Surround view detection method and system with natural surface near vegetation as relay surface

Abstract

This invention relates to a method and system for look-through detection using natural surfaces near vegetation as relay surfaces, and pertains to the fields of lidar remote sensing and single-photon detection imaging technology. The purpose of this application is to address the problem that existing vegetation penetration techniques can only utilize direct echoes and cannot handle semi-transparent, time-varying obstruction layers. This application illuminates a point on the relay surface using a fixed-direction point illumination method, allowing diffuse reflection light from this point to bypass vegetation and illuminate the obscured target. A single-photon array detector receives the echo photons, and a photon time-of-flight histogram is generated for each pixel of the single-photon array detector. Indirect echo peaks that are diffusely reflected from the relay surface to the obscured target and then return through vegetation gaps are identified and extracted from the photon time-of-flight histograms corresponding to each pixel. Based on the photon time-of-flight corresponding to the indirect echo peaks of the target and the geometric information of the relay surface, the spatial distribution of the obscured target is reconstructed.

Description

Technical Field

[0001] This application belongs to the field of lidar remote sensing and single-photon detection imaging technology. Background Technology

[0002] In reconnaissance, patrol, search and rescue, and forest operations, targets such as personnel, vehicles, and equipment are often obscured by tree canopies and shrubs. Conventional photoelectric detection methods, such as visible light cameras and infrared cameras, are completely ineffective at imaging targets obscured by vegetation because light waves cannot penetrate dense foliage. In contrast, lidar, as an active detection technology, uses laser pulses to scan the scene and record the echoes, and has become the primary detection method in this field.

[0003] The basic principle of existing lidar for detecting targets obscured by vegetation is to use laser pulses to penetrate the gaps between branches and leaves to reach the ground or target behind the vegetation and generate echoes. By recording multiple echoes of the same laser beam, distance information at multiple depth levels can be obtained, including the vegetation canopy, branches, and the obscured ground or target. Chinese patent CN107479045A (filed June 29, 2017), which describes a method for removing low-lying vegetation based on full-waveform point clouds, and Chinese patent CN104915958A (filed June 4, 2015), which describes a method for identifying vegetation from airborne laser point clouds, both follow this approach. They classify and process multi-echo point clouds or full-waveform data to remove vegetation echoes and retain echoes from the ground or target. However, this type of technology, which relies on direct penetration, has significant limitations. When vegetation is dense and there are no light-transmitting gaps in the line-of-sight between the detector and the target, the laser pulse is completely intercepted and reflected by the vegetation canopy before reaching the target, failing to generate any direct echo from the obscured target. In this situation, existing penetration methods based on multi-echo or full-waveform analysis collapse and become completely ineffective. More fundamentally, existing methods are all post-processing operations, only able to improve the quality of observational data when gaps naturally exist. They inherently lack the ability to actively bypass vegetation obstruction, and therefore cannot cope with extreme obstruction situations where gaps are completely absent.

[0004] Closely related to, yet distinct from, the aforementioned direct-penetration technology is non-line-of-sight (NLOS) imaging technology. The typical principle of this technology is to utilize a wall at a corner as a light relay surface. A laser first illuminates the wall, and the diffused light from the wall bypasses the corner to illuminate the target obscured by the wall. The scattered light from the target returns to the wall and is received by the detector. After photons undergo three bounces—from the wall to the target and back to the wall—the obscured target is reconstructed using a back-projection algorithm. Chinese patent CN106526612A (application date: December 15, 2016), entitled "Scanning Photon Counting Non-Line-of-Sight 3D Imaging Device," and Chinese patent CN120831675A (application date: September 18, 2025), entitled "Adaptive Windowing Low-Photon Non-Line-of-Sight Imaging Method," both employ this basic principle. However, existing NLOS imaging technologies have several shortcomings, preventing their direct application to outdoor detection scenarios obscured by vegetation. First, existing non-line-of-sight (NLOS) imaging techniques assume that the relay surface is a rigid, static, ideal diffuse reflective wall. This assumption is largely valid in urban corner imaging scenarios, but in outdoor vegetated environments, only natural surfaces such as the ground, tree trunks, and rocks can be used as relay surfaces. The reflective characteristics of these natural surfaces differ significantly from those of artificial walls, requiring independent modeling to accurately describe photon reflection behavior. Second, existing NLOS imaging techniques assume that obstructions are completely opaque walls. Therefore, the echo signal only contains the indirect path component from the wall to the target and back to the wall, eliminating the difficulty of signal source differentiation. However, in vegetated obstruction scenarios, vegetation is a semi-transparent layer. When there are gaps in the branches and leaves along the line of sight between the detector and the target, both direct echoes and indirect echoes from natural surfaces will coexist in the received signal. These two components from different physical sources are mixed together, and existing NLOS imaging methods have never considered the separation of different path components in the mixed echo. Therefore, it is impossible to extract effective target information from such complex echoes. Third, existing non-line-of-sight imaging technologies generally employ single-point detectors combined with mechanical scanning to acquire data point-to-point from different points on the relay surface. This serial acquisition method is time-consuming and requires the entire scene to remain stationary during the acquisition process. However, outdoor vegetation sways irregularly due to wind, causing the length of the indirect propagation path to change over time. If a scanning acquisition method with a long cumulative time is used, path mismatch will occur between signals acquired at different times, and the cumulative imaging will result in blurry reconstruction results.

