CN105603912A - High-efficiency intelligent zebra stripe guardrail - Google Patents
High-efficiency intelligent zebra stripe guardrail Download PDFInfo
- Publication number
- CN105603912A CN105603912A CN201610014794.6A CN201610014794A CN105603912A CN 105603912 A CN105603912 A CN 105603912A CN 201610014794 A CN201610014794 A CN 201610014794A CN 105603912 A CN105603912 A CN 105603912A
- Authority
- CN
- China
- Prior art keywords
- contour
- points
- noise
- curvature
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 241000283070 Equus zebra Species 0.000 title claims abstract description 54
- 238000001914 filtration Methods 0.000 claims abstract description 43
- 238000000034 method Methods 0.000 claims abstract description 17
- 230000011218 segmentation Effects 0.000 claims abstract description 17
- 238000012937 correction Methods 0.000 claims description 35
- 230000000694 effects Effects 0.000 claims description 27
- 238000009499 grossing Methods 0.000 claims description 26
- 239000000654 additive Substances 0.000 claims description 21
- 230000000996 additive effect Effects 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000012935 Averaging Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 2
- 230000014759 maintenance of location Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Classifications
-
- E—FIXED CONSTRUCTIONS
- E01—CONSTRUCTION OF ROADS, RAILWAYS, OR BRIDGES
- E01F—ADDITIONAL WORK, SUCH AS EQUIPPING ROADS OR THE CONSTRUCTION OF PLATFORMS, HELICOPTER LANDING STAGES, SIGNS, SNOW FENCES, OR THE LIKE
- E01F13/00—Arrangements for obstructing or restricting traffic, e.g. gates, barricades ; Preventing passage of vehicles of selected category or dimensions
- E01F13/04—Arrangements for obstructing or restricting traffic, e.g. gates, barricades ; Preventing passage of vehicles of selected category or dimensions movable to allow or prevent passage
Landscapes
- Engineering & Computer Science (AREA)
- Architecture (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a high-efficiency intelligent zebra stripe guardrail. The guardrail comprises a common intelligent zebra stripe guardrail body and a target identifying device mounted on the intelligent zebra stripe guardrail body, and the identifying device comprises a modeling module, a segmentation module, a merging module and a filtration module. According to the guardrail, the target identifying device is additionally mounted on the intelligent zebra stripe guardrail body, identification capability of the intelligent zebra stripe guardrail can be effectively enhanced, targets are identified through the target identifying device, and target contour noise can be effectively filtered out in the identification process, so that the intelligent zebra stripe guardrail makes judgment for pedestrians and vehicles.
Description
Technical Field
The invention relates to the field of intelligent guardrails, in particular to a high-efficiency zebra crossing intelligent guardrail.
Background
The zebra crossing ensures that the safety of pedestrians is guaranteed on a road, however, more and more 'Chinese type road crossing' appears in recent years, and the intelligent zebra crossing guardrail is a device capable of guaranteeing that the pedestrians safely cross the zebra crossing and aims to standardize the passage of citizens on the road and restrain the 'Chinese type road crossing'. In recent years, the intelligent zebra crossing guardrail is widely applied, and particularly has a great effect at intersections with heavy traffic. People also have higher and higher requirements on the 'intelligence' of the intelligent guardrail, and the intelligent guardrail with the high efficiency is the zebra crossing intelligent guardrail.
The target contour recognition is an important means of target recognition, and because the target contour is influenced by factors such as noise, quantization error and the like in practical application, the target contour inevitably generates distortion, and filtering and smoothing processing of the target contour is necessary to accurately describe contour characteristics. At present, scholars propose a plurality of filtering smoothing algorithms of noisy contours, but the problems of huge calculation amount, non-ideal noise reduction effect, target distortion caused by excessive filtering and the like are generally existed.
Disclosure of Invention
Aiming at the problems, the invention provides a high-efficiency intelligent zebra crossing guardrail.
The purpose of the invention is realized by adopting the following technical scheme:
a high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile;
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: target profile G (t) and noisy profile GN(t) the curvatures correspond to k (t) and k, respectivelyN(t); selecting a window function W (n) with width D for curvature kN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T1Comparing, and determining the curvature k 'of the noisy contour according to the comparison result'N(t), namely:
when | k1N(t)-k2N(t)|＞T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TKSelf-adaptive decision T by judging the amount of the target contour featuresKWhen k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1.
