Optical domain depth detection system and method thereof

Abstract

The invention provides an optical domain depth detection system and a method thereof. The optical domain depth detection system comprises a laser light source, a first lens, an all-optical Gaussian Laplacian calculator and a signal detection system, and the laser light source is used for emitting laser and irradiating the laser on a three-dimensional object to be detected; the first lens is used for imaging a three-dimensional object to obtain a first imaging light field; the all-optical Gaussian-like Laplacian calculator is used for completing Gaussian-like filtering and Laplacian operation on the first imaging light field in an optical domain so as to obtain a second imaging light field after Gaussian-like Laplacian operation; and the signal detection system is used for detecting the second imaging light field so as to obtain depth information of the three-dimensional object. According to the invention, the real-time performance of deep recognition can be greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computational optics, and in particular to an optical domain depth detection system and method thereof. BACKGROUND

[0002] The Laplacian of Gaussian (LoG) operator was first proposed by Marr as a means for image edge detection. The idea is to first apply a Gaussian filter to the image to suppress noise interference, and then perform a Laplacian operation to enhance regions with significant intensity changes in the image. In this way, the edge structure with sharp gray level changes in the image can be effectively highlighted.

[0003] Currently, the LoG operator is basically implemented in the digital domain, which has high computational cost and insufficient real-time performance. Specifically, the LoG is essentially a convolution operation, which usually requires a large computational overhead, especially for high-pixel images to be processed and large-size convolution kernels. For edge computing scenarios with low computing resources and high response requirements, how to reduce the computational energy consumption and latency of the LoG operator becomes a key technical problem to be solved.

[0004] In addition, in the traditional depth recognition task of three-dimensional objects, LoG filtering needs to be performed on each frame, further occupying a large amount of computing resources, leading to speed bottlenecks and high power consumption, and it is difficult to achieve high-speed real-time depth information extraction. If the Laplacian operation is directly performed on the original data, a large number of false edges will often be generated due to noise. Therefore, a larger Gaussian kernel is usually used to smooth and denoise first, and then the Laplacian operator is applied. Although this "smoothing first and then differentiating" method reduces noise interference, it also introduces new problems: a large-size convolution kernel is needed for Gaussian filtering to effectively suppress noise, and an excessively large convolution kernel not only has higher computational overhead, but also causes excessive blurring of image details and textures. SUMMARY

[0005] The purpose of the embodiments of the present application is to provide an optical domain depth detection system and method thereof, which can at least solve at least one of the technical problems mentioned in the prior art above.

[0006] One aspect of the present application provides an optical depth detection system. The optical depth detection system comprises a laser light source, a first lens, a full-optical Gaussian Laplacian calculator, and a signal detection system. The laser light source is configured to emit laser light and irradiate on a three-dimensional object to be measured. The first lens is configured to image the three-dimensional object to obtain a first imaging light field. The full-optical Gaussian Laplacian calculator is configured to perform Gaussian filtering and Laplacian operation on the first imaging light field in an optical domain to obtain a second imaging light field after Gaussian Laplacian operation. The signal detection system is configured to detect the second imaging light field to obtain depth information of the three-dimensional object.

[0007] Further, the full-optical Gaussian Laplacian calculator comprises a second lens and a Laplacian solver. The second lens is configured to perform Gaussian smoothing filtering on the first imaging light field. The Laplacian solver is configured to perform second-order differential operation on the Gaussian smoothing filtered light field.

[0008] Further, the aperture size of the second lens is continuously adjusted according to the noise interference of the first imaging light field to adjust the smoothing degree of the image. The larger the noise interference of the first imaging light field, the smaller the aperture size of the second lens is adjusted. Conversely, the smaller the noise interference of the first imaging light field, the larger the aperture size of the second lens is adjusted.

[0009] Further, the object distance and image distance of the second lens and the focal length of the second lens satisfy the following relationship: 1 / z3+1 / z4=1 / f2, where z3 represents the object distance of the second lens, z4 represents the image distance of the second lens, and f2 represents the focal length of the second lens.

