CN104573180A - Real-person shoe type copying device and shoe tree manufacturing method based on single-eye multi-angle-of-view robot vision - Google Patents

Abstract

The invention discloses a real-person shoe type copying device based on single-eye multi-angle-of-view robot vision. The device comprises a single-eye multi-angle-of-view 3D (Three-dimensional) vision box and a computer, wherein the single-eye multi-angle-of-view 3D vision box is used for shooting images of a foot of a real person at five different angles of view; the computer is used for achieving the 3D reconstruction of the foot of the real person and automatically generating a 3D printing file; a plane mirror rectangular bucket-type cavity is formed inside the single-eye multi-angle-of-view 3D vision box and consists of four trapezoidal mirror planes; the upper part of a mirror body is large, and the lower part of the mirror body is small; the mirror planes face the inner side of a cavity body; a light source used for providing uniform soft lighting for the foot of the real person is arranged at the lower part of the plane mirror rectangular bucket-type cavity; cameras used for obtaining the images of the foot of the real person at multiple angles of view according to the refraction and reflection principle of the mirror planes are further arranged in the single-eye multi-angle-of-view 3D vision box; the computer comprises a single-eye multi-angle-of-view 3D vision calibrating unit, an image division, conversion and correction unit, a real person foot surface shape measurement unit and an automatic STL (Standard Template Library) file generation unit. The invention further discloses a shoe tree manufacturing method based on the single-eye multi-angle-of-view robot vision.

Description

Based on true man's footwear type reproducing unit and the shoe tree manufacture method of monocular multi-view machine vision

Technical field

The present invention relates to the application of monocular multi-view machine vision in shoe tree digitizing, belong to the shoe pattern and shoe tree personalized design, shoe tree/true man's foot integrated treatment, shoe tree and the field such as automatically to make, particularly relate to a kind of true man's footwear type reproducing unit based on monocular multi-view machine vision and shoe tree manufacture method.

Background technology

People usually need the shoes trying different model on to find the footwear being applicable to oneself foot type as far as possible when buying shoes, to such an extent as in most cases, people's pin must adapt to the fixing size and dimension of shoes passively.In fact, the long interval foot type shape difference of pin belonged to corresponding to same code length is very large, even the left and right pin of same person is also not complete specular, also may there is the difference in various degree such as length, fat or thin, arch height, even the long interval situation of pin that do not belong to corresponding to same shoes code length of the left and right pin of same person is also comparatively common.

Put on the footwear fitted to allow people, the ideal that the footwear meeting human body personalization features are shoe industry manufacturers is always produced in design, and the measurement fast and automatically realizing foot type is exactly realize this desirable first step.Its second step is exactly shoe tree Digital manufacturing.

The design related in shoemaking process is divided into three major types: the design of shoe last or hat block figure modelling, upper pattern design, bottom parts and the design of end mould.In upper pattern design, no matter using graphic design method, three dimensional designs method, or computer-implemented method, all must carry out according to the surface size of shoe tree.In shoemaking process, shoe tree is the support of footwear design and manufacture, is to require to design the shoe-making die processed according to the foot type of people and shoemaking.Therefore, the design of shoe tree and digitizing thereof are the keys of shoe industry.

Domestic and international many manufacturers are all devoted to the research of advanced shoe making apparatus, greatly advance robotization and the personalized customization of shoe making apparatus.Make a general survey of current shoe tree digitizer, there is following characteristics: (1) have employed advanced digital control system in open type, is that underlying platform carries out shoe tree digitizing with digital control system; (2) size detection sensor is made with high-precision digital raster or photoelectric encoder; (3) low-end system still adopts contact type measurement, and high-grade system adopts laser measurement; (4) do not re-use female shoe last or hat block, directly by the shoe last or hat block type data of design, cut out required shoe last or hat block type; (5) there is the special CAD/CAM system of supporting shoemaking; (6) shoe tree digitizer and shoe tree process equipment are integrated into complete support equipment by footwear manufacturing equipment manufacturer, as shoe tree high-speed digitization scanning machine, shoe tree numerical control machine, shoe tree numerical control proof press etc.

Utilize 3 D laser scanning method for personalized shoe industry.Special foot type spatial digitizer is made by existing Inc., during measurement, pin to be measured is put into the assigned address of scanner, obtain the relevant three-dimensional information of pin, be connected to computer, use special software to scanning the data analysis obtained, reconstructing the three-dimensional model of pin.The method is swift to operate, and result precision is higher, and shortcoming is that scanner is expensive, is difficult to be widely used.The state-of-the-art producer of shoe tree digital processing equipment concentrates on Italy mostly in the world at present, and equipment price is expensive, and as shoe tree digital scanning machine adds numerical control machine, a set of price is 1,200,000 ~ 1,500,000 Renminbi.This kind of technology Chinese invention patent application number is 01130405.7 disclose a kind of personalized shoe tree mold designing method, first the method utilizes spatial digitizer to obtain the three-dimensional data of people pin and modeling, then itself and standard shoe last model are compared, and standard shoe tree is adjusted, to generate the specific shoe last model meeting specific consumers foot type.

Some scholar utilized laser sheet to carry out light to pin to cut afterwards, the face profile figure of the section in layer of pin that what light obtained after cutting is, the photo of pin taken by auxiliary and video camera, by analyzing the two-dimensional silhouette curve of every one deck, curves the most all is integrated, and can obtain three-dimensional foot type profile.Although the result that the method obtains is comparatively accurate, operation is comparatively complicated.

For such Ge Chan footwear big country of China, the requirement for shoe tree processing is that efficient, quality reaches a standard and has low cost.The cost of current three-dimensional laser scanner is also higher, and the price of the point type laser scanning head that precision is similar exceeds tens times even tens times of contact contour wheel, and the technology based on three-dimensional laser scanner measurement shoe tree of this restrict is in the development of China.

Footwear mould comprises shoe tree and people's foot, has following characteristics: (1) people foot is essentially convex closure from profile; (2) shoe tree adopts various material, as wooden, plastic cement etc.; (3) color is single, variation of image grayscale is little, and grey scale change has reacted the normal direction change of curved surface; (4) surface is irregular geometry, continuous and derivable; (5) compared with technical grade three-dimensionalreconstruction, the accuracy requirement of digitizing shoe tree is not high, can at grade.These features make machine vision can be applied in footwear mould digitizing aspect well.

Shoe tree is the mould of shoemaking.The first generation shoe tree mode of production, namely traditional shoe tree mode of production, be that female shoe last or hat block is rule of thumb made by hand by carving shoe last or hat block teacher, shoe tree process equipment is shoe-last machine tool, adopts the mode of pattern to copy female shoe last or hat block and obtains finished product shoe last or hat block; The second generation shoe tree mode of production, introduces manufacturing and designing of shoe tree by CAD/CAM technology, completes the design effort of shoe tree on computers, and generate shoe tree nc program, controls digital controlled block cutting machine, completes and obtains finished product shoe last or hat block to the high-efficient cutting processing of blank shoe last or hat block; This process will change the design production process of shoe tree greatly, greatly adapt to the changeable market demand; The third generation shoe tree mode of production, the 3D reproduction technology of true man's foot and 3D printing technique are merged, the shoe tree directly printing various personalization with 3D printer becomes possibility, and this will reduce personalized shoe tree cost of manufacture and cycle greatly.

For the second generation shoe tree mode of production, U.S., moral, Yi Deng state have carried out large quantity research at shoe tree CAM technical elements.With the own various places Shoes Factory all over the world of the CAM system such as upper stretching, upper base of conputer controlled; As the USM of Britain, the GOC of Germany, the DZA of Czech and the CMASOC Deng Xie industry department of the U.S. develop interactively computer-aided design (CAD) shoe system and corresponding secondary process system in succession, have also emerged in large numbers a large amount of new technology and patent in digital controlled block cutting.Generally, digital controlled block cutting machine unit is more than 200,000 dollars, and system is at 500,000 dollars even more than 1,000,000 dollars, and domestic shoemaking producer is generally difficult to bear.This kind of technology Chinese invention patent application number is the 200710068032.5 personalized shoe last model generating methods disclosing a kind of comprehensive three-dimensional foot type global parameter and partial cross-section adjustment, and this technology belongs to the second generation shoe tree mode of production.

About the third generation shoe tree mode of production, the 3D of true man's foot is copied to print two kinds of technology with shoe tree 3D and carry out fusion and dock, by monocular multi-view machine vision, 3D is carried out to true man's foot and copy, generate the three dimensional point cloud of true man's foot; When shoe tree 3D prints, the three dimensional point cloud that software reads true man's foot by CAD (computer aided design) technology (CAD) completes a series of digital slices, and these information of cutting into slices are sent on 3D printer, slim for continuous print aspect is stacked up by 3D printer, until a solid body is shaping, final formation and the on all four shoe tree of true man's foot type shape size.Shoe tree 3D is printed to the technology adopting " fused deposition modeling ", whole flow process is molten plastic (ABS/PLA or nylon) in shower nozzle, then just form thin layer by the mode of depositing plastic material fiber, 3D printer maximum printing size commercially available at present has met the actual demand that shoe tree 3D prints.