[0005] In summary, existing vegetation penetration techniques can only utilize the direct echo of laser pulses passing through multiple layers of vegetation. When the vegetation is dense, with few or no gaps, the signal-to-noise ratio of the direct echo drops sharply or even disappears completely. Furthermore, existing non-line-of-sight imaging techniques rely on rigid, static artificial walls and completely opaque obstructions, failing to address issues such as variations in the reflectivity of natural surfaces in the field, the mixing of direct and indirect echoes caused by semi-transparent vegetation, and time-varying paths due to vegetation movement. Therefore, how to utilize the reflection from natural surfaces near vegetation to create an indirect light propagation path under vegetation-covered conditions, allowing photons to bypass the vegetation and illuminate the obscured target, enabling the target echo to return to the detector through vegetation gaps, and separating the indirect echo component carrying target information from the complex mixed echo containing direct vegetation echoes and relay surface echoes, thereby reconstructing the obscured target, has become a pressing technical problem to be solved in this field. Summary of the Invention

[0006] This application aims to address the problems of existing vegetation penetration technologies, which can only utilize direct echoes, resulting in a sharp drop in the signal-to-noise ratio of direct echoes when vegetation is dense and gaps are sparse; and the limitations of existing NLOS technologies, which rely on rigid, static walls and opaque obstructions, in dealing with semi-transparent and time-varying obstruction layers like vegetation. The application provides a look-around detection method that uses a natural surface near vegetation (ground, tree trunks, rocks, etc.) as a light relay surface, illuminates this relay surface with point illumination, and uses diffuse reflection light from the relay surface to illuminate the target. The target echo returns through the gaps in the vegetation, forming an indirect path (look-around path) for detecting targets obscured by vegetation. A single-photon array detection system is also provided to implement this method.

[0007] The first aspect of this application provides a look-around detection method using a natural surface near vegetation as a relay surface, including:

[0008] Illumination points on the relay surface are illuminated with fixed-point illumination, so that the diffuse reflected light from the illumination points can bypass the vegetation and illuminate the obscured target. The relay surface is the bare natural surface near the vegetation in the area in front of the target.

[0009] A single-photon array detector is used to receive echo photons, and a photon time-of-flight histogram is generated for each pixel of the single-photon array detector.

[0010] Identify and extract the indirect echo peaks that are diffusely reflected from the relay surface to the occluded target and then return through the vegetation gaps from the photon time-of-flight histogram corresponding to each pixel.

[0011] Based on the photon flight time corresponding to the indirect echo peak of the target and the geometric information of the relay surface, the spatial distribution of the obscured target is reconstructed.

[0012] In one possible design, the photon time-of-flight histogram corresponding to each pixel includes:

[0013] Direct echo peaks from the vegetation cover layer, direct echo peaks from the relay surface, and indirect echo peaks from the covered target;

[0014] The echo peaks are classified according to the photon flight time, and the echo peak that appears latest is identified as the indirect echo peak of the obscured target.

[0015] In one possible design, the classification of echo peaks according to photon flight time includes:

[0016] The height difference of each echo peak relative to its left and right sides of the base is used as the significance criterion, and echo peaks that are lower than the preset significance threshold are removed.

[0017] Based on the geometric prior of the scene, three time windows are divided: direct echo from vegetation, direct echo from relay surface, and indirect echo from the target. The retained echo peaks are then classified according to their respective time windows.

