CN103306186B - Algorithm for detecting structure depth of cement concrete pavement - Google Patents

Abstract

The invention discloses an algorithm for detecting the structure depth of a cement concrete pavement. The algorithm specifically comprises the following steps of: inputting image three-dimensional data matrixes, and filtering the data; sequentially taking the filtered three-dimensional data matrixes, obtaining each-line-corresponding structure depth line by line to obtain the structure depths C1, C2, -, Cm line by line, and averaging the m structure depths to obtain the structure depth of the pavement within an image capturing region. The algorithm disclosed by the invention is simple to compute, is short in running time, and can be carried out without labors. After the surface measurement is adopted, the structure depth of the pavement can be detected only by imputing the acquired three-dimensional data matrixes of the cement concrete pavement, so that the detecting algorithm is high in efficiency, and exact in detection.

Description

A kind of detection algorithm of cement concrete pavement structure depth

Technical field

The invention belongs to field of road, refer to a kind of detection algorithm of cement concrete pavement structure depth especially.

Background technology

Along with the increase of the traffic volume, improving constantly of car speed, causes traffic accident constantly to occur, and the antiskid problem on road surface is become increasingly conspicuous.The generation of traffic accident and the cling property on road surface have very large relation, and have good correlation between the average texture on road surface and road surface section construction depth.Construction depth, also referred to as texture depth, refers to the mean depth of the rough open pores of the road surfaces of certain area.The construction depth of cement concrete surface has reacted the macrostructure of cement concrete surface, is the important indicator evaluating cement concrete pavement antiskid and drainage performance.In general, the cement pavement that surface texture depth is larger, can provide higher frictional resistance, in the area that amount of precipitation is larger, if road surfaces does not have enough construction depths to come retaining and draining, is easy to form moisture film at road surfaces and traffic accident occurs.Therefore to the detection of cement concrete pavement structure depth and evaluation very important.

In present stage, the detection method of pavement structural depth and device are a lot, and conventional method has and manually spreads micromicrofarad, digital picture detection method, laser range finder.Wherein, artificial paving micromicrofarad efficiency is low and test structure is subject to the factor of artificial disturbance a lot, poor reproducibility, be not easy to wet weather measure, for the highway that mileage is longer, some section can only be selected to carry out sample investigation, reduce the evaluation of test result to the pavement structural depth in whole section; Digital picture detection method is subject to the impact of the intensity of illumination of external environment and lighting angle very large, and image processing algorithm needs to be improved further; The result that laser range finder obtains is discontinuous, can not the microscopic textural features on true reappearance road surface, so error is larger.Although existing construction depth algorithm calculates simple, running time is short, is also adapted at adopting in real-time system, and the precision of algorithm is not high enough.In sum, existing detection technique also exists the problems such as error is large, efficiency is low, therefore studies that a kind of automaticity is high, easy to operate, efficiency is high, detects accurately high cement concrete pavement structure depth checkout gear and be extremely necessary.

Summary of the invention

For the defect existed in above-mentioned prior art or deficiency, main purpose of the present invention is, a kind of detection algorithm of cement concrete pavement structure depth is provided, this algorithm have employed correction function when calculating construction depth and revises, can not only to the real-time detection of cement concrete pavement structure depth, and algorithm calculates simple, efficient detection, and testing result is accurate.

In order to achieve the above object, the present invention adopts following technical scheme:

A detection algorithm for cement concrete pavement structure depth, specifically comprises the steps:

Step 1: computer reads 3 d image data matrix O _{m × n};

Step 2: carry out filtering process to 3 d image data matrix, comprises two-way standard deviation filtering and morphologic filtering two parts, obtains the 3 d image data matrix after filtering process;

Step 3: the construction depth asking every a line;

Get the i-th row data of the 3 d image data matrix after filtering process, by corresponding in three one-dimension array A1, A2, A3 after the data trisection of this row, data in A1, A2, A3 are carried out fitting a straight line respectively, by corresponding in three one-dimension array B1, B2, B3 for the corresponding match value obtained, respectively the data in A1, A2, A3 and the corresponding data in B1, B2, B3 are done difference and obtain three differences, and the data in C, by it sequentially stored in an one-dimension array C, are averaged and are designated as C by the absolute value of these three differences _i, then C _iit is the construction depth of the i-th row;

Step 4: the construction depth asking image acquisition region road surface;

Average to the construction depth of all row, brought in correction function y=a*x+b by this average and revise, x is this average, and y is correction value, is pavement structural depth value.