A merging module for eliminating the false characteristic points generated by noise interference and merging the characteristic points and non-characteristic points which can not form continuous areas to obtain effective characteristic areas and non-characteristic areas, wherein a starting point O is selected, the outline starting points extend to two sides to merge adjacent points, the starting point type is used as the preset type of the area, and each S × mu extends to two sides0Stopping, wherein S is a preset minimum length,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing the mean radius of curvature of the O points obtained from the window function, the real-time curvature correction factor mu0The method is used for automatically correcting the extension length according to different curvatures of different points, and can effectively reduce the distortion phenomenon after combination; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; adjacent areas of the same type are combined to obtain a continuous characteristic area and a non-characteristic area;
the filtering module adopts wiener filtering to carry out primary filtering because the multiplicative noise is related to the image signal and changes along with the change of the image signal, at the moment, the image information also contains residual multiplicative noise, and the F filter F (x, y) is q × exp- (x, y)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, containThe parametric equation for the arc length of the noisy object profile is denoted as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' represents coordinates of each point on the noisy contour after removing the multiplicative noise, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsSmoothing the noise-containing contour, named as K filter, and classifying contour points and dividing regions to obtain noise-containing contour GN(t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationIn the non-characteristic region, in order to raise the effect of suppressing noise Where σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
According to the invention, the target identification device is additionally arranged on the intelligent zebra crossing guardrail, so that the identification capability of the intelligent zebra crossing guardrail can be effectively enhanced, the intelligent zebra crossing guardrail identifies the target through the target profile, and the target profile noise can be effectively filtered in the identification process, so that the pedestrians and vehicles can be judged.
Drawings
The invention is further illustrated by means of the attached drawings, but the embodiments in the drawings do not constitute any limitation to the invention, and for a person skilled in the art, other drawings can be obtained on the basis of the following drawings without inventive effort.
Fig. 1 is a block diagram of the high-efficiency intelligent guardrail for the zebra crossing of the invention.
Detailed Description
The invention is further described with reference to the following examples.
FIG. 1 is a block diagram of the present invention, which includes: the device comprises a modeling module, a segmentation module, a combination module and a filtering module.
Example 1: a high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile;
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: target profile G (t) and noisy profile GN(t) the curvatures correspond to k (t) and k, respectivelyN(t); due to the influence of noise, the noisy contour GN(t) curvature value k of upper part feature pointN(t) cannot accurately represent the contour information, and in order to obtain an accurate curvature, a window function W (n) having a width D ∈ {7, 9} is selected, and a curvature k is calculatedN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T1Comparing the two values to 0.24, and determining the curvature k 'of the noisy contour according to the comparison result'N(t), namely:
when | k1N (T) -k2N (T) | > T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TKSelf-adaptive decision T by judging the amount of the target contour featuresKWhen k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1;
the distribution of the characteristic points and the non-characteristic points obtained after classification is not continuous, and a filter cannot be selected to carry out effective contour smoothing on the characteristic points and the non-characteristic points. In order to obtain a good contour smoothing effect, it is necessary to perform merging processing on contour points of the same type.
A merging module for eliminating the false characteristic points generated by noise interference and merging the characteristic points and non-characteristic points which can not form continuous areas to obtain effective characteristic areas and non-characteristic areas, wherein a starting point O is selected, the outline starting points extend to two sides to merge adjacent points, the starting point type is used as the preset type of the area, and each S × mu extends to two sides0Where S is a preset minimum length, in this embodiment, S is 17,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing O points obtained by the above-mentioned window functionMean radius of curvature, real-time curvature correction factor mu0The method is used for automatically correcting the extension length according to different curvatures of different points, wherein the part with large curvature needs smaller length, and the part with small curvature needs larger length, so that the distortion phenomenon after combination can be effectively reduced; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; and combining adjacent regions of the same type to obtain continuous characteristic regions and non-characteristic regions.
The filtering module adopts wiener filtering to carry out primary filtering because the multiplicative noise is related to the image signal and changes along with the change of the image signal, at the moment, the image information also contains residual multiplicative noise, and the F filter F (x, y) is q × exp- (x, y)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, the arc length parameterized equation of the noisy target contour is expressed as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' separately representing points on the noisy contour after removal of multiplicative noiseMark, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsSmoothing the noise-containing contour, named as K filter, and classifying contour points and dividing regions to obtain noise-containing contour GN(t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationIn the non-characteristic region, the effect of suppressing noise is focused onWhere σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
In this embodiment, S is 17, threshold T10.24, window function width D ∈ {7, 9}, there is better smooth effect to the noisy image of noise intensity I ∈ {10dB, 20dB }, and target contour recognition target is passed through to zebra crossing intelligent guardrail, and the recognition process can effectively filter target contour noise, can make accurate discernment to pedestrian and vehicle, improves crossing road efficiency and security.