[0010] Further, the signal detection system comprises a third lens and a detector. The third lens is configured to image the second imaging light field processed by the full-optical Gaussian Laplacian calculator onto the detector. The detector is configured to convert the second imaging light field from an optical domain signal to an electrical domain signal.

[0011] Further, the optical depth detection system further comprises a post-processing module. The post-processing module is configured to store multiple frames of detection results of the signal detection system, and determine the depth information of the three-dimensional object according to the multiple frames of detection results. Each time the parameter of the first lens is changed, a frame of detection result is obtained.

[0012] Another aspect of this application provides a method for optical depth detection. The method includes: illuminating a three-dimensional object to be measured with a laser; imaging the three-dimensional object through a first lens to obtain a first imaging light field; performing Gaussian-like filtering and Laplace operation on the first imaging light field in the optical domain using an all-optical Gaussian-Laplacian calculator to obtain a second imaging light field after Gaussian-Laplacian operation; detecting the second imaging light field using a signal detection system; and obtaining the depth information of the three-dimensional object based on the detection results.

[0013] Furthermore, the detection method further includes: changing the parameters of the first lens to obtain multiple frames of the detection results, wherein each change of the parameters of the first lens corresponds to obtaining one frame of the detection results, and obtaining the depth information of the three-dimensional object based on the detection results includes: performing post-processing on the multiple frames of the detection results to determine the depth information of the three-dimensional object.

[0014] Furthermore, changing the parameters of the first lens includes: driving the all-optical Gaussian Laplace calculator and the signal detection system behind it to move within a predetermined range via an electronic control device to change the image distance of the first lens; or keeping the image distance of the first lens constant while continuously changing the focal length of the first lens.

[0015] Furthermore, the detection results include sharpness data characterizing the degree of blur in the imaging of the three-dimensional object. The post-processing of the multiple frames of the detection results to determine the depth information of the three-dimensional object includes: for each pixel in the imaging, finding the frame index value with the highest imaging sharpness corresponding to that pixel from the multiple frames of the detection results; and obtaining the depth distribution information of the three-dimensional object based on the detection results corresponding to the frame index value with the highest imaging sharpness of all pixels.

[0016] Furthermore, the all-optical Gaussian-Laplace calculator includes a second lens and a Laplace solver. The step of performing Gaussian-like filtering and Laplace operation on the first imaging light field in the optical domain using the all-optical Gaussian-Laplace calculator to obtain the second imaging light field after Gaussian-like Laplace operation includes: performing Gaussian-like smoothing filtering on the first imaging light field through the second lens; and performing second-order differential operation on the Gaussian-like smoothed light field through the Laplace solver to obtain the second imaging light field after Gaussian-like Laplace operation.

[0017] Furthermore, the detection method further includes: continuously adjusting the aperture size of the second lens according to the noise interference received by the first imaging light field to adjust the smoothness of the image, wherein the greater the noise interference received by the first imaging light field, the smaller the aperture size of the second lens is adjusted; conversely, the less the noise interference received by the first imaging light field, the larger the aperture size of the second lens is adjusted.

[0018] The optical domain depth detection system and method of one or more embodiments of this application complete the depth extraction of objects in an all-optical manner, performing most of the computations in the depth extraction in the optical domain, thereby significantly improving the real-time performance of depth recognition. Since the calculations are mainly performed using the propagation and modulation of light, no additional power supply is required to drive the computing unit, resulting in extremely low overall system power consumption. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the optical path of an optical domain depth detection system according to an embodiment of this application.

[0020] Figure 2 This is a flowchart of an optical domain depth detection method according to an embodiment of this application. Detailed Implementation

[0021] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses consistent with some aspects of this application as detailed in the appended claims.

[0022] The optical domain depth detection system and method of this application will now be described in detail with reference to the accompanying drawings. Unless otherwise specified, the features of the following embodiments and implementations can be combined with each other.

[0023] Figure 1 A schematic diagram of the optical path of an optical domain depth detection system 100 according to an embodiment of this application is shown. Figure 1 As shown, an embodiment of the optical depth detection system 100 of this application includes a laser source (not shown), a first lens 110, an all-optical Gaussian Laplace calculator 120, and a signal detection system 130.

