CN106127187A - A kind of have the intelligent robot identifying function - Google Patents

Abstract

nullThe invention provides a kind of intelligent robot having and identifying function，Including intelligent robot and the Target Identification Unit that is connected with intelligent robot，Target is identified following the tracks of by Target Identification Unit view-based access control model feature，Including the suspection Target Acquisition module being sequentially connected with、Colouring information processing module、Profile information processing module、Feature evaluation module，Wherein described original hardwood image is carried out from RGB color to the conversion in hsv color space and builds the described suspection target tonal color model in hsv color space by colouring information processing module，Profile information processing module for carrying out the area type division of characteristic area and non-characteristic area to the TP of described original hardwood image、Adjacent same type area is merged，And the wave filter choosing different parameters be combined after characteristic area be smoothed respectively with non-characteristic area.The present invention has the advantage that accuracy of identification is high, recognition speed is fast.

Description

A kind of have the intelligent robot identifying function

Technical field

The present invention relates to field in intelligent robotics, be specifically related to a kind of intelligent robot having and identifying function.

Background technology

In correlation technique, intelligent robot use radar target be tracked in positioning precision and follow the tracks of success rate On possess the biggest advantage, but be only difficult to clarification of objective be made a distinction, especially in target from the range information that radar obtains It is blocked and under multi-target condition, is difficulty with target effective recognition and tracking.Use the visual information of target (such as color, profile Deng) portray target characteristic, target is identified following the tracks of the effective way being to solve the problems referred to above based on target visual feature.

Summary of the invention

For the problems referred to above, the present invention provides a kind of intelligent robot having and identifying function.

The purpose of the present invention realizes by the following technical solutions:

A kind of have the intelligent robot identifying function, knows including intelligent robot and the target being connected with intelligent robot Other device, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal are respectively It is connected with described processor；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described process Device is connected.

Preferably, also include that sheet metal and capacitance touch chip, described sheet metal are by capacitance touch chip and institute State processor to be connected.

Preferably, described sheet metal is arranged at the head of described old man's intelligent robot.

Preferably, described Target Identification Unit includes:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects target Original hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by defeated The picture signal gone out is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out from RGB color to hsv color space by it Conversion and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

$s=\left\{\begin{array}{cc}0& Max(r,g,b)=0\\ \frac{Max(r,g,b)-Min(r,g,b)}{Max(r,g,b)}& Max(r,g,b)\≠0\end{array}\right.$

V=Max (r, g, b)

Wherein, (r, g b) are the pixel RGB coordinate figure at RGB color of original hardwood image, virtual value model Enclose and be (0,1)；H is pixel form and aspect component in hsv color space, and s is saturated in hsv color space of pixel Degree component, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

$Q_{hue}=\{Q_{hue}\left(w\right),w=1,\mathrm{...}n;\underset{w=1}{\overset{n}{\Σ}}{Q}_{hue}\left(w\right)b_w=1\}$

Herein

$Q_{hue}\left(w\right)=\frac{1}{\underset{i=1}{\overset{m}{\Σ}}{e}^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\′}\mathrm{\σ}}}}\×\underset{i=1}{\overset{m}{\Σ}}\[e^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\′}\mathrm{\σ}}}\×\mathrm{\δ}\[d\left(x_i\right)-w\]\]$

$h^{\&prime;}=\left\{\begin{array}{cc}h\&times;0.9+0.05& s\&times;v\&GreaterEqual;0.4\\ v\&times;0.1& s\&times;v<0.04,v<0.2\\ v\&times;0.1-0.01& s\&times;v<0.04,0.2\&le;v<0.8\\ v\&times;0.1-0.05& s\&times;v<0.04,v\&GreaterEqual;0.8\end{array}\right.$

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the rope in space Draw, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image；

Preferably, described Target Identification Unit also includes:

(3) profile information processing module, for carrying out characteristic area and non-spy to the TP of described original hardwood image The area type levying region divides, merges adjacent same type area, and the wave filter choosing different parameters is combined After characteristic area be smoothed respectively with non-characteristic area, the selection area in described TP is judged as feature The decision condition in region is:

Herein

$f\left(t\right)=\left\{\begin{array}{cc}0& \left|{k^{\&prime;}}_N\left(t\right)\right|<T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\\ 1& \left|{k^{\&prime;}}_N\left(t\right)\right|\&GreaterEqual;T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\end{array}\right.$

