Laser point cloud data processing method, device and equipment

Abstract

The embodiment of the invention relates to the technical field of laser radars, and discloses a laser point cloud data processing method, device and equipment, and the method comprises the steps: obtaining a weather state parameter of a current environment; adaptively configuring detection parameters of the laser radar according to the weather state parameters; wherein the detection parameters comprise transmitting parameters and receiving parameters, the transmitting parameters at least comprise a plurality of laser transmitting power distribution parameters of wave bands with different weather penetration characteristics, and the receiving parameters at least comprise the direction of a polaroid; controlling the laser radar to execute detection based on the configured detection parameters, and obtaining original point cloud data of the target object; wherein the original point cloud data comprises point cloud data corresponding to a plurality of wave bands respectively; and performing data completion and fusion processing according to the original point cloud data and the weather state parameters, and determining enhanced point cloud data of the target object. According to the technical scheme, the detection performance of the laser radar under the severe weather condition can be improved.

Description

Technical Field

[0001] This application relates to the field of lidar technology, specifically to a method, apparatus, and device for processing laser point cloud data. Background Technology

[0002] In high-precision 3D perception scenarios such as intelligent driving, low-altitude security, and industrial automation, LiDAR is widely used as a core sensor for environmental modeling, target detection, and obstacle avoidance decision-making.

[0003] However, under adverse weather conditions such as rain, fog, and snow, precipitation particles (raindrops, fog droplets, or snowflakes, etc.) strongly scatter and absorb laser signals, leading to a significant attenuation of echo intensity, a shortened detection range, and even detection blind zones. Simultaneously, stray light reflection from precipitation particles can mask target echoes, resulting in significant gaps in point cloud data, causing broken target outlines and posing risks of missed detections and false detections. In conclusion, the detection performance of lidar under adverse weather conditions such as rain, fog, and snow needs further improvement. Summary of the Invention

[0004] In view of the above problems, this application provides a laser point cloud data processing method, apparatus and equipment to solve the problem that the detection performance of lidar needs to be further improved under adverse weather conditions such as rain, fog and snow.

[0005] According to one aspect of the embodiments of this application, a laser point cloud data processing method is provided, the method comprising:

[0006] Obtain the current weather status parameters;

[0007] The detection parameters of the lidar are adaptively configured according to the weather state parameters; wherein, the detection parameters include transmission parameters and reception parameters, the transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction;

[0008] The lidar is controlled to perform detection based on the configured detection parameters to acquire the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data corresponding to multiple wavebands respectively;

[0009] Based on the original point cloud data and the weather state parameters, data completion and fusion processing are performed to determine the enhanced point cloud data of the target object.

[0010] According to another aspect of the embodiments of this application, another laser point cloud data processing method is provided, the method comprising:

[0011] The target object is detected by a dual-band lidar to obtain the original point cloud data of the target object; wherein the original point cloud data includes point cloud data based on near-infrared and point cloud data based on mid-infrared.

[0012] The original point cloud data is subjected to weighted fusion processing to determine the enhanced point cloud data of the target object.

[0013] According to another aspect of the embodiments of this application, a laser point cloud data processing apparatus is provided, comprising:

[0014] The acquisition unit is used to acquire the weather state parameters of the current environment;

[0015] A configuration unit is used to adaptively configure the detection parameters of the lidar according to the weather state parameters; wherein the detection parameters include transmission parameters and reception parameters, the transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction;

[0016] The detection unit is used to control the lidar to perform detection based on the configured detection parameters and acquire the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data corresponding to multiple wavebands respectively;

[0017] The processing unit is used to perform data completion and fusion processing based on the original point cloud data and the weather state parameters to determine the enhanced point cloud data of the target object.

[0018] According to another aspect of the embodiments of this application, an electronic device is provided, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0019] The memory is used to store at least one executable instruction that causes the processor to perform the laser point cloud data processing method as described above.

[0020] According to another aspect of the embodiments of this application, another laser point cloud data processing apparatus is provided, including:

[0021] The first processing unit is used to detect the target object using a dual-band lidar and acquire the original point cloud data of the target object; wherein the original point cloud data includes near-infrared point cloud data and mid-infrared point cloud data.

[0022] The second processing unit is used to perform weighted fusion processing on the original point cloud data to determine the enhanced point cloud data of the target object.

[0023] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided, the storage medium storing at least one executable instruction that causes an electronic device / apparatus to perform the operation of the laser point cloud data processing method as described above.