[0018] In one possible design, defining the area in space where the occluded target is located includes:

[0019] Based on the flight time corresponding to the direct echo peak of the relay surface and combined with the illumination direction, the three-dimensional coordinates of the sampling points corresponding to each pixel on the relay surface are calculated.

[0020] Using the illumination point and the sampling point corresponding to each pixel as spatial constraint endpoints, and the flight time corresponding to the indirect echo peak of the occluded target as a constraint condition, the area where the occluded target is located is defined in space.

[0021] In one possible design, the method of look-around detection as described in claim 1 is executed from multiple different stations, and the reconstruction results of each station are unified into the same coordinate system for joint processing, and the fused spatial distribution of the occluded target is output.

[0022] In one possible design, the point illumination is: focusing a laser into a dotted spot to illuminate the illuminated point, with the laser direction fixed and no scanning performed during the illumination and acquisition process.

[0023] The second aspect of this application provides a look-around detection system that uses the natural surface near vegetation as a relay surface, including: a single-photon array detection unit, a pulsed laser emission unit, and a processing unit;

[0024] The single-photon array detection unit is used to receive echo photons and generate a photon flight time histogram for each pixel.

[0025] The pulsed laser emitting unit is used to illuminate the illumination point on the relay surface in a fixed-point illumination manner, so that the diffuse reflected light of the illumination point can bypass the vegetation and illuminate the obscured target. The relay surface is the bare natural surface near the vegetation in the area in front of the target.

[0026] The processing unit is used to identify and extract the indirect echo peaks that are diffusely reflected from the relay surface to the occluded target and then return through the vegetation gaps from the photon time-of-flight histogram corresponding to each pixel, and to reconstruct the spatial distribution of the occluded target based on the photon time of flight corresponding to the indirect echo peaks of the target and the geometric information of the relay surface.

[0027] In one possible design, the single-photon array detection unit includes: an optical receiving system and a single-photon array detector.

[0028] In one possible design, the pulsed laser emitting unit includes: a pulsed laser and an optical emitting system;

[0029] The optical emission system includes a beam collimator, a focusing lens group, and a reflector, used to focus the laser beam emitted by the pulsed laser into a centimeter-scale spot at the output end.

[0030] In one possible design, the around-view detection system, which uses the natural surface near the vegetation as a relay surface, also includes a dual-axis motorized turntable;

[0031] The dual-axis electric turntable is used to support the pulsed laser emitting unit and adjust the fixed orientation of the pulsed laser emitting unit.

[0032] The beneficial effects of this application are:

[0033] 1. This application's point illumination concentrates laser energy into a small spot, forming a high-brightness local light source at the illumination point. The diffuse reflection light from this illumination point propagates in all directions of space with a large solid angle, thus geometrically bypassing the vegetation obstruction layer and illuminating the target from the side or rear. Since this indirect illumination path is completely independent of the existence of gaps between vegetation branches and leaves through which the laser can penetrate in a straight line, even when the vegetation is dense and the line of sight between the detector and the target is completely blocked by branches and leaves, eliminating any direct path, the target surface can still receive effective illumination from the diffuse reflection light from the natural surface. The target echo returns through the gaps in the vegetation and is received by the single-photon array detector, ensuring that even when the direct echo has completely disappeared, the completely obscured target still possesses a detectable echo signal independent of the direct path. Simultaneously, the fixed-point illumination method eliminates the need for any scanning operations during the entire illumination and acquisition process, overcoming the increased acquisition time and dependence on scene stillness inherent in point-by-point scanning illumination.

[0034] 2. Each pixel histogram contains photon echo peaks from different propagation paths. The peak with the shortest flight time comes from direct reflection from the vegetation layer, the peak with the middle flight time comes from the straight-line return echo of the laser at the illumination point on the natural surface, and the peak with the latest flight time comes from the complete around-the-look path of the photon after diffuse reflection from the natural surface to the target and then back through the vegetation gaps. The processing unit first performs background subtraction and matched filtering enhancement on the histogram to suppress noise and improve the detection capability of weak signals. Then, it uses the height difference of each echo peak relative to the base on its left and right sides as the significance criterion to retain weak peaks that are relatively prominent on the local base even if the absolute amplitude is low. Then, based on the system emission pointing angle and scene geometric prior, it delineates three time windows: direct echo from vegetation, direct echo from relay surface, and indirect echo from the target. Each echo peak is then classified according to its flight time interval. By combining peak saliency screening with geometric time window division, the weak target indirect echo peaks, originally submerged in the strong echo tail of vegetation, are effectively separated from the mixed signal, providing clean look-around path time-of-flight data for subsequent target reconstruction. This signal separation mechanism enables this application to function normally in the most general case where direct and indirect echoes coexist, without pre-assuming that the obstruction is a completely opaque wall to avoid the mixed signal problem.