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Interior initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial The distance of profile point,The real-time curvature revising described development length s for being used at initial profile point is repaiied Positive coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target set in the colouring information processed and profile information and data base Feature is compared coupling calculate matching degree, described matching degree reach to judge during default matching threshold described suspection target as Follow the tracks of target and export result of determination；

Preferably, described profile information processing module includes the first wave filter being smoothed all characteristic areas With the second wave filter that all non-characteristic areas are smoothed, a length of institute of the confidence interval of described first wave filter Have 1/2 of minimum development length in characteristic area, a length of all non-characteristic areas of the confidence interval of described second wave filter 1/2 of minimum development length in territory；Curvature according to difference is different, development length correspondingly automatic adaptive change, has Effect reduces the distortion phenomenon after merging, it is simple to be more accurately identified target.

The invention have the benefit that

1, the mode using colouring information and profile information to combine describes tracking target, and the change of illumination has to external world The strongest robustness, it is to avoid target is described by use single features, improves the precision identified；

2, revised color space conversion formula more conforms to the visual effect of the mankind, it is possible to reflect more rich letter Breath, it is simple to realize quick recognition and tracking, introduces space weight in tonal color model and divides, repeatedly filter, make model more Science, practicality is higher；

3, profile information processing module is set, for the TP of described original hardwood image is carried out characteristic area with non- The area type of characteristic area divides, merges adjacent same type area, and the wave filter choosing different parameters is involutory Characteristic area after and is smoothed respectively with non-characteristic area, and amount of calculation is the most uncomplicated, smooth effective except making an uproar, it is considered to Profile diversity between dissimilar region, at suppression noise with retain and obtain good balance between details, according to The curvature of difference is different, development length correspondingly automatic adaptive change, effectively reduces the distortion phenomenon after merging, it is simple to More accurately target is identified.

Accompanying drawing explanation

The invention will be further described to utilize accompanying drawing, but the embodiment in accompanying drawing does not constitute any limit to the present invention System, for those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to obtain according to the following drawings Other accompanying drawing.

Fig. 1 is the Target Identification Unit module connection diagram of the present invention.

Fig. 2 is the intelligent robot schematic diagram of the present invention.

Detailed description of the invention

The invention will be further described with the following Examples.

Embodiment 1

Seeing Fig. 1, Fig. 2, the present embodiment is a kind of has the intelligent robot identifying function, including intelligent robot and and intelligence The Target Identification Unit that energy robot is connected, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal are respectively It is connected with described processor；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described process Device is connected.

Preferably, also include that sheet metal and capacitance touch chip, described sheet metal are by capacitance touch chip and institute State processor to be connected.

Preferably, described sheet metal is arranged at the head of described old man's intelligent robot.

Preferably, described Target Identification Unit includes:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects target Original hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by defeated The picture signal gone out is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out from RGB color to hsv color space by it Conversion and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

$s=\left\{\begin{array}{cc}0& Max(r,g,b)=0\\ \frac{Max(r,g,b)-Min(r,g,b)}{Max(r,g,b)}& Max(r,g,b)\&NotEqual;0\end{array}\right.$

V=Max (r, g, b)

Wherein, (r, g b) are the pixel RGB coordinate figure at RGB color of original hardwood image, virtual value model Enclose and be (0,1)；H is pixel form and aspect component in hsv color space, and s is saturated in hsv color space of pixel Degree component, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

$Q_{hue}=\{Q_{hue}\left(w\right),w=1,\mathrm{...}n;\underset{w=1}{\overset{n}{\&Sigma;}}{Q}_{hue}\left(w\right)b_w=1\}$

Herein

$Q_{hue}\left(w\right)=\frac{1}{\underset{i=1}{\overset{m}{\&Sigma;}}{e}^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}}\&times;\underset{i=1}{\overset{m}{\&Sigma;}}\&lsqb;e^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}\&times;\mathrm{\&delta;}\&lsqb;d\left(x_i\right)-w\&rsqb;\&rsqb;$