Example 2: a high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile;
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: target profile G (t) and noisy profile GNCurvature corresponding to (t)Are k (t) and k respectivelyN(t); due to the influence of noise, the noisy contour GN(t) curvature value k of upper part feature pointN(t) cannot accurately represent the contour information, and in order to obtain an accurate curvature, a window function W (n) having a width D ∈ {10, 12} is selected, and a curvature k is calculatedN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T1Comparing the two values to 0.24, and determining the curvature k 'of the noisy contour according to the comparison result'N(t), namely:
when | k1N(t)-k2N(t)|T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TNSelf-adaptive decision T by judging the amount of the target contour featuresKWhen k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1;
the distribution of the characteristic points and the non-characteristic points obtained after classification is not continuous, and a filter cannot be selected to carry out effective contour smoothing on the characteristic points and the non-characteristic points. In order to obtain a good contour smoothing effect, it is necessary to perform merging processing on contour points of the same type.
A merging module: the method is used for eliminating the pseudo feature points generated by noise interference and carrying out merging operation on the feature points and the non-feature points which cannot form a continuous area, so that an effective feature area and a non-feature area are obtained: selecting a starting point O, extending the outline starting point to two sides, merging adjacent points, taking the starting point type as the preset type of the area, and moving the area to two sidesLaterally extending each S × mu0A stop, where S is a preset minimum length, in this embodiment S19,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing the mean radius of curvature of the O points obtained from the window function, the real-time curvature correction factor mu0The method is used for automatically correcting the extension length according to different curvatures of different points, wherein the part with large curvature needs smaller length, and the part with small curvature needs larger length, so that the distortion phenomenon after combination can be effectively reduced; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; and combining adjacent regions of the same type to obtain continuous characteristic regions and non-characteristic regions.
The filtering module adopts wiener filtering to carry out primary filtering because the multiplicative noise is related to the image signal and changes along with the change of the image signal, at the moment, the image information also contains residual multiplicative noise, and the F filter F (x, y) is q × exp- (x, y)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, the arc length parameterized equation of the noisy target contour is expressed as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' represents coordinates of each point on the noisy contour after removing the multiplicative noise, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsSmoothing the noise-containing contour, named as K filter, and classifying contour points and dividing regions to obtain noise-containing contour GN(t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationAt a position other thanCharacteristic region, focusing on the effect of suppressing noise, andwhere σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
In this embodiment, S is 19, threshold T10.24, window function width D ∈ {10, 12}, there is better smooth effect to the noisy image of noise intensity I ∈ {20dB, 30dB }, and zebra crossing intelligent guardrail passes through target profile recognition target, can effectively filter target profile noise in the identification process, can make accurate discernment to pedestrian and vehicle, improves crossing road efficiency and security.
Example 3: a high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile;
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: target profile G (t) and noisy profile GN(t) the curvatures correspond to k (t) and k, respectivelyN(t); due to the influence of noise, the noisy contour GN(t) curvature value k of upper part feature pointN(t) cannot accurately represent the contour information, and in order to obtain an accurate curvature, a window function W (n) having a width D ∈ {13, 14} is selected, and a curvature k is calculatedN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T1Comparing the two parameters to 0.26, and determining the curvature k 'of the profile containing noise according to the comparison result'N(t), namely:
when | k1N(t)-k2N(t)|＞T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TKSelf-adaptive decision T by judging the amount of the target contour featuresKWhen k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1;
the distribution of the characteristic points and the non-characteristic points obtained after classification is not continuous, and a filter cannot be selected to carry out effective contour smoothing on the characteristic points and the non-characteristic points. In order to obtain a good contour smoothing effect, it is necessary to perform merging processing on contour points of the same type.
