Moving object real-time three-dimensional reconstruction system based on fringe projection

Abstract

A moving object real-time three-dimensional reconstruction system based on fringe projection comprises a processing system (PS end) used for communicating with a camera and a projector and managing acquisition and processing of image data; the programmable logic (PL end) is matched with the processing system (PS end) and is specially used for accelerating three-dimensional reconstruction, the programmable logic (PL end) comprises a three-dimensional reconstruction algorithm hardware acceleration unit, and a real-time three-dimensional reconstruction algorithm of a moving object in a dynamic scene is executed in a hardware acceleration mode; wherein the programmable logic (PL end) adopts a streaming computing architecture, the received image data is processed through a cache-free streaming design, the image data is analyzed immediately when each frame of image arrives, and a high-quality moving object real-time three-dimensional reconstruction result is continuously generated. When the method is used for processing three-dimensional reconstruction in a dynamic scene, high precision can be kept, high reconstruction speed can be realized, and challenges on reconstruction precision and reconstruction speed in the prior art are effectively solved.

Description

Technical Field

[0001] The invention relates to a real-time three-dimensional reconstruction technology, and in particular to a real-time three-dimensional reconstruction system for a moving object based on fringe projection. Background Art

[0002] In the field of computer vision and image processing, real-time 3D reconstruction technology is an important research direction and is widely used in many fields such as virtual reality, augmented reality, robot navigation and industrial quality inspection. Fringe projection is a commonly used 3D reconstruction method, which uses a projection device to project a known fringe pattern onto the surface of an object, and obtains the deformed fringe image through a camera device, and then calculates the 3D information of the object through image analysis technology.

[0003] The goal of current 3D reconstruction systems is to achieve efficient reconstruction speed and superior reconstruction accuracy. Most systems use chips such as FPGA for data processing to significantly speed up the calculation speed. In these systems, image acquisition is completed by high-speed cameras, which can capture rapidly changing scenes at a high frame rate. The projection device is responsible for accurately projecting complex stripe patterns onto the surface of the object. The data is quickly processed by FPGA to calculate high-precision 3D reconstruction data.

[0004] However, the current fringe projection-based 3D reconstruction system still faces some challenges, mainly in the following two aspects: reconstruction accuracy and reconstruction speed. First, most current 3D reconstruction systems are mainly optimized for static objects. When the system is applied to moving objects, the reconstruction accuracy is significantly reduced. This is because the surface features of moving objects change in a short period of time, and the speed of fringe pattern deformation exceeds the camera's capture capability, resulting in inaccurate 3D reconstruction results. Secondly, existing 3D reconstruction systems generally adopt a design of caching images first and then performing 3D reconstruction. Although this strategy can ensure relatively comprehensive data collection, it also brings significant problems: on the one hand, this method takes up a lot of storage resources, especially at high frame rates and high resolution settings, the data volume is even larger; on the other hand, the data handling and processing links often become the bottleneck of system performance, slowing down the overall 3D reconstruction speed. Therefore, in application scenarios with fast feedback and real-time processing, this design greatly limits the effectiveness and scope of application of the system.

[0005] It should be noted that the information disclosed in the above background technology section is only used for understanding the background of the present application, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the invention

[0006] The main purpose of the present invention is to overcome the defects existing in the above-mentioned background technology and provide a real-time three-dimensional reconstruction system of moving objects based on fringe projection.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A real-time 3D reconstruction system for moving objects based on fringe projection, comprising:

[0009] The processing system (PS side) is used to communicate with the camera and projector and manage the acquisition and processing of image data;

[0010] A programmable logic (PL side), in cooperation with the processing system (PS side), is specifically used for accelerating 3D reconstruction. The programmable logic (PL side) includes a 3D reconstruction algorithm hardware acceleration unit, which executes a real-time 3D reconstruction algorithm for moving objects in a dynamic scene by means of hardware acceleration;

[0011] Among them, the programmable logic (PL side) adopts a streaming computing architecture, processes the received image data through a cacheless streaming design, analyzes the image data instantly when each frame of the image arrives, and continuously generates high-quality real-time three-dimensional reconstruction results of moving objects.