[0024] This application's embodiments optimize detection performance under different weather conditions by real-time sensing of weather conditions and adaptively adjusting the multi-band transmission power and receiving polarization direction of the lidar accordingly. Furthermore, it utilizes weather state parameters to intelligently weight and fuse the original multi-band point clouds, generating a significantly enhanced fused point cloud. Through a complete technical loop encompassing environmental perception, hardware parameter adaptation, and data post-processing, it effectively solves the industry problem of severe point cloud quality degradation in adverse weather conditions such as rain, snow, and fog, significantly improving the reliability and accuracy of all-weather environmental perception systems.

[0025] The above description is merely an overview of the technical solutions of the embodiments of this application. In order to better understand the technical means of the embodiments of this application and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of this application more obvious and understandable, specific implementation methods of this application are described below. Attached Figure Description

[0026] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0027] Figure 1 A block diagram of the architecture of a dual-band lidar provided in an embodiment of this application;

[0028] Figure 2 A schematic flowchart illustrating a laser point cloud data processing method provided in an embodiment of this application;

[0029] Figure 3 A flowchart illustrating another laser point cloud data processing method provided in this application embodiment;

[0030] Figure 4 A flowchart illustrating another laser point cloud data processing method provided in this application embodiment;

[0031] Figure 5 A schematic diagram of the structure of a laser point cloud data processing device provided in an embodiment of this application;

[0032] Figure 6 A schematic diagram of another laser point cloud data processing device provided in an embodiment of this application;

[0033] Figure 7This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0034] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. Although exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein.

[0035] As a high-precision 3D detection device, LiDAR relies on the propagation and echo reception of laser signals. However, its detection performance faces severe challenges in adverse weather conditions such as rain, fog, and snow. For example, in autonomous driving scenarios, vehicles need to perceive the road and obstacles ahead in real time using LiDAR. However, raindrops, fog droplets, or snowflakes strongly scatter and absorb laser signals, resulting in a significant attenuation of echo intensity. This reduces the detection range by more than 50% compared to clear weather, and may even create blind spots. Simultaneously, stray light reflection from precipitation particles can obscure target echoes, leading to significant gaps in point cloud data. This results in broken vehicle outlines and blurred obstacle boundaries, potentially causing obstacle avoidance misjudgments or path planning errors. In low-altitude security, drones or monitoring equipment rely on LiDAR to identify unauthorized intrusion targets, but data loss in adverse weather conditions can lead to missed detections, threatening security. Furthermore, in industrial automation scenarios (such as logistics warehousing and robot navigation), the quality of LiDAR point clouds directly affects the positioning accuracy of robotic arms or the reliability of path planning.

[0036] To address the performance degradation of lidar under severe weather conditions, common solutions often involve increasing the power of a single band or simple filtering, such as increasing the power of near-infrared lasers to counteract attenuation. However, high power can easily lead to overheating and overload of the equipment, and cannot solve stray light interference. Some solutions introduce image-assisted completion, but these rely on external cameras, which also suffer from blurred vision under severe weather conditions, resulting in low reliability of data fusion.

[0037] To address the aforementioned technical issues, this application provides a laser point cloud data processing method. By integrating multi-dimensional technologies such as multi-band laser collaborative emission, polarization anti-interference, weather adaptive adjustment, and deep learning data completion, this method can solve the problems of laser radar detection attenuation and data loss under severe weather conditions, thereby achieving high-precision three-dimensional perception of laser radar under severe weather conditions.

[0038] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0039] It should be noted that the execution subject of the laser point cloud data processing method provided in this application embodiment can be a laser point cloud data processing device. The laser point cloud data processing device can be integrated into electronic devices such as lidar systems, computing units of autonomous vehicles, or independent point cloud processing devices. This application embodiment does not impose any restrictions, and the method of this application can be implemented by software, hardware, or a combination of software and hardware.

[0040] For example, Figure 1 This is a block diagram illustrating the architecture of a dual-band lidar as provided in an embodiment of this application. Figure 1 As shown, the lidar may include a dual-band transmitting module, a polarization receiving module, a multi-modal data fusion module, an adaptive control module, and a weather sensing module.

