Laser radar and detection method thereof

Abstract

The invention provides a laser radar comprising an emission device configured to emit a detection light beam for detecting an obstacle; the detection device comprises a plurality of detection units, each detection unit comprises a pixel array, and the pixel array is configured to respond to an echo light beam reflected by the detection light beam on an obstacle and convert the echo light beam into an electric signal; the processing device is coupled with the detection device and is configured to recombine the electric signals output by the pixel array of the at least one detection unit in different forms to obtain a plurality of recombined pixel arrays; and generating a laser radar point cloud according to the plurality of recombined pixel arrays. According to the laser radar provided by the invention, the electric signals output by the pixel array of at least one detection unit are recombined in different forms, the laser radar point cloud is further generated according to the plurality of recombined pixel arrays, and the distance, resolution and other information of an obstacle are calculated; the signal-to-noise ratio, the distance measuring capability, the accuracy of the detection result and other detection performances of the laser radar can be improved.

Description

Technical Field

[0001] The present invention generally relates to the field of laser radar, and in particular to a laser radar and a detection method of a laser radar. Background Art

[0002] LiDAR is a radar system that emits laser beams to detect the position, speed and other characteristic quantities of a target. It is an advanced detection method that combines laser technology with photoelectric detection technology. LiDAR is widely used in autonomous driving, transportation communications, drones, intelligent robots, resource exploration and other fields due to its advantages such as high resolution, good concealment, strong anti-active interference ability, good low-altitude detection performance, small size and light weight.

[0003] Due to the low intensity of weak echo signals, especially for echo signals of long-distance, low-reflectivity targets, weak echo signals have poor ability to trigger detection units, and the signal-to-noise ratio is relatively low. When analyzing the echo signal waveform, the echo signal may not be identified, thereby reducing the detection capability of weak echo signals, affecting the distance measurement performance of the lidar, and affecting the accuracy of the detection results. One solution to ensure the distance measurement capability of the lidar is to perform neighborhood accumulation of echo signals, but neighborhood accumulation may be accompanied by a certain loss of resolution. Therefore, how to improve the detection performance of the lidar (such as signal-to-noise ratio, distance measurement capability, accuracy, etc.) has always been a demand for continuous improvement in the lidar field, especially how to maximize the distance measurement capability of the lidar while ensuring a high signal-to-noise ratio and not sacrificing the spatial resolution of the target, is a technical problem that needs to be solved.

[0004] The contents of the background technology section are merely the technologies known to the inventors and do not necessarily represent the prior art in the field. Summary of the invention

[0005] In response to one or more problems existing in the prior art, the present invention provides a laser radar, which can improve at least one of the detection performances of the laser radar, such as the signal-to-noise ratio, the distance measurement capability, and the accuracy of the detection results, by reorganizing the electrical signals output by the pixel array in different forms, further generating a laser radar point cloud based on multiple reorganized pixel arrays, and calculating information such as the distance and resolution of the obstacle.

[0006] The laser radar comprises:

[0007] A transmitting device configured to transmit a detection beam for detecting obstacles;

[0008] A detection device, comprising a plurality of detection units, wherein each detection unit comprises a pixel array, wherein the pixel array is configured to respond to an echo beam reflected by the detection beam on an obstacle and convert the echo beam into an electrical signal; and

[0009] A processing device is coupled to the detection device and configured to:

[0010] Recombining the electrical signals output by the pixel array of at least one detection unit in different forms to obtain a plurality of recombined pixel arrays;

[0011] A lidar point cloud is generated based on the multiple recombined pixel arrays.

[0012] According to one aspect of the present invention, the processing device is configured to: perform convolution processing on the electrical signal output by the pixel array of the at least one detection unit using a plurality of convolution kernels of different sizes to obtain the plurality of recombined pixel arrays.

[0013] According to one aspect of the present invention, the processing device is configured to: select a recombined pixel array from the multiple recombined pixel arrays based on the waveform data of the multiple recombined pixel arrays, and generate a lidar point cloud based on the selected recombined pixel array.

[0014] According to one aspect of the present invention, the waveform data includes a signal-to-noise ratio and / or a pulse intensity.

[0015] According to one aspect of the present invention, the processing device is configured to: determine the signal-to-noise ratio of each recombined pixel array, and select a recombined pixel array having a signal-to-noise ratio greater than a signal-to-noise ratio threshold from the plurality of recombined pixel arrays.

[0016] According to one aspect of the present invention, the processing device is configured to: when the signal-to-noise ratio of multiple recombined pixel arrays is higher than the signal-to-noise ratio threshold, select the recombined pixel array with the smallest convolution kernel size.

[0017] According to one aspect of the present invention, the processing device is configured to: determine the pulse intensity of each reconstructed pixel array, and select a reconstructed pixel array having a pulse intensity greater than a pulse intensity threshold from the plurality of reconstructed pixel arrays.

[0018] According to one aspect of the present invention, the processing device is configured to: when the pulse intensity of multiple recombined pixel arrays is higher than the pulse intensity threshold, select the recombined pixel array with the smallest convolution kernel size.

[0019] According to one aspect of the present invention, the processing device is configured to expand the array of electrical signals output by the pixel array of the at least one detection unit.

[0020] According to one aspect of the present invention, the processing device is configured to expand the array of electrical signals output by the pixel array of the at least one detection unit according to the convolution kernel with the largest size.

[0021] According to one aspect of the present invention, the processing device is configured to: when the pixel array of the at least one detection unit responds to the echo light beam and converts it into an electrical signal, control the pixels around the pixel array of the at least one detection unit to simultaneously respond to the echo light beam and convert it into an electrical signal; and expand the array of electrical signals output by the pixel array of the at least one detection unit through the electrical signals output by the surrounding pixels.

[0022] According to one aspect of the present invention, each pixel array includes a plurality of pixels arranged in an array, wherein the pixels include single-photon avalanche diodes, and each single-photon avalanche diode can be independently gated and addressed.

[0023] The present invention also relates to a detection method of a laser radar, wherein the laser radar comprises a detection device, wherein the detection device comprises a plurality of detection units, and each detection unit comprises a pixel array; the detection method comprises:

[0024] S11: emits a detection beam to detect obstacles;

[0025] S12: responding to the echo beam reflected by the detection beam on the obstacle and converting it into an electrical signal;

[0026] S13: reorganizing the electrical signals output by the pixel array of at least one detection unit in different forms to obtain a plurality of reorganized pixel arrays; and

[0027] S14: Generate a laser radar point cloud according to the multiple recombined pixel arrays.

[0028] According to one aspect of the present invention, step S13 comprises: performing convolution processing on the electrical signal output by the pixel array of the at least one detection unit using a plurality of convolution kernels of different sizes to obtain the plurality of recombined pixel arrays.

[0029] According to one aspect of the present invention, step S14 includes: selecting a recombined pixel array from the multiple recombined pixel arrays according to the waveform data of the multiple recombined pixel arrays, and generating a lidar point cloud according to the selected recombined pixel array.

[0030] According to one aspect of the present invention, the waveform data includes a signal-to-noise ratio and / or a pulse intensity.

