CN1332430A - 3D tracking and measurement method of moving objects by 2D code - Google Patents

Abstract

Translated fromChinese

本发明涉及光学三维轮廓测量技术领域,它是用二维编码对运动物体进行三维跟踪测量的方法。用两个摄像机对被测物体同时进行图像采集,其中两个CCD摄像机之间保持一定的距离,利用人眼的双目立体视觉的原理进行三维测量,采用据有特定的排列次序的不同颜色的几何形状在被测物体表面做出标记,使两个CCD摄像机获得的图像之间容易找到匹配点。根据同一匹配点在两幅图像中的视差计算出该点在三维空间中的位置。本发明可以对人体的运动行为进行跟踪测量。在计算机三维动画制作,运动员的选拔与训练等领域有着广阔的应用前景。

The invention relates to the technical field of optical three-dimensional profile measurement, which is a method for three-dimensional tracking and measurement of moving objects by using two-dimensional codes. Use two cameras to collect images of the measured object at the same time, and keep a certain distance between the two CCD cameras, use the principle of binocular stereo vision of the human eye to perform three-dimensional measurement, and use different colors in a specific order The geometric shape marks the surface of the measured object, making it easy to find matching points between the images obtained by the two CCD cameras. Based on the disparity of the same matching point in the two images, the position of the point in 3D space is calculated. The invention can track and measure the movement behavior of the human body. It has broad application prospects in the fields of computer three-dimensional animation production, athlete selection and training, etc.

Description

With two-dimensional encoded method of moving object being carried out three-dimensional tracking measurement

Technical field

The present invention relates to the optical three-dimensional contour field of measuring technique.It is the two-dimensional encoded method of moving object being carried out three-dimensional tracking measurement of a kind of usefulness.

Technical background

Utilize the principle of binocular stereo vision to carry out three-dimensional observation and measure existing extensive studies and application.Many three-dimensional observations and measuring method have been proposed at present.Chinese patent CN1010466A is disclosed to be the striation band pattern that has monodrome coding with a series of, with and the optical axis of gamma camera between certain inclination angle project on the target, and accept the coded light signal that target reflects by the pixel in the gamma camera.Scale by suitable selection and coordinate system is chosen, and relevant striation band coding, to each pixel unit, just can measure the distribution of target by a coordinate that deducts this pixel from the binary coding light signal that reflexes to pixel simply.CN1230254A disclosed a kind of by optical imagery catch, projection pattern and trigonometric calculations carry out the method and apparatus that objective is measured.Graphic pattern projection device and image capture apparatus design respectively.In measuring process, can be located separately and be guided.CN1231724A has disclosed a kind of method that optical recording, projection sample and triangulation improve the conspicuousness of calculating the objective measurement of passing through, for this coded samples is projected on the target, to prevent repeatability during evaluation of image data.The measuring method that CN1141427A provides is pickup image to tested moving object marking the time by synchronous clock commander, identification marking and calculate the coordinate of marker characteristic point then calculates the testee movement locus, again along the length of track and current parameters such as movement velocity.

All adopted the method to testee projection specific pattern in above-mentioned first three kind document, this method can obtain the three-D profile of object or improve the conspicuousness of three dimensional viewing.But can't carry out three-dimensional to moving object follows the tracks of.Disclosed in last piece document is marking in moving object, calculates the movement locus of object by identification marking.This method can only be followed the tracks of the motion of simple objects, for some complicated moving object, then can't measure as the motion of human body.

Summary of the invention

The purpose of this invention is to provide that a kind of usefulness is two-dimensional encoded carries out the method for three-dimensional tracking measurement to moving object, it is that the human body that solves moving object, particularly motion carries out the problem of three-D profile tracking measurement.The present invention carries out image acquisition with two video cameras simultaneously to testee, wherein maintain a certain distance between two ccd video cameras, utilize the principle of the binocular stereo vision of human eye to carry out three-dimensional measurement, the geometric figure that employing is then arranged according to certain Regulations is formed the surface of two-dimensional encoded plane attached to testee, can find match point according to queueing discipline between the image that two video cameras are obtained.Calculate this position in three dimensions according to the parallaxometer of same match point in two width of cloth images.