Notice the difference of the third generation shoe tree mode of production and the second generation shoe tree mode of production, need in the second generation shoe tree mode of production to carry out modeling to shoe last model, by pin-shoe last or hat block comfort level mapping relations model, inference mechanism is utilized to draw the distortion parameter of standard shoe tree, finally with segmented adaptive algorithm, distortion is carried out to standard shoe tree and obtain customized shoe-last, then parametric curve is utilized to carry out the CAD modeling of shoe tree and foot type, calculate the processing cutter rail of this customized shoe-last on numerically controlled shoe tree machine, processed by numerically controlled shoe tree machine.And the third generation shoe tree mode of production, by monocular multi-view machine vision, 3D is carried out to true man's foot and copy, generate the three dimensional point cloud of true man's foot; No longer need the CAD modeling carrying out shoe tree and foot type, but the three dimensional point cloud directly reading true man's foot completes a series of digital slices, and these information of cutting into slices are sent on 3D printer, slim for continuous print aspect is stacked up by 3D printer, until a solid body is shaping, final formation and the on all four shoe tree of true man's foot type shape size.

In current machine vision, also not difficult by a compute depth information, difficult point is the registration of same place from the image of different viewing angles, i.e. characteristic matching.Due to the registration Algorithm more complicated put at present, consuming time longer, be also difficult to achieve satisfactory results.A large amount of documents describes method for registering a little, but not outer similarity measurement and continuity hypothesis.Picture neighborhood of a point should be similar in some physical quantity in two photos for spatial point P, and such as gray scale, geometric configuration etc., be thisly called similarity measurement; 2 points close on same object, its picture point also should be close, if a bit determined, then contiguous its picture point of spatial point also should just nearby, and the degree of depth is more or less the same, Here it is continuity hypothesis.Mainly there is following two problems in characteristic matching: one is for some regular shape, and superficial makings uniform true man foot is difficult to extract desirable unique point, especially as the vision measurement of true man's foot; Two is that matching algorithm all needs to expend a lot of computational resources.Be finally three-dimensional reconstruction, to the arbitrfary point on tested scenery surface, its world coordinates is obtained by the parameter matrix combined calculation of this point respective coordinates in two images and two video cameras, and its calculated amount is large and affect larger by calibrated error and matching error.

In order to reduce the difficulty of registration a little, some document proposes and allows gauger put on socks with Point Mesh outward appearance.This kind of technology Chinese invention patent application number is a kind of gauge point extracting method irrelevant with illumination for 200910153169.X discloses.It first takes the foreground image without the be taken background image of target and the target that is taken with gauge point respectively when keeping camera position and sight line is constant.Then measurement markers spot projection regional scale and gauge point projection neighborhood yardstick; Then, utilize gauge point view field yardstick and gauge point projection neighborhood yardstick to carry out neighborhood gray analysis to pixel each in foreground image and background image respectively, thus draw in foreground image the pixel belonging to gauge point view field, i.e. gauge point projected pixel; Afterwards, to gauge point projected pixel carry out packet identification publish picture picture in gauge point view field, and label taking note spot projection region center of gravity be gauge point projection Pixel-level precision position.Chinese invention patent application number be 200310108856.2 disclose a kind of based on the three-dimensional foot type measuring of specifiable lattice pattern and the method for modeling.It is through on the pin of people with the socks being printed on grid, then taken by multiple digital camera or video camera simultaneously, guarantee that the angle point (corner) of each grid was photographed by least two digital cameras or video camera, then based on the principle of stereoscopic vision and optical imagery, the three-dimensional position of all grid angle points is recovered; On this basis, the three-dimensional position of the angle point utilizing these to recover, simulates the 3D grid covered on people's pin again, finally according to 3D grid, the locus of standard pin is retrained, by deformable modeling method, set up the model of specific three-dimensional foot type, and calculate the parameter of foot type.Chinese invention patent application number be 200510061272.3 and 200510061271.9 a kind towards sparse grid based on standard pin distortion three-dimensional foot type fast acquiring method also belong to this kind of technology.

Clearly, true man's foot real-time three-dimensional is measured can reduce production cost, but its prerequisite is the requirement that computing velocity can meet on-the-spot real-time.If carry out registration to pixel each in the multiple pictures taken simultaneously, under current computer speed, the real-time of three-dimensional model reconfiguration system will be very poor, and can run into a lot of ill computational problem.How effectively to reduce computation complexity by the geometrical-restriction relation of sighting device, the processing speed of raising system, current a lot of morbid state are calculated and is converted into the calculating of good state, the speed of rebuilding shoe tree precision and calculating is reasonably balanced, theoretic research is changed into the yield-power of reality, this is a core content of the present invention.On the other hand, the three dimensional point cloud that true man's foot real-time three-dimensional is measured directly directly is docked with 3D printing technique, avoid the complex steps such as the CAD modeling of foot type, contribute to reducing the manufacturing cost of shoe tree, shortening the manufacturing cycle of shoe tree, really realize the shoe tree/design of true man's foot integration, manufacture, this is another core content of the present invention.Along with popularizing of 3D printing technique, people will have one's own shoe tree digitalized data from now on, shoe tree digitalized data is sent to shoemaking manufacturer by network, customize and oneself mate suitable shoes completely or print to one's name personalized shoe tree, the promotion conference of this technology produces significant impact to the business model of whole shoe industry, realizes the fast custom of shoe tree personalization completely.

A kind of stereo visual system of desirable true man's foot reconstruct wishes that the hardware unit of shooting is simple and reliable, with low cost, real-time is good, carry out that the software algorithm accuracy rate of true man's foot three-dimensionalreconstruction is high, real-time and calculated amount is little, the result of true man's foot reconstruct can be used for 3D and print.

Summary of the invention

In order to overcome exist in existing true man's foot Three Dimensional Reconfiguration poor real, registration difficulty, computation complexity is high, hardware cost is high, the shoe tree process-cycle is long, shoe tree manufacturing cost is high, safeguard and use the deficiencies such as difficulty, the invention provides a kind of real-time good, calculate simple, low in hardware cost, be convenient to the true man's footwear type reproducing unit based on monocular multi-view machine vision safeguarding and use and shoe tree manufacture method.

The technical solution adopted for the present invention to solve the technical problems is:

True man's footwear type reproducing unit of monocular multi-view machine vision and a shoe tree manufacture method, its core is the 3 D stereo size and shape that can reconstruct true man's foot efficiently, fast, accurately, simply.First the present invention make use of long in view in mechanical drawing aligning, high concordant, wide equal geometric projection principle, utilizes this geometric relationship as the three-dimensional measurement realized true man's foot and geometry constraint conditions during reconstruct; But here it should be noted that utilize long aligning, high concordant, wide equal geometric projection principle is ability meeting requirements on three-dimensional geometry constraint conditions in projection view situation just, and can only provide two-dimensional geometry constraint condition in non-projection view situation.

Therefore, how simply obtain efficiently and accurately from true man's foot view of different view directions be a gordian technique of the present invention.First the present invention proposes a kind of monocular multi-view stereoscopic vision case centered by true man's foot to obtain the vertical view of true man's foot, front view, left view, rear view and right view; Monocular multi-view stereoscopic vision case is made up of 1 high-definition camera, the first mirror surface, the second mirror surface, rectangular white LED mixed light sources and 1 rectangular-shaped bucket die cavity of level crossing, obtains the image of true man's foot in 5 different visual angles by the Polaroid of high-definition camera.Described level crossing rectangular-shaped bucket die cavity is made up of measure-alike 4 isosceles trapezoid minute surfaces, and the big up and small down and minute surface in mirror chamber is towards inside cavities.The optical axis of high-definition camera and the configuration at 45 ° of the first mirror surface, the first mirror surface and the configuration in 90 ° of the second mirror surface, the axle center of level crossing rectangular-shaped bucket die cavity and the configuration at 45 ° of the second mirror surface; The incident ray of the camera lens of high-definition camera is made up of through the second mirror surface and the first mirror surface the direct projection light in the rectangular-shaped bucket die cavity of level crossing and the 3rd mirror-reflection light.Wherein, be incident upon the middle section of high-definition camera imaging plane through the reflection of the second mirror surface and the first mirror surface by the direct projection light of cavity, be incident upon the neighboring area of high-definition camera imaging plane by the light of the 3rd mirror-reflection again through the reflection of the second mirror surface and the first mirror surface, just can obtain the image at the multiple visual angle of true man's foot according to the catadioptric principle of minute surface.5 different perspective projection points are had in described monocular multi-view stereoscopic vision case, be respectively the perspective projection point of virtual video camera in the perspective projection point of real camera and 4 level crossings, the image of imaging corresponds respectively to vertical view, front view, left view, rear view and right view.Therefore, from the effect of imaging, monocular multi-view stereoscopic vision case is equivalent to the high-definition camera being configured with 5 inner parameters and color system strict conformance.Each video camera, comprises real camera and 4 virtual video cameras, and locus between their coordinate systems keeps immobilizing, and relation in 90 ° between any two.