[0031] According to one aspect of the present invention, step S14 comprises:

[0032] determining a signal-to-noise ratio for each reconstructed pixel array; and

[0033] A reconstructed pixel array having a signal-to-noise ratio greater than a signal-to-noise ratio threshold is selected from the plurality of reconstructed pixel arrays.

[0034] According to one aspect of the present invention, the step S14 further comprises: when the signal-to-noise ratios of the multiple recombined pixel arrays are higher than the signal-to-noise ratio threshold, selecting the recombined pixel array with the smallest convolution kernel size.

[0035] According to one aspect of the present invention, step S14 includes: determining the pulse intensity of each recombined pixel array, and selecting a recombined pixel array having a pulse intensity greater than a pulse intensity threshold from the plurality of recombined pixel arrays.

[0036] According to one aspect of the present invention, step S14 comprises: when the pulse intensity of multiple recombined pixel arrays is higher than the pulse intensity threshold, selecting the recombined pixel array with the smallest convolution kernel size.

[0037] According to one aspect of the present invention, the step S13 further comprises: expanding the array of electrical signals output by the pixel array of the at least one detection unit.

[0038] According to one aspect of the present invention, the step of expanding the array of electrical signals output by the pixel array of at least one detection unit includes: expanding the array of electrical signals output by the pixel array of at least one detection unit according to a convolution kernel of the largest size.

[0039] According to one aspect of the present invention, the step of expanding the array of electrical signals output by the pixel array of at least one detection unit also includes: when the pixel array of the at least one detection unit responds to the echo light beam and converts it into electrical signals, controlling the pixels around the pixel array of the at least one detection unit to simultaneously respond to the echo light beam and convert it into electrical signals; and expanding the array of electrical signals output by the pixel array of the at least one detection unit through the electrical signals output by the surrounding pixels.

[0040] According to one aspect of the present invention, each pixel array includes a plurality of pixels arranged in an array, wherein the pixels include single-photon avalanche diodes, and each single-photon avalanche diode can be independently gated and addressed.

[0041] In summary, the scheme of the present invention has been introduced in detail. By adopting the scheme of the present invention, the electrical signals output by the pixel array of at least one detection unit are reorganized in different forms to obtain multiple recombined pixel arrays. A lidar point cloud is generated based on the multiple recombined pixel arrays, and the distance, resolution and other information of the obstacle are calculated, thereby improving the signal-to-noise ratio, distance measurement capability and accuracy of the detection results of the lidar.

[0042] Furthermore, by using convolution kernels of different sizes to perform different forms of convolution processing on the electrical signals output by the pixel array, multiple reconstructed pixel arrays can be obtained. The multiple reconstructed pixel arrays obtained in this way have waveform data with different characteristics, such as different resolutions, signal-to-noise ratios, echo pulse strengths, etc. According to the multiple reconstructed pixel arrays, the optimal waveform data can be selected to generate a laser radar point cloud, thereby improving the detection performance of the laser radar.

[0043] Furthermore, based on the waveform data of multiple recombined pixel arrays, one of the recombined pixel arrays can be selected to generate a lidar point cloud and calculate obstacle information. This can maximize the lidar's ranging capability while improving the high signal-to-noise ratio and spatial resolution of the lidar point cloud, thereby improving the accuracy of the detection results.

[0044] Furthermore, by expanding the array of electrical signals output by the pixel array of the at least one detection unit, the sizes of multiple recombined pixel arrays obtained after different forms of convolution processing can be made consistent, so that each recombined pixel has multiple convolution results after different forms of convolution processing, so that the optimal recombined pixel array can be selected according to the waveform data of the recombined pixel, which is beneficial to improving the adaptability and dynamic range of the lidar.

[0045] In the actual detection process, the laser radar can adaptively adjust the size of the convolution kernel according to the waveform data of the reconstructed signal and select the convolution result of the convolution kernel with the optimal size. This can improve the signal-to-noise ratio and ranging capability of the laser radar while taking into account the spatial resolution of the target, ensuring the accuracy of the detection results and dynamically maintaining better detection performance in the entire field of view. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1 A schematic diagram of a laser radar according to an embodiment of the present invention is shown;

[0048] Figure 2 A schematic diagram showing a detection device according to a preferred embodiment of the present invention is shown;

[0049] Figure 3 A schematic diagram showing a detection device according to another preferred embodiment of the present invention is shown;

[0050] Figure 4 shows an enlarged view of a detection unit according to a preferred embodiment of the present invention;

[0051] Figure 5 to Figure 7A schematic diagram showing convolution processing of the electrical signal output by the pixel array of the detection unit using convolution kernels of different sizes according to some preferred embodiments of the present invention is shown;

[0052] Figure 8 Shows Figure 5 to Figure 7 A comparison of signal waveforms of reconstructed pixels obtained by convolution processing using convolution kernels of different sizes in the embodiment;

[0053] Figures 9 to 11 A schematic diagram showing a method of reducing the accuracy of detection results due to convolution processing using a convolution kernel of too large a size according to a preferred embodiment of the present invention;

[0054] Fig.12 A schematic diagram showing a method of detecting a target by convolution using convolution kernels of different sizes according to a preferred embodiment of the present invention is shown;

[0055] Fig.13 A schematic diagram showing expansion of an array of electrical signals output by a pixel array of a detection unit according to a preferred embodiment of the present invention is shown;

[0056] Fig.14 A schematic diagram showing a laser radar adaptively adjusting the convolution kernel size according to a detected target according to a preferred embodiment of the present invention; and

[0057] Fig.15 A flow chart of a laser radar detection method according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0058] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0059] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0060] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, an electrical connection, or can communicate with each other; it can be directly connected, or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0061] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0062] The disclosure below provides many different embodiments or examples to realize different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplicity and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed. In addition, the present invention provides various specific examples of processes and materials, but those of ordinary skill in the art can be aware of the application of other processes and / or the use of other materials.

[0063] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0064] The present invention provides a laser radar. Figure 1 FIG. 1 shows a schematic diagram of a laser radar 100 according to an embodiment of the present invention. Figure 1 As shown, the laser radar 100 includes a transmitting device 110, a detecting device 120 and a processing device 130. The transmitting device 110 is configured to transmit a detection beam L for detecting an obstacle OB; the detecting device 120 includes a plurality of detecting units ( Figure 1 Schematically shows one of the detection units), wherein each detection unit includes a pixel array ( Figure 1 An 8×16 pixel array is shown as an example, the pixel array is configured to respond to the echo beam L' reflected by the detection beam L on the obstacle OB and convert it into an electrical signal; the processing device 130 is coupled to the detection device 120, and is configured to reorganize the electrical signal output by the pixel array of at least one detection unit in different forms to obtain multiple reorganized pixel arrays; based on the multiple reorganized pixel arrays, a laser radar point cloud is generated. The laser radar of the present invention can improve the detection performance of the laser radar, including at least one of the signal-to-noise ratio, the distance measurement capability, the accuracy of the detection result, etc. by reorganizing the electrical signal output by the pixel array of at least one detection unit in different forms and generating the laser radar point cloud. The following is a detailed introduction.

