CN102621546A

CN102621546A - 3D Information Acquisition Method Based on Correlation Imaging

Info

Publication number: CN102621546A
Application number: CN201210085368.3A
Authority: CN
Inventors: 刘旭; 李海峰; 张硕; 王杰; 王金成; 李东
Original assignee: Zhejiang University ZJU
Current assignee: Zhejiang University ZJU
Priority date: 2012-03-28
Filing date: 2012-03-28
Publication date: 2012-08-01
Anticipated expiration: 2032-03-28
Also published as: CN102621546B

Abstract

本发明公开了一种基于关联成像的三维信息获取方法，利用光源发出探测光，探测光经过空间光调制器处理后投射至目标物体，检测来自目标物体的反射光，并由预处理单元进行两次调制，分别进行存储，经过对缓存器中存储的随机图像和调制后的光强信息进行计算，从而获得目标物体的三维信息。本发明提出基于关联成像的三维信息获取方法，对于三维物体，在获取光强信号列之后，对其进行调制，仅需要两次二维信息的计算就可以获得三维信息，适用于远距离、快速的三维信息获取。The invention discloses a three-dimensional information acquisition method based on correlation imaging. A light source is used to emit detection light. The detection light is processed by a spatial light modulator and then projected to a target object. The reflected light from the target object is detected, and a preprocessing unit performs two The sub-modulation is stored separately, and the three-dimensional information of the target object is obtained by calculating the random image stored in the buffer and the modulated light intensity information. The present invention proposes a three-dimensional information acquisition method based on associated imaging. For a three-dimensional object, after obtaining the light intensity signal sequence, it is modulated, and only two calculations of two-dimensional information are required to obtain three-dimensional information, which is suitable for long-distance, fast 3D information acquisition.

Description

3D Information Acquisition Method Based on Correlation Imaging

技术领域 technical field

本发明属于三维信息获取领域，具体是一种基于关联成像原理的三维信息采集和处理方法，可用于远距离的三维成像。The invention belongs to the field of three-dimensional information acquisition, in particular to a three-dimensional information acquisition and processing method based on the principle of correlation imaging, which can be used for long-distance three-dimensional imaging.

背景技术 Background technique

三维信息的获取如今已经广泛用于三维建模，目标探测等领域，并且已经发展出来各种各样的方法，各有其优缺点及使用范围。如今比较流行的有基于计算机视觉的方法，可以从二维的图形获取三维信息，可以实现对大视场三维信息的获取；利用飞行时间法，在一定范围内对目标物体进行扫描，在获取二维信息的同时得到探测脉冲的飞行时间，从而获取目标物体的深度信息，其图像获取速度和分辨率受限于扫描速度和扫描点数；使用接触探针测量，可以对于较小体积的物体实现精确的测量，但是设备较为昂贵，获取速度慢，并且容易对目标物体的表面产生破坏；使用结构光对于目标物体进行主动式的测量，扫描速度快，测量精度高，比较适用于室内物体表面反射情况比较好的场合。The acquisition of 3D information has been widely used in 3D modeling, target detection and other fields, and various methods have been developed, each with its own advantages and disadvantages and scope of application. Nowadays, there are more popular computer vision-based methods, which can obtain three-dimensional information from two-dimensional graphics, and can realize the acquisition of three-dimensional information in a large field of view; use the time-of-flight method to scan the target object within a certain range, and obtain two-dimensional information. The time-of-flight of the detection pulse can be obtained at the same time as the three-dimensional information, so as to obtain the depth information of the target object. The image acquisition speed and resolution are limited by the scanning speed and the number of scanning points; the use of contact probe measurement can achieve accurate detection of smaller volume objects. measurement, but the equipment is expensive, the acquisition speed is slow, and it is easy to cause damage to the surface of the target object; the use of structured light for active measurement of the target object has fast scanning speed and high measurement accuracy, which is more suitable for indoor object surface reflection Better occasions.

基于计算机视觉的方法、探针接触测量法、结构光法都不适用于远距离三维信息的获取，而飞行时间法受限于扫描速度和扫描点数，并且容易受到传播过程中环境的影响，远距离三维信息的获取结果也不理想，而近些年发展出来的关联成像理论，对于远距离物体的三维信息的获取，具有其独特的优势。Methods based on computer vision, probe contact measurement, and structured light methods are not suitable for the acquisition of long-distance 3D information, while the time-of-flight method is limited by the scanning speed and number of scanning points, and is easily affected by the environment during the propagation process. The results of obtaining three-dimensional information at a distance are not ideal, and the correlative imaging theory developed in recent years has its unique advantages in obtaining three-dimensional information of distant objects.

关联成像算法，具有成像距离远，对于环境噪声的鲁棒性强，使用光谱范围广的优势。The correlative imaging algorithm has the advantages of long imaging distance, strong robustness to environmental noise, and wide spectral range.