[0024] A laser source is used to emit laser light and illuminate the three-dimensional object 200 to be measured. The three-dimensional object 200 may, for example, consist of a series of scattering particles dispersed in three-dimensional space.

[0025] The first lens 110 is used to image the three-dimensional object 200 to obtain a first imaging light field. The first imaging light field carries image information of the three-dimensional object 200.

[0026] The all-optical Gaussian-Laplace calculator 120 is used to perform Gaussian-like filtering and Laplace operation on the first imaging light field in the optical domain to obtain the second imaging light field after Gaussian-like Laplace operation.

[0027] The signal detection system 130 is used to detect the second imaging light field in order to obtain depth information of the three-dimensional object 200.

[0028] In some embodiments, the all-optical Gaussian-Laplace calculator 120 of this application includes a second lens 121 and a Laplace solver 122.

[0029] The second lens 121 is a small-aperture lens with adjustable parameters; its focal length and aperture size can be optimized according to specific system requirements. The second lens 121 is used to perform Gaussian-like smoothing filtering on the first imaging light field, achieving a Gaussian-like blur effect. The bandwidth of the filter can be controlled by adjusting the aperture size of the second lens 121, thereby achieving different degrees of smoothing processing on the image to be processed. This method can approximate the low-pass filtering effect of Gaussian convolution, thereby suppressing high-frequency noise in the image, smoothing surface textures, and retaining only the main structural information of the image, providing a smooth and stable input light field for subsequent Laplacian differentiation operations.

[0030] It is worth noting that, according to Fourier optics theory, for a lens with a finite aperture, when the imaging law is satisfied, its point spread function is the Sombrero function (the sinc function in two-dimensional polar coordinates). Therefore, the imaging result of this lens can be regarded as the convolution result of the input light field by a convolution kernel shaped like the Sombrero function, not a Gaussian convolution in the strict sense. However, the shape of the Sombrero function is approximately Gaussian, and after convolution by this kernel, it can still have a blurring effect on the input image. Therefore, in this application, a second lens 121 with a finite aperture is used to simulate the Gaussian filtering part in the LoG operator.

[0031] In some embodiments, the aperture size of the second lens 121 can be continuously adjusted according to the noise interference experienced by the first imaging light field, thereby adjusting the smoothness of the image.

[0032] The greater the noise interference to the first imaging light field, the smaller the aperture of the second lens 121 should be; conversely, the less noise interference to the first imaging light field, the larger the aperture of the second lens 121 should be. For example, if the noise interference to the first imaging light field carrying the image information of the three-dimensional object 200 is relatively small, the aperture of the second lens 121 can be increased to expand the passage range of low-frequency spatial components, thereby reducing the smoothness of the image. If the noise interference to the first imaging light field carrying the image information of the three-dimensional object 200 is relatively large, the aperture of the second lens 121 can be decreased to reduce the passage range of low-frequency spatial components, thereby increasing the smoothness of the image.

[0033] The aperture of the second lens 121 can be adjusted by designing apertures with different structures. The principle is to place a light-blocking metal plate in front of the second lens 121. The metal plate can have various shapes, with classic shapes including pin-type apertures, rotating apertures, and iris diaphragm blade apertures. These apertures can control the rotation and movement of the metal plate mechanically and / or electrically, thus conveniently controlling the size of the light-passing aperture.

[0034] The Laplacian solver 122 is an optical device that can realize the Laplacian operator in the optical domain. Placed behind the second lens 121, the Laplacian solver 122 performs Laplacian (i.e., second-order differential) operations on the Gaussian-smoothed light field, enhancing details in the image with drastic brightness variations (such as edges and textures). The Laplacian solver 122 is a specially designed filter whose transfer function is parabolic, thus enabling differential operations in two-dimensional space. Currently, there are many implementations of the Laplacian solver 122, with common structures including 4f systems, photonic crystals, and multilayer slabs. By optimizing the above optical structures, a specific transfer function can be realized in the optical spatial frequency domain, thereby achieving the Laplacian operation. It should be noted that this application does not limit the specific structure and implementation of the Laplacian solver 122; any optical device capable of realizing Laplacian calculations is within the scope of protection of this application.

