CN102538709A - Method for utilizing GPU (Graphics Processing Unit) concurrent computation in three-dimensional measurement system based on structured light - Google Patents
Method for utilizing GPU (Graphics Processing Unit) concurrent computation in three-dimensional measurement system based on structured light Download PDFInfo
- Publication number
- CN102538709A CN102538709A CN2012100035393A CN201210003539A CN102538709A CN 102538709 A CN102538709 A CN 102538709A CN 2012100035393 A CN2012100035393 A CN 2012100035393A CN 201210003539 A CN201210003539 A CN 201210003539A CN 102538709 A CN102538709 A CN 102538709A
- Authority
- CN
- China
- Prior art keywords
- coordinate
- gpu
- computing machine
- point
- width
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The invention discloses a method for utilizing GPU (Graphics Processing Unit) concurrent computation in a three-dimensional measurement system based on structured light, comprising the following steps of: transmitting a deformation fringe image collected from a camera and camera parameters obtained by binocular stereoscopic vision calibration into GPU equipment for one time; realizing image correction, phase computation, phase unwrapping, three-dimensional matching and computation of three-dimensional point coordinates on the GPU equipment; and finally, transmitting a computed result to a computer internal memory from the GPU equipment for one time. The invention provides a large-scale and fine-granularity rapid concurrent computation method and the method effectively utilizes CPU (Central Processing Unit) and GPU data transmission bandwidths, so that the loss of time of a plurality of times of reciprocated transmission of data is reduced, the advantage that the GPU is provided with a plurality of processors is sufficiently utilized, and the whole computation speed is improved.
Description
Technical field
The present invention relates to the optical three-dimensional measurement field, utilize the method for GPU parallel computation in particularly a kind of three-dimension measuring system based on structured light.
Background technology
The three-dimensional measurement technology is the master tool of record object appearance, has important value in a plurality of fields such as industry manufacturing, scientific research, biomedicine, cultural relic digitalization.Based on the three-dimensional measurement technology of the structured light of phase-shift method because advantages such as it is simple in structure, noncontact, high precision are widely used in the actual measurement.
The intensive calculations of big data quantity is the bottleneck of restriction structural light three-dimensional efficiency of measurement always.Because the intrinsic complicacy of multifrequency phase-shifted grating three-dimensional measurement algorithm, under the prerequisite of not sacrificing arithmetic accuracy and stability, its time complexity is difficult to reduce significantly.Present 1,000,000 grades of pixel camera machines are very general, 500 ten thousand, 800 ten thousand in addition millions pixel camera machine got into field of measurement, but all be faced with computing velocity slow, cause the not high problem of efficiency of measurement.With the coupling is example, and the only characteristic that left video camera obtains need find match point in the only characteristic of right video camera, retrain according to polar curve; Travel through the characteristic on the polar curve of right video camera; Both having needed 100 ten thousand to multiply by polar curve length, is example with the mega pixel camera, and the length that promptly will search for is 1,000 pixels; The calculating of being done is about billion-degree, and this does not also comprise interpolation calculation of looking for corresponding point etc.; Although core cpu quantity increases at present, its order set is more prone to Processing tasks, and the calculated performance of being brought promotes and be not obvious.GPU then is mainly used in figure, Flame Image Process and calculates, and GPU has the floating-point operation abilities that number decuples CPU, according to these characteristics, it optimizations utilization can be increased substantially the speed of three-dimensional measurement.
Applicant applies for and obtains patent right on July 29th, 2006; Name is called " vision measuring method of the object surface tri-dimensional profile of projecting multiple frequency grating "; The patent No. is: in 2,006 1 0010284.8 patents of invention; The phase unwrapping that discloses the frequency synthesis that utilizes grating is accomplished the three-dimensional coordinate calculating of pixel in the images acquired; The final three-dimensional measurement of accomplishing object, the present invention is according to above-mentioned its technology contents of patent disclosed method, the optimization implementation among the GPU of the computing machine in three-dimension measuring system.
Summary of the invention
The invention provides the method for utilizing the GPU parallel computation in a kind of three-dimension measuring system, can effectively utilize the computation capability of graphics processing unit, accelerated the measuring speed of optical measurement based on structured light.