After obtaining vertical view, front view, left view, rear view and right view respectively from different visual angles, further work how from each view, to be partitioned into true man's foot image file; Due to front view, left view, rear view and right view by catadioptric principle finally in real camera imaging, be not both direct perspective imaging with vertical view.Therefore, first will carry out the rotational transform of 180 ° to front view, left view, rear view and right view, each view after rotational transform meets perspective imaging principle; Then Threshold segmentation is carried out to each view after rotational transform, be partitioned into the true man's foot surfaces shape obtained from different certain viewing angles situation; Finally to carry out three-dimensional measurement to the true man's foot surfaces shape obtained in different certain viewing angles situation; Be stressed that different certain viewing angles here refer to vertical view, front view, left view, rear view and right view.

In the present invention, the true origin of world coordinate system is arranged on the point of intersection of the optical axis of 5 video cameras, by the Z axis of world coordinate system and the optical axis coincidence of real camera, the optical axis coincidence of 2 virtual video cameras in X-axis and Y-axis and 4 virtual video cameras; The areas imaging of true man's foot at vertical view, front view, left view, rear view and right view is just determined by such setting; In order to carry out three-dimensional measurement and three-dimensionalreconstruction to true man's foot surfaces shape exactly, need to demarcate described monocular multi-view stereoscopic vision case.

Further, problem below how from the view of variant certain viewing angles, to be partitioned into true man's foot edge shape exactly, the extraction of Image Segmentation Using and boundary contour is the basis of three-dimensional reconstruction, and it will directly have influence on the accuracy of model after three-dimensional reconstruction; Be employed herein overall Ostu algorithm to 5 width Image Segmentation Usings, after extraction true man foot prospect, also need to be normalized true man's foot image file at all visual angles.Because real camera is different from virtual video camera shooting distance, the image at vertical view and other visual angle can deposit difference in size, therefore need the size of the true man's foot adjusted in the view of variant certain viewing angles, position and direction, length between basic views aligns, high concordant, wide equal corresponding relation to make it meet.

In recent years, based on the method for the surface geometry modeling and rendering of sampled point unit, in field of Computer Graphics, increasing concern is received; Based on the topological structure that the geometric expression of point is consistent with the overall situation on process its surface of Maintenance free, and seem especially flexibly with simple; The principle of 3D printing technique is also extremely similar to unit's output simultaneously, and the essence of its each digital slices is exactly an element array; In view of above-mentioned factor, the present invention adopts the surface geometry modeling pattern of sampled point unit to carry out three-dimensional reconstruction to true man's foot; Specific practice is according to 5 different visual angles, namely from vertical view, front view, left view, rear view and right view, read the spatial information of each surface point on true man's foot, utilize between each view and meet long aligning, high concordant, wide equal space constraints limits the surface voxel geometric position on true man's foot.

Because the projection of true man's foot on a view only reflects its two-dimensional signal, therefore completely cannot determine true man's foot in a view.And the projection of true man's foot on two mutually perpendicular views can reflect its three-dimensional information, the 3D shape of true man's foot generally also can be determined.If but certain one side of true man's foot is all vertical with this two view plane, and eclipse phenomena can be caused to occur, usually the 3D shape of supplementary 3rd view projections ability complete reflection true man foot is needed in this case, and eclipse phenomena is also existed hardly for true man's foot, the 3D shape of true man's foot therefore just can be reconstructed by the projection on mutually perpendicular view between two.The projection system of conventional three mutually perpendicular plane composition objects in mechanical drawing.

In mechanical drawing, view is divided into basic views, direction view, partial view and oblique view.Object is called basic views to six Basic projection planes view obtained that projects, and is respectively front view, vertical view, left view, right view, upward view, rear view.What obtain from front projection is projected as front view, contains length and the elevation information of object outward appearance in visual angle, dead ahead in front view; What obtain from top projection is projected as vertical view, contains length and the width information of object outward appearance in vertical view; What obtain from left side projection is projected as left view, contains length and the elevation information of object outward appearance in visual angle, left side in left view; What obtain from right side projection is projected as right view, contains length and the elevation information of object outward appearance in visual angle, right side in right view; What obtain from below projection is projected as upward view, contains the length and width information of looking up object outward appearance in visual angle in upward view; What obtain from rear projection is projected as rear view, contains length and the elevation information of object outward appearance in visual angle, rear in rear view; Keep long between basic views and align, high concordant, wide equal corresponding relation, this obtains true man's foot surfaces dot information in the present invention, i.e. cloud data, and the basic premise carrying out three-dimensional reconstruction with this.

Because digitizing shoe tree does not need the colouring information of true man's foot surfaces point, the words that the cloud data Gaussian sphere coordinate of shoe tree is expressed, for (α, β, r), the point of intersection that the initial point of Gaussian sphere coordinate system is arranged on the optical axis of 5 video cameras is O, the Z axis of world coordinate system and the optical axis coincidence of real camera, the optical axis coincidence of 2 virtual video cameras in X-axis and Y-axis and 4 virtual video cameras; Shoe tree surface certain object point A is (x in the volume coordinate of cartesian coordinate system, y, z), A point is r to the distance of true origin O, angle between A point to the line and Z axis of true origin O is the angle between the projection line of the line of incident angle β, A point to true origin O in XY plane and X-axis is azimuth angle alpha.

Further, to illustrate on true man's foot surfaces the acquisition methods of cloud data a little.Suppose that the volume coordinate of a certain object point A of true man's foot surfaces is for (x, y, z); The so coordinate points (x, y) of the subpoint A1 of spatial point A in XOY plane necessarily in vertical view; The coordinate points (x, z) of subpoint A2 in XOZ plane necessarily in front view; The coordinate points (y, z) of subpoint A3 in YOZ plane necessarily in side view; According to above-mentioned constraint condition, by the mode traveled through in Gaussian sphere coordinate situation, with the surface of incident angle β, azimuth angle alpha and distance r whole true man's foot for variable travels through, obtain the cloud data of each object point in surface of true man's foot.

When traveling through whole true man's foot surfaces, first Gauss coordinate system to be set up, the initial point of coordinate system is set in the point of intersection of the optical axis of 5 video cameras, be XOZ plane in front view and rear view, be YOZ plane in left view and right view, be XOY plane in vertical view, be four individual by the cutting of true man's foot with XOZ plane and YOZ plane, be namely separated by 90 ° with azimuth angle alpha and carry out cutting; The scope traveling through view in different orientations α situation is obtained after carrying out such cutting, as shown in table 1;

The relation of the scope of table 1 azimuth angle alpha and the range of views of traversal

Some marginal information of true man's foot surfaces on each view is directly obtained by edge detection algorithm, as azimuth angle alpha in particular cases directly to calculate the marginal information of true man's foot surfaces 0 °, 90 °, 180 ° and 270 °; But be the marginal information that can not directly obtain in one party parallactic angle situation when except above-mentioned angle, this searches the algorithm of surface voxel under just needing to realize a no special situation in the surface process of traversal true man foot; More succinct in order to illustrate when realizing searching the algorithm of surface voxel, here three kinds of situations are divided into by according to a point spatially and true man's foot surfaces point relation, a point on internal point representation space is in the inside of true man's foot, a point on external point representation space is not or not inside and the edge of true man's foot, and a point on marginal point representation space is just in time on the surface of true man's foot; These are put us and are referred to as voxel, and our target obtains when traveling through position angle and incident angle the voxel that true man's foot surfaces passes under this position angle and incident angle situation.

Use discrete voxel to represent the geological information of true man's foot surfaces in the present invention, evenly being cut by azimuthal and incident angle makes whole true man's foot surfaces be a series of small cubes space by subdivision equably, according to the two-dimensional geometry scope long aligned, high concordant, wide equal principle defines true man's foot surfaces voxel, estimating the longest estimated distance R in certain incident angle β and certain azimuth angle alpha situation _e-outthe shortest estimated distance R _e-inthe small cubes space of a rear formation visual shell, is herein referred to as minimum bounding box.The size of minimum bounding box is Δ β × Δ α × (R _e-out-R _e-in), at this moment unique needs determine it is at the longest estimated distance R _e-outthe shortest estimated distance R _e-inbetween the voxel that passes of true man's foot surfaces.Therefore, the key of problem is at the longest estimated distance R _e-outthe shortest estimated distance R _e-inbetween scope in how to detect the voxel that true man's foot surfaces passes efficiently and accurately, be employed herein space engraving algorithm, its core is this evaluating of colour consistency.Colour consistency constraint is a step very crucial in most of voxel colouring algorithm, if colour consistency principle thinks that a voxel belongs to true man's foot surfaces, then this voxel all should be identical to the pixel color in the projection of its visual picture.