[0035] It is important to emphasize that the all-optical Gaussian Laplace calculator 120 of this application performs calculations entirely in the optical domain. All the optical components are passive optical devices, requiring no additional energy to complete the calculations. Furthermore, the entire calculation process is highly parallel: every pixel of the input image participates in the calculation simultaneously. Compared to LoG in the electrical digital domain, this application significantly improves computational efficiency.

[0036] This application achieves LoG-like computations in the optical domain by using the second lens 121 in conjunction with the Laplacian solver 122. The all-optical Gaussian-Laplacian calculator 120 of this application can process image information at extremely high speeds, directly performing convolution calculations based on the propagation and diffraction of light in the medium. The processing speed is close to the speed of light, eliminating pixel-by-pixel serial calculation delays and requiring only one light transmission process to complete the filtering operation. Therefore, it possesses ultra-high throughput and near-zero system latency.

[0037] Furthermore, the all-optical Gaussian Laplace calculator 120 of this application also has a significant advantage in ultra-low power consumption. Because it relies on passive optical elements such as lenses to modulate the light field, it does not require electrical power to drive itself and only relies on the photon energy of the input light to complete the calculation, resulting in power consumption far lower than traditional solutions that require digital processors.

[0038] It is worth mentioning that the all-optical Laplacian Gaussian calculator 120 of this application inherently possesses large-scale parallel processing capabilities. Regardless of the image resolution, the all-optical Laplacian Gaussian calculator 120 of this application can process all pixels in parallel at the same time, and its calculation speed does not decrease linearly with the increase of resolution, making it particularly suitable for the fast processing needs of real-time high-resolution scenes.

[0039] The all-optical Laplace-Gaussian calculator 120 of this application combines high speed, low power consumption, and parallel processing characteristics, and has broad application prospects in the fields of computational optics and artificial intelligence. In scenarios such as security monitoring and autonomous driving, the all-optical Laplace-Gaussian calculator 120 of this application can serve as a front-end processing unit to quickly extract target contours and edge information in the scene. In more complex tasks, the all-optical Laplace-Gaussian calculator 120 of this application can also effectively compress data in an analog manner, retaining only the key contour features of objects, thereby achieving more efficient artificial visual perception.

[0040] The optical depth detection system 100 of this application extracts optical depth based on an all-optical Laplace Gaussian calculator 120. By continuously changing the imaging parameters of the first lens 110, images with different degrees of blur can be obtained sequentially in time, forming a focal stack, which encodes the depth information of the three-dimensional object 200 into the blur of the image. The focal stack refers to a series of two-dimensional images captured by gradually changing the focal length or image distance of the first lens 110 for a static scene. This series of images maintains strict consistency in all other shooting parameters, with the only variable being the focal length or image distance of the first lens 110. Subsequently, the above imaging results are sequentially passed through the all-optical Laplace Gaussian calculator 120, and the intensity information of the spatial light field is recorded by the signal detection system 130. The recorded light intensity signal reflects the blur degree of the three-dimensional object 200 image. By finding the frame with the smallest blur in the focal stack as the criterion for clear imaging, the depth of the object is inferred based on the corresponding imaging parameters.

[0041] It is understood that the image mentioned in this application may refer to an imaging light field carrying an image of a three-dimensional object 200.

[0042] Compared with existing technologies, the optical domain depth detection system 100 of this application uses an all-optical approach to extract the depth of an object, performing most of the calculations in the depth extraction process in the optical domain, thereby significantly improving the real-time performance of depth recognition. Since the calculations are mainly performed using the propagation and modulation of light, no additional power supply is required to drive the computing unit, resulting in extremely low overall system power consumption.

[0043] Furthermore, the optical modules involved in the optical domain depth detection system 100 of this application have simple structures, making them easy to integrate with existing imaging sensors and optical systems. Based on the above advantages, the optical domain depth detection system 100 of this application is very suitable for application in three-dimensional perception scenarios with stringent requirements for speed and power consumption, such as environmental depth perception in autonomous driving, robot vision navigation, UAV mapping, and augmented reality, etc., and can provide real-time depth information with extremely high efficiency, thus having significant application value.