[0065] Figure 2 A schematic diagram of a detection device 120 according to a preferred embodiment of the present invention is shown. Figure 2 As shown, the detection device 120 includes a plurality of detection units, for example, N detection units R 1 , R 2 , R 3 , …, R N, where N is an integer greater than 1. Multiple detection units can be arranged as a one-dimensional linear array along the vertical direction to cover the vertical field of view of the laser radar. The scanning of the three-dimensional field of view can be achieved through a scanning device, such as a scanning mirror, a mechanical rotating laser radar, etc.

[0066] Figure 3 A schematic diagram of a detection device 120 according to another preferred embodiment of the present invention is shown. The detection device 120 includes a plurality of detection units R 11 , R 21 , …, R M1 , …, R 1N , R 2N , …, R MN , multiple detection units are arranged in a two-dimensional array to form an N×M detection unit array, that is, an array of detection units with N rows and M columns, where N and M are both integers greater than 1, and the present invention does not limit the size relationship between N and M. The detection units arranged in the array can cover a certain range of spatial field of view.

[0067] Figure 4 FIG. 4 shows an enlarged view of a detection unit according to a preferred embodiment of the present invention, as shown in FIG. Figure 4 As shown, each detection unit includes a pixel array, each pixel array includes a plurality of pixels arranged in an array, such as 8×16=128 pixels shown in the figure, each pixel includes one or more single photon avalanche diodes (SPAD), each single photon avalanche diode can be independently gated and addressed, that is, each single photon avalanche diode can respond to photons individually and output electrical signals. Furthermore, each detection unit can be powered on independently, so at the same time, only one detection unit in the detection device can detect echo signals, or multiple detection units can be enabled at the same time to detect echo signals.

[0068] The above embodiment has made a detailed introduction to the detection device 120 of the present invention. It should be noted that the present invention does not limit the number of detection units in the detection device 120 and their arrangement, the number of pixels in each detection unit and their arrangement, the size of each pixel, and the number and arrangement of SPADs in each pixel. In practical applications, they can be flexibly set according to needs. In addition, although not shown in the figure, the transmitting device 110 of the present invention includes at least one transmitting unit, and the transmitting unit includes a vertical cavity surface emitting laser (Vertical-Cavity Surface-Emitting Laser, VCSEL) or an edge emitting laser (Edge Emitting Laser, EEL), etc. When the transmitting device 110 includes multiple transmitting units, the multiple transmitting units can also be arranged as a one-dimensional linear array or a two-dimensional surface array in the manner of the above-mentioned detection units, which are all within the protection scope of the present invention.

[0069] A transmitting unit can usually correspond to one or more detection units. In each detection process, the detection beam L emitted by each transmitting unit can form an echo beam L' after being reflected by an obstacle and can be incident on the corresponding detection unit. The correspondence between the transmitting unit and the detection unit can be set and calibrated when the laser radar leaves the factory. In the detection process of the laser radar, the detection beam L emitted by the transmitting device 110 can be deflected by a scanning device (such as a rotating mirror), or the transmitting device 110 can be driven by the rotor to poll and emit the detection beam L at intervals of a certain horizontal angle (such as 0.05°, etc.). The pixel array of the detection device 120 can respond to the echo L' reflected by the detection beam L on the obstacle OB and convert it into an electrical signal. By moving the field of view, the laser radar 100 can complete the scanning of the entire scene of the surrounding environment, thereby realizing detection within a certain horizontal field of view. The processing device 130 obtains a three-dimensional point cloud of the surrounding environment by splicing the three-dimensional point cloud of each area. However, the laser radar of the present invention is not limited to the mechanical rotating laser radar or scanning laser radar introduced above, but can also be a solid-state laser radar, that is, it does not require any mechanical rotating structure, and can cover a certain spatial field of view through the array setting of the transmitting device and the detecting device.

[0070] In some laser radars, the processing device directly calculates the information of obstacles, including the distance and / or reflectivity of obstacles, based on the electrical signals output by the pixel array of each detection unit. This method has a weak detection capability for weak echo signals. The present invention has made improvements on this basis, and the electrical signals output by the pixel array of at least one detection unit 120 are reorganized multiple times (at least twice) in different forms to obtain multiple reorganized pixel arrays, and the point cloud of the laser radar is generated based on the multiple reorganized pixel arrays. The following describes the reorganization method according to an embodiment of the present invention.

[0071] According to a preferred embodiment of the present invention, the processing device 130 is configured to perform convolution processing on the electrical signal output by the pixel array of at least one detection unit 120 using a plurality of convolution kernels of different sizes, respectively, to obtain a plurality of reconstructed pixel arrays. The convolution kernel can also be called a convolution window or an accumulation window, and the size of the convolution kernel is P×Q (in the form of a matrix of P rows and Q columns), which can cover P×Q pixels, wherein P and Q are integers greater than or equal to 1, and the two can be equal or unequal. Preferably, the size of the convolution kernel can be 1×1 (refer to Figure 5 ), 3×3 (refer to Figure 6 ), 5×5 (refer to Figure 7 ), 3×2 (not shown), etc. During the convolution processing, the convolution kernel moves step by step in a preset direction with a determined convolution step size. Each time it moves, the electrical signals output by all pixels covered by the convolution kernel are accumulated (or weighted accumulation) to obtain a reconstructed pixel, and the entire pixel array of the at least one detection unit is traversed to obtain a reconstructed pixel array. In this way, convolution kernels of different sizes are used to perform convolution processing on the electrical signals output by the pixel array of the at least one detection unit 120, respectively, to obtain multiple reconstructed pixel arrays, and a lidar point cloud is generated based on the multiple reconstructed pixel arrays.

[0072] It should be noted that, in the solution of the present invention, when the electrical signal output by the pixel array is reorganized in different forms, the different sizes of the convolution kernels used are at least 2 different sizes. The following takes the convolution processing of the electrical signal output by the pixel array of a detection unit using 3 different sizes of convolution kernels as an example for detailed description.

[0073] Figure 5 A schematic diagram of performing convolution processing on the electrical signal output by the pixel array of the detection unit according to a preferred embodiment of the present invention is shown. In this embodiment, the detection unit includes an 8×16 pixel array, the size of the convolution kernel is 1×1, and the convolution step length stride=1. During the convolution processing, the convolution kernel moves step by step in a preset direction (for example, from left to right, from top to bottom) in units of convolution step length stride=1. For each movement, the electrical signals (for example, echo waveforms) output by all pixels (1 pixel) covered by the convolution kernel are accumulated to obtain a recombined pixel. By traversing the 8×16 pixel array of the entire detection unit, an 8×16 recombined pixel array can be obtained. Refer to Figure 5 , the dotted box frames one of the recombined pixels and its corresponding single pixel (original pixel), Figure 5The lower right part shows the signal waveform of a reconstructed pixel. It should be understood that when the size of the convolution kernel is equal to the size of a single pixel, the size and signal of the reconstructed pixel (array) are the same as the original pixel (array), and the reconstructed pixel (array) is equivalent to the original pixel (array).