A merging module for eliminating the false characteristic points generated by noise interference and merging the characteristic points and non-characteristic points which can not form continuous areas to obtain effective characteristic areas and non-characteristic areas, wherein a starting point O is selected, the outline starting points extend to two sides to merge adjacent points, the starting point type is used as the preset type of the area, and each S × mu extends to two sides0A time stop, where S is a preset minimum length, in this embodiment S21,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing the mean radius of curvature of the O points obtained from the window function, the real-time curvature correction factor mu0The method is used for automatically correcting the extension length according to different curvatures of different points, wherein the part with large curvature needs smaller length, and the part with small curvature needs larger length, so that the distortion phenomenon after combination can be effectively reduced; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; and combining adjacent regions of the same type to obtain continuous characteristic regions and non-characteristic regions.
The filtering module adopts wiener filtering to carry out primary filtering because the multiplicative noise is related to the image signal and changes along with the change of the image signal, at the moment, the image information also contains residual multiplicative noise, and the F filter F (x, y) is q × exp- (x, y)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, the arc length parameterized equation of the noisy target contour is expressed as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' represents coordinates of each point on the noisy contour after removing the multiplicative noise, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsThe noise-containing contour is smoothed, named as a K filter, and is subjected to contour point classification and region division, and the noise-containing contour GN (t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationIn the non-characteristic region, the effect of suppressing noise is focused onWhere σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
In this embodiment, S is 21, threshold T10.26, the window function width D ∈ {13, 14}, has better smoothing effect on noisy images of the noise intensity I ∈ {30dB, 40dB }, the calculated amount and the detail information retention condition are both in an acceptable interval and obtain better balance, the intelligent guardrail of the zebra crossing identifies the target through the target contour, the target contour noise can be effectively filtered in the identification process, and pedestrians and vehicles can be drivenAnd the road passing efficiency and the safety are improved by accurately identifying the road.
Example 4: a high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile;
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: target profile G (t) and noisy profile GN(t) the curvatures correspond to k (t) and k, respectivelyN(t); due to the influence of noise, the noisy contour GN(t) curvature value k of upper part feature pointN(t) cannot accurately represent the contour information, and in order to obtain an accurate curvature, a window function W (n) having a width D ∈ {15, 17} is selected, and a curvature k is calculatedN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T1Comparing the two values to 0.28, and determining the curvature k 'of the noisy contour according to the comparison result'N(t), namely:
when | k1N(t)-k2N(t)|＞T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TKSelf-adaptive decision T by judging the amount of the target contour featuresKWhen k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1;
the distribution of the characteristic points and the non-characteristic points obtained after classification is not continuous, and a filter cannot be selected to carry out effective contour smoothing on the characteristic points and the non-characteristic points. In order to obtain a good contour smoothing effect, it is necessary to perform merging processing on contour points of the same type.
A merging module for eliminating the false characteristic points generated by noise interference and merging the characteristic points and non-characteristic points which can not form continuous areas to obtain effective characteristic areas and non-characteristic areas, wherein a starting point O is selected, the outline starting points extend to two sides to merge adjacent points, the starting point type is used as the preset type of the area, and each S × mu extends to two sides0A stop, where S is a preset minimum length, in this embodiment S23,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing the mean radius of curvature, real-time curvature, of the O points obtained by the above-mentioned window functionCoefficient of rate correction mu0The method is used for automatically correcting the extension length according to different curvatures of different points, wherein the part with large curvature needs smaller length, and the part with small curvature needs larger length, so that the distortion phenomenon after combination can be effectively reduced; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; and combining adjacent regions of the same type to obtain continuous characteristic regions and non-characteristic regions.
The filtering module adopts wiener filtering to carry out primary filtering because the multiplicative noise is related to the image signal and changes along with the change of the image signal, at the moment, the image information also contains residual multiplicative noise, and the F filter F (x, y) is q × exp- (x, y)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, the arc length parameterized equation of the noisy target contour is expressed as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' represents coordinates of each point on the noisy contour after removing the multiplicative noise, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsSmoothing the noise-containing contour, named as K filter, and classifying contour points and dividing regions to obtain noise-containing contour GN(t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationIn the non-characteristic region, the effect of suppressing noise is focused onWhere σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
In this embodiment, S-23, threshold T10.28, window function width D ∈ {15, 17}, although the noise-containing image of noise intensity I ∈ {40dB, 50dB } has increased some calculated amount, but the image of this interval has excellent smooth effect, and detailed information keeps the condition better, and zebra crossing intelligent guardrail passes through target profile recognition target, and the effective filtering target profile noise of discernment in-process can make accurate discernment to pedestrian and vehicle, improves crossing street efficiency and security.