[0012] Furthermore, the three-dimensional reconstruction algorithm hardware acceleration unit includes: a fixed-point number calculation unit, which is used to replace floating-point number calculation with fixed-point number calculation, optimize the calculation resource occupancy and improve the calculation accuracy.

[0013] Furthermore, the 3D reconstruction algorithm hardware acceleration unit also includes: a data bit width allocation unit, which is used to accurately calculate and allocate the data bit width required for each value, and to represent the value with the minimum bit width while meeting the algorithm accuracy requirement.

[0014] Furthermore, the cache-free streaming design of the streaming computing architecture includes:

[0015] The image data rearrangement unit is used to rearrange the image data, gather pixels at the same position into a group, and optimize the streaming processing efficiency.

[0016] Furthermore, the image data rearrangement unit comprises:

[0017] Direct transfer module, used to transfer pixels at the same position of each image as a group directly to the programmable logic (PL side) in the processing system (PS side), without storing the image data in the BRAM cache;

[0018] A Gray code pattern projection unit is used to project a Gray code pattern onto the surface of a target object, where each pattern corresponds to a unique Gray code value, thereby roughly inferring the approximate depth range of the pixel point;

[0019] The phase shift code pattern projection unit is used to project a series of phase shift code patterns with gradually changing phase differences onto the surface of the target object after determining the approximate depth interval of the pixel point, and accurately calculate its depth information according to the changing rules.

[0020] Furthermore, the cache-free streaming design of the streaming computing architecture includes:

[0021] The single-row pixel window processing unit is used to convert the data stream from a multi-row window pixel form to a single-row pixel window form, eliminating the data dependency between rows and avoiding the bottleneck caused by waiting for the previous row of data during processing.

[0022] Furthermore, the single-row pixel window processing unit includes:

[0023] The jump point detection module is used to detect the jump point in the 3D reconstruction calculation process. The gray value in the window centered on the 3D reconstruction point is checked by using a 1*n window smoothing method to determine whether the point is greater than the average value of the remaining points to identify the jump point.

[0024] The jump point correction module, for the detected jump points, the PL end only performs jump checks in the horizontal direction to correct most of the jump points, while the jump points in the vertical direction are corrected in the processing system (PS end).

[0025] Furthermore, the streaming computing architecture adopts a multi-stage pipeline based on a cacheless streaming design, subdividing the reconstruction process into multiple processing stages, each of which independently processes a specified task to optimize the use of data flow and computing resources.

[0026] Furthermore, each processing stage of the multi-stage pipeline has a double cache mechanism, allowing data to be read in one cache area while data is written and processed in another cache area, so as to reduce processing delays and improve system throughput.

[0027] The present invention has the following beneficial effects:

[0028] The present invention aims at the problem of low accuracy of three-dimensional reconstruction in motion scenes, and proposes an efficient solution based on hardware implementation, which implements the three-dimensional reconstruction algorithm of motion scenes in hardware and maintains the high-precision advantage of the algorithm. For the problem of three-dimensional reconstruction speed, the present invention proposes a cache-free streaming computing architecture, which can effectively reduce data handling. When each frame of image arrives, the PL end can immediately analyze it, thereby continuously generating high-quality three-dimensional reconstruction results and avoiding the use of a large amount of on-chip storage resources. Compared with the traditional background technology, the present invention has the advantages of higher accuracy and faster speed. Since the three-dimensional reconstruction algorithm of motion scenes is implemented in hardware, and the data bit width is carefully designed, the accuracy of three-dimensional reconstruction in hardware is guaranteed to be equivalent to that of three-dimensional reconstruction in software. By using a cache-free streaming computing architecture, the image data of each frame is processed in real time, the data handling time is reduced, the speed of three-dimensional reconstruction is improved, and the resource usage is reduced.

[0029] Furthermore, in the hardware implementation process of the 3D reconstruction algorithm, fixed-point calculations are used to replace floating-point calculations in the software algorithm, accurately calculate the data bit width required for each number, and allocate a reasonable storage bit width to ensure that each value is represented with the minimum bit width while still meeting the accuracy requirements of the algorithm.

[0030] Furthermore, the multi-stage pipeline design allows each processing stage to work in parallel, thereby improving the overall throughput of the system. The double buffer design makes the processing of each stage almost unaffected by the processing time of the previous and next stages, greatly reducing processing delays.