$h^{\&prime;}=\left\{\begin{array}{cc}h\&times;0.9+0.05& s\&times;v\&GreaterEqual;0.4\\ v\&times;0.1& s\&times;v<0.04,v<0.2\\ v\&times;0.1-0.01& s\&times;v<0.04,0.2\&le;v<0.8\\ v\&times;0.1-0.05& s\&times;v<0.04,v\&GreaterEqual;0.8\end{array}\right.$

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the rope in space Draw, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image；

Preferably, described Target Identification Unit also includes:

(3) profile information processing module, for carrying out characteristic area and non-spy to the TP of described original hardwood image The area type levying region divides, merges adjacent same type area, and the wave filter choosing different parameters is combined After characteristic area be smoothed respectively with non-characteristic area, the selection area in described TP is judged as feature The decision condition in region is:

Herein

$f\left(t\right)=\left\{\begin{array}{cc}0& \left|{k^{\&prime;}}_N\left(t\right)\right|<T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\\ 1& \left|{k^{\&prime;}}_N\left(t\right)\right|\&GreaterEqual;T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\end{array}\right.$

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Interior initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial The distance of profile point,The real-time curvature revising described development length s for being used at initial profile point is repaiied Positive coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target set in the colouring information processed and profile information and data base Feature is compared coupling calculate matching degree, described matching degree reach to judge during default matching threshold described suspection target as Follow the tracks of target and export result of determination；

Wherein, described profile information processing module include the first wave filter that all characteristic areas are smoothed and The second wave filter being smoothed all non-characteristic areas, a length of of the confidence interval of described first wave filter owns 1/2 of minimum development length in characteristic area, a length of all non-characteristic areas of the confidence interval of described second wave filter In minimum development length 1/2；Curvature according to difference is different, development length correspondingly automatic adaptive change, effectively Reduce the distortion phenomenon after merging, it is simple to more accurately target is identified.

The mode that the present embodiment uses colouring information and profile information to combine describes tracking target, the to external world change of illumination Change and there is the strongest robustness, it is to avoid target is described by use single features, improve the precision identified；Revised face Colour space conversion formula more conforms to the visual effect of the mankind, it is possible to reflect more rich information, it is simple to realize quickly identifying with Track, introduces space weight in tonal color model and divides, repeatedly filter, make model more science, and practicality is higher；Wheel is set Wide message processing module, chooses the characteristic area after the wave filter of different parameters is combined and smooths respectively with non-characteristic area Process, it is contemplated that profile diversity between dissimilar region, obtain well between suppression noise and reservation details Balance, the curvature according to difference is different, development length correspondingly automatic adaptive change, effectively reduces the distortion after merging Phenomenon, it is simple to more accurately target is identified, wherein setsWidth be 3, the value of weights T is 0.2, identify essence Degree improves 2%, and recognition speed improves 1%.

Embodiment 2

Seeing Fig. 1, Fig. 2, the present embodiment is a kind of has the intelligent robot identifying function, including intelligent robot and and intelligence The Target Identification Unit that energy robot is connected, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal are respectively It is connected with described processor；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described process Device is connected.

Preferably, also include that sheet metal and capacitance touch chip, described sheet metal are by capacitance touch chip and institute State processor to be connected.

Preferably, described sheet metal is arranged at the head of described old man's intelligent robot.

Preferably, described Target Identification Unit includes:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects target Original hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by defeated The picture signal gone out is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out from RGB color to hsv color space by it Conversion and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

$s=\left\{\begin{array}{cc}0& Max(r,g,b)=0\\ \frac{Max(r,g,b)-Min(r,g,b)}{Max(r,g,b)}& Max(r,g,b)\&NotEqual;0\end{array}\right.$

V=Max (r, g, b)

Wherein, (r, g b) are the pixel RGB coordinate figure at RGB color of original hardwood image, virtual value model Enclose and be (0,1)；H is pixel form and aspect component in hsv color space, and s is saturated in hsv color space of pixel Degree component, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

$Q_{hue}=\{Q_{hue}\left(w\right),w=1,\mathrm{...}n;\underset{w=1}{\overset{n}{\&Sigma;}}{Q}_{hue}\left(w\right)b_w=1\}$