[0044] To ensure that the convolution process in the all-optical Gaussian Laplacian calculator 120 effectively achieves image smoothing, the object distance and image distance of the second lens 121 are set to satisfy the imaging law, namely:

[0045] 1 / z³ + 1 / z⁴ = 1 / f²

[0046] Where z3 represents the object distance of the second lens 121, z4 represents the image distance of the second lens 121, and f2 represents the focal length of the second lens 121.

[0047] In this configuration, the point spread function of the second lens 121 with a finite aperture is a Somprero function, and its shape is close to that of a Gaussian function. Their similarity is mainly reflected in three aspects: First, they both have a unique, highest central peak at the origin; second, their shapes are both radially symmetrical, and the function value depends only on the distance r from the center point; third, as the distance r from the center point increases, the function value rapidly decays to zero.

[0048] In some embodiments, the signal detection system 130 includes a third lens 131 and a detector 132. The third lens 131 is used to image the second imaging light field processed by the all-optical Gaussian Laplace calculator 120 onto the detector 132. The detector 132 is used to convert the second imaging light field from an optical domain signal to an electrical domain signal. The detector 132 may employ a high-resolution imaging sensor (such as a CCD / CMOS (Charge-Coupled Device / Complementary Metal-Oxide-Semiconductor) camera) to record the intensity distribution of the light field. The LoG calculation results of each image in the focus stack are acquired by the detector 132 and digitally stored.

[0049] In some embodiments, the optical depth detection system 100 of this application further includes a post-processing module 140. The post-processing module 140 is used to store multiple frames of detection results from the signal detection system 130 and determine the depth information of the three-dimensional object 200 based on the multiple frames of detection results, wherein each change in the parameters of the first lens 110 corresponds to obtaining one frame of detection results.

[0050] This application allows for the modification of the imaging parameters of the first lens 110 in two ways to obtain a focus stack.

[0051] The first method involves keeping the focal length of the first lens 110 constant while continuously changing the image distance z2 of the first lens 110, thereby sequentially obtaining focal stacks at different image distances. For each given image distance z2, the first lens 110 will generate a corresponding image plane 300 on its image side. If the object distance and image distance satisfy the imaging condition at this time, that is:

[0052] 1 / z1 + 1 / z2 = 1 / f1

[0053] Where z1 represents the object distance of the first lens 110, z2 represents the image distance of the first lens 110, and f1 represents the focal length of the first lens 110.

[0054] If the image on the image plane 300 is in sharp focus, then the image will be in sharp focus; otherwise, it will be out of focus and blurry. By gradually scanning the image distance z2, a multi-frame image sequence covering different depth planes of the three-dimensional object 200 can be obtained, where one frame corresponds to a sharp image of the object at a specific depth, while the other depths will be out of focus and blurry.

[0055] It is worth noting that, in the aforementioned optical calculation process, to ensure optimal filtering for each image plane 300, the all-optical Gaussian-like Laplacian calculator 120 and its subsequent signal detection system 130 need to move synchronously with the position of the image plane 300. Specifically, each time an image is formed on a new image plane 300, the second lens 121, along with the subsequent Laplacian solver 122 and signal detection system 130, translates along the optical axis, ensuring that the second lens 121 is always aligned with the current image plane 300 and accurately projects its image onto the Laplacian solver 122. Thus, regardless of the location of the image plane 300, the input of the all-optical Gaussian-like Laplacian calculator 120 is always at the conjugate position of that image plane 300, thereby ensuring the accuracy of Gaussian-like smoothing and Laplacian calculation. The relative positions of the components (from the second lens 121 to the signal detection system 130) inside the all-optical Gaussian Laplace calculator 120 remain unchanged. Therefore, the optical calculation parameters (such as the smoothing filtering characteristics of the small-aperture second lens 121, the differential characteristics of the Laplace solver 122, etc.) remain consistent throughout the scanning process, achieving a unified processing standard for image frames at different depths.

[0056] In the first approach, a high-precision and high-speed electronic control device (not shown) is required to drive the all-optical Gaussian Laplace calculator 120 at high speed, enabling it to move accurately and quickly within a certain range, thereby achieving the detection of LoG images at different image distances.