[0074] Figure 6 A schematic diagram of performing convolution processing on the electrical signal output by the pixel array of the detection unit according to another preferred embodiment of the present invention is shown. In this embodiment, the detection unit includes an 8×16 pixel array, the size of the convolution kernel is 3×3, and the convolution step length stride=1. During the convolution processing, the convolution kernel moves step by step in a preset direction (for example, from left to right, from top to bottom) in units of convolution step length stride=1. For each movement, the electrical signals (for example, echo waveforms) output by all pixels (9 pixels) in the convolution kernel are accumulated once to obtain a recombined pixel, and the entire 8×16 pixel array of the detection unit is traversed to obtain a 6×14 recombined pixel array. Refer to Figure 6 , the dotted box frames one of the recombined pixels and its corresponding 9 pixels (original pixels), Figure 6 The lower right part shows a comparison of the signal waveforms of a recombined pixel and a single pixel, where the upper waveform is the signal waveform of the recombined pixel, and the lower waveform is the signal waveform of a single pixel. It can be seen that the signal waveform of a single pixel has a lower peak value and a larger width, while the signal waveform of the recombined pixel has a higher peak value and a smaller width. It can be seen that when the size of the convolution kernel is larger than the size of a single pixel, the recombined pixel is equivalent to the signal accumulation of a single pixel and its neighboring pixels. Therefore, compared with the electrical signal output by a single pixel, the signal strength of the recombined pixel is stronger, which is more conducive to the detection of weak echo signals. Therefore, the recombined pixel can effectively enhance the echo signal strength, so that the echo signal can be clearly distinguished from the noise signal, which is conducive to improving the recognition probability of the echo signal, improving the signal-to-noise ratio, and improving the distance measurement capability of the laser radar.

[0075] Figure 7 A schematic diagram of performing convolution processing on the electrical signal output by the pixel array of the detection unit according to another preferred embodiment of the present invention is shown. In this embodiment, the detection unit includes an 8×16 pixel array, the size of the convolution kernel is 5×5, and the convolution step length stride=1. During the convolution process, the convolution kernel moves step by step in a preset direction (for example, from left to right, from top to bottom) in units of convolution step length stride=1. For each movement, the electrical signals (for example, echo waveforms) output by all pixels (25 pixels) in the convolution kernel are accumulated to obtain a recombined pixel. By traversing the 8×16 pixel array of the entire detection unit, a 4×12 recombined pixel array can be obtained. Refer to Figure 7 , the dotted box outlines one of the reconstructed pixels and its corresponding 25 pixels. Figure 7The lower right part shows the signal waveform comparison between a recombined pixel and a single pixel, where the upper waveform is the signal waveform of the recombined pixel and the lower waveform is the signal waveform of a single pixel. Figure 6 Similar to the situation described in the embodiment, compared with the electrical signal output by a single pixel, the signal waveform of the recombined pixel has a higher peak value and a smaller width, and the signal intensity of the recombined pixel is stronger and the signal-to-noise ratio is higher.

[0076] It should be noted that Figures 5 to 7 In the embodiments, convolution kernels of three sizes, 1×1, 3×3, and 5×5, are taken as examples, and the convolution step length adopted is stride=1, but the present invention is not limited to this. The convolution kernel may also adopt other sizes, and the convolution step length may also adopt other convolution step lengths, which are all within the protection scope of the present invention.

[0077] Figure 8 Shows Figures 5 to 7 In the embodiment of FIG. 1 , the signal waveforms of the reconstructed pixels obtained by convolution processing using convolution kernels of three sizes, 1×1, 3×3, and 5×5, are compared, as shown in FIG. Figure 8 As shown in the figure, the signal waveform peak of the reconstructed pixel (single pixel) obtained by convolution processing with a 1×1 convolution kernel is lower, that is, the echo pulse intensity is lower; the signal waveform peak of the reconstructed pixel obtained by convolution processing with a 3×3 convolution kernel is higher, that is, the echo pulse intensity is higher; the signal waveform peak of the reconstructed pixel obtained by convolution processing with a 5×5 convolution kernel is higher, that is, the echo pulse intensity is higher. Assuming that the convolution kernel size is P×P, and the echoes in the field of view covered by the convolution kernel correspond to the same obstacle and are evenly distributed, the signal-to-noise ratio of the reconstructed pixels obtained by convolution kernels of different sizes can be expressed as:

[0078] SNR=P×SNR 0

[0079] Among them, SNR 0 is the signal-to-noise ratio of a single pixel, and SNR is the signal-to-noise ratio of the reconstructed pixel. It can be seen that the larger the size of the convolution kernel, the stronger the signal intensity of the reconstructed pixel obtained after convolution processing, and the higher the signal-to-noise ratio, that is, the signal-to-noise ratio (or signal intensity) of the reconstructed pixel is positively correlated with the size of the convolution kernel.

[0080] It can be seen that in the actual detection process of LiDAR, in order to improve the signal-to-noise ratio, obtain better distance measurement performance and more accurate detection results, a larger convolution kernel can be used to perform convolution processing on the electrical signal output by the pixel array of the detection unit, which is especially beneficial for the detection of long-distance and weak targets.

[0081] However, if the LiDAR uses a convolution kernel that is too large for convolution processing, the spatial resolution of the LiDAR point cloud may be reduced, and it may also cause problems in accurately identifying the detected target, especially at the boundary of obstacles. The following describes the specific reasons why the use of a convolution kernel that is too large for convolution processing leads to inaccurate detection results.

[0082] Reference Figures 9 to 11 , the laser radar emits a detection beam for detection. Assume that the distance between target 1 and the laser radar is d1, the distance between target 2 and the laser radar is d2, and there is a radial distance difference Δd between target 1 and target 2 (that is, the distances from the laser radar are different), where Δd = |d1-d2|, and the radial distance difference Δd is between 0 and the pulse width of the detection beam. The spot of the detection beam is incident on target 1 and target 2 at the same time (for example, at the boundary between target 1 and target 2), where the echo beam reflected by the part of the detection beam incident on target 1 triggers part of the pixels of the detection unit at time t1 (refer to Fig. 9 The left shadow part) responds and converts it into an electrical signal, and its signal waveform refers to Fig.10 The six histograms shown above; the echo beam reflected by the partial detection beam incident on the target 2 triggers the partial pixels of the detection unit at time t2 (refer to Fig. 9 The shaded area on the right responds and converts it into an electrical signal. The signal waveform is referenced Fig.10 The four histograms shown below are when the electrical signal output by the pixel array of the detection unit is convolved. Assuming that a 3×3 convolution kernel is used, when the convolution kernel moves to the pixel boundary corresponding to target 1 and target 2, the convolution kernel covers the pixels corresponding to target 1 and target 2 at the same time, and accumulates the electrical signals output by the pixels corresponding to target 1 and target 2 to obtain a reconstructed pixel. The signal waveform of the reconstructed pixel is shown in FIG. Fig.11As shown, since the pixel array covered by the convolution kernel is for different targets, the electrical signals output by the covered pixel array correspond to different distances and the energy distribution is uneven. Therefore, the signal waveform of the recombined pixel will cause the leading edge distortion of target 1 and the pulse width distortion of target 2, causing the recombined pixel array and the laser radar point cloud to be distorted, and the target cannot be accurately identified. If the signal waveform (such as the rising edge) of the recombined pixel is used to calculate the obstacle information, the calculation result will be inaccurate, reducing the accuracy of the laser radar detection result. It can be seen that when a convolution kernel of too large size is used for convolution processing, when convolving to the pixel corresponding to the target boundary, the signal waveform of the recombined pixel obtained by the convolution processing will be distorted, causing distance ambiguity and affecting the accuracy of the detection result. Therefore, a convolution kernel of too large size may reduce the spatial resolution of the laser radar point cloud, resulting in inaccurate detection results. It should be noted that this embodiment only takes a 3×3 convolution kernel as an example for explanation, which does not mean that the size of the 3×3 convolution kernel is too large. The specific size of the oversized convolution kernel needs to be determined according to the actual situation such as the measurement target.