Herein

$Q_{hue}\left(w\right)=\frac{1}{\underset{i=1}{\overset{m}{\&Sigma;}}{e}^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}}\&times;\underset{i=1}{\overset{m}{\&Sigma;}}\&lsqb;e^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}\&times;\mathrm{\&delta;}\&lsqb;d\left(x_i\right)-w\&rsqb;\&rsqb;$

$h^{\&prime;}=\left\{\begin{array}{cc}h\&times;0.9+0.05& s\&times;v\&GreaterEqual;0.4\\ v\&times;0.1& s\&times;v<0.04,v<0.2\\ v\&times;0.1-0.01& s\&times;v<0.04,0.2\&le;v<0.8\\ v\&times;0.1-0.05& s\&times;v<0.04,v\&GreaterEqual;0.8\end{array}\right.$

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the rope in space Draw, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image；

Preferably, described Target Identification Unit also includes:

(3) profile information processing module, for carrying out characteristic area and non-spy to the TP of described original hardwood image The area type levying region divides, merges adjacent same type area, and the wave filter choosing different parameters is combined After characteristic area be smoothed respectively with non-characteristic area, the selection area in described TP is judged as feature The decision condition in region is:

Herein

$f\left(t\right)=\left\{\begin{array}{cc}0& \left|{k^{\&prime;}}_N\left(t\right)\right|<T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\\ 1& \left|{k^{\&prime;}}_N\left(t\right)\right|\&GreaterEqual;T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\end{array}\right.$

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Interior initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial The distance of profile point,The real-time curvature revising described development length s for being used at initial profile point is repaiied Positive coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target set in the colouring information processed and profile information and data base Feature is compared coupling calculate matching degree, described matching degree reach to judge during default matching threshold described suspection target as Follow the tracks of target and export result of determination；

Wherein, described profile information processing module include the first wave filter that all characteristic areas are smoothed and The second wave filter being smoothed all non-characteristic areas, a length of of the confidence interval of described first wave filter owns 1/2 of minimum development length in characteristic area, a length of all non-characteristic areas of the confidence interval of described second wave filter In minimum development length 1/2；Curvature according to difference is different, development length correspondingly automatic adaptive change, effectively Reduce the distortion phenomenon after merging, it is simple to more accurately target is identified.

The mode that the present embodiment uses colouring information and profile information to combine describes tracking target, the to external world change of illumination Change and there is the strongest robustness, it is to avoid target is described by use single features, improve the precision identified；Revised face Colour space conversion formula more conforms to the visual effect of the mankind, it is possible to reflect more rich information, it is simple to realize quickly identifying with Track, introduces space weight in tonal color model and divides, repeatedly filter, make model more science, and practicality is higher；Wheel is set Wide message processing module, chooses the characteristic area after the wave filter of different parameters is combined and smooths respectively with non-characteristic area Process, it is contemplated that profile diversity between dissimilar region, obtain well between suppression noise and reservation details Balance, the curvature according to difference is different, development length correspondingly automatic adaptive change, effectively reduces the distortion after merging Phenomenon, it is simple to more accurately target is identified, wherein setsWidth be 4, the value of weights T is 0.3, identify essence Degree improves 1%, and recognition speed improves 2%.

Embodiment 3

Seeing Fig. 1, Fig. 2, the present embodiment is a kind of has the intelligent robot identifying function, including intelligent robot and and intelligence The Target Identification Unit that energy robot is connected, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal are respectively It is connected with described processor；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described process Device is connected.

Preferably, also include that sheet metal and capacitance touch chip, described sheet metal are by capacitance touch chip and institute State processor to be connected.

Preferably, described sheet metal is arranged at the head of described old man's intelligent robot.

Preferably, described Target Identification Unit includes:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects target Original hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by defeated The picture signal gone out is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out from RGB color to hsv color space by it Conversion and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

$s=\left\{\begin{array}{cc}0& Max(r,g,b)=0\\ \frac{Max(r,g,b)-Min(r,g,b)}{Max(r,g,b)}& Max(r,g,b)\&NotEqual;0\end{array}\right.$

V=Max (r, g, b)

Wherein, (r, g b) are the pixel RGB coordinate figure at RGB color of original hardwood image, virtual value model Enclose and be (0,1)；H is pixel form and aspect component in hsv color space, and s is saturated in hsv color space of pixel Degree component, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