Example 5: a high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile;
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: eyes of a userMarking profile G (t) and noisy profile GN(t) the curvatures correspond to k (t) and k, respectivelyN(t); due to the influence of noise, the noisy contour GN(t) curvature value k of upper part feature pointN(t) cannot accurately represent the contour information, and in order to obtain an accurate curvature, a window function W (n) having a width D ∈ {17, 19} is selected, and a curvature k is calculatedN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T１=0.26, and determines a noise-containing contour curvature k 'from the comparison result'N(t), namely:
when | k1N(t)-k2N(t)|＞T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TKSelf-adaptive decision T by judging the amount of the target contour featuresKWhen k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1;
the distribution of the characteristic points and the non-characteristic points obtained after classification is not continuous, and a filter cannot be selected to carry out effective contour smoothing on the characteristic points and the non-characteristic points. In order to obtain a good contour smoothing effect, it is necessary to perform merging processing on contour points of the same type.
A merging module: the method is used for eliminating the pseudo feature points generated by noise interference and carrying out merging operation on the feature points and the non-feature points which cannot form a continuous area, so that an effective feature area and a non-feature area are obtained: selecting a starting point O, extending the outline starting point to two sides and merging the adjacent outline starting pointsWith the starting point type as the preset type of the region, extend each S × mu to both sides0A stop, where S is a preset minimum length, in this embodiment 25,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing the mean radius of curvature of the O points obtained from the window function, the real-time curvature correction factor mu0The method is used for automatically correcting the extension length according to different curvatures of different points, wherein the part with large curvature needs smaller length, and the part with small curvature needs larger length, so that the distortion phenomenon after combination can be effectively reduced; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; and combining adjacent regions of the same type to obtain continuous characteristic regions and non-characteristic regions.
A filtering module: the multiplicative noise is related to the image signal and changes along with the change of the image signal, the wiener filtering is adopted to carry out primary filtering, and the image information also comprisesWith residual multiplicative noise, pass F (x, y) filter as q × exp (- (x)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, the arc length parameterized equation of the noisy target contour is expressed as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' represents coordinates of each point on the noisy contour after removing the multiplicative noise, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsThe noise-containing contour is smoothed, named as a K filter, and is subjected to contour point classification and region division, and the noise-containing contour GN (t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationIn the non-characteristic region, the effect of suppressing noise is focused onWhere σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
In this embodiment, S is 25, threshold T10.26, window function width D ∈ {17, 19}, there is the smoothing effect of preferred to the noisy image of noise intensity I ∈ {50dB, 60dB }, and the detail information keeps the condition better, and zebra crossing intelligent guardrail passes through target profile recognition target, can effectively filter target profile noise in the identification process, can make accurate discernment to pedestrian and vehicle, improves crossing road efficiency and security.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the protection scope of the present invention, although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Data emulation
This zebra crossing intelligence guardrail's beneficial effect does: aiming at the diversity of noise types and the singleness of the current denoising method, a novel multi-time filtering device is adopted, and a novel contour segmentation and combination means and a filtering function are provided; the calculated amount is relatively uncomplicated, the factors of global characteristics and local characteristics are considered, and the smooth denoising effect is good; the difference of the contour among different types of regions is considered, and good balance is achieved between noise suppression and detail retention; according to the different curvatures of different points, the extension length is correspondingly and automatically changed in an adaptive manner, and the distortion phenomenon after combination is effectively reduced.