而压缩感知算法，将数据采集和压缩相结合，可以在远低于奈奎斯特采样频率的情况下重构出目标物体的图像。对于长度为N的实数信号x(n)，可以进行稀疏变换

或者x＝Ψθ。Ψ为相应的稀疏基矩阵。压缩感知并不直接对信号x(n)进行测量，而是通过一个随机投影矩阵Φ进行测量y＝Φx。Φ是一个M×N维的矩阵，每一行是一个基向量表示对信号x(n)进行一次线性的测量。M表示测量次数，并且满足M＜N。由于x可以在Ψ域进行稀疏表示，所以上式也可以表示为y＝Φx＝ΦΨ^Tθ。求解此方程的问题可以表示为求最小1范数的优化问题：

subject to y＝Φx＝ΦΨ^Tθ＝Θθ。可用的算法有基追踪算法，贪婪追踪算法，凸松弛法，组合算法等。The compressed sensing algorithm, which combines data acquisition and compression, can reconstruct the image of the target object at a rate much lower than the Nyquist sampling frequency. For a real signal x(n) of length N, sparse transformation can be performed

Or x = Ψθ. Ψ is the corresponding sparse basis matrix. Compressed sensing does not measure the signal x(n) directly, but measures y=Φx through a random projection matrix Φ. Φ is an M×N-dimensional matrix, each row is a basis vector Represents a linear measurement of the signal x(n). M represents the number of measurements, and M<N is satisfied. Since x can be represented sparsely in the Ψ domain, the above formula can also be expressed as y=Φx=ΦΨ ^T θ. The problem of solving this equation can be expressed as an optimization problem of finding the minimum 1-norm:

subject to y=Φx=ΦΨ ^T θ=Θθ. The available algorithms are basis pursuit algorithm, greedy pursuit algorithm, convex relaxation method, combinatorial algorithm, etc.

发明内容 Contents of the invention

本发明提出基于关联成像的一种新型三维信息的获取方法，对于三维物体，在获取光强信号列之后，对其进行调制，仅需要两次二维信息的计算就可以获得三维信息。The present invention proposes a new method for obtaining three-dimensional information based on correlation imaging. For three-dimensional objects, after obtaining the light intensity signal sequence, it is modulated, and only two calculations of two-dimensional information are required to obtain three-dimensional information.

一种基于关联成像的三维信息获取方法，包括利用光源发出探测光，探测光经过空间光调制器处理后投射至目标物体，检测来自目标物体的反射光，根据该反射光确定目标物体的三维信息，具体包括如下步骤：A method for obtaining three-dimensional information based on correlated imaging, including using a light source to emit detection light, the detection light is processed by a spatial light modulator and then projected to a target object, detecting reflected light from the target object, and determining the three-dimensional information of the target object according to the reflected light , including the following steps:

(1)设定随机图像的总像素数N，设定两个光强信号的调制函数f(x)和f′(x)；(1) Set the total number of pixels N of the random image, set the modulation functions f(x) and f'(x) of the two light intensity signals;

N为随机图像的总像素数；x泛指调制函数中的变量。N is the total number of pixels of the random image; x generally refers to the variable in the modulation function.

在设定随机图像的总像素数N时，主要根据目标物体的分辨率要求和空间光调制器规格设定。When setting the total number of pixels N of the random image, it is mainly set according to the resolution requirement of the target object and the specification of the spatial light modulator.

(2)设定目标物体的检测次数M，并生成一个M×N维的观测矩阵，利用该观测矩阵生成M幅随机图像，对于第i幅随机图像，可以表示为R_i，其中i≤M，每一幅随机图像的总像素数为N，R_i，n表示第i幅随机图像第n个像素点的灰度值，其中n≤N；(2) Set the detection times M of the target object, and generate an M×N-dimensional observation matrix, use the observation matrix to generate M random images, for the i-th random image, it can be expressed as R _i , where i≤M , the total number of pixels of each random image is N, R _{i, n} represents the gray value of the nth pixel of the i random image, where n≤N;

检测次数M根据所需要的测量精度确定，要求精度越高，则M值越大，观测矩阵可使用经过处理的满足伯努利分布的-1，1随机矩阵，或满足高斯分布的-1～1随机矩阵，或随机采样的Hadamard矩阵。The number of detections M is determined according to the required measurement accuracy. The higher the accuracy required, the greater the value of M. The observation matrix can be a processed -1, 1 random matrix that satisfies the Bernoulli distribution, or -1~1 that satisfies the Gaussian distribution. 1 random matrix, or randomly sampled Hadamard matrix.

生成M幅随机图像的一般做法是提取观测矩阵的每一行，对应的转换为一个二维的随机图像，共M行总计对应M幅随机图像。The general method of generating M random images is to extract each row of the observation matrix, and convert it into a two-dimensional random image, and a total of M rows correspond to M random images.

(3)空间光调制器依次向光源发出的探测光中加载M幅随机图像中的一幅，得到带有随机图像信息的探测光，而后投射至目标物体；(3) The spatial light modulator sequentially loads one of the M random images into the probe light emitted by the light source to obtain the probe light with random image information, and then project it to the target object;

(4)检测来自目标物体的反射光的光强，每次检测时间为T，在检测时间T内得到信号列s₁，其中，第j个数据可以记为s_1，j，其中1表示第副随机图像，即第一次测量，j为信号列内各个数据点的序号；(4) Detect the light intensity of the reflected light from the target object. Each detection time is T, and the signal sequence s ₁ is obtained within the detection time T. Among them, the jth data can be recorded as s _1,j , where 1 represents the first Secondary random image, that is, the first measurement, j is the serial number of each data point in the signal column;

不同的检测时间T之间可以通过预定的触发信号或来自脉冲激光器的信号进行切换或触发。Different detection times T can be switched or triggered by a predetermined trigger signal or a signal from a pulsed laser.