The ultimate principle of utilizing binocular stereo vision to carry out three-dimensional measurement is a triangulation method.When about after the relative position of two video cameras determines, according to the same point on the testee about in two video cameras the difference of image space just can calculate the locus of this point.In computer vision field, generally adopt the characteristic matching mode in order to guarantee measuring accuracy, promptly select the unique point on the object, calculate as angle point, boundary line etc.But to relatively slicker and more sly object, for example human body is looked for quite difficulty of match point.The present invention adopts the geometric figure with obvious characteristic to rearrange two-dimensional encoded plane according to certain rule on two dimensional surface, and with two-dimensional encoded plane attached to the testee surface, thereby formed a series of match point on the testee surface.This just makes and can find match point according to queueing discipline between the image of two video cameras acquisitions.

The geometric figure that is used to encode can be that square, triangle, hexagon, circle or other have the geometric figure of obvious characteristic.As long as the figure that computing machine can accurately be discerned all can adopt.Here it should be noted, because the geometric figure on the testee often becomes certain included angle with the plane of delineation, its projection meeting in image deforms geometric figure, for example, square is along with the difference of projection angle can be rectangle, rhombus etc., if a foursquare part is blocked, just might be trapezoidal or triangle.Therefore, select different geometric figures to carry out to avoid using the geometric figure of obscuring easily when two-dimensional encoded.The geometric figure that coding is adopted preferably can be desired consistent with three-dimensional reconstruction software.Can make data processing convenient like this.

In the image with the colored CCD acquisition, three kinds of colors of red, green, blue are independent mutually, are called three primary colors.When active computer carried out Flame Image Process, we can separate these three kinds of chromatic zoneses at an easy rate, thereby color characteristic is than the easier identification of geometric properties.We form one group with adjacent geometric figure in order to obtain big code plane with limited color, and the ordering in each group is all inequality.For example, encode with three kinds of colors of RGB, 4 geometric figures are one group, according to the permutation and combination principle, 81 kinds of different permutation and combination methods can be arranged altogether.Arrangement mode in each group can be that the geometric figure of delegation is formed one group, also can be that the geometric figure of row is formed one group, can also be that adjacent several geometric figures are formed one group.Geometric number can be determined according to the real needs of measuring in each group.Can both be determined uniquely that in order to make in the code plane each match point in a code period, the ordering of each group only occurs once.

In actual measurement, we can not get single redness, green or blue.Because a variety of causes always has some parasitic lights.In measurement, belong to noise.In order to make measuring system have certain anti-interference capability, make the difference between each color reach maximum as far as possible.Under 24 color modes, three kinds of each colors of color of RGB can have 256 GTGs.For the difference that makes color reaches maximum, each color is only got 0 and 255 two value.At this moment make up with three kinds of colors, can obtain white, red, green, blue, and their complementary color is black, green grass or young crops, product, Huang.If do back of the body background color, also have 7 kinds of colors can be used for encoding with black.

Under many circumstances, do not need very big code plane, and need decoding speed faster.At this moment can adopt three kinds of colors of RGB to encode, perhaps adopt four kinds of colors of red, green, blue and white to encode.Only encode with these several colors, fastest when carrying out Computer Processing.

Geometric number is many more in each group, and is just many more according to its arrangement mode of permutation and combination principle, and the two-dimensional encoded plane of being formed is just big more.But will consider also that simultaneously if there is the part element to be blocked in one group, can't determine uniquely that for the encoded radio of this group this just causes the blind area of measurement, geometric number is many more in one group, the scope of blocking the blind area of being caused is also just big more.When dynamic human body was measured, people's limbs blocked certain part of trunk through regular meeting, if block the excessive result that will influence measurement in the blind area of causing.

To block the blind area of being caused and reduce to minimum in order to make, the present invention forms one group with four adjacent squares, and when carrying out encoding process, the square on each non-edge all is an element in adjacent four groups simultaneously, as shown in Figure 1.Wherein R, G, B represent the square of three kinds of colors of red, green, blue respectively.With any 1 G is in the code planes (Fig. 1 e) formed of 3 * 3 at center square, can regard as by four 2 * 2 coding unit is staggered and form.Wherein, first group (Fig. 1 a) in the element in the lower right corner simultaneously also be the element in the lower left corner in second group (Fig. 1 b), it still be the element in upper right corner in the 3rd group (Fig. 1 c) and the element in the 4th group (Fig. 1 d) middle upper left corner.Encode in the two-dimensional encoded plane that obtains according to this rule, four adjacent elements can be formed one group arbitrarily, and their ordering is unique in a code period.When carrying out encoding process according to this rule, any one element is subjected to blocking and can have any impact to other element.