Further, the 3 d image data matrix O in described step 1 _{m × n}as follows:

$O_{m\×n}=\left[\begin{array}{ccccccc}{z}_{11}& z_{12}& z_{13}& \·\·\·& z_{1j}& \·\·\·& z_{1n}\\ z_{21}& z_{22}& z_{23}& \·\·\·& z_{2j}& \·\·\·& z_{2n}\\ \·& \·& \·& & \·& & \·\\ \·& \·& \·& & \·& & \·\\ \·& \·& \·& & \·& & \·\\ z_{i1}& z_{i2}& z_{i3}& \·\·\·& z_{\mathrm{ij}}& \·\·\·& z_{\mathrm{in}}\\ \·& \·& \·& & \·& & \·\\ \·& \·& \·& & \·& & \·\\ \·& \·& \·& & \·& & \·\\ z_{m1}& z_{m2}& z_{m3}& \·\·\·& z_{\mathrm{mj}}& \·\·\·& z_{\mathrm{mn}}\end{array}\right],(i=\mathrm{1,2}\·\·\·m,j=\mathrm{1,2}\·\·\·n)$

Z _ijexpression line number is i, the picture altitude data of row number corresponding to j.

Further, described step 2 specifically comprises the steps:

(1) two-way standard deviation filtering: 1> processes line by line: to data calculation art average and the standard deviation of every a line of 3 d image data matrix, then each data of this line are handled as follows: the value obtained divided by standard deviation with the absolute value of the difference of former data and arithmetic mean of instantaneous value and the threshold value of setting compare, if this value is greater than threshold value, then former data arithmetic mean of instantaneous value is replaced, otherwise keep former data constant; Described threshold value gets 3 ~ 8; 2> processes by column: on the basis processed line by line, process by column again, to data calculation art average and the standard deviation of each row of matrix, then each data of these row are handled as follows: the value obtained divided by standard deviation with the absolute value of the difference of former data and arithmetic mean of instantaneous value and the threshold value of setting compare, if value is greater than threshold value, then former data arithmetic mean of instantaneous value is replaced, otherwise keep former data constant; Described threshold value gets 3 ~ 8;

(2) morphologic filtering: carry out morphologic filtering on the basis of two-way standard deviation filtering, choice structure element carries out opening operation to matrix, and then choice structure element carries out expansion process to matrix; Obtain the 3 d image data matrix after filtering process.

The method that the present invention proposes has the following advantages:

1, without the need to artificial participation, overcome that the labour intensity that manual detection method has is large, safety is low, driving is disturbed, inefficiency and the lower shortcoming of detection accuracy.

2, adopt planar survey, only need input the 3 d image data collecting cement concrete pavement, can complete the detection of road pavement construction depth, therefore, this detection algorithm efficiency is high, detection is accurate, is adapted at adopting in real-time system.

3, carry out the two-way noise spot of row, column by two-way standard deviation filtering to image data matrix to eliminate; The noise spot in image data matrix has effectively been taken out by morphologic filtering, thus can the impact of effective stress release treatment point, applied widely.Being drawn by test can stress release treatment point better in conjunction with these two kinds of filtering methods, and the speed of service is very fast.

4, in this algorithm, employ correction function and the data calculated are revised, thus make calculated value closer to actual value, also namely improve the accuracy of algorithm.

5, for the maintenance management of cement concrete pavement provides strong Informational support, improve highway maintenance and managerial skills, simultaneously, for developing highway checkout equipment further, change the present situation of highway engineering in China checkout equipment overwhelming majority dependence on import, saving resource, the research and development technical force of cultivating oneself has laid manpower and technical foundation.

Below in conjunction with the drawings and specific embodiments, explanation is further explained to the present invention.

Accompanying drawing explanation

Fig. 1 is the general flow chart of algorithm of the present invention.

Fig. 2 is two-way standard deviation filtering algorithm flow chart.

Fig. 3 is morphologic filtering algorithm flow chart.

Fig. 4 calculates the i-th row data construction depth algorithm flow chart.