(5)在信号列数据s_1，j中取有效范围内的数据点，所述的有效范围为：j1≤j≤j2，对所得到的有效范围内的数据点进行调制，得到调制后的信号列；(5) Take the data points in the effective range in the signal sequence data s1 _{, j} , the described effective range is: j1≤j≤j2, modulate the data points in the obtained effective range to obtain the modulated signal column;

调制时设Y′(1)＝0，Y″(1)＝0，当j＝j1的时候开始，进行Y′(1)＝Y′(1)+f(d_j)×s_1，j，Y″(1)＝Y″(1)+f′(d_j)×s_1，j运算，直到j＝j2为止；When modulating, set Y'(1)=0, Y"(1)=0, start when j=j1, and perform Y'(1)=Y'(1)+f(d _j )×s _{1, j} , Y "(1)=Y "(1)+f'(d _j )×s _{1, j} operation, until j=j2;

d_j＝c(Δt·j+Δt′)，j1≤j≤j2；d _j = c(Δt·j+Δt′), j1≤j≤j2;

Δt为探测器每次采集信号的间隔时间；Δt is the interval time of each signal acquisition by the detector;

Δt′为探测器的延迟时间；Δt' is the delay time of the detector;

c为光速；c is the speed of light;

j为信号列内各个数据点的序号；j is the serial number of each data point in the signal column;

j1和j2为有效范围内的数据点序号阈值；j1 and j2 are the data point serial number thresholds within the valid range;

f和f′为步骤(1)所述的调制函数；f and f' are the modulation functions described in step (1);

T时间内由于探测器会多次间隔的进行信号采集，因此该信号列s_1，j内包含了多个信号(数据点)，本发明仅取用有效范围内的数据点，符合条件j1≤j≤j2的j认为是有效信号。During the T time, since the detector will collect signals at multiple intervals, the signal column s _{1, j} contains multiple signals (data points), and the present invention only uses data points within the effective range, which meets the condition that j1≤ j ≤ j2 is considered to be a valid signal.

例如j的最大值为1000，则意味着在T时间内探测器采集了1000个数据点，若j1和j2分别设定为100和500，则有效范围内的数据点从第100个数据点开始，直至第500个数据点。For example, the maximum value of j is 1000, which means that the detector collects 1000 data points within T time, if j1 and j2 are set to 100 and 500 respectively, the data points within the effective range start from the 100th data point , until the 500th data point.

由于三维的目标物体的不同部位与探测器的距离不同，因此不同的序号的数据点对应来自于不同距离d的光信号。Since different parts of the three-dimensional target object have different distances from the detector, data points with different serial numbers correspond to optical signals from different distances d.

(6)循环步骤(3)～步骤(5)直至M幅随机图像都发送完毕，对于第i副随机图像，对应得到信号列s_i，其中，第j个数据可以记为s_i，j，i为随机图像的序号；并进行Y′(i)＝Y′(i)+f(d_j)×s_i，j，Y″(i)＝Y″(i)+f′(d_j)×s_i，j的调制，完成M次检测，对应的得到M个调制后的信号列且表示为Y′和Y″：(6) Repeat step (3) to step (5) until all M random images are sent. For the i-th random image, the corresponding signal sequence s _i is obtained, where the j-th data can be recorded as s _{i, j} , i is the serial number of the random image; and Y′(i)=Y′(i)+f(d _j )×s _{i, j} , Y″(i)=Y″(i)+f′(d _j ) The modulation of ×s _{i, j} completes M times of detection, and correspondingly obtains M modulated signal columns, which are denoted as Y′ and Y″:

${Y Y}^{' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \end{matrix});;$

${Y Y}^{' '' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \end{matrix});;$

Y′和Y″中，每一行对应了一个步骤(5)中所述的调制后的信号列，由于是M次检测，因此Y′和Y″中均为M行；In Y' and Y ", each row corresponds to a modulated signal column described in step (5), because it is M times of detection, so Y' and Y " are M rows;

(7)对调制后的数据列Y′进行关联成像或者压缩感知的计算，得到包含有目标物体纹理信息和距离信息的数据列X′；(7) Perform associated imaging or compressed sensing calculations on the modulated data column Y' to obtain a data column X' containing the texture information and distance information of the target object;

对调制后的数据列Y″进行关联成像或者压缩感知的计算，得到同时包含有目标物体纹理信息和距离信息的数据列X″；Carry out associated imaging or compressed sensing calculations on the modulated data column Y″ to obtain a data column X″ containing both texture information and distance information of the target object;

由于调制函数f(x)和f′(x)已知，将X′和X″中的对应元进行运算，可求出纹理信息和距离信息，即为所述的三维信息。Since the modulation functions f(x) and f'(x) are known, the corresponding elements in X' and X" can be calculated to obtain texture information and distance information, which is the three-dimensional information.