Encode according to above rule with three kinds of colors of red, green, blue, can form one 9 * 9 code plane, as shown in Figure 2; When the testee surface area is big, can adopt more color to encode, also can be spliced into bigger code plane with the several encoding cycle.

Concrete measuring process of the present invention is described in detail as follows:

1) makes the color 2 D code plane according to two-dimensional encoded ordering shown in Figure 2 with the square of red green blue three kinds of colors.If human body is measured, then make the coding clothes.

2) with the surface of color 2 D code plane attached to testee.If human body is measured, the clothes of then will encoding are through on measured's the body.

3) form three-dimensional tracing measurement system with two video cameras.The relative position of two video cameras remains unchanged, but can whole move and object is carried out tracking measurement.

4) before actual measurement, system is proofreaied and correct with the object of known three-D profile.Tracing measurement system through overcorrect will keep constant in whole measuring process.

5) with three-dimensional tracing measurement system measuring to the foregoing three-dimensional body that has adhered to the color 2 D code plane.

6) image of measuring is sent into computing machine.

7) match point about computing machine is found out by image processing program in two width of cloth images, and calculate the pairing locus of match point according to the parallaxometer of match point in two width of cloth images.

8) according to the position of the spatial point that measures three-dimensional model with 3 d modeling software reconstruction of objects in computing machine.The difficulty of obtaining object dimensional information maximum with the principle of binocular vision is to look for match point.The present invention adopts color 2 D plane geometric figure Methods for Coding to solve this difficult problem effectively.The clothes made from color 2 D plane geometric figure coding are through the person when going up, and not only can measure the three-D profile of human body, and can carry out tracking measurement to the motor behavior of human body.Because whole measuring process is to carry out under white light conditions, thereby can adapt to the condition of various in-site measurements.The present invention makes at computer three-dimensional animation, and there is wide application prospect in fields such as athletic choice and training.

Description of drawings

The two-dimensional encoded unit figure of Fig. 1 RGB.

The two-dimensional encoded Pareto diagram of Fig. 2.

The installation drawing of Fig. 3 measuring system, 1-left side video camera, the right video camera of 2-, 3-testee, 4-world coordinate system x direction of principal axis.

The cylindrical left video camera experiment measuring figure of Fig. 4.

The cylindrical right video camera experiment measuring figure of Fig. 5.

Fig. 6 cylindrical measurement is three-dimensional reconstruction figure as a result.

The left video camera experiment measuring figure of Fig. 7 manikin.

The left video camera experiment measuring figure of Fig. 8 manikin.

Fig. 9 manikin measurement result three-dimensional reconstruction figure.

Embodiment

Embodiment 1

At first make the color 2 D code plane according to two-dimensional encoded ordering shown in Figure 2 with the square of three kinds of colors of RGB with computer drawing software, and make two-dimensional encoded plane with color printer output, the wherein foursquare length of side is 10mm, and the interval between each square also is 10mm.Is 135mm with two-dimensional encoded plane attached to diameter, on the right cylinder of high 200mm.Form measuring system with two video cameras according to device shown in Figure 3.Two video cameras are placed on 1,2 position respectively, and make them be in same level height.Video output signals is sent into computing machine by image pick-up card, and the resolution of image pick-up card is 688 * 516 pixels, 24 true color patterns.About making as far as possible, the configuration of light throws light on evenly, to reduce the shade on the testee.

Before carrying out actual measurement, proofread and correct measuring system.Because cylindrical radius and height all are known, can be used for system is proofreaied and correct.Respectively right cylinder is sampled and send into computing machine through image pick-up card withvideo camera 1,2, the sample graph that obtains as shown in Figure 4 and Figure 5.In computing machine, by image processing software each foursquare color is discerned, and obtained this foursquare encoded radio according to the color of its adjacent square.