[0031] Other beneficial effects of the embodiments of the present invention will be further described below. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 4 is an architecture diagram of a system for real-time 3D reconstruction of moving objects based on fringe projection according to an embodiment of the present invention.

[0033] Figure 2 FIG. 4 is a schematic diagram of an image data rearrangement unit according to an embodiment of the present invention.

[0034] Figure 3 Schematic diagram of a single-row pixel window processing unit according to an embodiment of the present invention. Figure 4 FIG. 4 is a schematic diagram of a Gray code pattern projection according to an embodiment of the present invention. Figure 5 FIG. 4 is a schematic diagram of a phase shift code pattern projection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following is a detailed description of the embodiments of the present invention. It should be emphasized that the following description is only exemplary and is not intended to limit the scope and application of the present invention.

[0036] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0037] The principle of fringe projection 3D reconstruction is mainly based on projecting fringe patterns onto the surface of an object and obtaining the deformation effect of the object surface through a camera. By analyzing the changes in the projected fringes, the 3D shape of the object can be restored. However, in moving scenes, due to the rapid movement of objects and environmental changes, the accuracy of 3D reconstruction is often reduced, and traditional algorithms repeatedly move data when processing data, resulting in slow reconstruction speed. In order to overcome these challenges, the present invention addresses the problem of reduced 3D reconstruction accuracy in moving scenes, proposes to implement a 3D reconstruction algorithm for moving scenes in hardware, and maintains the high-precision advantage of the algorithm. In addition, for the problem of 3D reconstruction speed, the present invention proposes a cache-free streaming computing architecture design, which can effectively reduce data handling. When each frame of the image arrives, the PL end can immediately analyze it, thereby continuously generating high-quality 3D reconstruction results and avoiding the use of a large amount of on-chip storage resources.

[0038] See also Figure 1 The embodiment of the present invention provides a real-time 3D reconstruction system for moving objects based on fringe projection, including: a processing system (PS end), which is used to communicate with a camera and a projector to manage the collection and processing of image data; a programmable logic (PL end), which cooperates with the processing system (PS end) and is specifically used for accelerating 3D reconstruction. The programmable logic (PL end) includes a 3D reconstruction algorithm hardware acceleration unit, which executes the real-time 3D reconstruction algorithm for moving objects in dynamic scenes by hardware acceleration; wherein the programmable logic (PL end) adopts a streaming computing architecture, processes the received image data by a cacheless streaming design, analyzes the image data in real time when each frame of the image arrives, and continuously generates high-quality real-time 3D reconstruction results for moving objects.

[0039] In a preferred embodiment, the 3D reconstruction algorithm hardware acceleration unit may include a fixed-point number calculation unit, which is used to replace floating-point number calculation with fixed-point number calculation, optimize the computing resource usage and improve the computing accuracy. The 3D reconstruction algorithm hardware acceleration unit may also include a data bit width allocation unit, which is used to accurately calculate and allocate the data bit width required for each value, and represent the value with the minimum bit width while meeting the algorithm accuracy requirements.

[0040] In a preferred embodiment, the cacheless streaming design of the streaming computing architecture may include an image data rearrangement unit for rearranging the image data, concentrating pixels at the same position in a group, and optimizing the streaming processing efficiency. Particularly preferably, the image data rearrangement unit includes: a direct transmission module for transmitting pixels at the same position of each image as a group directly to the programmable logic (PL side) in the processing system (PS side), without storing the image data in the BRAM cache; a Gray code pattern projection unit for projecting Gray code patterns onto the surface of the target object, each pattern corresponding to a unique Gray code value, so as to roughly infer the approximate depth interval of the pixel point; a phase shift code pattern projection unit for projecting a series of phase shift code patterns with gradually changing phase differences onto the surface of the target object after determining the approximate depth interval of the pixel point, and accurately calculating its depth information according to the changing law.