$Q_{hue}=\{Q_{hue}\left(w\right),w=1,\mathrm{...}n;\underset{w=1}{\overset{n}{\&Sigma;}}{Q}_{hue}\left(w\right)b_w=1\}$

Herein

$Q_{hue}\left(w\right)=\frac{1}{\underset{i=1}{\overset{m}{\&Sigma;}}{e}^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}}\&times;\underset{i=1}{\overset{m}{\&Sigma;}}\&lsqb;e^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}\&times;\mathrm{\&delta;}\&lsqb;d\left(x_i\right)-w\&rsqb;\&rsqb;$

$h^{\&prime;}=\left\{\begin{array}{cc}h\&times;0.9+0.05& s\&times;v\&GreaterEqual;0.4\\ v\&times;0.1& s\&times;v<0.04,v<0.2\\ v\&times;0.1-0.01& s\&times;v<0.04,0.2\&le;v<0.8\\ v\&times;0.1-0.05& s\&times;v<0.04,v\&GreaterEqual;0.8\end{array}\right.$

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the rope in space Draw, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image；

Preferably, described Target Identification Unit also includes:

(3) profile information processing module, for carrying out characteristic area and non-spy to the TP of described original hardwood image The area type levying region divides, merges adjacent same type area, and the wave filter choosing different parameters is combined After characteristic area be smoothed respectively with non-characteristic area, the selection area in described TP is judged as feature The decision condition in region is:

Herein

$f\left(t\right)=\left\{\begin{array}{cc}0& \left|{k^{\&prime;}}_N\left(t\right)\right|<T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\\ 1& \left|{k^{\&prime;}}_N\left(t\right)\right|\&GreaterEqual;T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\end{array}\right.$

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Interior initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial The distance of profile point,The real-time curvature revising described development length s for being used at initial profile point is repaiied Positive coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target set in the colouring information processed and profile information and data base Feature is compared coupling calculate matching degree, described matching degree reach to judge during default matching threshold described suspection target as Follow the tracks of target and export result of determination；

Wherein, described profile information processing module include the first wave filter that all characteristic areas are smoothed and The second wave filter being smoothed all non-characteristic areas, a length of of the confidence interval of described first wave filter owns 1/2 of minimum development length in characteristic area, a length of all non-characteristic areas of the confidence interval of described second wave filter In minimum development length 1/2；Curvature according to difference is different, development length correspondingly automatic adaptive change, effectively Reduce the distortion phenomenon after merging, it is simple to more accurately target is identified.

The mode that the present embodiment uses colouring information and profile information to combine describes tracking target, the to external world change of illumination Change and there is the strongest robustness, it is to avoid target is described by use single features, improve the precision identified；Revised face Colour space conversion formula more conforms to the visual effect of the mankind, it is possible to reflect more rich information, it is simple to realize quickly identifying with Track, introduces space weight in tonal color model and divides, repeatedly filter, make model more science, and practicality is higher；Wheel is set Wide message processing module, chooses the characteristic area after the wave filter of different parameters is combined and smooths respectively with non-characteristic area Process, it is contemplated that profile diversity between dissimilar region, obtain well between suppression noise and reservation details Balance, the curvature according to difference is different, development length correspondingly automatic adaptive change, effectively reduces the distortion after merging Phenomenon, it is simple to more accurately target is identified, wherein setsWidth be 5, the value of weights T is 0.4, identify essence Degree improves 2%, and recognition speed improves 3%.

Embodiment 4

Seeing Fig. 1, Fig. 2, the present embodiment is a kind of has the intelligent robot identifying function, including intelligent robot and and intelligence The Target Identification Unit that energy robot is connected, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal are respectively It is connected with described processor；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described process Device is connected.

Preferably, also include that sheet metal and capacitance touch chip, described sheet metal are by capacitance touch chip and institute State processor to be connected.

Preferably, described sheet metal is arranged at the head of described old man's intelligent robot.