On two-dimensional encoded plane, each foursquare four angle can be used as four match points.When measurement image is handled, accurately determine the position of each angle point, and determine its position in the space by the parallax of each angle point in the image of the left and right sides.

Fig. 6 is the spatial point that records according to experiment, carries out the result of three-dimensional reconstruction in computing machine with VRML2.0.The benefit of carrying out three-dimensional reconstruction with VRML2.0 is can carry out three-dimensional by browser to browse.From the effect of three-dimensional reconstruction, basic is consistent with actual object.

Embodiment 2

Measuring system is proofreaied and correct the back throughembodiment 1 and is kept constant, and we carry out three-dimensional measurement to manikin.Square with three kinds of colors of red, green, blue is attached to the manikin surface according to ordering shown in Figure 2, with measuring system the model is measured.Fig. 7, Fig. 8 are respectively the sample graph that left and right cameras obtains.By computer programs process, can obtain the volume coordinate of four angle points of each square.According to the spatial point that experiment records, in computing machine, carry out three-dimensional reconstruction with VRML2.0.Fig. 9 is the result by manikin three-dimensional measurement value three-dimensional reconstruction in computing machine.

Claims (7)

Translated fromChinese

1.一种用二维编码对运动物体进行三维跟踪测量的方法，其特征在于用两个摄像机对被测物体同时进行图像采集，利用双目立体视觉的原理进行三维测量，采用具有明显特征的几何图形在二维平面上进行排列组成二维编码平面，将二维编码平面附着在被测物体表面，两个摄像机根据排列的二维编码平面获得图像之间的匹配点。1. A method for carrying out three-dimensional tracking and measurement of a moving object with two-dimensional coding, characterized in that two cameras are used to collect images of the object to be measured simultaneously, utilize the principle of binocular stereo vision to carry out three-dimensional measurement, and adopt a distinctive feature The geometric figures are arranged on a two-dimensional plane to form a two-dimensional coding plane, and the two-dimensional coding plane is attached to the surface of the measured object, and the two cameras obtain matching points between images according to the arranged two-dimensional coding plane.2.根据权利要求1所述的用二维编码对运动物体进行三维跟踪测量的方法，其特征在于所述的几何图形是正方形、三角形、六边形或其它具有可识别特征的图形。2. The method for three-dimensional tracking and measurement of moving objects with two-dimensional coding according to claim 1, characterized in that said geometric figures are squares, triangles, hexagons or other figures with identifiable features.3.根据权利要求2所述的用二维编码对运动物体进行三维跟踪测量的方法，其特征在于所述的几何图形具有颜色进行排列。3. The method for three-dimensional tracking and measurement of moving objects with two-dimensional coding according to claim 2, characterized in that said geometric figures are arranged in colors.4.根据权利要求1所述的用二维编码对运动物体进行三维跟踪测量的方法，其特征在于若干个相邻的几何图形组成一组，每一组内的排列次序各不相同。4. The method for three-dimensional tracking and measurement of moving objects with two-dimensional coding according to claim 1, characterized in that several adjacent geometric figures form a group, and the order of arrangement in each group is different.5.根据权利要求4所述的用二维编码对运动物体进行三维跟踪测量的方法，其特征在于所述的几何图形每一组的排列次序在一个编码周期范围内是唯一的。5. The method for three-dimensional tracking and measurement of moving objects by two-dimensional coding according to claim 4, characterized in that the sequence of each group of said geometric figures is unique within one coding period.6.根据权利要求3所述的用二维编码对运动物体进行三维跟踪测量的方法，其特征在于所述的几何图形采用分立颜色进行排列。6. The method for three-dimensional tracking and measurement of moving objects with two-dimensional coding according to claim 3, characterized in that said geometric figures are arranged in discrete colors.7.根据权利要求1至6所述的用二维编码对运动物体进行三维跟踪测量的方法，其特征在于所述的几何图形是由四个不同颜色的相邻正方形组成一组，每一个非边缘上的正方形都同时是相邻四个组中的一个元素。7. The method for carrying out three-dimensional tracking and measurement of moving objects with two-dimensional coding according to claims 1 to 6, characterized in that the geometric figure is composed of four adjacent squares of different colors, each of which is not The squares on the edge are all simultaneously an element in four adjacent groups.

Images