Detailed description of the invention

Be below the specific embodiment that inventor provides, it should be noted that, the embodiment provided illustrates further explanation of the present invention, and protection scope of the present invention is not limited to given embodiment.

See Fig. 1-Fig. 4, the detection algorithm of the cement concrete pavement structure depth of the present embodiment, specifically comprises the steps:

Step 1: input picture three-dimensional data matrix;

By the 3 d image data matrix O that image capture device (adopting CCD area array cameras in the present invention) collects _{m × n}input computer, computer reads 3 d image data matrix, in the present embodiment, m=1000, n=1536.

Step 2: carry out filtering process to 3 d image data matrix, comprises two-way standard deviation filtering and morphologic filtering two parts:

(1) two-way standard deviation filtering: as shown in Figure 2,1> gets 3 d image data matrix O line by line _{m × n}every data line, as the i-th row R _i=(z _i1, z _i2... z _{i, 1536}), i=(1,2 ... 1000), R is obtained _iarithmetic mean of instantaneous value with standard deviation S _i; Get each the data z in this row successively _ijif met then use arithmetic mean of instantaneous value replace this point data value zi _j, otherwise z _ijremain unchanged, k is row filter factor, is also threshold value, here k=3; 2> is line by line after removal of images noise spot, and use the same method and process by column, stress release treatment point, complete to all column processing again, obtains 3 d image data matrix Q;

(2) morphologic filtering: as shown in Figure 3, morphologic filtering is carried out on the basis of two-way standard deviation filtering, choice structure element S E1=strel (' line', 11,9) opening operation is carried out to matrix Q, obtain three-dimensional data matrix Q1, and then choice structure element S E2SE2=strel (' ball', 5,5) expansion process is carried out to matrix Q1, obtain 3 d image data matrix I.Structural element is utilized to carry out to matrix the conventional means that opening operation and expansion process are Digital Image Processing, but the conventional means in not three-dimensional matrice field, and this is innovative point of the present invention.Opening operation and expansion process play the effect of removing noise spot.Noise spot be reflective due to road surface smoother or laser beat less than place, or machinery equipment vibrations etc. cause.The object effectively removing noise is reached through step 2.

Step 3: the construction depth asking every a line;

As shown in Figure 4, get the i-th row data of the 3 d image data matrix I after filtering process, by corresponding in three one-dimension array A1, A2, A3 after the data trisection of this row, data in A1, A2, A3 are carried out fitting a straight line respectively, by corresponding in three one-dimension array B1, B2, B3 for the match value obtained, difference data in A1, A2, A3 and the corresponding data in B1, B2, B3 is asked to obtain three differences respectively, and by the absolute value of these three differences order stored in one-dimension array C, the data in C are averaged and is designated as C _i, then C _iit is the construction depth of the i-th row.

Step 4: the construction depth asking image acquisition region road surface;

Average to the construction depth of all row, then substituted in correction function y=a*x+b by this average and revise, x is described average, and y is correction value, is the pavement structural depth value that the 3-D view of step 1 input is corresponding.

The acquisition of described correction function y=a*x+b: the calculating respectively polylith road surface being carried out to construction depth value with sand patch method and this algorithm.The construction depth value of the every block road obtained by this algorithm forms one dimension matrix hh from small to large ord, and the construction depth value of the every block road obtained by sand patch method forms one dimension matrix zz from small to large ord; Then with matrix hh be abscissa, matrix zz obtains correction function for ordinate carries out fitting a straight line.Rule between the construction depth value that this correction function reflects the road surface calculated by this algorithm and the construction depth value on this road surface obtained by sand patch method.

In the present embodiment, choose 23 block roads, the construction depth value of the every block road obtained by this algorithm rearranges one dimension matrix hh=[0.648982,0.668880 from small to large ord, ..., 2.192089,2.311696], the construction depth value of the every block road obtained by sand patch method forms one dimension matrix zz=[0.97396 from small to large ord, 0.56185, ..., 1.11815,1.78051]; With matrix hh for abscissa, matrix zz is ordinate, and the polyfit () function in Calling MATLAB just can obtain correction function and be: y=0.3736*x+0.5774.

Obtain 1000 construction depth C line by line ₁, C ₂..., C ₁₀₀₀, and to these 1000 construction depths average into: avg=1.1095, substitutes into this average in above-mentioned correction function y=0.3736*x+0.5774, and the construction depth obtaining collection road surface corresponding to this image data matrix is: avg1=0.9919.