对于信号列数据s_i，j进行调制并进行关联成像或者压缩感知的投影关系如下：The projection relationship of modulating and correlating imaging or compressed sensing for signal column data s _{i, j} is as follows:

${Y Y}^{' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot \cdot f f (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot \cdot \\ \cdot \cdot \\ \cdot &Center Dot; \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot \cdot f f (({d d}_{j j 22})) \end{matrix})$

$= = Φ Φ (\begin{matrix} {x x}_{11,, j j 11} \cdot \cdot f f (({d d}_{j j 11})) + + {x x}_{11,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{11,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ {x x}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ {x x}_{N N,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{N N,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{N N,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \end{matrix}) = = Φ Φ {X x}^{' '}$

${Y Y}^{' '' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \\ \cdot \cdot \\ \cdot &Center Dot; \\ \cdot \cdot \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \end{matrix})$

$= = Φ Φ (\begin{matrix} {x x}_{11,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {x x}_{11,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{11,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \\ {x x}_{22,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {x x}_{22,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{22,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot \cdot \\ {x x}_{N N,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {x x}_{N N,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{N N,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \end{matrix}) = = Φ Φ {X x}^{' '' '}$

可以记为：can be recorded as:

Y′(i)＝s_i，j1·f(d_j1)+s_i，j1+1·f(d_j1+1)+...+s_i，j2·f(d_j2)，Y'(i)=s _{i, j1} f(d _j1 )+s _{i, j1+1} f(d _j1+1 )+...+s _{i, j2} f(d _j2 ),

Y″(i)＝s_i，j1·f′(d_j1)+s_i，j1+1·f′(d_j1+1)+...+s_i，j2·f′(d_j2)，Y″(i)=si _,j1 ·f′(d _j1 )+s _i,j1+1 ·f′(d _j1+1 )+...+si _,j2 ·f′(d _j2 ),

使用关联成像的方式对目标物体的信息进行计算，需要计算缓存单元中存储的调制图像R_i中每一个像素的灰度值R_i，n，利用R_i，n与调制后的数据Y′(i)和Y″(i)的相关系数，即可得到X′和X″。其中：Using the method of associated imaging to calculate the information of the target object, it is necessary to calculate the gray value R _i,n of each pixel in the modulated image R _i stored in the buffer unit, and use R _i,n and the modulated data Y'( i) and the correlation coefficient of Y"(i), X' and X" can be obtained. in:

${X x}^{' '} ((n no)) = = \frac{\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '} ((i i)) {R R}_{i i,, n no} - - \frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '} ((i i)) \cdot &Center Dot; \frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {R R}_{i i,, n no}}{\sqrt{\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' ' 22} ((i i)) - - {((\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '} ((i i))))}^{22}} \sqrt{\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {R R}_{i i,, n no}^{22} - - {((\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {R R}_{i i,, n no}^{22}))}^{22}}}$

${X x}^{' '' '} ((n no)) = = \frac{\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '' '} ((i i)) {R R}_{i i,, n no} - - \frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '' '} ((i i)) \cdot \cdot \frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {R R}_{i i,, n no}}{\sqrt{\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '' ' 22} ((i i)) - - {((\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {Y Y}^{' '' '} ((i i))))}^{22}} \sqrt{\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {R R}_{i i,, n no}^{22} - - {((\frac{11}{M m} {Σ Σ}_{i i = = 11}^{M m} {R R}_{i i,, n no}^{22}))}^{22}}}$

压缩感知也是一种对图像进行重构的方式，可以大大降低采样次数，用远远低于传统的采样数目还原原始信号，通过M次的测量可以还原N维的信号(M＜N)。Compressed sensing is also a way to reconstruct images, which can greatly reduce the number of sampling times, restore the original signal with a number far lower than the traditional sampling number, and restore N-dimensional signals (M<N) through M measurements.

由于将待恢复的图像信号O展开成为N维行向量x与每一次测量到的光强信号组成M维列向量y之间存在关系：y＝Φx，x的求解过程可以转化为凸规划问题：

subject to y＝Φx＝ΦΨ^Tθ，其中Ψ为x的稀疏变换矩阵。利用现有的算法即可求解出x。投影矩阵Φ的元素可以是满足伯努利分布的-1，1随机矩阵或者随机采样的Hadamard矩阵，也可以是满足高斯分布的-1～1的随机矩阵。但是由于空间光调制器不能实现负值的加载，这时需要先使用一定的方法将Φ中的元素变为正值。Since there is a relationship between expanding the image signal O to be restored into an N-dimensional row vector x and each measured light intensity signal to form an M-dimensional column vector y: y=Φx, the solution process of x can be transformed into a convex programming problem:

subject to y=Φx=ΦΨ ^T θ, where Ψ is the sparse transformation matrix of x. Existing algorithms can be used to solve x. The elements of the projection matrix Φ can be -1, 1 random matrices satisfying Bernoulli distribution or randomly sampled Hadamard matrices, or -1~1 random matrices satisfying Gaussian distribution. However, since the spatial light modulator cannot be loaded with negative values, it is necessary to use a certain method to change the elements in Φ into positive values.

位于不同距离处的平面物体或者连续的物体，如果只用上述的方法，计算量将是很大的。以位于不同距离处的平面物体为例，探测器探测到的将是有多个脉冲的数据列，这时要进行多次峰值和位置的读取并且进行多次压缩感知的计算。而对于一个连续物体，探测器探测到的将是一个展宽的数据列，需要根据到达探测器的时间不同对数据进行切片，并将每个切片内的光强数值进行相加。For planar objects or continuous objects located at different distances, if only the above method is used, the amount of calculation will be very large. Taking planar objects at different distances as an example, what the detector detects will be a data series with multiple pulses. At this time, it is necessary to read the peak value and position multiple times and perform multiple compressed sensing calculations. For a continuous object, what the detector detects will be a widened data column, and the data needs to be sliced according to the time of reaching the detector, and the light intensity values in each slice are added.