At first to the grating of three (or a plurality of) frequencies of body surface projection, the grating fringe by the video camera shot object surface modulation that is installed in grating device for projecting both sides is saved to calculator memory with the image that is obtained; Obtain the camera calibration parameter through camera calibration; With view data in the internal memory and the disposable GPU device storage district that is delivered to of calibrating parameters.At first the image that passes over is proofreaied and correct, adopted texture storage in the trimming process, avoided the data of common storage area to read once more effectively, and linear interpolation calculating does not take programmable unit according to the video camera confidential reference items; Image to after proofreading and correct carries out phase calculation and phase unwrapping, finally obtains left and right cameras independent phase separately; According to the basis matrix parameter in the calibrating parameters, mate according to the polar curve constraint; By the coupling corresponding point to the Calculation of Three Dimensional point.Net result is returned the stored district.
According to the GPU own characteristic, adopt the GPU equipment end data structure organization mode of a kind of overall situation, full width face, location records data, efficiently solve the use of GPU equipment end data, storage problem.With record match point coordinate is example; Because write down left and right video camera corresponding point; Represent the location point of left video camera to scheme image width, high position, right video camera corresponding point horizontal coordinate is stored in the X coordinate memory block correspondence position of image size, vertical coordinate storage Y coordinate memory block correspondence position; As X coordinate memory block [100] location storage be match point X coordinate, Y coordinate memory block [100] location storage be match point Y coordinate, both (100,100) and (X [100], Y [100]) were that match point is right.This mode has been saved the left matched position of record, left matched position is lain in the memory block sequence number of right matched position.Application program is made up of two parts: a part runs on the CPU, is called main frame (Host) end; Another part runs on the GPU, is called equipment (Device) end.
The present invention realizes that the technical scheme that goal of the invention adopts is:
Utilize the method for GPU parallel computation in a kind of three-dimension measuring system based on structured light; This method by means of: the grating device for projecting in the three-dimension measuring system throws striped to measured object; Camera acquisition in the three-dimension measuring system is through the measured object image of grating projection; And be kept in the calculator memory in the three-dimension measuring system; And the camera intrinsic parameter
, the video camera that obtain video camera are joined
and
, camera lens distortion parameter
and
outward; The GPU of computing machine realizes the parallel computation of the three-dimensional point of testee in the three-dimension measuring system, and the step of this GPU parallel calculating method is:
⑴, all images that is obtained is stored in the internal memory of computing machine according to the order of sequence,, exists in the calculator memory in the three-dimension measuring system with the form of array with whole parameters of the video camera that obtains;
, in the equipment end of computing machine GPU, calculating needs the storage space that uses, and opens up this storage space, the image that the step ⑴ in the internal memory of computing machine is preserved and the supplemental characteristic of video camera be disposable to be copied in the common storage of computing machine GPU equipment end;
⑶, divide thread block (block) size, select thread block (block) is designed to two dimension according to pending data scale; Generally, can adopt following method to calculate the block quantity on certain dimension: the block quantity on the x direction=(data at the size on the x direction+each block in the size on the x direction-the 1)/size of each block on the x direction; The y direction in like manner.