Preferably, described Target Identification Unit includes:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects target Original hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by defeated The picture signal gone out is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out from RGB color to hsv color space by it Conversion and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

$s=\left\{\begin{array}{cc}0& Max(r,g,b)=0\\ \frac{Max(r,g,b)-Min(r,g,b)}{Max(r,g,b)}& Max(r,g,b)\&NotEqual;0\end{array}\right.$

V=Max (r, g, b)

Wherein, (r, g b) are the pixel RGB coordinate figure at RGB color of original hardwood image, virtual value model Enclose and be (0,1)；H is pixel form and aspect component in hsv color space, and s is saturated in hsv color space of pixel Degree component, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

$Q_{hue}=\{Q_{hue}\left(w\right),w=1,\mathrm{...}n;\underset{w=1}{\overset{n}{\&Sigma;}}{Q}_{hue}\left(w\right)b_w=1\}$

Herein

$Q_{hue}\left(w\right)=\frac{1}{\underset{i=1}{\overset{m}{\&Sigma;}}{e}^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}}\&times;\underset{i=1}{\overset{m}{\&Sigma;}}\&lsqb;e^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}\&times;\mathrm{\&delta;}\&lsqb;d\left(x_i\right)-w\&rsqb;\&rsqb;$

$h^{\&prime;}=\left\{\begin{array}{cc}h\&times;0.9+0.05& s\&times;v\&GreaterEqual;0.4\\ v\&times;0.1& s\&times;v<0.04,v<0.2\\ v\&times;0.1-0.01& s\&times;v<0.04,0.2\&le;v<0.8\\ v\&times;0.1-0.05& s\&times;v<0.04,v\&GreaterEqual;0.8\end{array}\right.$

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the rope in space Draw, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image；

Preferably, described Target Identification Unit also includes:

(3) profile information processing module, for carrying out characteristic area and non-spy to the TP of described original hardwood image The area type levying region divides, merges adjacent same type area, and the wave filter choosing different parameters is combined After characteristic area be smoothed respectively with non-characteristic area, the selection area in described TP is judged as feature The decision condition in region is:

Herein

$f\left(t\right)=\left\{\begin{array}{cc}0& \left|{k^{\&prime;}}_N\left(t\right)\right|<T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\\ 1& \left|{k^{\&prime;}}_N\left(t\right)\right|\&GreaterEqual;T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\end{array}\right.$

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Interior initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial The distance of profile point,The real-time curvature revising described development length s for being used at initial profile point is repaiied Positive coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target set in the colouring information processed and profile information and data base Feature is compared coupling calculate matching degree, described matching degree reach to judge during default matching threshold described suspection target as Follow the tracks of target and export result of determination；

Wherein, described profile information processing module include the first wave filter that all characteristic areas are smoothed and The second wave filter being smoothed all non-characteristic areas, a length of of the confidence interval of described first wave filter owns 1/2 of minimum development length in characteristic area, a length of all non-characteristic areas of the confidence interval of described second wave filter In minimum development length 1/2；Curvature according to difference is different, development length correspondingly automatic adaptive change, effectively Reduce the distortion phenomenon after merging, it is simple to more accurately target is identified.

The mode that the present embodiment uses colouring information and profile information to combine describes tracking target, the to external world change of illumination Change and there is the strongest robustness, it is to avoid target is described by use single features, improve the precision identified；Revised face Colour space conversion formula more conforms to the visual effect of the mankind, it is possible to reflect more rich information, it is simple to realize quickly identifying with Track, introduces space weight in tonal color model and divides, repeatedly filter, make model more science, and practicality is higher；Wheel is set Wide message processing module, chooses the characteristic area after the wave filter of different parameters is combined and smooths respectively with non-characteristic area Process, it is contemplated that profile diversity between dissimilar region, obtain well between suppression noise and reservation details Balance, the curvature according to difference is different, development length correspondingly automatic adaptive change, effectively reduces the distortion after merging Phenomenon, it is simple to more accurately target is identified, wherein setsWidth be 5, the value of weights T is 0.5, identify essence Degree improves 2%, and recognition speed improves 2.5%.

Embodiment 5

Seeing Fig. 1, Fig. 2, the present embodiment is a kind of has the intelligent robot identifying function, including intelligent robot and and intelligence The Target Identification Unit that energy robot is connected, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal are respectively It is connected with described processor；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described process Device is connected.

Preferably, also include that sheet metal and capacitance touch chip, described sheet metal are by capacitance touch chip and institute State processor to be connected.

Preferably, described sheet metal is arranged at the head of described old man's intelligent robot.