如下式所示，对于纵向分布在d_j1，d_j1+1，...，d_j2范围内的物体，公式y＝Φx可表示为：As shown in the following formula, for objects longitudinally distributed in the range of d _j1 , d _j1+1 , ..., d _j2 , the formula y=Φx can be expressed as:

其中，s_i，j是第i次测量时对应于距离d_j的物体分布探测器接收到的光强，为了减少计算量，需要对探测到的信号进行预处理，我们可以在上式两端都乘以与距离有关的调制函数矩阵F＝(f(d_j1)，f(d_j1+1)，...，f(d_j2))^T，相乘的结果是如下式所示：Among them, s _{i, j} is the light intensity received by the object distribution detector corresponding to the distance d _j at the i-th measurement. In order to reduce the amount of calculation, it is necessary to preprocess the detected signal. We can use both ends of the above formula Both are multiplied by the distance-related modulation function matrix F=(f(d _j1 ), f(d _j1+1 ),..., f(d _j2 )) ^T , and the multiplication result is shown in the following formula:

${Y Y}^{' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \end{matrix}) - - - - - - ((22))$

$= = Φ Φ (\begin{matrix} {x x}_{11,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{11,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{11,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ {x x}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ {x x}_{N N,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{N N,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{N N,, j j 22} \cdot \cdot f f (({d d}_{j j 22})) \end{matrix}) = = Φ Φ {X x}^{' '}$

(Y″同理)(Same for Y″)

假设X′(i)＝x_i，j1·f(d_j1)+x_i，j1+1·f(d_j1+1)+...+x_i，j2·f(d_j2)，由空间物体的分布特征可知，矩阵X中每一行最多只有一个非零值。假设第i行存在非零值x_i，j≠0，i∈{1，2，...，N}，j1≤j≤j2。那么X′第i行的元素值必然为x_i，j·f(d_j)。经过一次关联成像或者压缩感知的计算，可以得到矩阵X′，矩阵X′中每一个非零值都包含了空间物体对应位置的深度信息。Suppose X′(i)=xi _{, j1} ·f(d _j1 )+xi _{, j1+1} ·f(d _j1+1 )+...+ _{xi, j2} ·f(d _j2 ), by space According to the distribution characteristics of the objects, each row in the matrix X has at most one non-zero value. Suppose there is a non-zero value x _{i, j} ≠ 0 in the i-th row, i∈{1,2,...,N}, j1≤j≤j2. Then the element value of the i-th row of X' must be x _i,j ·f(d _j ). After a correlation imaging or compressed sensing calculation, the matrix X' can be obtained, and each non-zero value in the matrix X' contains the depth information of the corresponding position of the space object.

然后，创建另外一个调制函数矩阵Then, create another modulation function matrix

F′＝(f′(d_j1)，f′(d_j1+1)，...，f′(d_j2))^T，用上面相同的方法，可以求得矩阵X″，对应于X′中第i行的非零值

X″的第i行值为

既然调制函数f(x)和f′(x)已知，就可以该求解出深度信息d_j。F'=(f'(d _j1 ), f'(d _j1+1 ),..., f'(d _j2 )) ^T , using the same method above, the matrix X" can be obtained, corresponding to X' A non-zero value at row i in

The value of row i of X″ is

Now that the modulation functions f(x) and f′(x) are known, the depth information d _j can be solved.

例如，可以令f(x)≡1，f′(x)＝x，那么计算出来的X′包含了目标物体的二维信息，而f′(d_j)/f(d_j)＝d_j，又可以求取对应元的深度信息，从而获取了目标物体的三维信息。For example, f(x)≡1 can be set, f'(x)=x, then the calculated X' contains the two-dimensional information of the target object, and f'(d _j )/f(d _j )=d _j , and the depth information of the corresponding element can be obtained, thereby obtaining the three-dimensional information of the target object.

本发明方法，将关联成像和压缩感知的计算次数降低到2，大大地降低了计算量，对于三维信息的获得，有很大的优势。The method of the invention reduces the number of calculations of associated imaging and compressed sensing to 2, greatly reduces the amount of calculation, and has great advantages in obtaining three-dimensional information.

附图说明Description of drawings

图1是实现本发明方法的系统示意图。Fig. 1 is a schematic diagram of a system for realizing the method of the present invention.

图2是本发明方法的流程图。Figure 2 is a flow chart of the method of the present invention.

图3是本发明方法中调制过程的流程图。Fig. 3 is a flowchart of the modulation process in the method of the present invention.

具体实施方式 Detailed ways

为实现本发明一种基于关联成像的三维信息获取方法，可以利用图1所示的系统，图1中系统包括脉冲激光器，透镜组1，反射镜2，空间光调制器3，透镜组4、三维物体(即目标物体)、探测器，以及由缓存单元、预处理单元、存储器、计算单元组成的外围电路。In order to realize a method for obtaining three-dimensional information based on correlated imaging of the present invention, the system shown in Figure 1 can be utilized, and the system in Figure 1 includes a pulsed laser, a lens group 1, a mirror 2, a spatial light modulator 3, a lens group 4, A three-dimensional object (that is, a target object), a detector, and a peripheral circuit composed of a cache unit, a preprocessing unit, a memory, and a computing unit.