[0057] The second method involves keeping the image distance z2 of the first lens 110 constant while continuously changing the focal length f1 of the first lens 110. This also yields a series of focus stacks with varying degrees of blur. Similarly, when the focal length f1 of the first lens 110 is adjusted to satisfy the imaging condition 1 / z1 + 1 / z2 = 1 / f1, the image on the image plane 300 is in sharp focus; otherwise, it is out of focus and blurred. By progressively scanning the focal length f1 of the first lens 110, a multi-frame image sequence covering different depth planes of the three-dimensional object 200 can be obtained. One frame corresponds to a sharp image of the object at a specific depth, while the other depths are out of focus and blurred.

[0058] In the second approach, the focal length f1 of the first lens 110 can be adjusted through various methods. For example, liquid lens technology can be used. Specifically, this involves adjusting the focal length by changing the shape or pressure of the liquid, providing rapid and precise adjustment. Another method is to adjust the focal length by changing the temperature and influencing the refractive index of the material. This application does not limit the specific method for adjusting the focal length f1 of the first lens 110; any possible adjustment method is within the scope of protection of this application.

[0059] Compared to the first method, the second method does not require moving the all-optical Gaussian Laplace calculator 120 and the signal detection system 130 behind it. The positions of the components in the entire optical domain depth detection system 100 are relatively fixed, making it relatively easy to implement and resulting in a more compact system size.

[0060] like Figure 1 As shown, a laser source illuminates a three-dimensional object 200, and the light is imaged at the image plane 300 through a first lens 110. The focal length f1 or image distance z2 of the first lens 110 is an adjustable parameter. Each change in the parameter of the first lens 110 yields an image result on the image plane 300. By scanning the focal length f1 of the first lens 1101 or changing the image distance z2, multiple frames of image results can be obtained at the image plane 300, forming a focus stack. Note that the results of this focus stack are not stored; they are merely an intermediate result of the optical system. Each imaging signal in the focus stack is then sequentially fed into an all-optical Gaussian-Laplace calculator 120 for optical domain LoG calculation. The all-optical Gaussian-Laplace calculator 120 includes a second lens 121 and a Laplace solver 122, which respectively perform Gaussian kernel convolution calculation and Laplace operation. Through the optical domain LoG processing method, the sharpness measure of each pixel in each frame of the image in the focus stack is obtained. The measurement result was detected by the signal detection system 130.

[0061] The optical depth detection system 100 of this application is ultimately able to record the sharpness result at each pixel of all imaging results in the focus stack. For example, the detection results of detector 132 are stored as a three-dimensional array of shape (h, w, N), where h and w are the number of pixels in the height and width directions of the image, respectively, and N is the number of image frames in the focus stack. This three-dimensional array is then fed into the post-processing module 140 to obtain the depth prediction result of the three-dimensional object 200.

[0062] The post-processing module 140 processes the following: First, for each pixel in the image, it iterates pixel by pixel to find the frame index value with the highest sharpness corresponding to each pixel in the N frames of sharpness measurement results. Then, based on the detection result corresponding to the frame index value with the highest sharpness corresponding to each pixel and the configuration parameters of the third lens 131, the corresponding focal length or image distance is obtained. Finally, the object distance is obtained through the imaging formula, which is the depth distribution information of the three-dimensional object 200.

[0063] This application also provides a method for optical domain depth detection. Figure 2 A flowchart illustrating an embodiment of the optical domain depth detection method of this application is provided. Figure 2 As shown, an embodiment of the optical domain depth detection method of this application may include steps S1 to S5.

[0064] In step S1, the laser is irradiated onto the three-dimensional object 200 to be tested.

[0065] In step S2, the three-dimensional object 200 is imaged through the first lens 110 to obtain the first imaging light field.

[0066] In step S3, the all-optical Gaussian-Laplace calculator 120 performs Gaussian-Laplace filtering and Laplace operation on the first imaging light field in the optical domain to obtain the second imaging light field after Gaussian-Laplace operation.