[0083] Therefore, how to determine the appropriate size of the convolution kernel so that the lidar can maintain high signal-to-noise ratio while maintaining high resolution, ensuring the distance measurement capability and the accuracy of the detection results is crucial in actual detection. The following is a detailed introduction.

[0084] According to a preferred embodiment of the present invention, the processing device 130 is configured to select a recombined pixel array from the plurality of recombined pixel arrays according to the waveform data of the plurality of recombined pixel arrays, and generate a laser radar point cloud according to the selected recombined pixel array. Figures 5 to 7 As described in the embodiment of the invention, the electrical signals output by the pixel array of the at least one detection unit are convolved with convolution kernels of different sizes to obtain multiple reconstructed pixel arrays.

[0085] According to a preferred embodiment of the present invention, the waveform data includes a signal to noise ratio (SNR). The processing device 130 is configured to determine the SNR of each reconstructed pixel array and select from the plurality of reconstructed pixel arrays a pixel array having a SNR greater than a SNR threshold T. SNR When the signal-to-noise ratio of the plurality of recombined pixel arrays is higher than the signal-to-noise ratio threshold T SNR When , the reconstructed pixel array with the smallest convolution kernel size is selected. Specifically, for example, convolution kernels with sizes of 1×1, 3×3, and 5×5 are used to perform convolution processing on the electrical signal output by the at least one detection unit, and the signal-to-noise ratios of the signal waveform of the reconstructed pixel array obtained after the convolution processing are T1, T3, and T5 respectively. If T1 <T SNR , T3 <T SNR, T5 > T SNR , then the convolution result (reorganized pixel array) obtained by selecting a 5×5 convolution kernel is used to generate lidar point cloud, and obstacle information is calculated based on the convolution result (reorganized pixels) of the 5×5 convolution kernel; if T1 < T SNR , T3 > T SNR , T5 > T SNR , then the convolution result (reorganized pixel array) obtained by selecting a 3×3 convolution kernel is used to generate lidar point cloud, and obstacle information is calculated based on the convolution result (reorganized pixels) of the 3×3 convolution kernel; if T1 > T SNR , T3 > T SNR , T5 > T SNR , then the convolution result (reorganized pixel array, i.e., the original pixel array) obtained by selecting a 1×1 convolution kernel is used to generate lidar point cloud, and obstacle information is calculated based on the convolution result (reorganized pixels, i.e., the original single pixel) of the 1×1 convolution kernel; if T1 < T SNR , T3 < T SNR , T5 < T SNR , it indicates that the signal-to-noise ratio of the signal is too low and no target is detected, so no lidar point cloud is generated.

[0086] Therefore, when processing the detection data output by multiple pixels in the detection unit, convolution processing is respectively performed using convolution kernels with side lengths of n0, n1, n2,... sizes (where n0 < n1 < n2 <...). Assuming that the signal-to-noise ratios of the echo signals of the reorganized pixels obtained are SNR(n0), SNR(n1), SNR(n2),..., and so on, and setting the signal-to-noise ratio threshold as T SNR , then the finally selected convolution kernel size is:

[0087] n = argmin(SNR(ni) > T SNR )

[0088] That is, the smallest convolution kernel size with a signal-to-noise ratio exceeding the threshold T SNR is selected as the finally selected convolution kernel size, and the lidar point cloud is generated based on the reorganized pixel array of the selected size convolution kernel.

[0089] According to another preferred embodiment of the present invention, the waveform data further includes pulse intensity. The processing device 130 is configured to determine the pulse intensity of each reorganized pixel array, and select the reorganized pixel array with a pulse intensity greater than the pulse intensity threshold T PS from the multiple reorganized pixel arrays. When the pulse intensities of multiple reorganized pixel arrays are higher than the pulse intensity threshold T PSWhen, select the recombined pixel array with the smallest convolution kernel size among them. Specifically, for example, convolution processing is respectively performed on the electrical signals output by the at least one detection unit using convolution kernels with sizes of 1×1, 3×3, and 5×5. The pulse intensities of the signal waveforms of the recombined pixel arrays obtained after convolution processing are P1, P3, and P5 respectively. If P1 < T PS , P3 < T PS , P5 > T PS , then select the convolution result (recombined pixel array) obtained by the 5×5 convolution kernel to generate lidar point cloud, and calculate obstacle information based on the convolution result (recombined pixel) of the 5×5 convolution kernel; if P1 < T PS , P3 > T PS , P5 > T PS , then select the convolution result (recombined pixel array) obtained by the 3×3 convolution kernel to generate lidar point cloud, and calculate obstacle information based on the convolution result (recombined pixel) of the 3×3 convolution kernel; if P1 > T PS , P3 > T PS , P5 > T PS , then select the convolution result (recombined pixel array, that is, the original pixel array) obtained by the 1×1 convolution kernel to generate lidar point cloud, and calculate obstacle information based on the convolution result (recombined pixel, that is, the original single pixel) of the 1×1 convolution kernel; if P1 < T PS , P3 < T PS , P5 < T PS , then no lidar point cloud is generated.

[0090] Therefore, when processing the detection data output by multiple pixels in the detection unit, convolution processing is respectively performed using convolution kernels with side lengths of n0, n1, n2, …… (where n0 < n1 < n2 < ……). Assume that the pulse intensities of the echo signals of the recombined pixels obtained are PS(n0), PS(n1), PS(n2) ……, and so on. Set the pulse intensity threshold as T PS , then the finally selected convolution kernel size is:

[0091] n = argmin(PS(ni) > T PS )

[0092] That is, select the smallest convolution kernel size whose pulse intensity exceeds the threshold T PS as the finally selected convolution kernel size, and generate lidar point cloud based on the recombined pixel array obtained by the convolution kernel of the selected size.