[0041] The weather sensing module collects ambient light spectral characteristics and precipitation particle scattering signals from the atmosphere, matching features to output weather state parameters to the adaptive control module and the multimodal data fusion module. The adaptive control module adjusts the power ratio and scanning frequency of the near-infrared and mid-infrared bands of the dual-band infrared transmitter in real time based on the weather state parameters, while simultaneously adjusting the polarizer direction of the polarization receiver module. After the dual-band transmitter module adjusts the dual-band ratio, it emits a dual-source laser to detect the reflected echo from the target object. The polarization receiver module receives the target polarized echo and outputs point cloud data to the multimodal data fusion module. The multimodal data fusion module first performs dB4 wavelet denoising on the dual-band digital signal, decomposes it into three layers, and then applies soft thresholding suppression (threshold = 0.02 × maximum signal value) to the high-frequency noise coefficients. It retains the target signal coefficients and reconstructs them, removing point cloud noise caused by raindrop stray noise. Subsequently, data integrity is assessed by calculating the point cloud missing rate (valid points / theoretical points). A threshold is set; when the missing rate exceeds this threshold, dual-band point cloud data and weather condition parameters are used as input to fuse and calculate the missing areas, improving the point cloud density. Finally, by adjusting the weight of mid-infrared data in echo data under different weather conditions, and combining near-infrared details with mid-infrared contours, a complete 3D point cloud of the target is generated. Through dual-band laser collaboration, polarization anti-interference, weather adaptive adjustment, and deep learning completion, the problems of lidar detection attenuation and data loss under severe weather conditions are solved, improving the lidar's detection performance in adverse weather conditions such as rain, fog, and snow.

[0042] Figure 2 This is a flowchart illustrating a laser point cloud data processing method provided in an embodiment of this application. This method can be executed by a laser point cloud data processing device. Figure 2 As shown, the method may include the following steps:

[0043] Step 210: Obtain the current weather status parameters.

[0044] For example, weather state parameters refer to physical quantities that can quantify the current atmospheric conditions of the detection environment. These parameters can be monitored and quantified in real time by an integrated environmental sensing unit, allowing the lidar to detect and quantify the external meteorological conditions. For instance, real-time weather information can be received from a meteorological service network via vehicle-mounted weather sensors (such as visibility meters, humidity sensors, rain gauges, etc.), multispectral imaging equipment fixed to the lidar, or a vehicle-mounted communication module. Real-time weather information may include one or more of visibility, relative humidity, rainfall intensity, snowfall intensity, and haze concentration. This real-time weather information can be used as weather state parameters, or weather can be classified based on the real-time weather information to determine the weather state parameters. This application does not impose any limitations on these embodiments.

[0045] For example, spectral signals of the atmospheric environment can be collected using multispectral sensors. Multispectral sensors can detect atmospheric aerosol concentrations, identify cloud morphology and distribution, detect precipitation intensity in clouds, and atmospheric vertical water vapor content. Subsequently, weather classification units (including but not limited to support vector machine models) are used to classify the weather based on the relevant atmospheric data detected by the multispectral sensors, and finally output the current weather type (such as sunny, rainy, foggy, snow) and its intensity level (such as light rain, moderate rain, heavy rain) as weather state parameters.

[0046] By proactively and in real-time acquiring quantified weather state parameters, rather than relying on fixed thresholds or unreliable external data, the system receives accurate environmental perception input. This transforms lidar from passive adaptation to proactive optimization, forming the decision-making basis for achieving full-scene adaptive performance.

[0047] Step 220: Adaptively configure the detection parameters of the lidar according to the weather state parameters; wherein, the detection parameters include transmission parameters and reception parameters, the transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction.

[0048] For example, in this embodiment, the lidar is a multi-band adjustable lidar capable of emitting at least two laser beams of different wavelengths, such as a dual-source design of "near-infrared + mid-infrared". Based on the acquired weather state parameters, the lidar's detection parameters (i.e., operating parameters) are adaptively and dynamically adjusted through a control logic or algorithm model preset within the lidar point cloud data processing device. This transforms the lidar's detection mode from a fixed state to a closed-loop system linked to the environmental conditions, optimizing signal transmission and reception at the hardware level and laying the foundation for acquiring higher-quality raw data under adverse weather conditions.

[0049] Understandably, lasers of different wavelengths are scattered and absorbed to varying degrees by particles such as rain, fog, and snow when propagating in the atmosphere, with longer wavelengths (mid-infrared) generally having better penetration capabilities. Laser emission power allocation parameters refer to the strategy of dynamically distributing the total output power among these two or more bands. For example, in clear weather, higher power can be allocated to the near-infrared band, which offers better resolution; in dense fog or heavy rain, the power proportion of the mid-infrared band is significantly increased, prioritizing detection range and basic contour information at the expense of some detail.