[0035] 3. This application employs a single-photon area array detector to receive echo photons in parallel. Each pixel in the array independently generates its own photon time-of-flight histogram. Each pixel corresponds to a different sampling point on the natural surface in space. Therefore, a single laser emission and acquisition can simultaneously acquire the multi-channel look-around path information corresponding to all pixels. The processing unit independently completes echo peak separation and extracts the time of flight of the relay surface sampling point in the histogram of each pixel. The signal acquisition of each channel of the entire array is parallel in time, and there is no serial time accumulation caused by point-by-point scanning. Since the single-frame acquisition time is compressed to the microsecond to millisecond level, which is much smaller than the time scale of significant positional changes of vegetation due to wind swaying, the spatial configuration of vegetation branches and leaves can be regarded as approximately static during the acquisition process, and the propagation path length of the target indirect echo remains stable within the acquisition window. This acquisition speed is thousands of times faster than the serial acquisition method of single-point detector combined with mechanical scanning in existing non-line-of-sight imaging technologies, fundamentally alleviating the path time-varying and cumulative ambiguity problems caused by vegetation swaying, making look-around detection in dynamic environments in the field feasible.

[0036] 4. In this application, photons in the direct path travel only in a straight line between the laser and the target, resulting in the shortest propagation distance and flight time. Photons in the bypass path, however, must traverse a complete path from the laser to the illumination point on the natural surface, through diffuse reflection from the illumination point to the target, and finally back to the detector via vegetation gaps. This path has a significantly longer propagation distance and the longest flight time. The two types of paths are naturally separated in terms of flight time and can be extracted independently through multi-peak separation processing. Therefore, the bypass detection channel implemented in this application and the existing direct penetration detection channel through vegetation gaps do not interfere with each other at the signal level. The reconstruction results of both can be fused and output in the same coordinate system. When gaps exist, the direct channel provides high-precision target information, while the bypass channel supplements the side view. When gaps are missing, the bypass channel independently provides target detection capability, thus forming a complete detection capability for vegetation-covered scenes, from gap penetration to bypass coverage. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the optical path geometry for around-the-eye detection;

[0038] Figure 2 The flowchart shows a look-around detection method that uses the natural surface near vegetation as a relay surface.

[0039] Figure 3 This is a schematic diagram of a single-pixel histogram with a "multi-peak distribution". Detailed Implementation

[0040] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0041] Specific Implementation Method 1: This implementation method proposes a look-around detection method using the natural surface near vegetation as a relay surface, including:

[0042] Illumination points on the relay surface are illuminated with fixed-point illumination, so that the diffuse reflected light from the illumination points can bypass the vegetation and illuminate the obscured target. The relay surface is the bare natural surface near the vegetation in the area in front of the target.

[0043] A single-photon array detector is used to receive echo photons, and a photon time-of-flight histogram is generated for each pixel of the single-photon array detector.

[0044] Identify and extract the indirect echo peaks that are diffusely reflected from the relay surface to the occluded target and then return through the vegetation gaps from the photon time-of-flight histogram corresponding to each pixel.

[0045] Based on the photon flight time corresponding to the indirect echo peak of the target and the geometric information of the relay surface, the spatial distribution of the obscured target is reconstructed.

[0046] In one embodiment, the photon time-of-flight histogram corresponding to each pixel includes:

[0047] Direct echo peaks from the vegetation cover layer, direct echo peaks from the relay surface, and indirect echo peaks from the covered target;

[0048] The echo peaks are classified according to the photon flight time, and the echo peak that appears latest is identified as the indirect echo peak of the obscured target.

[0049] In one implementation, classifying the echo peaks according to photon flight time includes:

[0050] The height difference of each echo peak relative to its left and right sides of the base is used as the significance criterion, and echo peaks that are lower than the preset significance threshold are removed.

[0051] Based on the geometric prior of the scene, three time windows are divided: direct echo from vegetation, direct echo from relay surface, and indirect echo from the target. The retained echo peaks are then classified according to their respective time windows.

