;

(63) utilize structural element template, n closing operation of mathematical morphology is carried out to the two-value vertical edge characteristic pattern through above-mentioned process;

(64) Bock Altitude is greater than the connected region of predetermined threshold value.

7. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (7), angle of inclination, described acquisition sidewalk for visually impaired people, comprises the following steps:

(71) based on principle of least square method, adopt following formula to carry out fitting a straight line to the data of each connected region foreground target point respectively, calculate the angle of inclination of each connected region, and result of calculation is corrected:

$\mathrm{\&alpha;}\left[j\right]=\arctan \left(\frac{\mathrm{N\&Sigma;}{x}_i{y}_i-\mathrm{\&Sigma;}{x}_i\mathrm{\&Sigma;}{y}_i}{\mathrm{N\&Sigma;}{x_i}^2-{\left(\mathrm{\&Sigma;}{x}_i\right)}^2}\right)$

${\mathrm{\&alpha;}}^{\&prime;}\left[j\right]=\left\{\begin{array}{cc}\mathrm{\&alpha;}\left[j\right]+\mathrm{\&pi;},& \mathrm{\&alpha;}\left[j\right]<0\\ \mathrm{\&alpha;}\left[j\right],& \mathrm{\&alpha;}\left[j\right]\&GreaterEqual;0\end{array}\right.$

Wherein, α [j] represents the angle of inclination of the connected region calculated, x _i, y _irepresent horizontal ordinate, the ordinate of foreground target point in connected region, N represents the number of foreground target point in connected region; α ' [j] represents the angle of inclination of the connected region after correcting;

(72) following formula is adopted to add up the angle of inclination histogram of connected region:

$\left\{\begin{array}{c}\mathrm{hist}\left[\mathrm{bin}\right]=\mathrm{hist}\left[\mathrm{bin}\right]+1\\ \mathrm{bin}=\mathrm{floor}\left(\frac{{\mathrm{\&alpha;}}^{\&prime;}\left[j\right]*30}{\mathrm{\&pi;}}\right)\end{array}\right.$

Wherein, hist [bin] represents the angle of inclination histogram of connected region, function representation is chosen and is not more than maximum integer, the span of bin is 0 ~ 29;

(73) maximum histogram dimension bi is obtained;

(74) mean value at all angles of inclination fallen in hist [bi] is calculated be angle of inclination, sidewalk for visually impaired people.

8. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (8), described in calculate working direction need revise angular metric, realize especially by following formula:

Wherein, d_ α represents the angular metric that working direction needs are revised, γ represents the working direction angle of previous frame, represent the angle of inclination, sidewalk for visually impaired people of present frame, d_ α > 0 represents that needing angle correction d_ α, d_ α < 0 left to represent needs angle correction to the right | d_ α | and, d_ α=0 represents and does not need to carry out working direction correction.

9. the sidewalk for visually impaired people direction detection method based on video according to claim 1, is characterized in that, in step (9), the working direction angle of described renewal present frame, realizes especially by following formula:

${\mathrm{\&gamma;}}^{\&prime;}=\mathrm{\&gamma;}*0.4+\stackrel{\&OverBar;}{\mathrm{\&alpha;}}*0.6$

Wherein, the working direction angle of γ ' expression present frame, γ represents the working direction angle of previous frame, represent the angle of inclination, sidewalk for visually impaired people of present frame.