透镜组1位于脉冲激光器前端，发射的脉冲激光束(探测光)经过透镜组1、反射镜2进入空间光调制器3，空间光调制器3依次向光源发出的探测光中加载随机图像，得到带有随机图像信息的探测光，而后经过透镜组4投射至目标物体。The lens group 1 is located at the front end of the pulsed laser, and the emitted pulsed laser beam (probe light) enters the spatial light modulator 3 through the lens group 1 and the reflector 2, and the spatial light modulator 3 sequentially loads random images into the probe light emitted by the light source to obtain The probe light with random image information is then projected to the target object through the lens group 4 .

探测器收集不同时间由目标物体返回的光强列，并交由外围电路处理。探测器收集到的信号列不是直接进行存储以进行计算，而是先经过预处理单元对其进行调制，添加距离信息。The detector collects the light intensity series returned by the target object at different times, and sends them to the peripheral circuit for processing. The signal column collected by the detector is not directly stored for calculation, but modulated by the preprocessing unit to add distance information.

空间光调制器3可以是反射型的数字微镜阵列(DMD)，也可以是反射型的硅上液晶器件(LCOS)，空间光调制器3与缓存单元相连接。位于一旁的探测器可用隔板与上述装置隔开，防止发射光强的干扰，探测器需要与预处理单元相连接，用以处理收集到的光强。缓存单元，预处理单元，存储单元和计算单元构成的外围电路可通过电脑或单片机来实现相应的功能。The spatial light modulator 3 may be a reflective digital micromirror array (DMD), or a reflective liquid crystal on silicon device (LCOS). The spatial light modulator 3 is connected to the buffer unit. The detector located on the side can be separated from the above-mentioned device by a partition to prevent the interference of emitted light intensity. The detector needs to be connected with a preprocessing unit to process the collected light intensity. A peripheral circuit composed of a cache unit, a preprocessing unit, a storage unit and a calculation unit can realize corresponding functions through a computer or a single-chip microcomputer.

参见图2、3，本发明一种基于关联成像的三维信息获取方法包括：Referring to Figures 2 and 3, a method for obtaining three-dimensional information based on associated imaging in the present invention includes:

(2)设定目标物体的检测次数M，并生成一个M×N维的观测矩阵，利用该观测矩阵生成M幅随机图像；(2) Set the detection times M of the target object, and generate an M×N-dimensional observation matrix, and use the observation matrix to generate M random images;

观测矩阵可使用经过处理的满足伯努利分布的-1，1随机矩阵，或满足高斯分布的-1～1随机矩阵，或随机采样的Hadamard矩阵。The observation matrix can use processed -1, 1 random matrix satisfying Bernoulli distribution, or -1~1 random matrix satisfying Gaussian distribution, or randomly sampled Hadamard matrix.

由于目标物体的横向分辨率由加载的随机图像所决定，每一副随机图像都可以展开成为有N个元素的一维向量。Since the horizontal resolution of the target object is determined by the loaded random images, each random image can be expanded into a one-dimensional vector with N elements.

(3)作为光源的脉冲激光器空间发射脉冲激光束(探测光)经过透镜组1、反射镜2进入空间光调制器3，空间光调制器3依次向光源发出的探测光中加载M幅随机图像中的一幅，得到带有随机图像信息的探测光，经过透镜组4投射至目标物体；(3) The pulsed laser as the light source spatially emits the pulsed laser beam (probe light) through the lens group 1 and the mirror 2 and enters the spatial light modulator 3, and the spatial light modulator 3 sequentially loads M random images into the probe light emitted by the light source In one of them, the probe light with random image information is obtained, and is projected to the target object through the lens group 4;

(4)利用探测器检测来自目标物体的反射光的光强，每次检测时间为T，在检测时间T内得到信号列s₁，其中，第j个数据可以记为s_1，j，其中1表示第一副随机图像，j为信号列内各个数据点的序号；(4) Use the detector to detect the light intensity of the reflected light from the target object. Each detection time is T, and the signal sequence s ₁ is obtained within the detection time T, where the jth data can be recorded as s _1,j , where 1 means the first random image, and j is the serial number of each data point in the signal column;

不同的检测时间T之间可以通过预定的触发信号、或来自脉冲激光器的信号进行切换或触发。Different detection times T can be switched or triggered by a predetermined trigger signal or a signal from a pulsed laser.

(5)在信号列数据s_1，j中取有效范围内的数据点，有效范围为：j1≤j≤j2，在预处理单元中对所得到的有效范围内的数据点进行调制，得到调制后的信号列；(5) Take the data points within the effective range in the signal sequence data s _{1, j} , the effective range is: j1≤j≤j2, and modulate the obtained data points within the effective range in the preprocessing unit to obtain modulation after the signal column;

d_j＝c(Δt·j+Δt′)，j1≤j≤j2；d _j = c(Δt·j+Δt′), j1≤j≤j2;

Δt′为探测器的延迟时间；Δt' is the delay time of the detector;

c为光速；c is the speed of light;

T时间内由于探测器会多次间隔的进行信号采集，因此该信号列数据s_1，j内包含了多个信号(数据点)，本发明仅取用有效范围内的数据点，符合条件j1≤j≤j2的j认为是有效信号。In the T time, because the detector will carry out signal acquisition at multiple intervals, so the signal sequence data s1 _{, j} contains a plurality of signals (data points), the present invention only uses the data points within the effective range, which meets the condition j1 ≤j≤j2 is considered to be a valid signal.

