CN103822581B - A kind of irregularly shaped object volume measuring method based on compressed sensing - Google Patents

Abstract

A kind of irregularly shaped object volume measuring method based on compressed sensing, first, demarcates system, obtains the pixel equivalent of system；Secondly, utilize three photographic head to obtain object under test respectively to overlook, left view, the right original image regarding three directions, and the maximum peripheral lengths image coordinate parameter of object under test is obtained by top view, utilize simultaneously and overlook greatest length pixel count for according to the reconstruct scope corresponding to the view of segmentation left and right, respectively left view and right view are carried out three-dimensionalreconstruction by shape from shading method, it is randomly chosen the partial coordinates point that shape from shading method obtains, with partial coordinates point as measured value, select compression calculation matrix and sparse base, rebuild the original coordinates of left and right view, improve the accuracy of coordinate figure, the volume of left and right view restructuring object is calculated respectively under image coordinate, then sue for peace；Finally, calculating the actual volume of object under test in world coordinate system according to pixel equivalent, the present invention can realize the cubing of irregularly shaped object.

Description

A kind of irregularly shaped object volume measuring method based on compressed sensing

Technical field

The present invention relates to a kind of method utilizing digital image processing techniques to realize irregularly shaped object cubing, It is specifically related to a kind of irregularly shaped object volume measuring method based on compressed sensing.

Background technology

In real life and industrial circle, some irregular materials are had to need to measure its volume, such as industry Raw-material on-line detecting system.Along with the development of the technology of measurement, the measurement circumstance complication of irregularly shaped object Changeable, measuring method is the most constantly reformed.

Traditional measuring method is that people measures the geometry of rule objects with tape measure scene, although simple, Easy, but object of observation need to be arranged, manual site measures poor in timeliness, and precision is the highest, uncomfortable Close the measurement of large-sized object, it is also difficult to realize non-contact measurement.Existing based on computer vision measurement master If visually being felt by double (many) and irregularly shaped object being carried out three-dimensionalreconstruction, then calculate three-dimensionalreconstruction The volume of object.Double (many) mesh vision measurements are the object image information obtaining different angles, and multiple stage is taken the photograph Relative position between camera to be fixed accurately, and multiple cameras to be ensured is same shooting Step property and stability, control is difficult to, substantially can only realize the application of little scope, needs to find For the characteristic point of coupling between image, and Feature Points Matching is the difficult point in image procossing, hence with many It is complicated, computationally intensive that width image carries out the calculating operation of the volume after three-dimensional reconstruction, is not suitable for dynamic scene etc. Problem.There is not the coupling of image characteristic point in monocular vision measuring method, is to directly utilize in scene Know geometric properties, under the meaning of one affine transformation of difference, obtain the three dimensional structure of object, to equipment Require simple, it is not necessary to the coupling of multiple image, with low cost, it is easy to accomplish, saved the time, improve Efficiency, shape from shading method (SFS) method is the most frequently used monocular vision method, is according to objects in images table The light and shade change in face, the relative altitude of recovery body surface each point, but as it is assumed that the reconstruction of condition object Precision is the highest, and existing irregularly shaped object volume measuring method limits it to a certain extent in industry mistake Application in journey.

Utilizing compressed sensing (CS) that object carries out three-dimensional reconstruction is the portion obtained by shape from shading method Divided data carries out secondary recovery to object, improves the reconstruction precision of object, and the method can reduce binocular vision Three-dimensional reconstruction carries out the complexity of volume calculating, and the volume reducing irregularly shaped object calculates error.

Summary of the invention

In order to overcome the shortcoming of above-mentioned prior art, it is an object of the invention to provide a kind of based on compressed sensing Irregularly shaped object volume measuring method, in commercial production, irregular material cubing is as background, utilize The part data that shape from shading obtains carry out secondary reconstruct to irregularly shaped object, are used for realizing irregularity The hard measurement of body volume, accelerates measuring speed, improves production efficiency.

To achieve these goals, the technical solution used in the present invention is:

A kind of irregularly shaped object volume measuring method based on compressed sensing, comprises the steps:

Three photographic head are installed and are positioned at three directions left and right, upper composition volume survey of irregularly shaped object to be measured Amount system, first demarcates volume measuring system, obtains the pixel equivalent of system；

Secondly, three photographic head are utilized to obtain object under test vertical view, left view, the right side respectively regarding three directions Original image, and obtained the maximum peripheral lengths image coordinate parameter of object under test by top view, simultaneously Utilize the reconstruct scope overlooking greatest length pixel count corresponding to the view of foundation segmentation left and right, extensive by light and shade Complex shape method carries out three-dimensionalreconstruction to left view and right view respectively, is randomly chosen shape from shading method The partial coordinates point obtained, with partial coordinates point as measured value, selects compression calculation matrix and sparse base, Rebuild the original coordinates of left and right view, improve the accuracy of coordinate figure, under image coordinate, calculate a left side respectively The volume of right view reconstruction of objects, then sues for peace；

Finally, the actual volume of object under test in world coordinate system is calculated according to pixel equivalent.