Preferably, described Target Identification Unit includes:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects target Original hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by defeated The picture signal gone out is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out from RGB color to hsv color space by it Conversion and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

$s=\left\{\begin{array}{cc}0& Max(r,g,b)=0\\ \frac{Max(r,g,b)-Min(r,g,b)}{Max(r,g,b)}& Max(r,g,b)\&NotEqual;0\end{array}\right.$

V=Max (r, g, b)

Wherein, (r, g b) are the pixel RGB coordinate figure at RGB color of original hardwood image, virtual value model Enclose and be (0,1)；H is pixel form and aspect component in hsv color space, and s is saturated in hsv color space of pixel Degree component, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

$Q_{hue}=\{Q_{hue}\left(w\right),w=1,\mathrm{...}n;\underset{w=1}{\overset{n}{\&Sigma;}}{Q}_{hue}\left(w\right)b_w=1\}$

Herein

$Q_{hue}\left(w\right)=\frac{1}{\underset{i=1}{\overset{m}{\&Sigma;}}{e}^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}}\&times;\underset{i=1}{\overset{m}{\&Sigma;}}\&lsqb;e^{-\frac{\left|\right|x_i-x_c|{|}^2}{\sqrt{2}{h}^{\&prime;}\mathrm{\&sigma;}}}\&times;\mathrm{\&delta;}\&lsqb;d\left(x_i\right)-w\&rsqb;\&rsqb;$

$h^{\&prime;}=\left\{\begin{array}{cc}h\&times;0.9+0.05& s\&times;v\&GreaterEqual;0.4\\ v\&times;0.1& s\&times;v<0.04,v<0.2\\ v\&times;0.1-0.01& s\&times;v<0.04,0.2\&le;v<0.8\\ v\&times;0.1-0.05& s\&times;v<0.04,v\&GreaterEqual;0.8\end{array}\right.$

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the rope in space Draw, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image；

Preferably, described Target Identification Unit also includes:

(3) profile information processing module, for carrying out characteristic area and non-spy to the TP of described original hardwood image The area type levying region divides, merges adjacent same type area, and the wave filter choosing different parameters is combined After characteristic area be smoothed respectively with non-characteristic area, the selection area in described TP is judged as feature The decision condition in region is:

Herein

$f\left(t\right)=\left\{\begin{array}{cc}0& \left|{k^{\&prime;}}_N\left(t\right)\right|<T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\\ 1& \left|{k^{\&prime;}}_N\left(t\right)\right|\&GreaterEqual;T\&times;max\left|{k^{\&prime;}}_N\left(t\right)\right|\end{array}\right.$

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Interior initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial The distance of profile point,The real-time curvature revising described development length s for being used at initial profile point is repaiied Positive coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target set in the colouring information processed and profile information and data base Feature is compared coupling calculate matching degree, described matching degree reach to judge during default matching threshold described suspection target as Follow the tracks of target and export result of determination；

Wherein, described profile information processing module include the first wave filter that all characteristic areas are smoothed and The second wave filter being smoothed all non-characteristic areas, a length of of the confidence interval of described first wave filter owns 1/2 of minimum development length in characteristic area, a length of all non-characteristic areas of the confidence interval of described second wave filter In minimum development length 1/2；Curvature according to difference is different, development length correspondingly automatic adaptive change, effectively Reduce the distortion phenomenon after merging, it is simple to more accurately target is identified.

The mode that the present embodiment uses colouring information and profile information to combine describes tracking target, the to external world change of illumination Change and there is the strongest robustness, it is to avoid target is described by use single features, improve the precision identified；Revised face Colour space conversion formula more conforms to the visual effect of the mankind, it is possible to reflect more rich information, it is simple to realize quickly identifying with Track, introduces space weight in tonal color model and divides, repeatedly filter, make model more science, and practicality is higher；Wheel is set Wide message processing module, chooses the characteristic area after the wave filter of different parameters is combined and smooths respectively with non-characteristic area Process, it is contemplated that profile diversity between dissimilar region, obtain well between suppression noise and reservation details Balance, the curvature according to difference is different, development length correspondingly automatic adaptive change, effectively reduces the distortion after merging Phenomenon, it is simple to more accurately target is identified, wherein setsWidth be 4, the value of weights T is 0.3, identify essence Degree improves 2.5%, and recognition speed improves 3.5%.