[0093] More intuitively, referring to Fig.12 , Fig.12 shows a schematic diagram of using 1×1 and 3×3 convolution kernels for convolution detection of the target. Assume Fig.12The targets shown are the same target, for Fig.12 At the pixel position shown in (a) (the pixel position shown by the white square in the figure), the detection data of the target can be convolved using 1×1 and 3×3 convolution kernels, so T1>T SNR , T3>T SNR , then select the reconstructed pixel array obtained by the 1×1 convolution kernel to generate the lidar point cloud. Fig.12 At the pixel position shown in (b), the target cannot be detected by using a 1×1 convolution kernel, but the target can be detected by using a 3×3 convolution kernel, i.e., T1 <T SNR , T3>T SNR , then the laser radar point cloud is generated based on the recombined pixel array obtained by the 3×3 convolution kernel. Fig.12 At the pixel position shown in (c), the target cannot be detected by using 1×1 or 3×3 convolution kernels, that is, T1 <T SNR , T3 <T SNR , no laser radar point cloud is generated. It should be noted that this embodiment is introduced by taking the signal-to-noise ratio as an example, and the pulse intensity is similar to it, which will not be described here.

[0094] According to another preferred embodiment of the present invention, the size of the convolution kernel may be determined comprehensively based on the signal-to-noise ratio and pulse intensity of the reconstructed pixel array.

[0095] Specifically, for example, the signal-to-noise ratio is selected to exceed the threshold T SNR And the pulse intensity exceeds the threshold T PS The minimum convolution kernel size is used as the final selected convolution kernel size, as follows:

[0096] n=argmin(SNR(ni)>T SNR &PS(ni)>T PS )

[0097] For example, if the signal-to-noise ratio exceeds the threshold T SNR Or the pulse intensity exceeds the threshold T PS The minimum convolution kernel size is used as the final selected convolution kernel size, as follows:

[0098] n=argmin(SNR(ni)>T SNR |PS(ni)>T PS )

[0099] The above embodiment introduces how to determine the appropriate convolution kernel size based on waveform data (signal-to-noise ratio and / or pulse intensity), while maximizing the laser radar's distance measurement capability, maintaining a high signal-to-noise ratio and high spatial resolution, and taking into account the accuracy of the detection results. In addition, when performing convolution processing in a single detection unit area, the size of the convolution kernel does not need to be consistent in the entire detection unit area, and can be adaptively adjusted according to the signal waveform data of the reorganized pixels. Different convolution kernels can be used for pixel reorganization in different areas, which is beneficial to improving the laser radar's adaptive capability and dynamic adjustment range. It should be noted that although the above embodiment is introduced with the convolution kernel with a side length of n0, n1, n2..., that is, P=Q, as an example, the case of P≠Q is similar to it. It only needs to replace n0, n1, n2... with P1×Q1, P2×Q2, P3×Q3..., which will not be repeated here.

[0100] Since the convolution kernel has a certain size, the convolution operation of the detection unit boundary area will be affected. Specifically, after the electrical signal output by the pixel array of the same detection unit is convolved with convolution kernels of different sizes and the same step length, the size of the reconstructed pixel array obtained is inconsistent. More specifically, refer to Figures 5 to 7 , the electrical signal output by the 8×16 pixel array of the detection unit is also subjected to convolution processing with a convolution step length of stride=1, wherein Figure 5 In the embodiment of , the size of the convolution kernel is 1×1, and the size of the resulting reconstructed pixel array is 8×16; Figure 6 In the embodiment of , the size of the convolution kernel is 3×3, and the size of the resulting reconstructed pixel array is 6×14; Figure 7 In the embodiment, the size of the convolution kernel is 5×5, and the size of the resulting reconstructed pixel array is 4×12.

[0101] In order to reduce the influence of the size of the convolution kernel on the convolution operation of the border area of ​​the detection unit, so that the reconstructed pixel array of the same size can be obtained after the convolution processing using the convolution kernels of different sizes, according to a preferred embodiment of the present invention, the processing device 130 is configured to expand the array of the electrical signal output by the pixel array of the at least one detection unit. Taking a detection unit including an 8×16 pixel array and the size of the convolution kernel P=Q as an example, the processing device 130 not only collects the electrical signal output by the 8×16 pixel array, but also collects the electrical signals output by more pixels, that is, the number of pixels of the data actually collected by the processing device 130 is greater than 8×16, for example, it can be 10×18, or 12×20, etc.

[0102] Preferably, the processing device 130 is configured to expand the array of electrical signals output by the pixel array of the at least one detection unit according to the largest convolution kernel. Taking a detection unit including an 8×16 pixel array as an example, assuming that the maximum size of the convolution kernel is 5×5, 4 rows and 4 columns of pixels can be expanded around the detection unit; assuming that the maximum size of the convolution kernel is 3×3, 2 rows and 2 columns of pixels can be expanded around the detection unit, and so on. The present invention does not limit the specific method of expansion. Fig.13 As shown in (a), the expansion method can be uniform expansion around the detection unit, such as expansion upward, downward, left, and right. In addition, the expansion method can also be non-uniform expansion around the detection unit, such as expansion only to the left and upward, or right and downward. Fig.13 As shown in (b), for example, only 4 rows and 4 columns of pixels are extended downward and to the right around the detection unit. It should be noted that the detection unit periphery should be understood in a broad sense, and can refer to the four sides of the detection unit, or one side, two sides, or three sides of the detection unit.

[0103] In a specific implementation, the processing device 130 can simultaneously collect data by expanding some pixels around the at least one detection unit. Specifically, when the pixel array of the at least one detection unit responds to the echo light beam and converts it into an electrical signal, the processing device 130 controls the pixels around the pixel array of the at least one detection unit to simultaneously respond to the echo light beam and convert it into an electrical signal; and expands the array of electrical signals output by the pixel array of the at least one detection unit through the electrical signals output by the surrounding pixels.

[0104] By expanding the array of electrical signals output by the pixel array of the at least one detection unit, after performing convolution processing with the same step length using convolution kernels of different sizes on the at least one detection unit, multiple reconstructed pixel arrays of the same size can be obtained, and each reconstructed pixel at the same position has multiple convolution data. Figure 6 , the size of the convolution kernel is 3×3, which can extend 2 rows and 2 columns of pixels downward and to the right around the detection unit; refer to Figure 7 The size of the convolution kernel is 5×5, which can expand 4 rows and 4 columns of pixels downward and to the right around the detection unit. Figures 5 to 7After the electrical signals output by the 8×16 pixel array of the detection unit are convolved with convolution kernels of 1×1, 3×3, and 5×5 with a convolution step length of stride=1, an 8×16 reconstructed pixel array can be obtained, which effectively alleviates the influence of the size of the convolution kernel on the convolution operation in the boundary area of ​​the detection unit, and each reconstructed pixel at the same position has 3 data values, including 1×1 convolution data, 3×3 convolution data, and 5×5 convolution data. Therefore, each reconstructed pixel at the same position has waveform data with different characteristics, so that multiple reconstructed pixel arrays have waveform data with different characteristics, such as different resolutions, signal-to-noise ratios, echo pulse intensity and other characteristics, which makes it convenient for the processing device 130 to select the optimal reconstructed pixel array according to the waveform data of multiple reconstructed pixel arrays, and generate a laser radar point cloud according to the selected optimal reconstructed pixel array, which is conducive to further improving the adaptive ability and dynamic range of the laser radar and improving the detection performance of the laser radar.