[0050] Configuring the polarizer direction refers to controlling the angle of the polarizing filter (such as an electrically controlled rotating polarizer with a response time <5μs) in the receiving optical path to match the polarization direction of the emitted laser or to be at the optimal angle for suppressing stray light. Since the reflections of target objects (such as vehicles and buildings) differ in polarization characteristics from those of spherical raindrops and fog droplets, this configuration effectively filters out a large amount of non-target echoes (noise) generated by precipitation particles whose polarization state has changed, thereby effectively suppressing interference from weather background scattering noise and significantly improving the signal-to-noise ratio of the received signal. The maximum scanning angle of the MEMS scanning unit is ±30°.

[0051] Furthermore, the transmission parameters can include the scanning frequency (e.g., the scanning frequency of a MEMS galvanometer can be adaptively adjusted from 20Hz to 100Hz), and the reception parameters can include the detector gain, etc. In adverse weather conditions, the scanning frequency can be appropriately reduced to increase the residence time of the single-point laser and accumulate more echo energy; at the same time, the detector gain can be increased to amplify weak target signals.

[0052] Through closed-loop control from environmental perception to hardware parameter adjustment, the transmission and reception parameters can be adaptively configured for different weather conditions. At the physical level, the energy input and noise suppression of the signal are optimized to the maximum extent, providing a fundamental guarantee for obtaining the highest possible quality raw endpoint cloud data under harsh conditions, while avoiding the energy consumption and heat generation problems caused by a single high-power solution.

[0053] Step 230: Control the lidar to perform detection based on the configured detection parameters and obtain the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data corresponding to multiple bands respectively.

[0054] For example, after configuring the lidar detection parameters, the lidar is further controlled to scan and detect according to the configured parameters. The target object can be a vehicle, pedestrian, road infrastructure, or other object to be sensed. The raw point cloud data refers to the set of point clouds directly generated by the lidar without subsequent fusion processing. Due to the use of multi-band detection, the final raw point cloud data consists of multiple sub-point cloud datasets. Each subset corresponds to a specific lidar band and contains the three-dimensional coordinates of the points detected under that band (and may also include intensity, polarization information, etc.). These multi-band point cloud data are synchronized and calibrated spatially and temporally, providing a multi-view, multi-characteristic data source for subsequent fusion processing.

[0055] For example, the lidar's transmitting module drives near-infrared and mid-infrared lasers to emit laser pulses based on configured power distribution parameters. The two laser beams are coaxially combined using a beam combiner to ensure they have identical optical paths and fields of view. Then, they are projected onto the target area via a MEMS scanning mirror at a configured scanning frequency. The echo signal is filtered by a polarizer set at a configured angle to remove most of the stray light (mainly from precipitation particles) that is not polarized to the target. Subsequently, the filtered light signal is separated by wavelength by a beam splitter and received by photodetectors of the corresponding band (e.g., InGaAs detectors for near-infrared and HgCdTe detectors for mid-infrared), converting them into weak electrical signals. This electrical signal is then amplified by a variable gain amplifier configured with appropriate gain, and finally converted into a digital signal by an analog-to-digital converter. The lidar's processing unit (e.g., an FPGA) then processes the digital signal for each band, calculating the three-dimensional coordinates (distance, azimuth, and elevation angles) and reflection intensity information of each detection point using principles such as time-of-flight (ToF), thus forming the raw point cloud data for that band. Furthermore, the final output is raw point cloud data, which contains two or more sets of point clouds from the near-infrared and mid-infrared bands. Each set of point clouds reflects the three-dimensional morphology and reflection characteristics of the target in that band.

[0056] By executing the aforementioned collaboratively optimized detection process, multi-band, spatially aligned raw point cloud data can be acquired simultaneously. Even if a single-band data is severely degraded, the two-band data can still complement each other, providing a rich information foundation for subsequent data restoration. The coaxial design avoids parallax issues and simplifies the complexity of data fusion.

[0057] Step 240: Based on the original point cloud data and weather condition parameters, perform data completion and fusion processing to determine the enhanced point cloud data of the target object.

[0058] For example, due to weather interference, the original point cloud data of the target object may contain missing points (sparse points) and noise. Therefore, further processing is required to improve the point cloud quality. For instance, for a point cloud missing area in a certain band due to severe weather attenuation, the relatively complete point cloud information of another band in that area can be used for interpolation or model-driven point cloud generation to complete the point cloud morphology of the target object.