[0041] In a preferred embodiment, the cacheless streaming design of the streaming computing architecture may also include a single-row pixel window processing unit, which is used to convert the data stream from a multi-row window pixel form to a single-row pixel window form, eliminate inter-row data dependencies, and avoid bottlenecks caused by waiting for the previous row of data during processing. Particularly preferably, the single-row pixel window processing unit includes: a jump point detection module, which is used to detect jump points during the three-dimensional reconstruction calculation process, using a 1*n window smoothing method to check the grayscale value in the window centered on the three-dimensional reconstruction point, and determine whether the point is greater than the average value of the remaining points to identify the jump point; a jump point correction module, for the detected jump point, the PL end only performs a jump check in the horizontal direction to correct most of the jump points, and the jump points in the vertical direction are corrected in the processing system (PS end).

[0042] In a preferred embodiment, the streaming computing architecture adopts a multi-stage pipeline based on a cacheless streaming design, subdividing the reconstruction process into multiple processing stages, each of which independently processes a specified task to optimize the use of data flow and computing resources. Furthermore, each processing stage of the multi-stage pipeline has a double cache mechanism, allowing data to be read in one cache while data is written and processed in another cache, so as to reduce processing delays and improve system throughput.

[0043] The present invention provides an efficient real-time 3D reconstruction system for moving objects based on fringe projection, which significantly improves the speed and accuracy of 3D reconstruction through the design of hardware acceleration and streaming computing architecture. The programmable logic (PL end) is introduced to work closely with the processing system (PS end) to execute the real-time 3D reconstruction algorithm of moving objects in dynamic scenes through hardware acceleration. The programmable logic (PL end) uses a non-cached streaming computing architecture to instantly analyze the image data arriving at each frame and continuously generate high-quality 3D reconstruction results, avoiding the delay and resource consumption caused by caching a large amount of image data in traditional designs.

[0044] Furthermore, the use of fixed-point calculations instead of traditional floating-point calculations not only optimizes the use of computing resources, but also improves the calculation accuracy. Through precise data bit width allocation, each value is ensured to be represented with the minimum bit width, which meets the algorithm accuracy requirements while reducing the demand for storage resources.

[0045] Furthermore, the rearrangement of image data and single-row pixel window processing further optimizes the data flow, eliminates inter-row data dependencies, reduces processing bottlenecks, and improves processing efficiency.

[0046] In addition, the introduction of a multi-stage pipeline design and a double-buffer mechanism allows each processing stage to work in parallel, greatly reducing processing delays and improving the overall throughput of the system.

[0047] A three-dimensional reconstruction system architecture of a specific embodiment is as follows: Figure 1 As shown in the figure, it is implemented based on the Zynq development board. The system is divided into two main parts: the processing system (PS side) and the programmable logic (PL side). On the PS side, the controller is responsible for communicating with the camera and projector, managing the acquisition and processing of image data. The camera captures the stripe projection images in the moving scene in real time and continuously transmits these images to the PL side for further processing. The PL side focuses on the acceleration of 3D reconstruction, uses the dynamic scene 3D reconstruction algorithm for hardware acceleration, and processes the received image data through a cacheless streaming design.

[0048] In the hardware implementation of 3D reconstruction algorithms, achieving a balance between high precision and low resource usage is a key challenge. To this end, fixed-point calculations are used to replace floating-point calculations in software algorithms, accurately calculate the data bit width required for each number, and allocate a reasonable storage bit width to ensure that each value is represented with the minimum bit width while still meeting the accuracy requirements of the algorithm.

[0049] In the streaming design of the 3D reconstruction algorithm, more efficient streaming processing is achieved by rearranging the image data and reducing data dependencies at the algorithm level. The image rearrangement design groups pixels at the same position into a group, such as Figure 3As shown in Figure 1, it avoids storing large blocks of image data in the cache of the PL. Gray code is a special binary encoding method, such as Figure 4 As shown, the Gray code pattern is projected onto the surface of the target object. Each pattern corresponds to a unique Gray code value, so as to roughly infer the approximate depth interval of the pixel point. After determining the approximate depth interval of the pixel point, a series of phase shift code patterns with gradually changing phase differences are used to project onto the surface of the target object, such as Figure 5 As shown, the depth information is accurately calculated according to the change rule. The image rearrangement method used in the prior art is to divide the large image into small blocks, such as Figure 2 As shown, several rows of each image are extracted as a group and cached in the BRAM at the PL end. Each time, the pixels at the same position of each image are read from the BRAM as a group for subsequent 3D reconstruction calculation. The present invention adopts an image rearrangement method as shown in FIG. Figure 3 , in the processing system (PS side), the pixels at the same position of each image are grouped together and transmitted to the PL side for direct subsequent 3D reconstruction calculations without the need for BRAM cache.