3 d image data is gathered to the cement concrete pavement sampling in section to be assessed, obtain multiple 3-D view, for each 3-D view, obtain the pavement structural depth value of its correspondence according to above-mentioned algorithm, after then arithmetic average being asked to the pavement structural depth value of collected multiple 3 d image data matrixes, just can obtain the construction depth of pickup area cement concrete pavement.

Claims (2)

1. a detection algorithm for cement concrete pavement structure depth, is characterized in that, specifically comprises the steps:

Step 1: computer reads 3 d image data matrix O _{m × n};

Step 2: carry out filtering process to 3 d image data matrix, comprises two-way standard deviation filtering and morphologic filtering two parts, obtains the 3 d image data matrix after filtering process; Described step 2 specifically comprises the steps:

(1) two-way standard deviation filtering: 1> processes line by line: to data calculation art average and the standard deviation of every a line of 3 d image data matrix, then each data of this line are handled as follows: the value obtained divided by standard deviation with the absolute value of the difference of former data and arithmetic mean of instantaneous value and the threshold value of setting compare, if value is greater than threshold value, then former data arithmetic mean of instantaneous value is replaced, otherwise keep former data constant; Described threshold value gets 3 ~ 8; 2> processes by column: on the basis processed line by line, process by column again, to data calculation art average and the standard deviation of each row of matrix, then each data of these row are handled as follows: the value obtained divided by standard deviation with the absolute value of the difference of former data and arithmetic mean of instantaneous value and the threshold value of setting compare, if this value is greater than threshold value, then former data arithmetic mean of instantaneous value is replaced, otherwise keep former data constant; Described threshold value gets 3 ~ 8;

(2) morphologic filtering: carry out morphologic filtering on the basis of two-way standard deviation filtering, choice structure element carries out opening operation to matrix, and then choice structure element carries out expansion process to matrix; Obtain the 3 d image data matrix after filtering process;

Step 3: the construction depth asking every a line;

Get the i-th row data of the 3 d image data matrix after filtering process, by corresponding in three one-dimension array A1, A2, A3 after the data trisection of this row, data in A1, A2, A3 are carried out fitting a straight line respectively, by corresponding in three one-dimension array B1, B2, B3 for the corresponding match value obtained, respectively the data in A1, A2, A3 and the corresponding data in B1, B2, B3 are done difference and obtain three differences, and the data in C, by it sequentially stored in an one-dimension array C, are averaged and are designated as C by the absolute value of these three differences _i, then C _iit is the construction depth of the i-th row;

Step 4: the construction depth asking image acquisition region road surface;

Average to the construction depth of all row, substituted in correction function y=a*x+b by this average and revise, x is this average, and y is correction value, is the construction depth on this image acquisition region road surface;

The acquisition of described correction function y=a*x+b: the construction depth value of the every block road obtained by above-mentioned steps forms one dimension matrix hh from small to large ord, the construction depth value of the every block road obtained by sand patch method forms one dimension matrix zz from small to large ord; Then with matrix hh be abscissa, matrix zz obtains correction function for ordinate carries out fitting a straight line.

2. the method for claim 1, is characterized in that, the 3 d image data matrix O that described step 1 obtains _{m × n}as follows:

$Q_{m\×n}=\left[\begin{array}{ccccccc}{z}_{11}& z_{12}& z_{13}& ...& z_{1j}& ...& z_{1n}\\ z_{21}& z_{22}& z_{23}& ...& z_{2j}& ...& z_{2n}\\ .& .& .& & .& & .\\ .& .& .& & .& & .\\ .& .& .& & .& & .\\ z_{i1}& z_{i2}& z_{i3}& ...& z_{\mathrm{ij}}& ...& z_{\mathrm{in}}\\ .& .& .& & .& & .\\ .& .& .& & .& & .\\ .& .& .& & .& & .\\ z_{m1}& z_{m2}& z_{m3}& ...& z_{\mathrm{mj}}& ...& z_{\mathrm{mn}}\end{array}\right],(i=\mathrm{1,2}...m,j=\mathrm{1,2}...n)$

Z _ijexpression line number is i, the picture altitude data of row number corresponding to j.