[0059] After obtaining the completed multi-band point cloud, it is then fused based on weather condition parameters (especially intensity levels). For example, an adaptive weighted fusion algorithm is used, assigning fusion weights to the point cloud data of each band according to the current weather condition parameters. The more severe the weather conditions, the higher the weight is given to the point cloud data of bands with stronger penetration (such as mid-infrared). For example, on a clear day, the weight of near-infrared point cloud can be set to 0.8 and mid-infrared to 0.2 to retain more details; on a foggy day, the weights may be adjusted to 0.3 for near-infrared and 0.7 for mid-infrared to rely on more reliable contour information. Finally, the weighted dual-band point clouds are synthesized to generate the final enhanced point cloud data.

[0060] By combining physical perception and data analysis, not only are data gaps filled, but the advantages of multi-band information are also integrated through intelligent weighted fusion. The final output enhanced point cloud has higher point cloud density, more complete target contours, and more accurate 3D information, thereby significantly improving the perception reliability of LiDAR in harsh environments such as rain, fog, and snow, directly supporting safety decisions in applications such as autonomous driving and robot navigation.

[0061] In this embodiment, the provided laser point cloud data processing method optimizes detection performance under different weather conditions from the source by real-time sensing of weather conditions and adaptively adjusting the multi-band transmission power and receiving polarization direction of the lidar accordingly. Furthermore, it utilizes weather state parameters to perform intelligent weighted fusion and completion of the original multi-band point cloud, generating a significantly enhanced fused point cloud. This method forms a complete technical closed loop from environmental perception and hardware parameter adaptation to data post-processing, effectively solving the industry problem of severe point cloud quality degradation in existing lidar systems under adverse weather conditions such as rain, snow, and fog, and significantly improving the reliability and accuracy of all-weather environmental perception systems.

[0062] Figure 3 This is a flowchart illustrating another laser point cloud data processing method provided in this application embodiment. Based on the previous embodiment, this method further refines the acquisition method of weather state parameters, introduces a data quality assessment and intelligent completion mechanism, and expands the adaptively configurable detection parameters, thereby forming a more precise and robust closed-loop processing flow. This method can be executed by a laser point cloud data processing device, such as... Figure 3 As shown, the method may include the following steps:

[0063] Step 310: Collect the environmental spectral signal of the current environment using a multispectral sensor.

[0064] For example, unlike ordinary cameras or single-point light intensity detectors, multispectral sensors can simultaneously or rapidly time-divisionally detect multiple discrete, specific spectral bands with high sensitivity. By actively collecting spectral signals from the atmospheric environment, these signals carry crucial information such as the composition, concentration, and particle size distribution of aerosols, clouds, and precipitation particles (e.g., raindrops, fog droplets, ice crystals). By analyzing the intensity, wavelength distribution, and specific absorption / scattering peak characteristics of these spectral signals, a precise understanding of the atmospheric physical state can be derived.

[0065] The use of multispectral sensors provides richer and more physical environmental data than single visible light images or simple humidity sensors. This provides a reliable data source for subsequent accurate weather classification and quantification, and is the first cornerstone for achieving high-precision adaptive control.

[0066] Step 320: Input the environmental spectral signal into the support vector machine model to analyze the position of the scattering peak and the ambient light intensity, and output the weather state parameters of the current environment; wherein, the weather state parameters include at least the weather type and intensity level.

[0067] For example, the Support Vector Machine (SVM) model is a pre-trained classification and regression model whose training data comes from a large number of environmental spectral signal samples collected under different weather conditions (such as sunny, foggy, rainy, snowy, hazy, etc.) and different intensity levels.

[0068] The environmental spectral signal of the current environment, collected by a multispectral sensor, is input into the SVM model. The SVM model analyzes the positions of scattering peaks in the input spectrum and the intensity attenuation of ambient light in key bands. For example, raindrops or fog droplets of different sizes exhibit different Mie scattering characteristics for specific wavelengths of laser light, forming identifiable spectral peaks. Simultaneously, the overall light intensity attenuation is directly related to visibility. Based on the attenuation ratio of ambient light intensity in different bands, the current weather type (such as drizzle, moderate rain, dense fog, haze, etc.) can be accurately identified and its intensity level (such as rainfall level and visibility range) can be quantified. Based on the above feature analysis, the SVM model outputs structured weather state parameters. These parameters include at least two parts: 1) Weather type: such as discrete categories like "sunny," "rain," "fog," and "snow"; 2) Intensity level: such as "light rain," "moderate rain," and "heavy rain" for precipitation, or quantified levels like "light fog," "moderate fog," and "dense fog" for fog. This parameter provides a precise and quantitative "diagnostic report" of the current environment.

[0069] By using machine learning models to intelligently analyze spectral signals, an automatic conversion from raw data to semantic parameters that can be directly used for control decisions is achieved. This method is more accurate and adaptable than rule-based judgments based on fixed thresholds, and can distinguish between similar weather phenomena with different optical characteristics, providing a more accurate key basis for subsequent refined adaptive configuration.