[0050] In addition, the data stream originally processed in the form of multi-row window pixels is converted into a single-row pixel window form. This optimization effectively eliminates the data dependency between rows, ensuring that there is no bottleneck during processing due to waiting for the previous row of data, thus avoiding multi-row cache problems such as Figure 3 As shown. In the 3D reconstruction calculation process, some reconstructed 3D points will be separated from the original surface, which are called jump points. In order to reduce the occurrence of jump points, n*n window smoothing is often used, that is, to detect whether the point is greater than the average value of the remaining points in each window centered on the 3D reconstruction point. If it is greater, it is a jump point, so there is inter-row data dependency. In this regard, the present invention optimizes the n*n window smoothing at the PL end to 1*n window smoothing. Since the jump point is easy to calculate at the junction of light and dark stripes, Gray code and phase shift code are both vertical stripes. Generally, the grayscale value is continuous in the vertical direction and discontinuous in the horizontal direction. PL only needs to perform horizontal jump checks to correct most of the jump points, and the vertical jump points are corrected at the PS end.

[0051] The present invention proposes an efficient solution combined with hardware implementation to solve the problems of reduced accuracy and slow speed of three-dimensional reconstruction in motion scenes. Through the three-dimensional reconstruction system built on the Zynq development board, the stripe projection images in the motion scene are acquired and processed in real time by utilizing the division of labor between the processing system (PS end) and the programmable logic (PL end). In the design, fixed-point number calculation and precise data bit width allocation are adopted to achieve a balance between high accuracy and low resource occupancy. At the same time, a cache-free streaming computing architecture is introduced to reduce data handling and improve processing speed. In addition, the efficiency of streaming processing is optimized by rearranging image data and eliminating inter-row data dependencies.

[0052] In order to further improve the efficiency and accuracy when processing 3D reconstruction in dynamic scenes, a multi-stage pipeline design scheme is proposed based on the streaming design. This scheme divides the 3D reconstruction process into multiple stages, each of which independently handles specific tasks to optimize the use of data flow and computing resources. On the basis of the 3D reconstruction process being disassembled into multiple independent processing stages, a double buffer mechanism is further adopted for each processing stage. This means that there are two buffers for processing intermediate data in each stage, so that data can be read in one buffer while data is written and processed in the other buffer, ensuring that there is no waiting or delay during processing. The advantage of this scheme is that the multi-stage pipeline design allows each processing stage to work in parallel, thereby improving the overall throughput of the system. The double buffer design makes the processing of each stage almost unaffected by the processing time of the previous and next stages, greatly reducing the processing delay.

[0053] In summary, the present invention proposes an efficient solution based on hardware implementation for fringe projection 3D reconstruction technology, which has the advantages of higher accuracy and faster speed compared to traditional background technologies. First, since the 3D reconstruction algorithm for motion scenes is implemented in hardware, and the data bit width is carefully designed, the 3D reconstruction accuracy of the hardware is guaranteed to be equivalent to that of the software. Second, a non-cached streaming computing architecture is adopted to process the image data of each frame in real time, reduce data transfer time, increase the 3D reconstruction speed, and reduce resource usage.

[0054] When processing three-dimensional reconstruction in dynamic scenes, the present invention can not only maintain high precision but also achieve fast reconstruction speed, effectively solving the challenges of reconstruction accuracy and speed faced by the prior art, especially in application scenarios of fast feedback and real-time processing, significantly improving the effectiveness and application scope of the system.

[0055] The above content is a further detailed description of the present invention in combination with specific / preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, without departing from the concept of the present invention, it can also make several substitutions or modifications to these described embodiments, and these substitutions or modifications should be regarded as belonging to the protection scope of the present invention. In the description of this specification, the description of the reference terms "an embodiment", "some embodiments", "preferred embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in combination with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily target the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In the absence of mutual contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples. Although the embodiments of the present invention and their advantages have been described in detail, it should be understood that various changes, substitutions and modifications can be made herein without departing from the scope of protection of the patent application.