[0067] The all-optical Gaussian-Laplacian calculator 120 includes a second lens 121 and a Laplacian solver 122. Therefore, in some embodiments, step S3, which involves performing Gaussian-like filtering and Laplacian calculation on the first imaging light field in the optical domain using the all-optical Gaussian-Laplacian calculator 120 to obtain the second imaging light field after Gaussian-Laplacian calculation, may further include steps S31 and S32.

[0068] In step S31, the first imaging light field is subjected to Gaussian-like smoothing filtering through the second lens 121.

[0069] In step S32, the second-order differential operation is performed on the Gaussian-like smoothed light field by the Laplacian solver 122 to obtain the second imaging light field after the Gaussian-like Laplacian operation.

[0070] In some embodiments, the optical domain depth detection method of this application may further include step S33.

[0071] In step S33, the aperture size of the second lens 121 can be continuously adjusted according to the noise interference received by the first imaging light field to adjust the smoothness of the image. The greater the noise interference received by the first imaging light field, the smaller the aperture size of the second lens 121 is adjusted; conversely, the smaller the noise interference received by the first imaging light field, the larger the aperture size of the second lens 121 is adjusted.

[0072] In step S4, the second imaging light field is detected by a signal detection system.

[0073] In step S5, the depth information of the three-dimensional object 200 is obtained based on the detection results.

[0074] In some embodiments, the optical domain depth detection method of this application further includes steps S6 and S7.

[0075] In step S6, it is determined whether the predetermined number of probes has been completed. If the result of the determination is "yes", the process proceeds to step S5. Otherwise, the process continues to step S7.

[0076] In step S7, the parameters of the first lens 110 are changed, and then steps S2 to S4 above are continued until a predetermined number of detections are completed, thereby obtaining multiple frames of detection results. Each time the parameters of the first lens 110 are changed, one frame of detection result is obtained.

[0077] Step S5, obtaining the depth information of the three-dimensional object 200 based on the detection results, may include: post-processing the detection results of multiple frames to determine the depth information of the three-dimensional object 200.

[0078] In some embodiments, changing the parameters of the first lens 110 in step S7 may include: driving the all-optical Gaussian Laplace calculator 120 and the signal detection system 130 behind it to move within a predetermined range via an electronic control device to change the image distance of the first lens 110.

[0079] In other embodiments, the step S7 of changing the parameters of the first lens 110 may include: keeping the image distance of the first lens 110 constant while continuously changing the focal length of the first lens 110.

[0080] The detection results include sharpness data used to characterize the degree of blur in the imaging of the three-dimensional object 200. Post-processing of the multi-frame detection results to determine the depth information of the three-dimensional object 200 may further include steps S51 and S52.

[0081] In step S51, for each pixel (x, y) in the image, the frame index value with the highest imaging sharpness corresponding to that pixel is found from the multi-frame detection results.

[0082] In step S52, the depth distribution information of the three-dimensional object 200 is obtained based on the detection result corresponding to the frame index value with the highest imaging clarity among all pixels.

[0083] The optical domain depth detection method of this application has beneficial technical effects that are generally similar to those of the optical domain depth detection system 100 described above, so it will not be repeated here.

[0084] The optical domain depth detection system and method provided in the embodiments of this application have been described in detail above. Specific examples have been used to illustrate the optical domain depth detection system and method of this application. The descriptions of the embodiments above are only for helping to understand the core ideas of this application and are not intended to limit this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the spirit and principles of this application, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A light-domain depth detection system, characterized in that, It includes a laser source, a first lens, an all-optical Gaussian-Laplace calculator, and a signal detection system, among which, The laser source is used to emit laser light and illuminate the three-dimensional object to be measured; The first lens is used to image the three-dimensional object to obtain a first imaging light field; The all-optical Gaussian-Laplacian calculator is used to perform Gaussian-like filtering and Laplacian operation on the first imaging light field in the optical domain to obtain the second imaging light field after Gaussian-Laplacian operation. The signal detection system is used to detect the second imaging light field in order to obtain the depth information of the three-dimensional object.

2. The optical domain depth detection system as described in claim 1, characterized in that, The all-optical Gaussian-Laplace calculator includes a second lens and a Laplace solver, wherein... The second lens is used to perform Gaussian-like smoothing filtering on the first imaging light field; The Laplace solver is used to perform second-order differential operations on the light field after Gaussian-like smoothing filtering.