[0070] Step 330: Adaptively configure the detection parameters of the lidar based on the weather condition parameters.

[0071] The detection parameters include transmission parameters and reception parameters. The transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction. Optionally, the transmission parameters also include the scanning frequency, and the reception parameters also include the detector gain. When adaptively configuring the lidar's detection parameters according to weather state parameters, the lidar's detection parameters are specifically configured according to the weather type and intensity level indicated by the weather state parameters.

[0072] For example, the emission parameters include at least the allocation parameters of laser emission power for multiple bands (such as near-infrared and mid-infrared). The power ratio for each band can be determined by looking up a table or by calculating using a function, depending on the weather type and intensity level. For instance, in moderate rain conditions, the ratio could be set to near-infrared:mid-infrared = 4:6 (e.g., near-infrared 1.6W, mid-infrared 2.4W).

[0073] The receiving parameters include at least the polarizer orientation. Similarly, based on the weather type and intensity level, the optimal polarization angle that best suppresses stray light from the primary source of interference (such as raindrops of a specific size) can be determined by looking up a table or by function calculation, and the electrically controlled rotating polarizer is driven to adjust to that angle. For example, in moderate rain, the polarizer of the polarization receiving module is controlled to rotate to 45°.

[0074] In addition, transmission parameters may include the scanning frequency. For example, in adverse weather conditions, to increase the signal energy at a single point, the MEMS scanning mirror can be controlled to reduce the scanning frequency (e.g., from 100Hz to 50Hz); in clear weather, it can be restored to a higher scanning frequency to obtain higher time resolution. Reception parameters may also include the detector gain. The detector gain can be dynamically adjusted according to the expected echo signal strength. When the predicted signal is weak (e.g., at long distances or in strongly attenuated weather), the gain is increased to amplify the effective signal; when the predicted signal is too strong and may saturate (e.g., near highly reflective targets), the gain is reduced to protect the detector and maintain a linear response. For example, in moderate rain, to compensate for signal attenuation, the gain value of the variable gain amplifier can be increased (e.g., from 40dB to 60dB).

[0075] By coordinating and configuring multiple parameters (power, polarization, frequency, gain), the detection behavior of lidar can be adjusted more precisely and in a more three-dimensional way, thereby improving performance while taking into account energy consumption, heat dissipation and equipment lifespan, and coping with complex weather challenges with optimal configuration.

[0076] Step 340: Control the lidar to perform detection based on the configured detection parameters and obtain the original point cloud data of the target object.

[0077] The raw point cloud data includes point cloud data corresponding to multiple bands.

[0078] For example, the specific implementation of step 340 can be referred to the description of step 230, which will not be repeated here.

[0079] Optionally, in one possible embodiment, controlling the lidar to perform detection based on configured detection parameters to acquire raw point cloud data of the target object may include:

[0080] S1, The laser radar emission module controls the emission of lasers in multiple bands based on the configured laser emission power allocation parameters;

[0081] S2, The scanning module of the control lidar performs coaxial scanning of multiple laser bands based on the configured scanning frequency;

[0082] S3. Receive the echo signal reflected by the target object, perform polarization filtering on the echo signal based on the configured polarizer direction, and amplify and convert the filtered signal based on the configured detector gain to obtain the digital signal corresponding to each band.

[0083] S4. Perform wavelet denoising on the digital signals corresponding to each band and use soft thresholding for suppression to generate point cloud data corresponding to each band in the original point cloud data.

[0084] For example, the transmitting module of the lidar in this application includes multiple lasers, each emitting laser light in a different wavelength band corresponding to different weather penetration characteristics. For instance, the transmitting module is a dual-source assembly comprising a near-infrared laser and a mid-infrared laser. The driver of the transmitting module receives and executes instructions from the adaptive control module to precisely control the output power of the two lasers, ensuring that it conforms to the ratio indicated by the laser emission power allocation parameters. This guarantees that the emitted light signal has been energy-optimized for the current weather conditions, laying the energy foundation for obtaining effective echoes under adverse conditions.