, in the equipment end of computing machine GPU, accomplish image rectification, the image rectification step is following:
A, the data that will proofread and correct and texture are bound, and use the normalization coordinate of texture storage device;
B, the linear interpolation of using the texture storage device to be provided are obtained the eigenwert of non-integer position;
C, establish
Be desirable imaging surface coordinate (unit: mm),
Be the coordinate of distortion, then
, wherein
, k
iBe radial distortion parameter, p
iBe the tangential distortion parameter; Obtain distorted position thus, accomplish image rectification;
, in the equipment end of computing machine GPU, accomplish phase calculation and phase unwrapping, phase calculation and phase unwrapping may further comprise the steps:
A, acquisition present level directional ray program number tx and hard Nogata are to thread sequence number ty;
B, acquisition current thread handled picture position sequence number i=ty * picture traverse Width+tx;
C, according to grating fringe number calculated rate coefficient;
The phase place of d, calculating current location point, and according to frequency synthesis expansion phase place;
The unique phase place of unfolded of e, each video camera of finally obtaining;
, in the equipment end of computing machine GPU, accomplish the solid coupling of phase place, the solid coupling of phase place may further comprise the steps:
A, acquisition present level directional ray program number tx=blockIdx.x * blockDim.x+threadIdx.x;
With hard Nogata to thread sequence number ty=blockIdx.y * blockDim.y+threadIdx.y;
B, with the current thread sequence number as the position to be matched of left video camera phase place;
C, calculate polar curve, in right video camera phase place, look for corresponding point along polar curve according to basis matrix;
D, for the corresponding point that find; The horizontal direction coordinate is stored in position, match point X coordinate memory block [ty*Width+tx]; The vertical direction coordinate is stored in match point Y coordinate memory block YM [ty*Width+tx]; What do not obtain match point then is changed to zero with this location storage district level, vertical direction [ty*Width+tx] coordinate, and expression does not obtain match point;
, in the equipment end of computing machine GPU; The completion three-dimensional point is calculated; With current thread tx and ty as left side matched position point; According to the right match point coordinate that is stored in correspondence position in X coordinate memory block and the Y coordinate memory block, Calculation of Three Dimensional coordinate x, y, z are stored in respectively in position, three-dimensional point X coordinate memory block [ty*Width+tx], position, Y coordinate memory block [ty*Width+tx], the position, Z coordinate memory block [ty*Width+tx].
⑻, the three-dimensional point coordinate that step ⑺ is accomplished be disposable to be back in the internal memory of computing machine, discharges the storage space that computing machine GPU equipment end is opened up.
The invention has the beneficial effects as follows; On a large scale, the fast parallel computing method of fine granularity, this method has effectively been utilized CPU and GPU data transfer bandwidth, has reduced the repeatedly loss in reciprocal transmission time of data; Make full use of the advantage of GPU multiprocessor, improved whole computing velocity.
Embodiment
1. throw striped by the grating device for projecting in the three-dimension measuring system to measured object.
2. in the three-dimension measuring system, camera acquisition is through the measured object image of grating projection, and is kept in the internal memory of computing machine.
3. the parameter of video camera was both participated in outer ginseng in the video camera in the acquisition three-dimension measuring system; Camera intrinsic parameter
; Video camera is joined
and
outward, camera lens distortion parameter
and
; I=1,2 ..., i representes camera number.
A. camera intrinsic parameter
B. rotation matrix
C. translation vector
D. radial distortion parameter
E. tangential distortion parameter
4, will be kept at the image and the disposable GPU of the passing to equipment end of calibrating parameters of the host side of computing machine in the three-dimension measuring system.
The all images that is obtained is stored in according to the order of sequence in the internal memory of computing machine; Calibrating parameters also is stored in the internal memory with the form of array.
In the GPU equipment end, calculating needs the storage space of use, and opens up this space; Copy in the equipment end common storage the data in the host side internal memory are disposable.
5, divide thread block (block) size according to the maximum thread amount and the pending data scale of GPU chipset; Because pending data are two dimensional image, select thread block (block) is designed to two dimension.
6,, accomplish image rectification work in the GPU equipment end;
A. the data that will proofread and correct and texture are bound, and use the normalization coordinate of texture storage device;
B. the linear interpolation of using the texture storage device to be provided is obtained the eigenwert of non-integer position.
C. establishing
is desirable imaging surface coordinate (unit: mm);
is the coordinate of distortion, then
Wherein
, k
iBe radial distortion parameter, p
iFor the tangential distortion parameter can obtain distorted position thus, proofread and correct thereby accomplish.
7,, accomplish phase calculation and phase unwrapping work in the GPU equipment end; Frequency synthesis phase unwrapping method according to being proposed can be known; The phase place of each pixel is definitely independent; Be in every width of cloth image, the phase value of certain point only with following 4 width of cloth phase shifted images of same frequency in same position relevant, and with same width of cloth image in any other point this characteristics that have nothing to do.These characteristics have guaranteed in phase calculation and the phase unwrapping process, and each thread can directly calculate the gray-scale value of certain any phase place through this position in the associated picture, and this mode can be good at using these characteristics of GPU multithreading.The only left and right sides phase result of final acquisition.