Colour consistency is judged by the color variance and threshold value comparing voxel projected pixel collection, if Pro (v) is the set of pixels of voxel v back projection on K width relevant view, the colour consistency function that Con (v) is Pro (v), λ is given threshold value; If v RGB triple channel of projected pixel on view m is expressed as (R _m, G _m, B _m), m ∈ 1,2 ..., K}; Color variance formula (9) calculates;

$\left\{\begin{array}{c}{\mathrm{\σ}}_x^2=\frac{1}{K-1}(\underset{m=1}{\overset{K}{\mathrm{\Σ}}}{X}_m^2-\frac{1}{K}{\left(\underset{m=1}{\overset{K}{\mathrm{\Σ}}}{X}_m\right)}^2);X\∈(R,G,B)\\ \mathrm{\σ}\frac{1}{3}\sqrt{{\mathrm{\σ}}_R^2+{\mathrm{\σ}}_G^2+{\mathrm{\σ}}_B^2}\end{array}\right.---\left(9\right)$

The colour consistency of voxel v is calculated with formula (10);

Con(v)＝σ-λ (10)

If Con (v) < 0, be then judged as that projected pixel collection meets colour consistency, this voxel is true man's foot surfaces voxel;

The concrete steps of ergodic algorithm are as follows:

STEP1: set up Gauss coordinate system, carries out cutting by the method shown in table 1 to true man's foot image file;

STEP2: initialization is carried out to incident angle β, azimuth angle alpha; Azimuth angle alpha=0 is set; Incident angle β=-90; Δ α=0.1; Δ β=0.1;

STEP3: determine the range of views traveled through according to azimuth angle alpha table 1;

STEP4: estimate the longest estimated distance R under certain incident angle β and certain azimuth angle alpha situation respectively with formula (4) and formula (5) _e-outthe shortest estimated distance R _e-in;

$R_{e-\mathrm{out}}=\sqrt{x_e^2+y_e^2+z_e^2}---\left(4\right)$

$R_{e-\mathrm{in}}=\min (\sqrt{x_e^2+y_e^2},\sqrt{x_e^2+z_e^2},\sqrt{z_e^2+y_e^2})---\left(5\right)$

In formula, x _e, y _e, z _edetermine by table 1 coordinate figure under traversal value incident angle β and azimuth angle alpha situation that the image in the range of views traveled through is obtained by rim detection; Obtain the bounding box Δ β × Δ α × (R of true man's foot surfaces voxel _e-out-R _e-in);

STEP5: from the shortest estimated distance R _e-instart traversal and search true man's foot surfaces voxel v _e, true man's foot surfaces voxel v _ejudgment mode carry out colour consistency detection by formula (10), if detected by colour consistency, then it can be used as the voxel v of true man's foot surfaces _e; If in bounding box, all voxels all do not meet formula (10) and carry out colour consistency detection in ergodic process, using the minimum value that calculates in formula (10) as true man's foot surfaces v _ecandidate; By the color average of set of pixels time computing formula (9) give this voxel, obtain the distance value r of this voxel to true origin simultaneously;

STEP6: (x, y, the z) value calculated under r with formula (6);

$\left\{\begin{array}{c}x=r\&times;\cos \mathrm{\&beta;}\&times;\cos \mathrm{\&alpha;}\\ y=r\&times;\cos \mathrm{\&beta;}\&times;\sin \mathrm{\&alpha;}\\ z=r\&times;\sin \mathrm{\&beta;}\end{array}\right.---\left(6\right)$

Calculate the cloud data of this object point preserve this cloud data;

STEP7: calculate β ← β+Δ β, judge β > 90, jumps to STEP4 if met;

STEP8: calculate α ← α+Δ α, judge α < 360, if met, β=-90; Jump to STEP3;

STEP9: terminate.

Minimum incident angleplied laminate and maximum incident angle layer outline data is obtained by the traversal of above-mentioned incident angle, the triangulation of outline data is obtained according to azimuthal traversal, then the tri patch of the profile of adjacent layer is connected, finally generate STL formatted file;

STL form is the common interface format that current 3D printed/increased the use of material manufacturing equipment, it is the interface protocol formulated in 1988 by 3Dsystems company of the U.S., be the three-dimensional picture file layout that a kind of 3D of being prints and increasing material manufacturing technology is served, become the de facto standards form that 3D printed/increased material manufacture at present.

STL form is a kind of straightforward procedure storing three-dimensional model information, and the digital model of complexity is carried out approximate expression with a series of three-dimensional triangulation shape dough sheet by it.According to ASC II format characteristic of stl file, STL three-dimensional model file is made up of a series of tri patch, and each tri patch is made up of three summits corresponding in three dimensions and the normal vector of plane that forms thereof.The form of following global variable is adopted to define summit and normal vector respectively.

typedef struct tagPoint3D

{

Double x; // store summit x coordinate;

Double y; // store summit y coordinate;

Double z; // store summit z coordinate;

CPoint3D; The dough sheet apex coordinate of // // store in STL;

typedef struct tagVector3D

{

Double dx; // storage method vector x coordinate;

Double dy; // storage method vector y coordinate;

Double dz; // storage method vector z coordinate;

CVector3D; // storage method vectorial coordinate;

In stl file, tri patch is the base unit of configuration file, definition class CTriClass storage cell tri patch, and structure is as follows:

class CTriClass:public CObject

{

public:

CPoint3D vex[3]；

CVector3D normal；

public:

CTriClass()；

virtual～CTriClass()；

Virtu virtual void Serialize (CArchive & ar); // serialization accesses;

}；

Beneficial effect of the present invention is mainly manifested in: beneficial effect of the present invention is mainly manifested in: 1) adopt a high-definition camera and 4 plane mirrors to obtain true man's foot image file from 5 different visual angles simultaneously, not only reduce the cost of three-dimensional reconstruction, also solve inner parameter skimble-scamble problem when multiple-camera is taken jointly simultaneously; 2) utilize the strict geometric relationship between basic views, add the constraint of the silhouettes of true man's foot, greatly simplifie the complexity of three-dimensional reconstruction; 3) three dimensional point cloud that true man's foot real-time three-dimensional is measured directly directly is docked with 3D printing technique, avoid the complex steps such as the CAD modeling of foot type, contribute to reducing the manufacturing cost of shoe tree, shortening the manufacturing cycle of shoe tree, really realize the Design and manufacture of shoe tree/true man's foot integration.

Accompanying drawing explanation

Fig. 1 is the design concept figure that a kind of true man's footwear type of monocular multi-view machine vision copies, wherein 1 is high-definition camera, 2 is the first mirror surface, 3 is the second mirror surface, and 4 is rectangular white LED mixed light sources inner circle camber mixed light face, and 5 is the 3rd mirror surface forming the rectangular-shaped bucket die cavity of level crossing, 6 is true man's foot, 7 is monocular multi-view machine vision case shell, and 8 is the subpoint of virtual video camera, and 9 is White LED light source;

Fig. 2 is true man's foot image file of the different visual angles adopting true man's footwear type reproduction technology of monocular multi-view machine vision to obtain;

Fig. 3 is the position view of true man's foot in world coordinate system;

Fig. 4 is the vertical view of true man's foot;

Fig. 5 is the front view of true man's foot;

Fig. 6 is the right view of true man's foot;

Fig. 7 is true man's foot replication processes flow process;

Fig. 8 is the relation explanation between the surface voxel of true man's foot and inside and outside voxel;

Fig. 9 is that 3D prints personalized shoe tree process;

Figure 10 is that 3D prints personalized shoe tree result;

Figure 11 is that 3D prints and increases the three-dimensional picture stl file form of material manufacturing technology service;

Figure 12 is the design concept figure of catadioptric minute surface in monocular multi-view stereoscopic vision case, and wherein 1 is true high-definition camera, and 5 is isosceles trapezoid minute surface, and 8 is virtual video camera, and 6 is true man's foot;

Figure 13 is the rectangular white LED mixed light sources design concept figure in monocular multi-view stereoscopic vision case, and wherein 4 is rectangular white LED mixed light sources inner circle camber mixed light face, and 9 is White LED light source, and 10 is the framework of rectangular white LED mixed light sources.

Embodiment

With reference to Fig. 1 ~ 11, true man's footwear type reproducing unit of monocular multi-view machine vision and a shoe tree manufacture method, by one for taking true man's foot at the monocular multi-view stereoscopic vision case of 5 different visual angles images and the computing machine that realizes true man's foot three-dimensionalreconstruction and automatically generate towards 3D printout stl file.

Monocular multi-view stereoscopic vision case is made up of 1 high-definition camera 1, first mirror surface 2, second mirror surface 3, rectangular white LED mixed light sources 9 and 1 rectangular-shaped bucket die cavity of level crossing, obtains the image of true man's foot in 5 different visual angles by the Polaroid of high-definition camera; The rectangular-shaped bucket die cavity of level crossing is made up of 4 isosceles trapezoid minute surfaces, and the big up and small down and minute surface in mirror chamber is towards inside cavities; The optical axis of high-definition camera and the configuration at 45 ° of the first mirror surface 2, the first mirror surface 2 and the configuration in 90 ° of the second mirror surface 3, the axle center of level crossing rectangular-shaped bucket die cavity and the configuration at 45 ° of the second mirror surface; The incident ray of the camera lens of high-definition camera 1 is formed through the reflection of the second mirror surface 3 and the first mirror surface 2 by the direct projection light in the rectangular-shaped bucket die cavity of level crossing and the 3rd mirror-reflection light; Wherein, be incident upon the middle section of high-definition camera imaging plane through the reflection of the second mirror surface 3 and the first mirror surface 2 by the direct projection light of cavity, be incident upon the neighboring area of high-definition camera 1 imaging plane by the light of the 3rd mirror-reflection again through the reflection of the second mirror surface 3 and the first mirror surface 2, just can obtain the image at the multiple visual angle of true man's foot according to the catadioptric principle of minute surface; Rectangular white LED mixed light sources provides evenly soft light source for true man's foot throws light on; 5 different perspective projection points are had in monocular multi-view stereoscopic vision case, be respectively the perspective projection point of virtual video camera in the perspective projection point of real camera and 4 level crossings, the image of imaging corresponds respectively to vertical view, front view, left view, rear view and right view, as shown in Figure 2.