[0085] The scanning module of a lidar system refers to a MEMS (Micro-Electro-Mechanical Systems) galvanometer or other beam deflection device. Based on the aforementioned scanning frequency, the operating speed of the scanning module is controlled to adjust the residence time of the laser beam in a single direction. Multiple laser beams of different wavelengths emitted by the transmitting module are combined into a completely identical optical path after passing through a beam combiner, and then reflected by the same scanning mirror to illuminate the target object. In other words, although the wavelengths are different, they point in the same spatial direction at the same time and follow the exact same scanning trajectory. This ensures that the instantaneous field of view of each laser wavelength is completely coincident, and the scanned light spots are strictly aligned in space. Based on this, the subsequently acquired point cloud data of different wavelengths can have inherent consistency and comparability in spatial location, avoiding registration errors caused by different scanning paths.

[0086] When laser light strikes a target object, it generates an echo signal. This echo signal first enters a polarization filter unit, where polarizers (such as electrically controlled rotating polarizers) are adjusted to a specific angle according to a configuration. This angle is set to maximize the transmission of target-reflected light with the same polarization state as the emitted laser, while blocking stray light whose polarization state has changed due to scattering by precipitation particles. The filtered light signal is then separated by wavelength by a beam splitter and fed into corresponding photodetectors (e.g., InGaAs detectors for near-infrared and HgCdTe detectors for mid-infrared) to be converted into weak current signals. These current signals are then input to a variable gain amplifier, whose gain value is pre-configured according to weather conditions (e.g., high gain mode under moderate rain conditions), amplifying the signal to a level suitable for sampling. Finally, an analog-to-digital converter converts the amplified analog electrical signal into a digital signal, with each band operating independently.

[0087] By optimizing the signal at each stage at the receiver, polarization filtering actively suppresses the main noise sources at the optical physics level; configuration-based gain amplification compensates for atmospheric attenuation of the signal at the electronic level. The combination of these two techniques results in the raw signal input to the digital processing unit having the optimal signal-to-noise ratio, providing clean raw data for generating high-quality point clouds.

[0088] Furthermore, multi-scale decomposition (such as 3-level decomposition) can be used to denoise the digital signal, employing techniques such as db4 wavelet denoising. Since the high-frequency coefficients after decomposition contain a large amount of random noise (including residual precipitation scattering noise), a "soft thresholding" function can be used to suppress these high-frequency coefficients. This involves significantly attenuating or zeroing coefficients with amplitudes below an adaptively calculated threshold (e.g., threshold = 0.02 × maximum signal value), while retaining coefficients exceeding the threshold that are more likely to originate from the real target. Finally, wavelet reconstruction is performed on the processed coefficients to obtain the purified time-domain signal.

[0089] Based on this purified signal, and using calculation principles such as Time-of-Flight (ToF), raw point cloud data, which has been preliminarily removed from high-frequency noise and corresponds one-to-one with each band, is finally generated. For example, each digital signal can be processed by a processing unit (such as an FPGA). Specifically, the time of flight of each laser pulse is calculated, and combined with the precise angles (azimuth and elevation) of the scanning mirror at that time, the three-dimensional coordinates (X, Y, Z) of the echo point relative to the radar are obtained through geometric calculation. At the same time, the amplitude information of the signal is recorded as the reflection intensity of that point. By performing the above calculations and collecting all valid echo points within one scanning cycle, the point cloud data for that band is formed. Since there are multiple signals, multiple spatially aligned but complementary point cloud sets are ultimately generated, which together constitute the raw point cloud data.

[0090] The detection performed using the configured detection parameters yields raw point cloud data that represents the optimal initial state achievable under current weather conditions. The multi-band data is spatially aligned and complementary in information (near-infrared may have richer details but more missing information, while mid-infrared may have more complete outlines but less detail), laying a solid foundation for subsequent depth information processing.

[0091] Step 350: Calculate the data missing rate of the original point cloud data; whereby the data missing rate is used to indicate the proportion of invalid echo points.

[0092] For example, instead of unconditionally initiating the computationally intensive completion algorithm, a validity assessment is performed first. The data missing rate is calculated as: (Theoretical number of point clouds - Actual number of valid echo points) / Theoretical number of point clouds. The theoretical number of point clouds is determined by the scanning mode (e.g., number of lines, number of points per line), while the actual number of valid echo points refers to the number of points identified as true target echoes after filtering by a basic signal-to-noise ratio threshold. The data missing rate directly reflects the degree of data incompleteness caused by weather attenuation and occlusion.

[0093] By introducing the data missing rate as a criterion, a quantitative and objective triggering mechanism is provided, which avoids unnecessary computational overhead when the data quality is still acceptable (energy saving) and ensures that the repair process can be started in a timely manner when the data quality deteriorates severely (reliability).

[0094] Step 360: Determine whether the data missing rate is greater than or equal to the preset threshold.

[0095] If yes, proceed to step 370; otherwise, proceed directly to step 380.