A. obtain current thread tx and ty, tx representes horizontal direction thread sequence number, and ty representes that hard Nogata is to the thread sequence number;
B. obtain the handled picture position of current thread sequence number i=ty * Width+tx; Wherein Width is a picture traverse;
C. according to grating fringe number calculated rate coefficient;
D. calculate the phase place of current location point, and launch phase place (see summary of the invention 2 in the patent (ZL ZL2006 1 0010284.8) for details, adopt the frequency synthesis technology to launch phase place) according to the frequency synthesis technology;
E. the unique phase place of the unfolded of each video camera in the three-dimension measuring system that finally obtains.
8,, accomplish three-dimensional coupling work in the GPU equipment end;
A. obtain current thread tx and ty;
B. with current thread as the position to be matched of left camera feature (phase place), the shooting of a both known left side (tx, the ty) phase value of position, in right video camera phase place, look for its phase value less than given threshold value and immediate point.
C. calculate polar curve according to basis matrix, in right video camera phase place, look for corresponding point along polar curve.
D. for the corresponding point that find, the horizontal direction coordinate is stored in position, match point X coordinate memory block [ty*Width+tx], the vertical direction coordinate is stored in position, match point Y coordinate memory block [ty*Width+tx].What do not obtain match point then is changed to zero with this location storage district level, vertical direction [ty*Width+tx] coordinate, and expression does not obtain match point.
9,, accomplish the three-dimensional point evaluation work in the GPU equipment end; With current thread tx and ty as left side matched position point; According to the right match point coordinate that is stored in correspondence position in X coordinate memory block and the Y coordinate memory block; Calculation of Three Dimensional coordinate x, y, z are stored in respectively in position, three-dimensional point X coordinate memory block [ty*Width+tx], position, Y coordinate memory block [ty*Width+tx], the position, Z coordinate memory block [ty*Width+tx].
10, be back in the host side internal memory storage space that the releasing arrangement end is applied for three-dimensional coordinate is disposable.
Below in conjunction with specific embodiment the present invention is further specified
This instance adopts two black and white industrial cameras, images acquired wide (Width) 1280 pixels, high (Height) 1024 pixels; GPU adopts GTX580, and CPU is Core i7.
1. the fringe number of utilizing the grating device for projecting to generate virtual three groups of phase-shifted gratings is respectively 96,90,85, utilizes the grating grenade instrumentation that grating is projected on the object.
2. two video camera while images acquired are kept at 24 width of cloth that obtained (each 12 width of cloth of left and right cameras) image in the calculator memory according to the order of sequence; Obtain calibrating parameters (comprising two camera intrinsic parameters, outer parameter, lens distortion parameter).
3. in equipment end, the space that equates with host side storage calibrating parameters is applied in the space that application equates with the host side memory image; In equipment end, press storage space, phase place storage space, match point storage space, three-dimensional point storage space that the back image is proofreaied and correct in image size application storage.
4. with the image and the disposable equipment end memory block that reaches of calibrating parameters (comprising two camera intrinsic parameters, outer parameter, lens distortion parameter) of host side.
5. divide the thread block size in equipment end according to equipment maximum thread amount and pending data scale; Adopt in the enforcement thread block is divided into two dimension, tx, ty represent horizontal direction, vertical direction thread sequence number respectively.
6. realization image rectification, interpolation method uses bilinear interpolation; Use the texture storage device, the data in the texture cache can be repeated to utilize, and when data that visit once needs are Already in the texture cache, just can avoid GPU is held reading once more of common storage area.And linear interpolation is calculated and is not taken programmable unit.