Therefore, from the effect of imaging, monocular multi-view stereoscopic vision case is equivalent to the high-definition camera being configured with 5 inner parameters and color system strict conformance; Each video camera, comprises real high-definition camera and 4 virtual video cameras, and locus between their coordinate systems keeps immobilizing, and relation in 90 ° between any two; High-definition camera adopts the video camera of the above resolution of 2,000,000 pixel.

In order to utilize long in view in mechanical drawing aligning, high concordant, wide equal geometric projection principle, increases the three-dimensional measurement to true man's foot and geometry constraint conditions during reconstruct.After obtaining vertical view, front view, left view, rear view and right view respectively from different visual angles, from each view, be partitioned into true man's foot image file; Due to front view, left view, rear view and right view by catadioptric principle finally in real camera imaging, be not both direct perspective imaging with vertical view; Therefore, first will carry out the rotational transform of 180 ° to front view, left view, rear view and right view, each view after rotational transform meets perspective imaging principle; Then Threshold segmentation is carried out to each view after rotational transform, be partitioned into the true man's foot surfaces shape obtained from different certain viewing angles situation; Finally to carry out three-dimensional measurement to the true man's foot surfaces shape obtained in different certain viewing angles situation; Be stressed that different certain viewing angles refers to vertical view, front view, left view, rear view and right view here.

In computing machine, software section includes monocular multi-view stereoscopic vision and demarcates unit, Iamge Segmentation, conversion and correcting unit, color image enhancement unit, sets up coordinate system and pretreatment unit, true man's foot surfaces shape measuring unit, automatically generates stl file unit and true man's foot surfaces Shape Reconstruction result display unit; Relation between whole treatment scheme and each unit as shown in Figure 7.

In the present invention, the true origin of world coordinate system is arranged on the point of intersection of the optical axis of 5 video cameras, the Z axis of the present invention by world coordinate system and the optical axis coincidence of real camera, the optical axis coincidence of 2 virtual video cameras in X-axis and Y-axis and 4 virtual video cameras; The areas imaging of true man's foot at vertical view, front view, left view, rear view and right view is just determined by such setting; In order to carry out three-dimensional measurement and three-dimensionalreconstruction to true man's foot surfaces shape exactly, need to demarcate monocular multi-view stereoscopic vision case.

Monocular multi-view stereoscopic vision demarcates unit, for demarcating monocular multi-view stereoscopic vision case; Be employed herein monocular multi-view stereoscopic vision case and use flat mirror reflects light, a width multiple views projected image of shooting comprises multiple images of scaling board; From flat mirror reflects principle, these scaling board images are equivalent to same scaling board on diverse location, project to several scaling board images that video camera projection plane obtains respectively; So, use the multiple views projected image calibration for cameras inner parameter of a width scaling board, be equivalent to the proving operation using several scaling board images, comprise 5 width independence images of scaling board in image, high-precision calibration result can be obtained; Calibrated result leaves in calibration result database, to call in subsequent calculations.

Further, problem below how from the view of variant certain viewing angles, to be partitioned into true man's foot edge shape exactly, the extraction of Image Segmentation Using and boundary contour is the basis of three-dimensional reconstruction, and it will directly have influence on the accuracy of model after three-dimensional reconstruction; Be employed herein overall Ostu algorithm to 5 width Image Segmentation Usings, after extraction true man foot prospect, also need to be normalized true man's foot image file at all visual angles.Because real camera is different from virtual video camera shooting distance, the image at vertical view and other visual angle can deposit difference in size, therefore need the size of the true man's foot adjusted in the view of variant certain viewing angles, position and direction, length between basic views aligns, high concordant, wide equal corresponding relation to make it meet.

Iamge Segmentation, conversion and correcting unit, for being partitioned into true man's foot edge shape in the view from variant certain viewing angles, and utilize the true man foot image file of calibration result parameter to all visual angles to be normalized, have employed size conversion, translation transformation, rotational transform and aberration correction algorithm in normalized, 5 width aligned from the image of the different visual angles length met between basic views, high concordant, wide equal principle; 5 width that Iamge Segmentation, conversion and correcting unit process obtain from true man's foot edge shape of different multi-view image as shown in Figure 2.

Because the true man's foot in left, main, the right side, rear view is by catadioptric imaging, the impact that the brightness of the image of true man's foot is subject to specular reflectance has some decay, and the irradiation that is greater than in other views of lighting source exposure intensity is in a top view strong in addition; Therefore, need the brightness of true man's foot image file in an adjustment left side, master, the right side, rear view, make the brightness after adjusting consistent with the vertical view of direct imaging; The mode of color image enhancement is adopted to process respectively image that is left, main, right, rear view; First, with formula (1), image that is left, main, right, rear view is transformed on HSI color space,

$\left\{\begin{array}{c}H=\mathrm{ar}\cos \left\{\frac{[(R-G)+(R-B)]/2}{{[{(R-G)}^2+(R-B)(G-B)]}^{\frac{1}{2}}}\right\}\\ S=1-\frac{3}{R+G+B}\left[\min (R,G,B)\right]\\ I=\frac{1}{3}(R+G+B)\end{array}\right.---\left(1\right)$

In formula, R, G, B are respectively the component of image on RGB color space that be left, main, right, rear view, and H, S, I are respectively the component of image on HSI color space that be left, main, right, rear view.Then, the specular reflectance provided according to minute surface manufacturer also considers the illumination intensity of lighting source in different visual angles, adjusts luminance component with formula (2),

I'＝I×ΔI (2)

In formula, Δ I be greater than 1 brightness adjustment coefficient, I' is the luminance component after brightness adjustment; The image of the left side after then using formula (3) brightness to be strengthened, master, the right side, rear view transforms on RGB color space again, makes on each view

True man's foot image file there is identical color and vein characteristic;

$\left\{\begin{array}{c}h=H\&times;\mathrm{\&pi;}/180;s=S/100;i=I^{\&prime;}/255\\ x=i\&times;(1-s)\\ y=i\&times;[1+\frac{s\&times;\cos \left(h\right)}{\cos (\mathrm{\&pi;}/3-h)}]\\ z=3i-(x+y)\\ r=y;g=z;b=\mathrm{xwhenh}<2\mathrm{\&pi;}/3\\ r=x;g=y;b=\mathrm{zwhen}2\mathrm{\&pi;}/3\&le;h<4\mathrm{\&pi;}/3,h=h-2\mathrm{\&pi;}/3\\ r=z;g=x;b=\mathrm{ywhen}4\mathrm{\&pi;}/3\&le;h<2\mathrm{\&pi;},h=h-4\mathrm{\&pi;}/3\end{array}\right.---\left(3\right)$

R, g, b that formula (3) calculates are the color components of normalized RGB color space, and the RGB color space that convert standard to also needs to be multiplied by 255 to the value of r, g, b, obtain brightness and strengthen rear R', G', B' value.

Color image enhancement unit, for carrying out enhancing pre-service to the image in left, main, the right side, rear view, making the color of the image in a left side, master, the right side, rear view consistent with the color of vertical view, establishing unified Color mode for carrying out three-dimensionalreconstruction from the image of different visual angles.

Set up coordinate system and pretreatment unit, for carrying out measuring basis for the dimensional visual measurement of true man's foot.

In recent years, based on the method for the surface geometry modeling and rendering of sampled point unit, in field of Computer Graphics, increasing concern is received; Based on the topological structure that the geometric expression of point is consistent with the overall situation on process its surface of Maintenance free, and seem especially flexibly with simple; The principle of 3D printing technique is also extremely similar to unit's output simultaneously, and the essence of its each digital slices is exactly an element array; In view of above-mentioned factor, the present invention adopts the surface geometry modeling pattern of sampled point unit to carry out three-dimensional reconstruction to true man's foot; Specific practice is according to 5 different visual angles, namely from vertical view, front view, left view, rear view and right view, read the spatial information of each surface point on true man's foot, utilize between each view and meet long aligning, high concordant, wide equal space constraints limits the surface voxel geometric position on true man's foot.