${Y Y}^{' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot \cdot f f (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot \cdot f f (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot \cdot \\ {s the s}_{M m,, j j 11} \cdot \cdot f f (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \end{matrix});;$

${Y Y}^{' '' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ \cdot \cdot \\ \cdot \cdot \\ \cdot \cdot \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \end{matrix});;$

Y′和Y″中，每一行对应了一个步骤(5)中所述的调制后的信号列，由于是M次检测，因此Y′和Y″中均为M行；调制后的信号列均传送至存储器中保存。In Y' and Y ", each row corresponds to the modulated signal row described in step (5), because it is M times of detection, so Y' and Y " are M rows; the modulated signal row is equal to Transfer to storage for storage.

(7)对调制后的数据列Y′进行压缩感知计算，得到包含有目标物体纹理信息和距离信息的数据列X′；(7) Perform compressed sensing calculation on the modulated data column Y' to obtain the data column X' containing the texture information and distance information of the target object;

${Y Y}^{' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot \cdot f f (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot \cdot \\ {s the s}_{M m,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot \cdot f f (({d d}_{j j 22})) \end{matrix})$

$= = Φ Φ (\begin{matrix} {x x}_{11,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{11,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{11,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ {x x}_{22,, j j 11} \cdot &Center Dot; f f (({d d}_{j j 11})) + + {x x}_{22,, j j 11 + + 11} \cdot &Center Dot; f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{22,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \\ \cdot \cdot \\ \cdot \cdot \\ \cdot \cdot \\ {x x}_{N N,, j j 11} \cdot \cdot f f (({d d}_{j j 11})) + + {x x}_{N N,, j j 11 + + 11} \cdot \cdot f f (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{N N,, j j 22} \cdot &Center Dot; f f (({d d}_{j j 22})) \end{matrix}) = = Φ Φ {X x}^{' '}$

对调制后的数据列Y″进行压缩感知的计算，得到同时包含有目标物体纹理信息和距离信息的数据列X″；Calculate the compressed sensing on the modulated data column Y″ to obtain the data column X″ containing both the texture information and the distance information of the target object;

${Y Y}^{' '' '} = = (\begin{matrix} {s the s}_{11,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{11,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{11,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ {s the s}_{22,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{22,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{22,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ \cdot &Center Dot; \\ {s the s}_{M m,, j j 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11})) + + {s the s}_{M m,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {s the s}_{M m,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \end{matrix})$

$= = Φ Φ (\begin{matrix} {x x}_{11,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {x x}_{11,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{11,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \\ {x x}_{22,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {x x}_{22,, j j 11 + + 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{22,, j j 22} \cdot \cdot {f f}^{' '} (({d d}_{j j 22})) \\ \cdot &Center Dot; \\ \cdot \cdot \\ \cdot \cdot \\ {x x}_{N N,, j j 11} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 11})) + + {x x}_{N N,, j j 11 + + 11} \cdot \cdot {f f}^{' '} (({d d}_{j j 11 + + 11})) + + . . . . . . + + {x x}_{N N,, j j 22} \cdot &Center Dot; {f f}^{' '} (({d d}_{j j 22})) \end{matrix}) = = Φ Φ {X x}^{' '' '}$

Claims

1. 3 D information obtaining method based on relevance imaging; Comprise and utilize light source to send detection light; Survey and be projected to target object after light is handled through spatial light modulator, detect reflected light, confirm the three-dimensional information of target object according to this reflected light from target object; It is characterized in that, specifically comprise the steps:

(1) the total pixel number N of setting random image sets the modulating function f (x) of two light intensity signals and f ' (x);

(2) the detection number of times M of target setting object, and generate the observing matrix that a M * N ties up, utilize this observing matrix to generate M width of cloth random image;

(3) spatial light modulator loads the width of cloth in the M width of cloth random image successively in the detection light that light source sends, and obtains having the detection light of random image information, then is projected to target object;

(4) detection is from the catoptrical light intensity of target object, and be T each detection time, obtains signal train s in the T in detection time ₁, wherein, j data can be designated as s _{1, j}, wherein 1 representes the first secondary random image, j is the sequence number of each data points in the signal train;

(5) at signal train data s _{1, j}In get the data point in the effective range, described effective range is: j1≤j≤j2, the data point in the resulting effective range is modulated the signal train after obtaining modulating;

" Y ' (1)=Y ' (1)+f (d in the time of j=j1, is carried out in (1)=0 to establish Y ' (1)=0 during modulation, Y _j) * s _{1, j}, Y " (1)=Y " (1)+f ' (d _j) * s _{1, j}Computing is till j=j2;

d _j＝c(Δt·j+Δt′)，j1≤j≤j2；

Δ t is the interval time of the each acquired signal of detector;

Δ t ' is the time delay of detector;

C is the light velocity;

J is the sequence number of each data points in the signal train;

J1 and j2 are the data point sequence number threshold value in the effective range;

F and f ' are the described modulating function of step (1);