A. data to be corrected and texture are bound, use the normalization coordinate of texture storage device;
B. according to the c item content in (6) step in the summary of the invention, calculate mapping point;
C. use bilinear interpolation to obtain the eigenwert of location of interpolation.
D. the result after will proofreading and correct is stored in the correcting image memory block.
Explain: (linear interpolation that this part interpolation work also can not use the texture storage device to provide, and adopt other interpolation method to accomplish as required.）
7. realization phase unwrapping,
A. obtain current thread tx and ty;
B. obtain the handled picture position of current thread sequence number i
C. calculate the phase place of current location point according to image sequence;
D. phase place is stored in the phase place memory block.
8. realize three-dimensional coupling, this process is promptly sought the process of left and right sides phase place corresponding point;
A. with the current thread sequence number as left phase place position to be matched; Calculate basis matrix according to calibrating parameters; And then calculate its pairing polar curve in right phase place, phase place on the traversal polar curve in right phase place is when finding less than given threshold value Tp; And with right phase place near the time, think that this position is the match point position; That does not find corresponding matched position is changed to 0.
B. the match point that obtains is stored in the match point memory block.
9. realize three-dimensional point calculating, the Calculation of Three Dimensional point coordinate is stored in the three-dimensional point memory block with the result.Computation process finishes.
10. net result is back to host side.
11. discharge the memory block that the GPU equipment end takies.
Claims (1)
1. utilize the method for GPU parallel computation in the three-dimension measuring system based on structured light; This method by means of: the grating device for projecting in the three-dimension measuring system throws striped to measured object; Camera acquisition in the three-dimension measuring system is through the measured object image of grating projection; And be kept in the calculator memory in the three-dimension measuring system; And the camera intrinsic parameter
, the video camera that obtain video camera are joined
and
, camera lens distortion parameter
and
outward; The GPU of computing machine realizes the parallel computation of the three-dimensional point of testee in the three-dimension measuring system, and it is characterized in that: the step of this GPU parallel calculating method is:
⑴, all images that is obtained is stored in the internal memory of computing machine according to the order of sequence,, exists in the calculator memory in the three-dimension measuring system with the form of array with whole parameters of the video camera that obtains;
, in computing machine GPU equipment end, the storage space that calculating need to be used, and open up this storage space, the image that the step ⑴ in the internal memory of computing machine is preserved and the supplemental characteristic of video camera be disposable to be copied in the common storage of computing machine GPU equipment end;
⑶, divide thread block (block) size, select thread block (block) is designed to two dimension according to pending data scale;
, in computing machine GPU equipment end, accomplish image rectification, the image rectification step is following:
A, the data that will proofread and correct and texture are bound, and use the normalization coordinate of texture storage device;
B, the linear interpolation of using the texture storage device to be provided are obtained the eigenwert of non-integer position;
C, to establish
be desirable imaging surface coordinate (unit: mm);
is the coordinate of distortion, then
Wherein
, k
iBe radial distortion parameter, p
iBe the tangential distortion parameter
Obtain distorted position thus, accomplish image rectification;
, in computing machine GPU equipment end, accomplish phase calculation and phase unwrapping, phase calculation and phase unwrapping may further comprise the steps:
A, acquisition present level directional ray program number tx=blockIdx.x * blockDim.x+threadIdx.x;
With hard Nogata to thread sequence number ty=blockIdx.y * blockDim.y+threadIdx.y;
B, acquisition current thread handled picture position sequence number i=ty * Width picture traverse+tx;
C, according to grating fringe number calculated rate coefficient;