Because the projection of true man's foot on a view only reflects its two-dimensional signal, therefore completely cannot determine true man's foot in a view.And the projection of true man's foot on two mutually perpendicular views can reflect its three-dimensional information, the 3D shape of true man's foot generally also can be determined; If but certain one side of true man's foot is all vertical with this two view plane, and eclipse phenomena can be caused to occur, usually the 3D shape of supplementary 3rd view projections ability complete reflection true man foot is needed in this case, and eclipse phenomena is also existed hardly for true man's foot, the 3D shape of true man's foot therefore just can be reconstructed by the projection on mutually perpendicular view between two; The projection system of conventional three mutually perpendicular plane composition objects in mechanical drawing.

In mechanical drawing, view is divided into basic views, direction view, partial view and oblique view.Object is called basic views to six Basic projection planes view obtained that projects, and is respectively front view, vertical view, left view, right view, upward view, rear view.What obtain from front projection is projected as front view, contains length and the elevation information of object outward appearance in visual angle, dead ahead in front view; What obtain from top projection is projected as vertical view, contains length and the width information of object outward appearance in vertical view; What obtain from left side projection is projected as left view, contains length and the elevation information of object outward appearance in visual angle, left side in left view; What obtain from right side projection is projected as right view, contains length and the elevation information of object outward appearance in visual angle, right side in right view; What obtain from below projection is projected as upward view, contains the length and width information of looking up object outward appearance in visual angle in upward view; What obtain from rear projection is projected as rear view, contains length and the elevation information of object outward appearance in visual angle, rear in rear view; Keep long between basic views and align, high concordant, wide equal corresponding relation, this obtains true man's foot surfaces dot information in the present invention, i.e. cloud data, and the basic premise carrying out three-dimensional reconstruction with this.

Because digitizing shoe tree does not need the colouring information of true man's foot surfaces point, the words that the cloud data Gaussian sphere coordinate of shoe tree is expressed, for (α, β, r), the point of intersection that the initial point of Gaussian sphere coordinate system is arranged on the optical axis of 5 video cameras is O, the Z axis of world coordinate system and the optical axis coincidence of real camera, the optical axis coincidence of 2 virtual video cameras in X-axis and Y-axis and 4 virtual video cameras; Shoe tree surface certain object point A is (x in the volume coordinate of cartesian coordinate system, y, z), A point is r to the distance of true origin O, angle between A point to the line and Z axis of true origin O is the angle between the projection line of the line of incident angle β, A point to true origin O in XY plane and X-axis is azimuth angle alpha;

Further, to illustrate on true man's foot surfaces the acquisition methods of cloud data a little.Suppose that the volume coordinate of a certain object point A of true man's foot surfaces is for (x, y, z), as shown in Figure 3; The so coordinate points (x, y) of the subpoint A1 of spatial point A in XOY plane necessarily in vertical view, as shown in Figure 4; The coordinate points (x, z) of subpoint A2 in XOZ plane necessarily in front view, as shown in Figure 5; The coordinate points (y, z) of subpoint A3 in YOZ plane necessarily in side view, as shown in Figure 6; According to above-mentioned constraint condition, by the mode traveled through in Gaussian sphere coordinate situation, with the surface of incident angle β, azimuth angle alpha and distance r whole true man's foot for variable travels through, obtain the cloud data of each object point in surface of true man's foot.

When traveling through whole true man's foot surfaces, first Gauss coordinate system to be set up, the initial point of coordinate system is set in the point of intersection of the optical axis of 5 video cameras, be XOZ plane in front view and rear view, be YOZ plane in left view and right view, be XOY plane in vertical view, be four individual by the cutting of true man's foot with XOZ plane and YOZ plane, be namely separated by 90 ° with azimuth angle alpha and carry out cutting; The scope traveling through view in different orientations α situation is obtained after carrying out such cutting, as shown in table 1;

The relation of the scope of table 1 azimuth angle alpha and the range of views of traversal

True man's foot surfaces shape measuring unit, three-dimensional measurement is carried out for true man's foot surfaces shape, obtain the three dimensional point cloud of true man's foot surfaces shape, and the three dimensional point cloud of true man's foot surfaces shape is written in measured database, to call when automatically generating stl file.

The marginal information of the true man's foot surfaces on different views is obtained in Iamge Segmentation, conversion and correcting unit; The marginal information of true man's foot surfaces in particular cases directly can be calculated 0 °, 90 °, 180 ° and 270 ° in azimuth angle alpha; But be the marginal information that can not directly obtain in one party parallactic angle situation when except above-mentioned angle, this searches the algorithm of surface voxel under just needing to realize a no special situation in the surface process of traversal true man foot; More succinct in order to illustrate when realizing searching the algorithm of surface voxel, here three kinds of situations are divided into by according to a point spatially and true man's foot surfaces point relation, a point on internal point representation space is in the inside of true man's foot, a point on external point representation space is not or not inside and the edge of true man's foot, and a point on marginal point representation space is just in time on the surface of true man's foot; These are put us and are referred to as voxel, and our target obtains when traveling through position angle and incident angle the voxel that true man's foot surfaces passes under this position angle and incident angle situation, as shown in Figure 8.

Use discrete voxel to represent the geological information of true man's foot surfaces in the present invention, evenly being cut by azimuthal and incident angle makes whole true man's foot surfaces be a series of small cubes space by subdivision equably, according to the two-dimensional geometry scope long aligned, high concordant, wide equal principle defines true man's foot surfaces voxel, estimating the longest estimated distance R in certain incident angle β and certain azimuth angle alpha situation _e-outthe shortest estimated distance R _e-inthe small cubes space of a rear formation visual shell, is herein referred to as minimum bounding box.The size of minimum bounding box is Δ β × Δ α × (R _e-out-R _e-in), at this moment unique needs determine it is at the longest estimated distance R _e-outthe shortest estimated distance R _e-inbetween the voxel that passes of true man's foot surfaces; Therefore, the key of problem is at the longest estimated distance R _e-outthe shortest estimated distance R _e-inbetween scope in how to detect the voxel that true man's foot surfaces passes efficiently and accurately, be employed herein space engraving algorithm, its core is this evaluating of colour consistency; Colour consistency constraint is a step very crucial in most of voxel colouring algorithm, if colour consistency principle thinks that a voxel belongs to true man's foot surfaces, then this voxel all should be identical to the pixel color in the projection of its visual picture.

Colour consistency is judged by the color variance and threshold value comparing voxel projected pixel collection, if Pro (v) is the set of pixels of voxel v back projection on K width relevant view, the colour consistency function that Con (v) is Pro (v), λ is given threshold value; If v RGB triple channel of projected pixel on view m is expressed as (R _m, G _m, B _m), m ∈ 1,2 ..., K}; Color variance formula (9) calculates;

$\left\{\begin{array}{c}{\mathrm{\&sigma;}}_x^2=\frac{1}{K-1}(\underset{m=1}{\overset{K}{\mathrm{\&Sigma;}}}{X}_m^2-\frac{1}{K}{\left(\underset{m=1}{\overset{K}{\mathrm{\&Sigma;}}}{X}_m\right)}^2);X\&Element;(R,G,B)\\ \mathrm{\&sigma;}\frac{1}{3}\sqrt{{\mathrm{\&sigma;}}_R^2+{\mathrm{\&sigma;}}_G^2+{\mathrm{\&sigma;}}_B^2}\end{array}\right.---\left(9\right)$

The colour consistency of voxel v is calculated with formula (10);

Con(v)＝σ-λ (10)

If Con (v) < 0, be then judged as that projected pixel collection meets colour consistency, this voxel is true man's foot surfaces voxel;

The concrete steps of ergodic algorithm are as follows:

STEP1: set up Gauss coordinate system, carries out cutting by the method shown in table 1 to true man's foot image file;

STEP2: initialization is carried out to incident angle β, azimuth angle alpha; Azimuth angle alpha=0 is set; Incident angle β=-90; Δ α=0.1; Δ β=0.1;

STEP3: determine the range of views traveled through according to azimuth angle alpha table 1;

STEP4: estimate the longest estimated distance R under certain incident angle β and certain azimuth angle alpha situation respectively with formula (4) and formula (5) _e-outthe shortest estimated distance R _e-in;

$R_{e-\mathrm{out}}=\sqrt{x_e^2+y_e^2+z_e^2}---\left(4\right)$

$R_{e-\mathrm{in}}=\min (\sqrt{x_e^2+y_e^2},\sqrt{x_e^2+z_e^2},\sqrt{z_e^2+y_e^2})---\left(5\right)$

In formula, x _e, y _e, z _edetermine by table 1 coordinate figure under traversal value incident angle β and azimuth angle alpha situation that the image in the range of views traveled through is obtained by rim detection; Obtain the bounding box Δ β × Δ α × (R of true man's foot surfaces voxel _e-out-R _e-in);

STEP5: from the shortest estimated distance R _e-instart traversal and search true man's foot surfaces voxel v _e, true man's foot surfaces voxel v _ejudgment mode carry out colour consistency detection by formula (10), if detected by colour consistency, then it can be used as the voxel v of true man's foot surfaces _e; If in bounding box, all voxels all do not meet formula (10) and carry out colour consistency detection in ergodic process, using the minimum value that calculates in formula (10) as true man's foot surfaces v _ecandidate; By the color average of set of pixels time computing formula (9) give this voxel, obtain the distance value r of this voxel to true origin simultaneously;

STEP6: (x, y, the z) value calculated under r with formula (6);

$\left\{\begin{array}{c}x=r\&times;\cos \mathrm{\&beta;}\&times;\cos \mathrm{\&alpha;}\\ y=r\&times;\cos \mathrm{\&beta;}\&times;\sin \mathrm{\&alpha;}\\ z=r\&times;\sin \mathrm{\&beta;}\end{array}\right.---\left(6\right)$

Calculate the cloud data of this true man's foot surfaces point preserve this cloud data;

STEP7: calculate β ← β+Δ β, judge β > 90, jumps to STEP4 if met;

STEP8: calculate α ← α+Δ α, judge α < 360, if met, β=-90; Jump to STEP3;

STEP9: terminate.