(6) circulation step (3)～step (5) is all sent until M width of cloth random image and is finished, and for the secondary random image of i, correspondence obtains signal train s _i, wherein, j data can be designated as s _{I, j}, i is the sequence number of random image; And carry out Y ' (i)=Y ' (i)+f (d _j) * s _{I, j}, Y " (i)=Y " (i)+f ' (d _j) * s _{I, j}Modulation, accomplish M detection, corresponding obtain M after modulating signal train and be expressed as Y ' and Y ":

Y^{'} = (\begin{matrix} s_{1, j 1} \cdot f (d_{j 1}) + s_{1, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + s_{1, j 2} \cdot f (d_{j 2}) \\ s_{2, j 1} \cdot f (d_{j 1}) + s_{2, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + s_{2, j 2} \cdot f (d_{j 2}) \\ \cdot \\ \cdot \\ \cdot \\ s_{M, j 1} \cdot f (d_{j 1}) + s_{M, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + s_{M, j 2} \cdot f (d_{j 2}) \end{matrix});

Y^{''} = (\begin{matrix} s_{1, j 1} \cdot f^{'} (d_{j 1}) + s_{1, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + s_{1, j 2} \cdot f^{'} (d_{j 2}) \\ s_{2, j 1} \cdot f^{'} (d_{j 1}) + s_{2, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + s_{2, j 2} \cdot f^{'} (d_{j 2}) \\ \cdot \\ \cdot \\ \cdot \\ s_{M, j 1} \cdot f^{'} (d_{j 1}) + s_{M, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + s_{M, j 2} \cdot f^{'} (d_{j 2}) \end{matrix});

(7) the data rows Y ' after the modulation is carried out the calculating of relevance imaging or compressed sensing, obtain including the data rows X ' of target object texture information and range information;

Data rows Y to after the modulation " carries out the calculating of relevance imaging or compressed sensing, is included the data rows X of target object texture information and range information simultaneously ";

Because modulating function f (x) and f ' are (x) known, with X ' and X " in corresponding element carry out computing, can obtain texture information and range information, be described three-dimensional information.

2. the 3 D information obtaining method based on relevance imaging as claimed in claim 1 is characterized in that, the data rows Y ' after the modulation is carried out satisfying relation as follows when compressed sensing is calculated:

Y^{'} = (\begin{matrix} s_{1, j 1} \cdot f (d_{j 1}) + s_{1, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + s_{1, j 2} \cdot f (d_{j 2}) \\ s_{2, j 1} \cdot f (d_{j 1}) + s_{2, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + s_{2, j 2} \cdot f (d_{j 2}) \\ \cdot \\ \cdot \\ \cdot \\ s_{M, j 1} \cdot f (d_{j 1}) + s_{M, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + s_{M, j 2} \cdot f (d_{j 2}) \end{matrix})

= Φ (\begin{matrix} x_{1, j 1} \cdot f (d_{j 1}) + x_{1, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + x_{1, j 2} \cdot f (d_{j 2}) \\ x_{2, j 1} \cdot f (d_{j 1}) + x_{2, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + x_{2, j 2} \cdot f (d_{j 2}) \\ \cdot \\ \cdot \\ \cdot \\ x_{N, j 1} \cdot f (d_{j 1}) + x_{N, j 1 + 1} \cdot f (d_{j 1 + 1}) + . . . + x_{N, j 2} \cdot f (d_{j 2}) \end{matrix}) = Φ X^{'} .

3. the 3 D information obtaining method based on relevance imaging as claimed in claim 1 is characterized in that, to the data rows Y after the modulation " satisfy relation as follows when carrying out the calculating of compressed sensing:

Y^{''} = (\begin{matrix} s_{1, j 1} \cdot f^{'} (d_{j 1}) + s_{1, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + s_{1, j 2} \cdot f^{'} (d_{j 2}) \\ s_{2, j 1} \cdot f^{'} (d_{j 1}) + s_{2, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + s_{2, j 2} \cdot f^{'} (d_{j 2}) \\ \cdot \\ \cdot \\ \cdot \\ s_{M, j 1} \cdot f^{'} (d_{j 1}) + s_{M, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + s_{M, j 2} \cdot f^{'} (d_{j 2}) \end{matrix})

= Φ (\begin{matrix} x_{1, j 1} \cdot f^{'} (d_{j 1}) + x_{1, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + x_{1, j 2} \cdot f^{'} (d_{j 2}) \\ x_{2, j 1} \cdot f^{'} (d_{j 1}) + x_{2, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + x_{2, j 2} \cdot f^{'} (d_{j 2}) \\ \cdot \\ \cdot \\ \cdot \\ x_{N, j 1} \cdot f^{'} (d_{j 1}) + x_{N, j 1 + 1} \cdot f^{'} (d_{j 1 + 1}) + . . . + x_{N, j 2} \cdot f^{'} (d_{j 2}) \end{matrix}) = Φ X^{''} .

4. the 3 D information obtaining method based on relevance imaging as claimed in claim 1; It is characterized in that described observing matrix is to satisfy-1,1 stochastic matrix that Bernoulli Jacob distributes; Or satisfy-1～1 stochastic matrix of Gaussian distribution or the Hadamard matrix of stochastic sampling.

5. the 3 D information obtaining method based on relevance imaging as claimed in claim 1 is characterized in that, extracts each row of observing matrix in the step (2), the corresponding width of cloth random image that converts into.