[0105] According to an embodiment of the present invention, during the convolution process of the laser radar, the size of the convolution kernel does not need to remain consistent, and can be adaptively adjusted according to the waveform data of the reconstructed pixels, thereby maximizing the laser radar's range finding capability while also improving the high signal-to-noise ratio and the spatial resolution of the target. Fig.14 In the entire field of view of the laser radar, for the small field of view area corresponding to each detection unit, the size of the convolution kernel can be determined according to the waveform data of the recombined pixels of the detected target. For example, a smaller size (such as 1×1) convolution kernel is used for close-range, strong echo (high reflectivity) targets, and a larger size (such as 3×3) convolution kernel is used for long-range, low reflectivity or small target signals. This makes the strong echo signal area (high signal-to-noise ratio area R1) have a higher spatial resolution, while the weak echo signal area (low signal-to-noise ratio area R2) has a higher signal-to-noise ratio and a stronger distance measurement capability. This can ensure that long-range, low reflectivity, and small target signals can be detected, and that the detection of close-range, strong echo (high reflectivity) targets has a higher spatial resolution. It can also reduce the expansion problem of high reflectivity objects, which is beneficial to improving the accuracy of the detection results and enabling the laser radar to dynamically maintain better detection performance in the entire field of view.

[0106] In addition, the present invention also relates to a detection method 10 of a laser radar 100, referring to Figure 1 and Fig.15 , the laser radar 100 includes a detection device 120, the detection device 120 includes a plurality of detection units, each detection unit includes a pixel array; the detection method 10 includes steps S11-S14:

[0107] In step S11, a detection light beam L is emitted to detect an obstacle OB;

[0108] In step S12, respond to the echo beam L' reflected by the detection beam L on the obstacle OB and convert it into an electrical signal;

[0109] In step S13, the electrical signals outputted by the pixel array of at least one detection unit are reorganized in different forms to obtain a plurality of reorganized pixel arrays; and

[0110] In step S14, a lidar point cloud is generated based on the multiple recombined pixel arrays.

[0111] According to a preferred embodiment of the present invention, the step S13 comprises: performing convolution processing on the electrical signal output by the pixel array of the at least one detection unit using a plurality of convolution kernels of different sizes to obtain the plurality of recombined pixel arrays.

[0112] According to a preferred embodiment of the present invention, step S14 comprises: selecting a recombined pixel array from the multiple recombined pixel arrays according to the waveform data of the multiple recombined pixel arrays, and generating a lidar point cloud according to the selected recombined pixel array.

[0113] According to a preferred embodiment of the present invention, the waveform data includes a signal-to-noise ratio and / or a pulse intensity.

[0114] According to a preferred embodiment of the present invention, step S14 comprises:

[0115] determining a signal-to-noise ratio for each reconstructed pixel array; and

[0116] Select from the plurality of reconstructed pixel arrays a pixel with a signal-to-noise ratio greater than a signal-to-noise ratio threshold T SNR of the recombined pixel array.

[0117] According to a preferred embodiment of the present invention, the step S14 further comprises: when the signal-to-noise ratio of the plurality of recombined pixel arrays is higher than the signal-to-noise ratio threshold T SNR When , the reconstructed pixel array with the smallest convolution kernel size is selected.

[0118] According to a preferred embodiment of the present invention, the step S14 comprises: determining the pulse intensity of each recombined pixel array, and selecting from the plurality of recombined pixel arrays the pulse intensity of which is greater than a pulse intensity threshold T PS of the recombined pixel array.

[0119] According to a preferred embodiment of the present invention, the step S14 comprises: when the pulse intensity of the plurality of recombined pixel arrays is higher than the pulse intensity threshold T PS When , the reconstructed pixel array with the smallest convolution kernel size is selected.

[0120] According to a preferred embodiment of the present invention, the step S13 further comprises: expanding the array of electrical signals output by the pixel array of the at least one detection unit.

[0121] According to a preferred embodiment of the present invention, the step of expanding the array of electrical signals output by the pixel array of at least one detection unit includes: expanding the array of electrical signals output by the pixel array of at least one detection unit according to a convolution kernel of the largest size.

[0122] According to a preferred embodiment of the present invention, the step of expanding the array of electrical signals output by the pixel array of at least one detection unit also includes: when the pixel array of the at least one detection unit responds to the echo light beam and converts it into electrical signals, controlling the pixels around the pixel array of the at least one detection unit to simultaneously respond to the echo light beam and convert it into electrical signals; and expanding the array of electrical signals output by the pixel array of the at least one detection unit through the electrical signals output by the surrounding pixels.

[0123] According to a preferred embodiment of the present invention, each pixel array includes a plurality of pixels arranged in an array, wherein the pixels include single photon avalanche diodes (SPADs), and each single photon avalanche diode (SPAD) can be independently gated and addressed.

[0124] The present invention also provides a computer-readable storage medium, comprising computer-executable instructions stored thereon, wherein the executable instructions, when executed by a processor, implement the detection method 10 of the laser radar 100 as described above.

[0125] In some preferred embodiments, the computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it. The computer-readable storage medium can adopt any combination of one or more computer-readable media. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, or semiconductor form or device, and more specific examples (non-exhaustive list) include: an electrical connection with one or more wires, a portable computer hard disk, a hard disk, a random access memory (RAM), a non-volatile random access memory (NVRAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0126] In addition, the present invention also provides a computer device (not shown), comprising a memory and a processor, wherein the memory is configured to store computer executable instructions, and the executable instructions implement the detection method 10 described above when executed by the processor.

[0127] In some preferred embodiments, the processor / processing device 130 may include a central processing unit (CPU), and may also include other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The present invention is not limited to this and it depends on the specific circumstances.

[0128] In summary, the scheme of the present invention has been introduced in detail. By adopting the scheme of the present invention, the electrical signals output by the pixel array of at least one detection unit are reorganized in different forms to obtain multiple recombined pixel arrays. A lidar point cloud is generated based on the multiple recombined pixel arrays, and the distance, resolution and other information of the obstacle are calculated, thereby improving the signal-to-noise ratio, distance measurement capability and accuracy of the detection results of the lidar.

[0129] Furthermore, by using convolution kernels of different sizes to perform different forms of convolution processing on the electrical signals output by the pixel array, multiple reconstructed pixel arrays can be obtained. The multiple reconstructed pixel arrays obtained in this way have waveform data with different characteristics, such as different resolutions, signal-to-noise ratios, echo pulse strengths, etc. According to the multiple reconstructed pixel arrays, the optimal waveform data can be selected to generate a laser radar point cloud, thereby improving the detection performance of the laser radar.

[0130] Furthermore, based on the waveform data of multiple recombined pixel arrays, one of the recombined pixel arrays can be selected to generate a lidar point cloud and calculate obstacle information. This can maximize the lidar's ranging capability while improving the high signal-to-noise ratio and spatial resolution of the lidar point cloud, thereby improving the accuracy of the detection results.

[0131] Furthermore, by expanding the array of electrical signals output by the pixel array of the at least one detection unit, the sizes of multiple recombined pixel arrays obtained after different forms of convolution processing can be made consistent, so that each recombined pixel has multiple convolution results after different forms of convolution processing, so that the optimal size of the recombined pixel array can be selected according to the waveform data of the recombined pixel, which is beneficial to improving the adaptability and dynamic range of the lidar.