Claims

1. A real-time 3D reconstruction system for moving objects based on fringe projection, characterized in that: include: The processing system (PS side) is used to communicate with the camera and projector and manage the acquisition and processing of image data; A programmable logic (PL side), in cooperation with the processing system (PS side), is specifically used for accelerating 3D reconstruction. The programmable logic (PL side) includes a 3D reconstruction algorithm hardware acceleration unit, which executes a real-time 3D reconstruction algorithm for moving objects in a dynamic scene by means of hardware acceleration; Among them, the programmable logic (PL side) adopts a streaming computing architecture, processes the received image data through a cacheless streaming design, analyzes the image data instantly when each frame of the image arrives, and continuously generates high-quality real-time three-dimensional reconstruction results of moving objects.

2. The real-time 3D reconstruction system for moving objects based on fringe projection according to claim 1, characterized in that: The three-dimensional reconstruction algorithm hardware acceleration unit includes: a fixed-point number calculation unit, which is used to replace floating-point number calculation with fixed-point number calculation, optimize the calculation resource occupancy and improve the calculation accuracy.

3. The real-time 3D reconstruction system for moving objects based on fringe projection according to claim 2, characterized in that: The three-dimensional reconstruction algorithm hardware acceleration unit also includes: a data bit width allocation unit, which is used to accurately calculate and allocate the data bit width required for each value, and to represent the value with the minimum bit width while meeting the algorithm accuracy requirements.

4. The real-time 3D reconstruction system for moving objects based on fringe projection according to any one of claims 1 to 3, characterized in that: The cache-free streaming design of the streaming computing architecture includes: The image data rearrangement unit is used to rearrange the image data, gather pixels at the same position into a group, and optimize the streaming processing efficiency.

5. The real-time 3D reconstruction system for moving objects based on fringe projection according to claim 4, characterized in that: The image data rearrangement unit comprises: Direct transfer module, used to transfer pixels at the same position of each image as a group directly to the programmable logic (PL side) in the processing system (PS side), without storing the image data in the BRAM cache; A Gray code pattern projection unit is used to project a Gray code pattern onto the surface of a target object, where each pattern corresponds to a unique Gray code value, thereby roughly inferring the approximate depth range of the pixel point; The phase shift code pattern projection unit is used to project a series of phase shift code patterns with gradually changing phase differences onto the surface of the target object after determining the approximate depth interval of the pixel point, and accurately calculate its depth information according to the changing rules.

6. The real-time 3D reconstruction system for moving objects based on fringe projection according to any one of claims 1 to 5, characterized in that: The cache-free streaming design of the streaming computing architecture includes: The single-row pixel window processing unit is used to convert the data stream from a multi-row window pixel form to a single-row pixel window form, eliminating the data dependency between rows to avoid bottlenecks caused by waiting for the previous row of data during processing.

7. The real-time 3D reconstruction system for moving objects based on fringe projection according to claim 6, characterized in that: The single-row pixel window processing unit comprises: The jump point detection module is used to detect the jump point in the 3D reconstruction calculation process. The gray value in the window centered on the 3D reconstruction point is checked by using a 1*n window smoothing method to determine whether the point is greater than the average value of the remaining points to identify the jump point. The jump point correction module, for the detected jump points, the PL end only performs jump checks in the horizontal direction to correct most of the jump points, while the jump points in the vertical direction are corrected in the processing system (PS end).

8. The real-time 3D reconstruction system for moving objects based on fringe projection according to any one of claims 1 to 7, characterized in that: The streaming computing architecture adopts a multi-stage pipeline based on a cacheless streaming design, subdividing the reconstruction process into multiple processing stages, each of which independently processes a specified task to optimize the use of data flow and computing resources.

9. The real-time 3D reconstruction system for moving objects based on fringe projection according to claim 8, characterized in that: Each processing stage of the multi-stage pipeline has a double cache mechanism, allowing data to be read in one cache area while data is written and processed in another cache area, so as to reduce processing delay and improve system throughput.

Images