3. The optical domain depth detection system as described in claim 2, characterized in that, The aperture size of the second lens is continuously adjusted according to the noise interference experienced by the first imaging light field to adjust the smoothness of the image. The greater the noise interference to the first imaging light field, the smaller the aperture of the second lens should be; conversely, the less noise interference to the first imaging light field, the larger the aperture of the second lens should be.

4. The optical domain depth detection system as described in claim 2, characterized in that, The object distance and image distance of the second lens, as well as the focal length of the second lens, satisfy the following relationship: 1 / z³ + 1 / z⁴ = 1 / f² Where z3 represents the object distance of the second lens, z4 represents the image distance of the second lens, and f2 represents the focal length of the second lens.

5. The optical domain depth detection system as described in claim 1, characterized in that, The signal detection system includes a third lens and a detector, wherein... The third lens is used to image the second imaging light field processed by the all-optical Gaussian Laplace calculator onto the detector; The detector is used to convert the second imaging light field from an optical domain signal to an electrical domain signal.

6. The optical domain depth detection system as described in any one of claims 1 to 5, characterized in that, It also includes a post-processing module, which is used to store the multi-frame detection results of the signal detection system and determine the depth information of the three-dimensional object based on the multi-frame detection results. Each time the parameters of the first lens are changed, a corresponding frame of the detection result is obtained.

7. A method for optical domain depth detection, characterized in that, include: Illuminate the three-dimensional object to be tested with a laser; The three-dimensional object is imaged through the first lens to obtain a first imaging light field; The first imaging light field is subjected to Gaussian filtering and Laplace operation in the optical domain by an all-optical Gaussian Laplace calculator to obtain the second imaging light field after Gaussian Laplace operation. The second imaging light field is detected by a signal detection system; The depth information of the three-dimensional object is obtained based on the detection results.

8. The optical domain depth detection method as described in claim 7, characterized in that, Also includes: The parameters of the first lens are changed to obtain multiple frames of the detection results, wherein each change in the parameters of the first lens corresponds to obtaining one frame of the detection results. The process of obtaining the depth information of the three-dimensional object based on the detection results includes: The depth information of the three-dimensional object is determined by post-processing the detection results of multiple frames.

9. The optical domain depth detection method as described in claim 8, characterized in that, The changes to the parameters of the first lens include: The all-optical Gaussian Laplace calculator and the signal detection system behind it are moved within a predetermined range by an electronic control device to change the image distance of the first lens; or Keeping the image distance of the first lens constant while continuously changing the focal length of the first lens.

10. The optical domain depth detection method as described in claim 8, characterized in that, The detection results include sharpness data characterizing the degree of blur in the 3D object image, and the post-processing of multiple frames of the detection results to determine the depth information of the 3D object includes: For each pixel in the image, find the frame index value with the highest image clarity corresponding to that pixel from the detection results of multiple frames; The depth distribution information of the three-dimensional object is obtained based on the detection result corresponding to the frame index value with the highest imaging clarity among all pixels.

11. The optical domain depth detection method as described in claim 7, characterized in that, The all-optical Gaussian-Laplace calculator includes a second lens and a Laplace solver. The step of performing Gaussian-Laplace filtering and Laplace operations on the first imaging light field in the optical domain using the all-optical Gaussian-Laplace calculator to obtain the second imaging light field after Gaussian-Laplace operation includes: The first imaging light field is subjected to Gaussian-like smoothing filtering through the second lens; The second-order differential operation is performed on the Gaussian-like smoothed light field by the Laplacian solver to obtain the second imaging light field after Gaussian-like Laplacian operation.

12. The optical domain depth detection method as described in claim 11, characterized in that, Also includes: The aperture size of the second lens is continuously adjusted based on the noise interference received by the first imaging light field to adjust the smoothness of the image. The greater the noise interference to the first imaging light field, the smaller the aperture of the second lens should be; conversely, the less noise interference to the first imaging light field, the larger the aperture of the second lens should be.

Images