Automatic generation stl file unit, for automatically generating the three-dimensional picture file layout stl file meeting 3D and print and increase the service of material manufacturing technology according to the true man's foot surfaces three dimensional point cloud obtained;

STL form is the common interface format that current 3D printed/increased the use of material manufacturing equipment, it is the interface protocol formulated in 1988 by 3Dsystems company of the U.S., be the three-dimensional picture file layout that a kind of 3D of being prints and increasing material manufacturing technology is served, become the de facto standards form that 3D printed/increased material manufacture at present;

STL form is a kind of straightforward procedure storing three-dimensional model information, and the digital model of complexity is carried out approximate expression with a series of three-dimensional triangulation shape dough sheet by it; According to ASC II format characteristic of stl file, as shown in Figure 11; STL three-dimensional model file is made up of a series of tri patch, and each tri patch is made up of three summits corresponding in three dimensions and the normal vector of plane that forms thereof; The form of global variable is adopted to define summit and normal vector respectively; Minimum incident angleplied laminate and maximum incident angle layer outline data is obtained by the traversal of above-mentioned incident angle, the triangulation of outline data is obtained according to azimuthal traversal, then the tri patch of the profile of adjacent layer is connected, finally generate STL formatted file.

When actual 3D prints shoe tree, the shoe tree Material selec-tion ABS printed or nylon, automatically the STL formatted file generated is transferred to 3D printer by computing machine, 3D printer prints personalized shoe tree according to the 3D reconstruction result of the true man's foot copied, as shown in accompanying drawing 9 and accompanying drawing 10, accompanying drawing 9 illustrates shoe tree in print procedure, fig. 10 shows shoe tree and terminates in printing, really achieve shoe tree/true man's foot integrated treatment.

The design concept in rectangular-shaped bucket type mirror chamber as shown in Figure 12, in order to guarantee accurately to obtain the front view, left view, rear view and the right view that are reconstructed object in a high-definition camera simultaneously, in monocular multi-view stereoscopic vision case, the perspective projection point position of 4 virtual video cameras all must keep level consistent, and the spacious angle of the perspective projection of virtual video camera point position mainly with rectangular-shaped bucket type mirror chamber is relevant; By doing to analyze further to the imaging process of accompanying drawing 12, obtain the principle of design in rectangular-shaped bucket type mirror chamber; First in accompanying drawing 12, following parameter is defined: the central point of true high-definition camera is C, the central point of virtual video camera is D, true man's foot center is E, and the intersection point of straight line DE and minute surface is F, and the intersection point of straight line CD and minute surface extended line is G, ∠ CFE is γ, ∠ CFG is η, ∠ GFD is θ, line segment CE, namely the length of object distance is h, and the length of line segment EF is l; According to the catadioptric principle of minute surface, true high-definition camera is equal to the distance of minute surface with virtual video camera to the distance of minute surface, and namely straight line FG is the perpendicular bisector of line segment CD, therefore has η=θ; When virtual video camera is positioned at horizontal level shooting, there is straight line DE perpendicular to straight line CE, therefore can obtain the relational expression of formula (7):

tanγ＝h/l (7)

Utilize formula (7) with the η=θ obtained above and then can obtain:

Can know from the geometric relationship of accompanying drawing 12, θ is the spacious angle of minute surface; During design after the object distance of high-definition camera shooting and true man's foot are determined to the distance of level crossing, design the spacious angle in rectangular-shaped bucket type mirror chamber according to formula (7); The long limit of rectangle in rectangular-shaped bucket type mirror chamber is designed to L=L _footMax+ Δ L, wherein L _footMaxfor the length of maximum adult's foot, Δ L designs length in reserve; The rectangle minor face in rectangular-shaped bucket type mirror chamber is designed to W=W _footMax+ Δ W, wherein W _footMaxfor the width of maximum adult's foot, Δ W designs reserved width.

Rectangular white LED mixed light sources provides evenly soft light source for true man's foot throws light on; Design drawing as shown in Figure 13, evenly distributed rectangular white LED9 is fixed on the framework 10 of rectangular white LED mixed light sources facing to rectangular white LED mixed light sources inner circle camber mixed light face 4, produces a kind of evenly soft light source by the abundant mixing in rectangular white LED mixed light sources inner circle camber mixed light face 4; The physical dimension of the framework of rectangular white LED mixed light sources is consistent with the long limit of rectangle in rectangular-shaped bucket type mirror chamber and minor face.

Claims (9)

1. the true man's footwear type reproducing unit based on monocular multi-view machine vision, it is characterized in that, comprising for taking true man's foot at the monocular multi-view stereoscopic vision case of 5 different visual angles images and the computing machine realizing true man's foot three-dimensionalreconstruction and automatic generation 3D print file;

Be provided with the rectangular-shaped bucket die cavity of level crossing in described monocular multi-view stereoscopic vision case, this level crossing rectangular-shaped bucket die cavity is made up of 4 trapezoidal minute surfaces, and the up big and down small and minute surface of mirror body is towards inside cavities; The bottom of level crossing rectangular-shaped bucket die cavity is provided with as true man's foot provides the light source of even soft illumination; The video camera obtaining the multiple perspective images of true man's foot according to the catadioptric principle of minute surface is also provided with in monocular multi-view stereoscopic vision case;

Described computing machine comprises:

Monocular multi-view stereoscopic vision demarcates unit, for demarcating monocular multi-view stereoscopic vision case;

Iamge Segmentation, conversion and correcting unit, for being partitioned into true man's foot edge shape in the view from variant certain viewing angles, and utilize the true man foot image file of calibration result parameter to all visual angles to be normalized, make 5 different visual angles images meet projection relation;

Color image enhancement unit, for the brightness of image homogenising by true man's foot in 5 different visual angles images;

True man's foot surfaces shape measuring unit, carries out three-dimensional measurement for true man's foot surfaces shape, obtains the three dimensional point cloud of true man's foot surfaces shape;

Automatic generation stl file unit, for according to the three dimensional point cloud obtained, generates 3D print file.

2. as claimed in claim 1 based on true man's footwear type reproducing unit of monocular multi-view machine vision, it is characterized in that, also be provided with the second mirror surface of the axle center configuration at 45 ° of the rectangular-shaped bucket die cavity with level crossing in described monocular multi-view stereoscopic vision case, and with the first mirror surface of the optical axis of video camera configuration at 45 °; And the first described mirror surface and described the second mirror surface configuration in 90 °.

3. as claimed in claim 1 based on true man's footwear type reproducing unit of monocular multi-view machine vision, it is characterized in that, described light source comprises framework, is fixed on the inner circle camber mixed light face on framework, and contrasts the evenly distributed rectangular white LED in described inner circle camber mixed light face.

4., as claimed in claim 1 based on true man's footwear type reproducing unit of monocular multi-view machine vision, it is characterized in that, the principle of design of described level crossing rectangular-shaped bucket die cavity is as follows:

First following parameter is defined: the central point of video camera is C, the central point of virtual video camera is D, true man's foot center is E, and the intersection point of straight line DE and minute surface is F, and the intersection point of straight line CD and minute surface extended line is G, ∠ CFE is γ, ∠ CFG is η, ∠ GFD is θ, line segment CE, namely the length of object distance is h, and the length of line segment EF is l; According to the catadioptric principle of minute surface, video camera is equal to the distance of minute surface with virtual video camera to the distance of minute surface, and namely straight line FG is the perpendicular bisector of line segment CD, therefore has η=θ; When virtual video camera is positioned at horizontal level shooting, there is straight line DE perpendicular to straight line CE, therefore can obtain the relational expression of formula (7):

tanγ＝h/l (7)

Utilize formula (7) with the η=θ obtained above and then can obtain:

In formula, θ is the spacious angle of minute surface;

After the object distance of video camera shooting and true man's foot are determined to the distance of minute surface, come the spacious angle of design plane mirror rectangular-shaped bucket die cavity according to formula (8); The long limit of rectangle of level crossing rectangular-shaped bucket die cavity is L=L _footMax+ Δ L, wherein L _footMaxfor the length of maximum adult's foot, Δ L designs length in reserve; Rectangle minor face is designed to W=W _footMax+ Δ W, wherein W _footMaxfor the width of maximum adult's foot, Δ W designs reserved width.