[0052] In one implementation, defining the area where the obscured target is located in space includes:

[0053] Based on the flight time corresponding to the direct echo peak of the relay surface and combined with the illumination direction, the three-dimensional coordinates of the sampling points corresponding to each pixel on the relay surface are calculated.

[0054] Using the illumination point and the sampling point corresponding to each pixel as spatial constraint endpoints, and the flight time corresponding to the indirect echo peak of the occluded target as a constraint condition, the area where the occluded target is located is defined in space.

[0055] In one embodiment, the look-around detection method using the natural surface near vegetation as a relay surface further includes executing the look-around detection method of claim 1 from multiple different stations, unifying the reconstruction results of each station into the same coordinate system for joint processing, and outputting the fused spatial distribution of the occluded target.

[0056] In one embodiment, the point illumination is: focusing a laser into a dotted spot to illuminate the illuminated point, wherein the laser direction is fixed and no scanning is performed during the illumination and acquisition process.

[0057] Specific implementation method two: This implementation method proposes a look-around detection system with the natural surface near the vegetation as the relay surface, including: a single photon array detection unit 1, a pulsed laser emission unit 2, a processing unit 3, and a pointing adjustment unit 7.

[0058] The single-photon array detection unit 1 consists of an optical receiving system and a 64×64 pixel InGaAs single-photon avalanche diode array camera. Each pixel of the array camera has an independent time-correlated single-photon counting function, which can record the arrival time of each photon and accumulate them into a histogram.

[0059] The pulsed laser emitting unit 2 uses a pulsed laser with a working wavelength of 1064nm, which has low background noise under sunlight. This pulsed laser requires an optical emitting system, which includes a beam collimator, focusing lens group, and reflector, to focus the laser beam into a small spot with a diameter on the order of centimeters at the output end for point illumination. The laser beam's direction remains fixed throughout the acquisition process, eliminating the need for any form of scanning of the relay surface.

[0060] The processing unit 3 is an embedded high-performance industrial control computer, which is connected to the single-photon array detection unit 1 via a cable. It is used to control the acquisition timing, receive histogram data and execute subsequent signal processing and reconstruction algorithms, and control the pointing angle of the turntable.

[0061] The pointing adjustment unit adopts a dual-axis electric turntable 7, which is used to support the pulsed laser emitting unit 2 and the single-photon array detection unit 1. The angular positioning accuracy of the turntable is better than 0.01 degrees. It is used to align the optical axis of the system with the target area and select a suitable relay surface illumination point before each acquisition begins, while the turntable remains stationary during the acquisition process.

[0062] The detection system is obscured by vegetation. In front of or to the side of the vegetation, there are often exposed natural surfaces, such as forest clearings, tree trunks, rocks, or roads, which are called relay surfaces. In this embodiment, the dual-axis motorized turntable 7 is first used for rough alignment, directing the laser of the optical emission system towards an exposed natural surface (mid-section) in front of the target area. This natural surface can be soil or gravel in a forest clearing, or it can be an exposed rock surface or the side of a large tree trunk. In this embodiment, a piece of exposed ground within a range of approximately 2m to 10m in front of the vegetation is selected as the relay surface.

[0063] The laser is directed to a specific illumination point on the bare ground using the turntable control software. This allows the laser pulse to undergo diffuse reflection at that point, enabling the reflected light to bypass the vegetation cover layer 4 at a large solid angle and illuminate the target area behind the vegetation. After the pointing adjustment is completed, the dual-axis motorized turntable 7 is locked, and the laser pointing will not change during the subsequent acquisition process.

[0064] During the acquisition process, the pulsed laser continuously illuminates the illumination point in a fixed direction, and the single-photon array detection unit 1 synchronously and continuously receives the echo photons. Each pixel in the 64×64 pixels independently generates a photon count time-of-flight histogram. The time window of the histogram covers the entire possible flight time range from the direct echo from the vegetation layer to the indirect echo from the target. The indirect echo from the target corresponds to the photon's path around the laser to the relay illumination point, diffuse reflection from the illumination point to the target, and scattering from the target back to the detector through the vegetation gaps.