Described demarcating system, the method for the pixel equivalent obtaining system is: utilize this system to one Object known to individual geometric parameter L measures, and obtains representing the pixel count M of its parameter, then utilizes Formula ε=L/M calculates pixel equivalent ε, obtains the actual size representated by unit picture element.

Utilize Da-Jin algorithm Threshold segmentation that top view is carried out binary conversion treatment, then obtain top view and push up up and down Point position, takes both lateral symmetry centrages, utilizes and overlooks greatest length pixel count for about according to segmentation Reconstruct scope corresponding to view.

The process respectively left view and right view being carried out three-dimensionalreconstruction by shape from shading method is:

First, left and right view is split with background, background gray levels is set to 0；

Secondly according to the half-tone information of irregularly shaped object left and right view, the smoothness condition of body surface is introduced, Solve the Grad of body surface；

The last height value being obtained object by Grad, and then obtain the three dimensional structure of object.

Test result indicate that, the present invention relates to measures irregularly shaped object volume based on compressed sensing, The cubing of irregularly shaped object can be realized, be the most contactless volume measuring method of one.If Apply the present invention to field of industrial production, can preferably solve industry initial material volume cannot accurately, Quickly, the problem of non-cpntact measurement, reduce manual operation, promote industrial development, improve national economy, tool There is the biggest market potential.

Accompanying drawing explanation

Fig. 1 is volume measuring method process chart of the present invention.

Detailed description of the invention

The present invention is using stone as measurand, and handling process as shown in Figure 1, is embodied as step such as Under:

Step1, three photographic head of installation are positioned at three directions left and right, upper composition of irregularly shaped object to be measured Volume measuring system, first demarcates volume measuring system, i.e. utilizes this system to a geometric parameters Object known to number L measures, and obtains representing the pixel count M of its parameter, then utilizes formula ε=L/M calculates pixel equivalent ε, obtains the actual size representated by unit picture element.

Step2, obtain the top view of determinand and left and right two width side view by CCD camera.

Step3, utilize Da-Jin algorithm Threshold segmentation that top view is carried out binary conversion treatment.

Step4, obtain the upper and lower vertex position of top view, take both lateral symmetry centrages, utilize and overlook Long length pixel count is according to the reconstruct scope corresponding to the view of segmentation left and right.

Step5, according to shape from shading method, left and right view is carried out three-dimensionalreconstruction, recovers left and right view 3-D view, obtain NL image coordinate set of threedimensional model of left view: (x_i,y_i,z_i)'；Right view NR image coordinate set of threedimensional model: (x_i,y_i,z_i)'。

Step6, NL (x to left view_i,y_i,z_i) randomly select ML coordinate points as measured value Y_ML×1, Wherein Y_ML×1Each element be 1 × 3 the coordinate figure of point, select condensation matrix Φ_ML×NL, sparse base Ψ_NL×NL, Utilize compressed sensing to recover the original coordinates (x of left view_i,y_i,z_i)。

Actual voxel block number included in Step7, statistics left view threedimensional model space, is calculated as sum_left。

Step8, NR (x to right view_i,y_i,z_i) randomly select MR coordinate points as measured value Y_MR×1, Wherein Y_MR×1Each element be 1 × 3 the coordinate figure of point, select condensation matrix Φ_MR×NR, sparse base Ψ_NR×NR, Utilize compressed sensing to recover the original coordinates (x of right view_i,y_i,z_i)。

Actual voxel block number included in Step9, statistics right view threedimensional model space, is calculated as sum_right。

In Step10, image coordinate system, the voxel block number that left and right threedimensional model comprises is sum=sum_left+sum_right。

Step11, utilize pixel equivalent, utilize formula V=sum × ε³, the volume in image coordinate system is changed Calculate and measure volume for reality.

Step12, output object actual volume.