5., based on a shoe tree manufacture method for true man's footwear type reproducing unit described in any one of Claims 1 to 4, it is characterized in that, comprising:

1) vertical view of monocular multi-view stereoscopic vision case acquisition true man foot, front view, left view, rear view and right view is utilized;

2) monocular multi-view stereoscopic vision case is demarcated;

3) from the view of variant certain viewing angles, be partitioned into true man's foot edge shape, and utilize the true man foot image file of calibration result parameter to all visual angles to be normalized, make 5 different visual angles images meet projection relation;

4) enhancing pre-service is carried out to the image in front view, left view, rear view and right view, make front view, left view, rear view consistent with the color of the image in right view and the color of vertical view;

5) according to the coordinate information of each object point in the surface of true man's foot in each view, obtain the cloud data of each object point of true man's foot surfaces, and generate 3D print file.

6. shoe tree manufacture method as claimed in claim 5, is characterized in that, in step 2) in, monocular multi-view stereoscopic vision case uses flat mirror reflects light, a width multiple views projected image of shooting comprises multiple images of scaling board; From flat mirror reflects principle, scaling board image is equivalent to same scaling board on diverse location, projects to several scaling board images that video camera projection plane obtains respectively; Use the multiple views projected image calibration for cameras inner parameter of a width scaling board, be equivalent to the proving operation using several scaling board images, comprise 5 width independence images of scaling board in image, obtain high-precision calibration result.

7. shoe tree manufacture method as claimed in claim 5, is characterized in that, described step 4) detailed process as follows:

With formula (1), the image of front view, left view, rear view and right view is transformed on HSI color space, $\left\{\begin{array}{c}H=\mathrm{ar}\cos \left\{\frac{[(R-G)+(R-B)]/2}{[{(R-G)}^2+(R-B){(G-B)]}^{\frac{1}{2}}}\right\}\\ S=1-\frac{3}{R+G+B}\left[\min (R,G,B)\right]\\ I=\frac{1}{3}(R+G+B)\end{array}\right.---\left(1\right)$

In formula, R, G, B are respectively the component of image on RGB color space of front view, left view, rear view and right view, and H, S, I are respectively the component of image on HSI color space of front view, left view, rear view and right view;

Then, the illumination intensity in different visual angles according to specular reflectance and lighting source, adjusts luminance component with formula (2),

I'＝I×ΔI (2)

In formula, Δ I be greater than 1 brightness adjustment coefficient, I' is the luminance component after brightness adjustment; The image of front view, left view, rear view and right view after then using formula (3) brightness to be strengthened transforms on RGB color space again, makes the true man's foot image file on each view have identical color and vein characteristic;

$\left\{\begin{array}{c}h=H\&times;\mathrm{\&pi;}/180;s=S/100;i=I^{\&prime;}/255\\ x=i\&times;(1-s)\\ y=i\&times;[1+\frac{s\&times;\cos \left(h\right)}{\cos (\mathrm{\&pi;}/3-h)}]\\ z=3i-(x+y)\\ r=y;g=z;b=\mathrm{x\; when\; h}<2\mathrm{\&pi;}/3\\ r=x;g=y;b=\mathrm{z\; when}2\mathrm{\&pi;}/3\&le;h<4\mathrm{\&pi;}/3,h=h-2\mathrm{\&pi;}/3\\ r=z;g=x;b=\mathrm{y\; when}4\mathrm{\&pi;}/3\&le;h<2\mathrm{\&pi;},h=h-4\mathrm{\&pi;}/3\end{array}\right.---\left(3\right)$

R, g, b that formula (3) calculates are the color components of normalized RGB color space, and the RGB color space that convert standard to also needs to be multiplied by 255 to the value of r, g, b, obtain brightness and strengthen rear R', G', B' value.

8. shoe tree manufacture method as claimed in claim 5, it is characterized in that, described cloud data acquisition process is as follows:

First Gauss coordinate system is set up, the initial point of coordinate system is set in the point of intersection of the optical axis of 5 video cameras, be XOZ plane in front view and rear view, be YOZ plane in left view and right view, it is XOY plane in vertical view, be four individual by the cutting of true man's foot with XOZ plane and YOZ plane, be namely separated by 90 ° with azimuth angle alpha and carry out cutting;

The scope traveling through view in different orientations α situation is obtained after carrying out such cutting:

α is 0 ° ~ 90 °, on the right of front view and the right view left side;

α is 90 ° ~ 180 °, on the right of the front view left side and left view;

α is 180 ° ~ 270 °, on the right of the left view left side and rear view;

α is 270 ° ~ 360 °, on the right of the rear view left side and right view;

Use discrete voxel to represent the geological information of true man's foot, evenly being cut by azimuthal and incident angle makes whole true man's foot be a series of small cubes space by subdivision equably, according to the two-dimensional geometry scope long aligned, high concordant, wide equal principle defines true man's foot surfaces voxel to be reconstructed, estimating the longest estimated distance R under certain incident angle β and certain azimuth angle alpha situation _e-outthe shortest estimated distance R _e-in, form the small cubes space of a visual outer shell process, be referred to as minimum bounding box; The size of minimum bounding box is Δ β × Δ α × (R _e-out-R _e-in); Adopt the voxel that the true man's foot surfaces to be reconstructed in space engraving algorithm detection minimum bounding box passes;

Colour consistency is judged by the color variance and threshold value comparing voxel projected pixel collection, if Pro (v) is the set of pixels of voxel v back projection on K width relevant view, the color compatibility function that Con (v) is Pro (v), λ is given threshold value; If v RGB triple channel of projected pixel on view m is expressed as (R _m, G _m, B _m), m ∈ 1,2 ..., K}; Color variance formula (9) calculates;

$\left\{\begin{array}{c}{\mathrm{\&sigma;}}_x^2=\frac{1}{K-1}(\underset{m=1}{\overset{K}{\mathrm{\&Sigma;}}}{X}_m^2-\frac{1}{K}{\left(\underset{m=1}{\overset{K}{\mathrm{\&Sigma;}}}{X}_m\right)}^2);X\&Element;(R,G,B)\\ \mathrm{\&sigma;}=\frac{1}{3}\sqrt{{\mathrm{\&sigma;}}_R^2+{\mathrm{\&sigma;}}_G^2+{\mathrm{\&sigma;}}_B^2}\end{array}\right.---\left(9\right)$

The colour consistency of voxel v is calculated with formula (10);

Con(v)＝σ-λ (10)

If Con (v) < 0, be then judged as that projected pixel collection meets colour consistency, this voxel is true man's foot surfaces voxel.

9. shoe tree manufacture method as claimed in claim 8, is characterized in that, described space engraving algorithm, and the concrete steps of its ergodic algorithm are as follows:

STEP1: set up Gauss coordinate system, carries out cutting by azimuth angle alpha to true man's foot image file;

STEP2: initialization is carried out to incident angle β, azimuth angle alpha; Azimuth angle alpha=0 is set; Incident angle β=-90; Δ α=0.1; Δ β=0.1;

STEP3: the range of views determined according to azimuth angle alpha;

STEP4: estimate the longest estimated distance R under certain incident angle β and certain azimuth angle alpha situation respectively with formula (4) and formula (5) _e-outthe shortest estimated distance R _e-in;

$R_{e-\mathrm{out}}=\sqrt{x_e^2+y_e^2+z_e^2}---\left(4\right)$

$R_{e-\mathrm{in}}=\min (\sqrt{x_e^2+y_e^2},\sqrt{x_e^2+z_e^2},\sqrt{z_e^2+y_e^2})---\left(5\right)$

In formula, x _e, y _e, z _eit is the coordinate figure under traversal value incident angle β and azimuth angle alpha situation that the image in the range of views of traversal is obtained by rim detection; Obtain the bounding box Δ β × Δ α × (R of true man's foot surfaces voxel _e-out-R _e-in);

STEP5: from the shortest estimated distance R _e-instart traversal and search true man's foot surfaces voxel v _e, true man's foot surfaces voxel v _ejudgment mode carry out colour consistency detection by formula (10), if detected by colour consistency, then it can be used as the voxel v of true man's foot surfaces _e; If in bounding box, all voxels all do not meet formula (10) and carry out colour consistency detection in ergodic process, using the minimum value that calculates in formula (10) as true man's foot surfaces v _ecandidate; By the color average of set of pixels time computing formula (9) give this voxel, obtain the distance value r of this voxel to true origin simultaneously;

STEP6: (x, y, the z) value calculated under r with formula (6);

$\left\{\begin{array}{c}x=r\&times;\cos \mathrm{\&beta;}\&times;\cos \mathrm{\&alpha;}\\ y=r\&times;\cos \mathrm{\&beta;}\&times;\sin \mathrm{\&alpha;}\\ z=r\&times;\sin \mathrm{\&beta;}\end{array}\right.---\left(6\right)$

Calculate the cloud data of this true man's foot surfaces point preserve this cloud data;

STEP7: calculate β ← β+Δ β, judge β > 90, jumps to STEP4 if met;

STEP8: calculate α ← α+Δ α, judge α < 360, if met, β=-90; Jump to STEP3;

STEP9: terminate.