[0065] After data acquisition, processing unit 3 performs echo peak separation processing on the histogram of each pixel. Specifically, background noise is first removed from the histogram, followed by matched filtering enhancement to improve the detection capability of weak signals. Subsequently, a multi-peak extraction pipeline is used, employing peak saliency (i.e., the height difference between the peak and its left and right bases) as the screening criterion to extract weak indirect echo peaks submerged in the strong echo tails of vegetation. The extracted echo peaks are then categorized according to their flight time from near to far. The categorization is based on the geometric priors of the system and the scene, thereby dividing the time windows into three categories: direct vegetation echo, relay surface echo, and target indirect echo. In this embodiment, the first peak with the shortest flight time is the direct vegetation echo peak, which is generated by the direct reflection of the laser pulse at the vegetation canopy. The second peak that appears next is the relay surface echo peak, which corresponds to the straight back-and-forth echo of the laser beam at the illumination point of the relay surface. This peak generally has strong energy. If there is a candidate peak that appears later than the relay surface echo peak in the histogram of the same pixel, and its significance meets the screening threshold, then the candidate peak is classified as the target indirect echo peak. This peak carries the distance information of the obscured target, has the longest flight time, and the weakest energy.

[0066] Processing unit 3 uses the flight time of the relay surface echo peaks, combined with the transmission pointing angle, to calculate the three-dimensional coordinates of the illumination point and its surrounding sampling points. Since each pixel in the array corresponds to a different sampling point on the relay surface, the relay surface echo peak data of the entire array of pixels naturally constitute the sampling grid of the relay surface, completing the geometric calibration of the relay surface. The accuracy of the relay surface calibration directly affects the accuracy of subsequent target reconstruction.

[0067] Subsequently, processing unit 3 uses the flight time of the target indirect echo peak extracted from each pixel as a constraint to determine the path length from the relay surface sampling point corresponding to that pixel to the target. Specifically, for each pixel out of all 64×64 pixels, processing unit 3 calculates an ellipsoidal shell constraint based on the flight time of the target indirect echo peak extracted from that pixel. The two foci of this ellipsoidal shell are the laser illumination point and the relay surface sampling point corresponding to that pixel, respectively. The target may be located at any position within this ellipsoidal shell. Processing unit 3 discretizes the entire detection space into a three-dimensional grid. For the ellipsoidal shell constraint of each pixel, it assigns a confidence increment to the grid points traversed by the ellipsoidal shell. The confidence increments of all 64×64 pixels are accumulated on the same three-dimensional grid to obtain the total confidence distribution of each grid point. The grid region with the highest confidence value represents the most likely spatial location and approximate shape range of the occluded target. Based on this, processing unit 3 can further sharpen the confidence distribution using optical cone transformation or virtual wave reconstruction algorithms to improve the reconstruction resolution and finally output a colored 3D point cloud of the occluded target. Throughout the acquisition and reconstruction process, the laser illumination direction is fixed and no scanning is required; the processing unit executes automatically, and the reconstruction result of the occluded target can be obtained in a single acquisition.

[0068] The specific method for target localization is as follows: For each pixel, the total optical path is obtained by multiplying the flight time corresponding to the indirect echo peak of the target by the speed of light. The optical path from the laser emitter to the illumination point and the optical path from the sampling point to the receiver of the pixel are subtracted to obtain the remaining optical path from the illumination point to the target and the sampling point. An ellipsoidal shell surface is constructed with the illumination point and the sampling point as two foci and the remaining optical path as a distance constant. The occluded target is constrained on the ellipsoidal shell surface. The confidence scores of the ellipsoidal shell surface corresponding to each pixel are accumulated on a discretized spatial grid, and the grid region with the highest confidence score is taken as the spatial position of the occluded target.

[0069] To further improve reconstruction resolution and resistance to swaying interference, this embodiment also employs a multi-station joint reconstruction method. A dual-axis motorized turntable or vehicle platform drives the detection system to repeatedly perform the aforementioned acquisition and reconstruction process on the same vegetation-covered area from multiple different angles or stations. The processing unit unifies the global pose information of each station into the world coordinate system. This pose information can come from the turntable encoder, global navigation satellite system, inertial measurement unit, or multi-station registration algorithm. Under the unified coordinate system, the reconstruction results obtained from the independent back-projection of each station are overlapped and combined, and finally, the fused reconstruction result of the covered target is output.