The phase place of d, calculating current location point, and according to frequency synthesis technology expansion phase place;
The unique phase place of unfolded of e, each video camera of finally obtaining;
, in computing machine GPU equipment end, accomplish the solid coupling of phase place, the solid coupling of phase place may further comprise the steps:
A, acquisition present level directional ray program number tx=blockIdx.x * blockDim.x+threadIdx.x;
With hard Nogata to thread sequence number ty=blockIdx.y * blockDim.y+threadIdx.y;
B, with current thread sequence number tx, the position to be matched of the left video camera phase place of ty conduct;
C, calculate polar curve, in right video camera phase place, look for corresponding point along polar curve according to basis matrix;
D, for the corresponding point that find; The horizontal direction coordinate is stored in position, match point X coordinate memory block [ty*Width+tx]; The vertical direction coordinate is stored in position, match point Y coordinate memory block [ty*Width+tx]; What do not obtain match point then is changed to zero with this location storage district level, vertical direction [ty*Width+tx] location conten, and expression does not obtain match point;
, in the equipment end of computing machine GPU; The completion three-dimensional point is calculated; With current thread tx, ty is as left side matched position point, according to the right match point coordinate that is stored in correspondence position in X coordinate memory block and the Y coordinate memory block; Calculation of Three Dimensional coordinate x, y, z are stored in respectively in position, three-dimensional point X coordinate memory block [ty*Width+tx], position, Y coordinate memory block [ty*Width+tx], the position, Z coordinate memory block [ty*Width+tx];
⑻, the three-dimensional point coordinate that step ⑺ is accomplished be disposable to be back in the internal memory of computing machine, discharges the storage space that computing machine GPU equipment end is opened up.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012100035393A CN102538709A (en) | 2012-01-09 | 2012-01-09 | Method for utilizing GPU (Graphics Processing Unit) concurrent computation in three-dimensional measurement system based on structured light |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012100035393A CN102538709A (en) | 2012-01-09 | 2012-01-09 | Method for utilizing GPU (Graphics Processing Unit) concurrent computation in three-dimensional measurement system based on structured light |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102538709A true CN102538709A (en) | 2012-07-04 |
Family
ID=46346212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012100035393A Pending CN102538709A (en) | 2012-01-09 | 2012-01-09 | Method for utilizing GPU (Graphics Processing Unit) concurrent computation in three-dimensional measurement system based on structured light |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102538709A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103335611A (en) * | 2013-06-13 | 2013-10-02 | 华中科技大学 | Method for GPU-based object three-dimensional shape measurement |
| CN103994732A (en) * | 2014-05-29 | 2014-08-20 | 南京理工大学 | Three-dimensional measurement method based on fringe projection |
| CN104990517A (en) * | 2015-06-30 | 2015-10-21 | 东莞市盟拓光电科技有限公司 | Inclined projection correction method and system in structured light three-dimensional measurement |
| CN105243280A (en) * | 2015-10-30 | 2016-01-13 | 西安电子科技大学 | Time domain physical optics algorithm based on CPU (Central Processing Unit) and GPU (Graphics Processing Unit) hybrid asynchronous parallel way |
| WO2017193477A1 (en) * | 2016-05-09 | 2017-11-16 | 中国科学院深圳先进技术研究院 | Method and device for processing three-dimensional medical image data |
| CN104103058B (en) * | 2013-04-01 | 2018-08-14 | 鸿富锦精密工业（深圳）有限公司 | Image matching system and method |
| CN111899326A (en) * | 2020-06-18 | 2020-11-06 | 苏州小优智能科技有限公司 | Three-dimensional reconstruction method based on GPU parallel acceleration |
| CN112258378A (en) * | 2020-10-15 | 2021-01-22 | 武汉易维晟医疗科技有限公司 | Real-time three-dimensional measurement system and method based on GPU acceleration |
| CN114765660A (en) * | 2020-12-30 | 2022-07-19 | 浙江舜宇智能光学技术有限公司 | Image calibration method for TOF camera module and FPGA chip |
| CN114937159A (en) * | 2022-07-27 | 2022-08-23 | 成都飞机工业（集团）有限责任公司 | Binocular matching method based on GPU acceleration |
-
2012
- 2012-01-09 CN CN2012100035393A patent/CN102538709A/en active Pending
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104103058B (en) * | 2013-04-01 | 2018-08-14 | 鸿富锦精密工业（深圳）有限公司 | Image matching system and method |