Through simulation, the device is adopted to carry out comparison under the noise intensity N, and the recognition rate of the target is as follows:
Claims (2)
1. A high-efficiency intelligent zebra crossing guardrail comprises a common intelligent zebra crossing guardrail and a target recognition device arranged on the intelligent zebra crossing guardrail, wherein the intelligent zebra crossing guardrail has strong recognition capability, and the target recognition device can recognize a target according to a target profile; wherein,
the modeling module is used for establishing a parameterized equation of the target contour: for a given target contour g (t), the arc length parameterization equation is expressed as g (t) ═ (x (t), y (t)), where x (t) and y (t) respectively represent the coordinates of the contour points, t represents the parameters of the contour curve equation, and t e [0, 1 ];
the parametric equation for the arc length of the noisy profile is expressed as: gN(t)＝G(t)+N1(t)+N2(t) G (t), wherein the additive noise part N1(t)＝N1(x1(t)，y1(t)), multiplicative noise part N2(t)＝N2(x2(t)，y2(t))；
A segmentation module for segmenting the contour: target profile G (t) and noisy profile GN(t) the curvatures correspond to k (t) and k, respectivelyN(t) selecting a window function W (n) of width D, D ∈ {7, 9}, for curvature kN(t) performing neighborhood averaging to obtain an average curvature k1N(t) simultaneously ordering the curvature values within the window, selecting a median curvature k2N(t) average curvature k1N(t) and median curvature k2N(T) the absolute value of the difference with a selected threshold T1Comparing, and determining the curvature k 'of the noisy contour according to the comparison result'N(t)，T10.2, i.e.:
when | k1N(t)-k2N(t)|＞T1K'N(t)＝k1N(t)
Otherwise, k'N(t)＝k2N(t)；
Since contour points with larger curvature values generally reflect the salient features of the target in terms of k'N(T) dividing all contour points in the contour into feature points or non-feature points, and setting variable weight TKSelf-adaptive decision T by judging the amount of the target contour featuresK，
When k'N(t)|＜TK*max|k′N(t) |, the characteristic function f (t) is 0
Otherwise, the feature function f (t) is 1.
2. The intelligent zebra crossing guardrail of claim 1, further characterized in that the merge module: for eliminating false characteristic points generated by noise interference and carrying out merging operation on characteristic points and non-characteristic points which cannot form continuous areasSelecting a starting point O, extending the outline starting point to two sides, merging adjacent points, using the starting point type as the preset type of the area, and extending each S × mu to two sides0Stopping, wherein S is a preset minimum length, S is 15,is the real-time curvature correction factor at point O,represents the radius of curvature of the O point,representing the mean radius of curvature of the O points obtained from the window function, the real-time curvature correction factor mu0The method is used for automatically correcting the extension length according to different curvatures of different points, and can effectively reduce the distortion phenomenon after combination; respectively calculating the number N +1 and the number N-1 of the different points in the two side areas, wherein if the number of the different points is less than the set minimum number of the different points of the type, the area is the same as the preset type, otherwise, the area is opposite to the preset type; then two stopping points O+1And point O-1Restart the calculation as the starting point, extend outward by S × muO+1Or S × muO-1Is stopped at a time of muO+1And muO-1Respectively represent point O+1And point O-1Real-time curvature correction coefficient of (d), O+1The number of the different points in the two side areas is N+2，O-1The number of the different points in the two side areas is N-2According to the judging conditions, sequentially determining the profile types of all the sections, calculating the number of different points of the part with the length less than S according to the ratio of the part with the length less than S, and counting the number into the corresponding characteristic area; adjacent areas of the same type are combined to obtain a continuous characteristic area and a non-characteristic area;
a filtering module: the multiplicative noise is related to the image signal and changes along with the change of the image signal, the wiener filtering is adopted to carry out primary filtering, and the image information also contains residualThe remaining multiplicative noise is filtered by an F filter F (x, y) to q × exp (- (x)2+y2)/β2Two-stage filtering is performed, where q is a coefficient that normalizes the function, i.e., (x ^ q × exp (- (x)2+y2)/β2) dxdy ═ 1, β are image template parameters;
after multiplicative noise filtering, the arc length parameterized equation of the noisy target contour is expressed as GN(t)’＝G(t)+N1(t); the additive noise is assumed to be white gaussian noise: x is the number ofN(t)’＝x(t)+g1(t，σ2)，yN(t)’＝y(t)+g2(t，σ2) Wherein x isN(t)' and yN(t)' represents coordinates of each point on the noisy contour after removing the multiplicative noise, g1(t，σ2) And g2(t，σ2) Respectively mean value of zero and variance of sigma2The Gaussian white noise is used for simulating additive noise in the noise-containing target contour;