[0070] While specific embodiments of this application have been described herein with reference to them, it should be understood that these embodiments are merely examples of the principles and applications of this application. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of this application as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for sight-around detection using natural surfaces near vegetation as relay surfaces, characterized in that, include: Illumination points on the relay surface are illuminated with fixed-point illumination, so that the diffuse reflected light from the illumination points can bypass the vegetation and illuminate the obscured target. The relay surface is the bare natural surface near the vegetation in the area in front of the target. A single-photon array detector is used to receive echo photons, and a photon time-of-flight histogram is generated for each pixel of the single-photon array detector. Identify and extract the indirect echo peaks that are diffusely reflected from the relay surface to the occluded target and then return through the vegetation gaps from the photon time-of-flight histogram corresponding to each pixel. Based on the photon flight time corresponding to the indirect echo peak of the target and the geometric information of the relay surface, the spatial distribution of the obscured target is reconstructed.

2. The look-around detection method according to claim 1, using the natural surface near vegetation as a relay surface, is characterized in that, The photon time-of-flight histogram corresponding to each pixel includes: Direct echo peaks from the vegetation cover layer, direct echo peaks from the relay surface, and indirect echo peaks from the covered target; The echo peaks are classified according to the photon flight time, and the echo peak that appears latest is identified as the indirect echo peak of the obscured target.

3. The look-around detection method using the natural surface near vegetation as a relay surface according to claim 2, characterized in that, The classification of echo peaks according to photon flight time includes: The height difference of each echo peak relative to its left and right sides of the base is used as the significance criterion, and echo peaks that are lower than the preset significance threshold are removed. Based on the geometric prior of the scene, three time windows are divided: direct echo from vegetation, direct echo from relay surface, and indirect echo from the target. The retained echo peaks are then classified according to their respective time windows.

4. The look-around detection method using the natural surface near vegetation as a relay surface according to claim 3, characterized in that, Defining the area where the obscured target is located in space includes: Based on the flight time corresponding to the direct echo peak of the relay surface and combined with the illumination direction, the three-dimensional coordinates of the sampling points corresponding to each pixel on the relay surface are calculated. Using the illumination point and the sampling point corresponding to each pixel as spatial constraint endpoints, and the flight time corresponding to the indirect echo peak of the occluded target as a constraint condition, the area where the occluded target is located is defined in space.

5. The look-around detection method according to claim 1, using the natural surface near vegetation as a relay surface, is characterized in that, It also includes executing the look-around detection method of claim 1 from multiple different stations, unifying the reconstruction results of each station into the same coordinate system for joint processing, and outputting the fused spatial distribution of the occluded target.

6. The look-around detection method according to claim 1, using the natural surface near vegetation as a relay surface, is characterized in that, The point illumination is achieved by focusing a laser into a dotted spot to illuminate the illumination point, with the laser direction fixed and no scanning performed during the illumination and acquisition process.

7. A look-around detection system that uses natural surfaces near vegetation as relay surfaces, characterized in that, include: Single-photon array detection unit, pulsed laser emission unit, and processing unit; The single-photon array detection unit is used to receive echo photons and generate a photon flight time histogram for each pixel. The pulsed laser emitting unit is used to illuminate the illumination point on the relay surface in a fixed-point illumination manner, so that the diffuse reflected light of the illumination point can bypass the vegetation and illuminate the obscured target. The relay surface is the bare natural surface near the vegetation in the area in front of the target. The processing unit is used to identify and extract the indirect echo peaks that are diffusely reflected from the relay surface to the occluded target and then return through the vegetation gaps from the photon time-of-flight histogram corresponding to each pixel, and to reconstruct the spatial distribution of the occluded target based on the photon time of flight corresponding to the indirect echo peaks of the target and the geometric information of the relay surface.

8. The look-around detection system using the natural surface near vegetation as a relay surface according to claim 1, characterized in that, The single-photon array detection unit includes: an optical receiving system and a single-photon array detector.

9. The look-around detection system according to claim 1, using the natural surface near vegetation as a relay surface, is characterized in that, The pulsed laser emitting unit includes: a pulsed laser and an optical emitting system; The optical emission system includes a beam collimator, a focusing lens group, and a reflector, used to focus the laser beam emitted by the pulsed laser into a centimeter-scale spot at the output end.

10. The look-around detection system using the natural surface near vegetation as a relay surface according to claim 1, characterized in that, It also includes a dual-axis electric rotary table; The dual-axis electric turntable is used to support the pulsed laser emitting unit and adjust the fixed orientation of the pulsed laser emitting unit.

Images