Last it should be noted that, above example is only in order to illustrate technical scheme, rather than the present invention is protected Protecting the restriction of scope, although having made to explain to the present invention with reference to preferred embodiment, those of ordinary skill in the art should Work as understanding, technical scheme can be modified or equivalent, without deviating from the reality of technical solution of the present invention Matter and scope.

Claims (6)

1. there is the intelligent robot identifying function, including intelligent robot and the target recognition that is connected with intelligent robot Device, is characterized in that, described intelligent robot includes:

Processor, infrared induction module and intelligent communication terminal, described infrared induction module and intelligent communication terminal respectively with institute State processor to be connected；Described intelligent communication terminal includes voice cue module, institute's voice cue module and described processor phase Connect.

The most according to claim 1 a kind of have identify function intelligent robot, it is characterized in that, also include sheet metal and Capacitance touch chip, described sheet metal are connected with described processor by capacitance touch chip.

A kind of intelligent robot with identification function the most according to claim 2, is characterized in that, described sheet metal is arranged Head in described old man's intelligent robot.

A kind of intelligent robot with identification function the most according to claim 3, is characterized in that, described target recognition fills Put and include:

(1) Target Acquisition module is suspected, for identifying that in monitor video suspecting that target reading comprise suspects the original of target Hardwood image, it includes the infrared CCD being connected with field computer, and described infrared CCD is by output Picture signal is input to field computer and is made whether to there is the differentiation process suspecting target；

(2) colouring information processing module, described original hardwood image is carried out turning to hsv color space from RGB color by it Changing and build the described suspection target tonal color model in hsv color space, conversion formula is as follows:

V=Max (r, g, b)

Wherein, (r, g, b) be the pixel RGB coordinate figure at RGB color of original hardwood image, and valid value range is equal For (0,1)；H is pixel form and aspect component in hsv color space, and s is that pixel saturation in hsv color space is divided Amount, v is pixel chrominance component in hsv color space；

Tonal color model is as follows:

Herein

Wherein, function δ [d (x_i)-w] it is pixel x_iProjection in w sub spaces region, w is characterized the index in space, b_wFor the weight of each subspace,It is with pixel x_cCentered by kernel function in two dimensional image.

A kind of intelligent robot with identification function the most according to claim 4, is characterized in that, described target recognition fills Put and also include:

(3) profile information processing module, for carrying out characteristic area and non-characteristic area to the TP of described original hardwood image After the area type in territory divides, merges adjacent same type area, and the wave filter choosing different parameters is combined Characteristic area is smoothed respectively with non-characteristic area, and the selection area in described TP is judged as characteristic area Decision condition be:

Herein

Wherein, t represents the profile point of the TP of described original hardwood image, t₀It is positioned at described selection area for default Initial profile point, s is default development length, and the value of development length is that the edge contour point of selection area is to described initial profile The distance of point,For the real-time curvature correction for revising described development length s at initial profile point Coefficient,For the radius of curvature of initial profile point,For being obtained by the window function that width range is [3,5] preset The mean radius of curvature of profile starting point；F (t) is for judging whether profile point is characterized characteristic function a little, and f (t)=1 represents This profile point is characterized a little, and f (t)=0 represents that this profile point is non-characteristic point, N_{F (t)=1}The spy being had in representing selection area Levy number a little, N_yFor the number needing the characteristic point included as characteristic area set, k'_NT () is by described window function pair The TP curvature that TP carries out neighborhood averaging and obtains, max | k'_N(t) | represent the absolute value of TP curvature Maximum, T be the span of weights and T be [0.2,0.5]；

(4) feature evaluation module, for the target characteristic set in the colouring information processed and profile information and data base Coupling of comparing also calculates matching degree, and described matching degree reaches to judge during default matching threshold that described suspection target is as following the tracks of Target also exports result of determination.

A kind of intelligent robot with identification function the most according to claim 5, is characterized in that, at described profile information Reason module includes the first wave filter being smoothed all characteristic areas and all non-characteristic areas carries out smooth place Second wave filter of reason, the minimum development length in a length of all characteristic areas of the confidence interval of described first wave filter 1/2,1/2 of the minimum development length in a length of all non-characteristic areas of the confidence interval of described second wave filter.