[0132] In the actual detection process, the laser radar can adaptively adjust the size of the convolution kernel according to the waveform data of the reconstructed signal and select the convolution result of the convolution kernel with the optimal size. This can improve the signal-to-noise ratio and ranging capability of the laser radar while taking into account the spatial resolution of the target, ensuring the accuracy of the detection results and dynamically maintaining better detection performance in the entire field of view.

[0133] It should be noted that, although several modules or submodules of the laser radar 100 are mentioned in the above detailed description, such division is not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules described above can be implemented in one module. Conversely, the features and functions of one module described above can be further divided into multiple modules to be embodied.

[0134] It should be noted that this specification provides method operation steps as described in the embodiments or schematic diagrams, but more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual system or device product is executed, it can be executed in the order or in parallel according to the method shown in the embodiments or flowcharts.

[0135] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A laser radar, comprising: A transmitting device configured to transmit a detection beam for detecting obstacles; A detection device, comprising a plurality of detection units, wherein each detection unit comprises a pixel array, wherein the pixel array is configured to respond to an echo beam reflected by the detection beam on an obstacle and convert the echo beam into an electrical signal; and A processing device is coupled to the detection device and configured to: Recombining the electrical signals output by the pixel array of at least one detection unit in different forms to obtain a plurality of recombined pixel arrays; A lidar point cloud is generated based on the multiple recombined pixel arrays.

2. The laser radar according to claim 1, wherein the processing device is configured to: perform convolution processing on the electrical signal output by the pixel array of the at least one detection unit using multiple convolution kernels of different sizes to obtain the multiple recombined pixel arrays.

3. The laser radar according to claim 2, wherein the processing device is configured to: select a recombined pixel array from the multiple recombined pixel arrays based on the waveform data of the multiple recombined pixel arrays, and generate a laser radar point cloud based on the selected recombined pixel array.

4. A laser radar according to claim 3, wherein the waveform data includes signal-to-noise ratio and / or pulse intensity.

5. The laser radar according to claim 4, wherein the processing device is configured to: determine the signal-to-noise ratio of each recombined pixel array, and select a recombined pixel array with a signal-to-noise ratio greater than a signal-to-noise ratio threshold from the multiple recombined pixel arrays.

6. The laser radar according to claim 5, wherein the processing device is configured to: when the signal-to-noise ratio of multiple recombined pixel arrays is higher than the signal-to-noise ratio threshold, select the recombined pixel array with the smallest convolution kernel size.

7. The laser radar according to claim 4, wherein the processing device is configured to: determine the pulse intensity of each recombined pixel array, and select a recombined pixel array with a pulse intensity greater than a pulse intensity threshold from the multiple recombined pixel arrays.

8. The laser radar according to claim 7, wherein the processing device is configured to: when the pulse intensity of multiple recombined pixel arrays is higher than the pulse intensity threshold, select the recombined pixel array with the smallest convolution kernel size.

9. The laser radar according to any one of claims 1-8, wherein the processing device is configured to expand the array of electrical signals output by the pixel array of the at least one detection unit.

10. The laser radar according to claim 9, wherein the processing device is configured to expand the array of electrical signals output by the pixel array of the at least one detection unit according to a convolution kernel of the largest size.

11. The laser radar according to claim 10, wherein the processing device is configured to: when the pixel array of the at least one detection unit responds to the echo light beam and converts it into an electrical signal, control the pixels around the pixel array of the at least one detection unit to simultaneously respond to the echo light beam and convert it into an electrical signal; and expand the array of electrical signals output by the pixel array of the at least one detection unit through the electrical signals output by the surrounding pixels.

12. The laser radar according to claim 11, wherein each pixel array comprises a plurality of pixels arranged in an array, wherein the pixels comprise single photon avalanche diodes, and each single photon avalanche diode can be independently enabled and addressed.

13. A detection method of a laser radar, the laser radar comprising a detection device, the detection device comprising a plurality of detection units, each detection unit comprising a pixel array; the detection method comprising: S11: emits a detection beam to detect obstacles; S12: responding to the echo beam reflected by the detection beam on the obstacle and converting it into an electrical signal; S13: Recombining the electrical signals output by the pixel array of at least one detection unit in different forms to obtain a plurality of recombined pixel arrays; and S14: Generate a laser radar point cloud according to the multiple recombined pixel arrays.

14. The detection method according to claim 13, wherein the step S13 comprises: The electrical signals output by the pixel array of the at least one detection unit are convolved using a plurality of convolution kernels of different sizes to obtain the plurality of recombined pixel arrays.

15. The detection method according to claim 14, wherein the step S14 comprises: According to the waveform data of the multiple recombined pixel arrays, a recombined pixel array is selected from the multiple recombined pixel arrays, and a laser radar point cloud is generated according to the selected recombined pixel array. The detection method according to claim 15 , wherein the waveform data comprises a signal-to-noise ratio and / or a pulse intensity.

17. The detection method according to claim 16, wherein the step S14 comprises: determining a signal-to-noise ratio for each reconstructed pixel array; and A reconstructed pixel array having a signal-to-noise ratio greater than a signal-to-noise ratio threshold is selected from the plurality of reconstructed pixel arrays.

18. The detection method according to claim 17, wherein the step S14 further comprises: When the signal-to-noise ratios of the multiple recombined pixel arrays are higher than the signal-to-noise ratio threshold, the recombined pixel array with the smallest convolution kernel size is selected.

19. The detection method according to claim 16, wherein the step S14 comprises: The pulse intensity of each reconstructed pixel array is determined, and a reconstructed pixel array having a pulse intensity greater than a pulse intensity threshold is selected from the plurality of reconstructed pixel arrays.

20. The detection method according to claim 19, wherein the step S14 comprises: When the pulse intensities of the multiple recombined pixel arrays are higher than the pulse intensity threshold, the recombined pixel array with the smallest convolution kernel size is selected.

21. The detection method according to any one of claims 13 to 20, wherein the step S13 further comprises: The array of electrical signals output by the pixel array of the at least one detection unit is expanded.

22. The detection method according to claim 21, wherein the step of expanding the array of electrical signals output by the pixel array of at least one detection unit comprises: The array of electrical signals output by the pixel array of the at least one detection unit is expanded according to the convolution kernel with the largest size.

23. The detection method according to claim 22, wherein the step of expanding the array of electrical signals output by the pixel array of at least one detection unit further comprises: When the pixel array of the at least one detection unit responds to the echo light beam and converts it into an electrical signal, the pixels around the pixel array of the at least one detection unit are controlled to respond to the echo light beam and convert it into an electrical signal at the same time; and the array of electrical signals output by the pixel array of the at least one detection unit is expanded through the electrical signals output by the surrounding pixels.

24. The detection method according to claim 23, wherein each pixel array comprises a plurality of pixels arranged in an array, wherein the pixels comprise single photon avalanche diodes, and each single photon avalanche diode can be independently enabled and addressed.

Images