| CN103335611A (en) * | 2013-06-13 | 2013-10-02 | 华中科技大学 | Method for GPU-based object three-dimensional shape measurement |
| CN103335611B (en) * | 2013-06-13 | 2016-02-24 | 华中科技大学 | Based on the object dimensional surface shape measurement method of GPU |
| CN103994732A (en) * | 2014-05-29 | 2014-08-20 | 南京理工大学 | Three-dimensional measurement method based on fringe projection |
| CN103994732B (en) * | 2014-05-29 | 2016-08-17 | 南京理工大学 | A kind of method for three-dimensional measurement based on fringe projection |
| CN104990517A (en) * | 2015-06-30 | 2015-10-21 | 东莞市盟拓光电科技有限公司 | Inclined projection correction method and system in structured light three-dimensional measurement |
| CN104990517B (en) * | 2015-06-30 | 2018-11-02 | 东莞市盟拓光电科技有限公司 | Oblique projection bearing calibration in structural light three-dimensional measurement and system |
| CN105243280A (en) * | 2015-10-30 | 2016-01-13 | 西安电子科技大学 | Time domain physical optics algorithm based on CPU (Central Processing Unit) and GPU (Graphics Processing Unit) hybrid asynchronous parallel way |
| CN105243280B (en) * | 2015-10-30 | 2018-02-02 | 西安电子科技大学 | Time domain physical optical computing method based on CPU Yu the mixed asynchronous parallel modes of GPU |
| WO2017193477A1 (en) * | 2016-05-09 | 2017-11-16 | 中国科学院深圳先进技术研究院 | Method and device for processing three-dimensional medical image data |
| CN111899326A (en) * | 2020-06-18 | 2020-11-06 | 苏州小优智能科技有限公司 | Three-dimensional reconstruction method based on GPU parallel acceleration |
| CN112258378A (en) * | 2020-10-15 | 2021-01-22 | 武汉易维晟医疗科技有限公司 | Real-time three-dimensional measurement system and method based on GPU acceleration |
| CN114765660A (en) * | 2020-12-30 | 2022-07-19 | 浙江舜宇智能光学技术有限公司 | Image calibration method for TOF camera module and FPGA chip |
| CN114937159A (en) * | 2022-07-27 | 2022-08-23 | 成都飞机工业（集团）有限责任公司 | Binocular matching method based on GPU acceleration |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102538709A (en) | Method for utilizing GPU (Graphics Processing Unit) concurrent computation in three-dimensional measurement system based on structured light | |
| CN110349251B (en) | Three-dimensional reconstruction method and device based on binocular camera | |
| CN109949899B (en) | Image three-dimensional measurement method, electronic device, storage medium, and program product | |
| CN107170043B (en) | A kind of three-dimensional rebuilding method | |
| Jeon et al. | Accurate depth map estimation from a lenslet light field camera | |
| CN100520285C (en) | Vision measuring method for projecting multiple frequency grating object surface tri-dimensional profile | |
| CN107917701A (en) | Measuring method and RGBD camera systems based on active binocular stereo vision | |
| CN106525004A (en) | Binocular stereo vision system and depth measuring method | |
| CN110223222B (en) | Image stitching method, image stitching device, and computer-readable storage medium | |
| WO2010133007A1 (en) | Techniques for rapid stereo reconstruction from images | |
| CN105654547A (en) | Three-dimensional reconstruction method | |
| CN108122280A (en) | The method for reconstructing and device of a kind of three-dimensional point cloud | |
| CN103335611A (en) | Method for GPU-based object three-dimensional shape measurement | |
| Li et al. | An efficient photogrammetric stereo matching method for high-resolution images | |
| CN106500625A (en) | A kind of telecentricity stereo vision measuring apparatus and its method for being applied to the measurement of object dimensional pattern micron accuracies | |
| CN109584157A (en) | Object plane degree measurement method and device, storage medium, electronic metering equipment | |
| Zicari | Efficient and high performance FPGA-based rectification architecture for stereo vision | |
| CN116129037A (en) | Visual touch sensor, three-dimensional reconstruction method, system, equipment and storage medium thereof | |
| Ding et al. | Improved real-time correlation-based FPGA stereo vision system | |
| CN115294277B (en) | Three-dimensional reconstruction method and device of object, electronic equipment and storage medium | |
| Hess et al. | FPGA-based phase measuring profilometry system | |
| Junger et al. | FPGA-accelerated phase rectification for a stereo-based phase measuring profilometry system | |
| CN111462199A (en) | Rapid speckle image matching method based on GPU | |
| Kong | 3D image reconstruction of marine plankton based on virtual reality | |
| Zhang et al. | Passive 3D reconstruction based on binocular vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20120704 |