using functionsSmoothing the noise-containing contour, named as K filter, and classifying contour points and dividing regions to obtain noise-containing contour GN(t)' is expressed as a combination of different types of contour segments:whereinRepresenting the segmentation of the contour that contains the region of the feature,representing contour segment containing non-feature region, selecting parameters of K filter according to contour feature distribution, and considering global feature and local feature factors, in the feature region, in order to retain detail informationIn the non-characteristic region, in order to raise the effect of suppressing noise Where σ' is the global variance, σ, estimated a priori1Estimating the variance, σ, a priori for selected feature regions0The variance is estimated a priori for the selected non-characteristic regions,the real-time curvature correction factor is averaged for the selected feature region,a real-time curvature correction factor for the mean of the selected non-characteristic region; in order to achieve a good smoothing effect, half of the minimum length S of each type of region is selected as the length of the 85% confidence interval of the K filter, so that the K filters with different parameters are self-adapted according to the lengths of the two types of regions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610014794.6A CN105603912A (en) | 2016-01-07 | 2016-01-07 | High-efficiency intelligent zebra stripe guardrail |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610014794.6A CN105603912A (en) | 2016-01-07 | 2016-01-07 | High-efficiency intelligent zebra stripe guardrail |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105603912A true CN105603912A (en) | 2016-05-25 |
Family
ID=55984221
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610014794.6A Pending CN105603912A (en) | 2016-01-07 | 2016-01-07 | High-efficiency intelligent zebra stripe guardrail |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105603912A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107905152A (en) * | 2017-12-11 | 2018-04-13 | 段云涛 | A kind of holography zebra stripes guardrail |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030156762A1 (en) * | 2001-10-15 | 2003-08-21 | Jonas August | Volterra filters for enhancement of contours in images |
| CN101673345A (en) * | 2009-07-01 | 2010-03-17 | 北京交通大学 | Method for extracting target closed contour based on shape prior |
| CN103903251A (en) * | 2012-12-30 | 2014-07-02 | 南京理工大学 | Night vision image salient contour extracting method based on non-classical receptive field composite modulation |
-
2016
- 2016-01-07 CN CN201610014794.6A patent/CN105603912A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030156762A1 (en) * | 2001-10-15 | 2003-08-21 | Jonas August | Volterra filters for enhancement of contours in images |
| CN101673345A (en) * | 2009-07-01 | 2010-03-17 | 北京交通大学 | Method for extracting target closed contour based on shape prior |
| CN103903251A (en) * | 2012-12-30 | 2014-07-02 | 南京理工大学 | Night vision image salient contour extracting method based on non-classical receptive field composite modulation |
Non-Patent Citations (1)
| Title |
|---|
| 史思琦: ""基于轮廓特征的目标识别研究"", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107905152A (en) * | 2017-12-11 | 2018-04-13 | 段云涛 | A kind of holography zebra stripes guardrail |
| CN107905152B (en) * | 2017-12-11 | 2020-12-08 | 段云涛 | Holographic zebra crossing guardrail |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107679520B (en) | Lane line visual detection method suitable for complex conditions | |
| CN106652465B (en) | Method and system for identifying abnormal driving behaviors on road | |
| TWI607901B (en) | Image inpainting system area and method using the same | |
| CN106022243B (en) | A kind of retrograde recognition methods of the car lane vehicle based on image procossing | |
| Xu et al. | Shadow removal with blob-based morphological reconstruction for error correction | |
| CN105678262A (en) | Tunnel geological monitoring equipment capable of conducting autonomous operation | |
| CN105869174B (en) | A kind of Sky Scene image partition method | |
| CN110909686A (en) | Low-illumination image enhancement system for driving assistance | |
| CN104331683A (en) | Facial expression recognition method with noise robust | |
| CN105528590A (en) | Quickly-alarming alarm device | |
| CN105414774A (en) | Laser cutting device capable of achieving autonomous cutting | |
| Pradeep et al. | An improved technique for night-time vehicle detection | |
| CN105603912A (en) | High-efficiency intelligent zebra stripe guardrail | |
| Wang et al. | Lane detection algorithm based on density clustering and RANSAC | |
| CN105469394B (en) | A kind of Intelligent target tracking based on complex environment | |
| CN105459901A (en) | Vehicle driving restriction warning device capable of autonomously operating | |
| CN105447485A (en) | Landscape modeling detection tool realizing rapid detection | |
| CN105678770A (en) | Wall crack detection device excellent in profile identification and filtering performances | |
| CN105488514A (en) | Welding device with autonomous welding function | |
| JP2008112349A (en) | Moving object detection device | |
| CN114463814B (en) | Quick certificate photo glasses detection method based on image processing | |
| CN105631460A (en) | Automobile auxiliary brake device capable of fast braking | |
| CN105677984A (en) | Bridge construction device good in contour recognition filtering performance | |
| CN114663860A (en) | Lane line recognition system and method for lane departure system | |
| CN105590105A (en) | Transverse limiting and error-connecting device with rapid response |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160525 |
|
| RJ01 | Rejection of invention patent application after publication |