The three-dimensionalreconstruction process of above-mentioned right view:

Step1: the right view after being processed by gray processing is split with background, and background gray levels is set to 0；

Step2: be that (x y), surface graded is z=z by the representation of a surface of the right view of irregularly shaped object (p, q), i.e. $p(x,y)=\frac{\∂z(x,y)}{\∂x},q(x,y)=\frac{\∂z(x,y)}{\∂y},$ Then the normal vector of curved surface is (p, q ,-1), light Incident vector be estimated as: (p_s,q_s,-1)。

Step3: shape from shading method assumes that the gray value E of object table cake is only by the cosine of this point source Determine: $E(x,y)=\frac{{\mathrm{pp}}_s+{\mathrm{qq}}_s+1}{\sqrt{1+p^2+q^2}\sqrt{1+p_s^2+q_s^2}},$

Step4: introduce integrability constraints solve surface graded (p, q).

Step5: according to (p q) solves partial differential equation and obtains z (x, y), for object right view midpoint (x_i,y_i), coordinate figure is (x_i,y_i,z_i(x_i,y_i)) ', i.e. (x_i,y_i,z_i) ', and then obtain the three dimensional structure of right view.

The three-dimensionalreconstruction process of left view:

Step1: the left view after being processed by gray processing is split with background, and background gray levels is set to 0；

Step2: be that (x y), surface graded is z=z by the representation of a surface of the left view of irregularly shaped object (p, q), i.e. $p(x,y)=\frac{\∂z(x,y)}{\∂x},q(x,y)=\frac{\∂z(x,y)}{\∂y},$ Then the normal vector of curved surface is (p, q ,-1), light source The estimation gradient in direction is (p_s,q_s).

Step3: shape from shading method assumes that the gray value E of object table cake is only by the cosine of this point source Determine: $E(x,y)=\frac{{\mathrm{pp}}_s+{\mathrm{qq}}_s+1}{\sqrt{1+p^2+q^2}\sqrt{1+p_s^2+q_s^2}},$

Step4: introduce integrability constraints solve surface graded (p, q).

Step5: according to (p q) solves partial differential equation and obtains z (x, y), for object left view midpoint (x_i,y_i), coordinate figure is (x_i,y_i,z_i(x_i,y_i)) ', i.e. (x_i,y_i,z_i) ', and then obtain the three dimensional structure of left view.

Claims (4)

1. an irregularly shaped object volume measuring method based on compressed sensing, comprises the steps:

Three photographic head are installed and are positioned at three directions left and right, upper composition volume survey of irregularly shaped object to be measured Amount system, first demarcates volume measuring system, obtains the pixel equivalent of system；

Secondly, three photographic head are utilized to obtain object under test vertical view, left view, the right side respectively regarding three directions Original image, and obtained the maximum peripheral lengths image coordinate parameter of object under test by top view, simultaneously Utilize the reconstruct scope overlooking greatest length pixel count corresponding to the view of foundation segmentation left and right, extensive by light and shade Complex shape method carries out three-dimensionalreconstruction to left view and right view respectively, it is characterised in that

It is randomly chosen the partial coordinates point that shape from shading method obtains, with partial coordinates point as measured value, Select compression calculation matrix and sparse base, rebuild the original coordinates of left and right view, improve the accurate of coordinate figure Degree, calculates the volume of left and right view restructuring object under image coordinate respectively, then sues for peace；

Finally, the actual volume of object under test in world coordinate system is calculated according to pixel equivalent.

Irregularly shaped object volume measuring method based on compressed sensing the most according to claim 1, its Being characterised by, described demarcate system, the method for the pixel equivalent obtaining system is: utilize this to be Object known to one geometric parameter L is measured by system, obtains representing the pixel count M of its parameter, so After utilize formula ε=L/M to calculate pixel equivalent ε, obtain the actual size representated by unit picture element.

Irregularly shaped object volume measuring method based on compressed sensing the most according to claim 1, its It is characterised by, utilizes Da-Jin algorithm Threshold segmentation that top view is carried out binary conversion treatment, then obtain top view Vertex position up and down, takes both lateral symmetry centrages, utilizes and overlooks greatest length pixel count for according to dividing Cut the reconstruct scope corresponding to the view of left and right.

Irregularly shaped object volume measuring method based on compressed sensing the most according to claim 1, its Being characterised by, the process respectively left view and right view being carried out three-dimensionalreconstruction by shape from shading method is:

First, left and right view is split with background, background gray levels is set to 0；

Secondly according to the half-tone information of irregularly shaped object left and right view, the smoothness condition of body surface is introduced, Solve the Grad of body surface；

The last height value being obtained object by Grad, and then obtain the three dimensional structure of object.