[0096] For example, the preset threshold can be set according to the requirements of downstream applications (such as target recognition, map building) for point cloud integrity, or it can be configured by the user, for example, set to 20%, 30%, etc., and this application embodiment does not impose any restrictions. The calculated data missing rate is compared with the preset threshold. If the data missing rate is greater than or equal to the preset threshold, it indicates that the point cloud quality is severely degraded, and the data completion step must be initiated (execute step 370) to restore usable information. If the missing rate is lower than the preset threshold, it indicates that the current point cloud quality can still meet the basic requirements, and the computationally intensive completion step can be skipped, and the system can directly enter the fusion stage (execute step 380) to improve the system's real-time performance.

[0097] For example, if the theoretical number of scan points is 10,000, the number of valid points is 7,500, and the preset threshold is 20%, since the data missing rate = (10,000-7,500) / 10,000 = 25% > 20%, then the data completion step will be executed.

[0098] Step 370: Input the original point cloud data and weather state parameters into the pre-trained point cloud completion model for data completion processing to obtain the completed point cloud data for each band.

[0099] For example, the pre-trained point cloud completion model is a deep learning-based neural network model, such as a variant of a point cloud-specific network model like PointNet++. During training, the model uses a large amount of paired data collected under various weather conditions: incomplete original dual-band point clouds with weather labels as input, and corresponding complete, labeled point clouds as learning targets. For instance, it is trained based on 5000 sets of point cloud data under special weather conditions such as rain and fog. During inference, the model receives the current incomplete dual-band original point cloud data and the weather state parameters of the current environment, and then uses its learned knowledge to more accurately predict and generate the missing point cloud parts, ultimately outputting more dense and complete point cloud data for each band.

[0100] By inputting weather parameters and point cloud data into the model, the completion process is guided by physical common sense and matched with specific weather conditions, rather than just mathematical interpolation, which greatly improves the rationality and accuracy of the completion results.

[0101] Step 380: Based on the weather state parameters, perform data fusion processing on the completed point cloud data corresponding to each band to determine the enhanced point cloud data of the target object.

[0102] For example, even after completion, the quality of point cloud data in different bands still varies. By performing data fusion processing on the completed point cloud data corresponding to each band based on weather state parameters, high-quality, highly complete enhanced point cloud data can be obtained. For instance, the fusion processing can employ an adaptive weighted fusion algorithm, assigning a weight to the point cloud data of each band according to the weather state parameters; the more severe the weather conditions, the higher the weight is given to point clouds in bands with stronger penetration (such as mid-infrared). Then, for multi-band point clouds with spatially aligned coordinates, a weighted average or selective fusion is performed based on their weights to generate the final, unified enhanced point cloud data. If step 260 determines "no" (the missing data is not severe), this fusion step is performed directly using the original point cloud or a point cloud that has only undergone simple filtering.

[0103] By employing an intelligent weighting strategy, the advantages of different frequency bands are dynamically combined, enabling the output of the most complete, accurate, and dense 3D environmental model possible regardless of weather conditions. This fundamentally ensures the reliability of subsequent applications (such as obstacle recognition for autonomous driving).

[0104] Optionally, in one possible embodiment, data fusion processing is performed on the completed point cloud data corresponding to each band based on weather state parameters to determine the enhanced point cloud data of the target object, which may include:

[0105] S10. Based on weather condition parameters, assign different fusion weights to the completed point cloud data of different bands; among them, the stronger the penetration of the band, the higher the weight assigned to its point cloud data under severe weather conditions.

[0106] S20. Based on the fusion weight, perform weighted fusion on the completed point cloud data corresponding to each band to obtain the enhanced point cloud data of the target object.

[0107] For example, fusion weight refers to the contribution coefficient assigned to each band of point cloud data during the fusion process. Weight values ​​are typically normalized so that the sum of the weights for each band is 1. The allocation of weights directly determines which band's data characteristics the final fusion result favors. The allocation can be based on a pre-defined weight mapping rule that reflects physical laws. The core logic of this rule is that the weight coefficients are positively correlated with the severity of weather and with the penetration capability of the band itself. Weight mapping can be a lookup table or a lightweight computational model. For example, by inputting weather state parameters indicating the weather type and intensity level, a weight vector can be output, allowing for real-time and smooth adjustment of weight allocation as the weather changes.

[0108] Furthermore, since the previous steps have ensured that the multi-band point clouds are spatially aligned (like coaxial scanning and model completion processing), points from different wave systems at the same spatial location (or the same target surface) can be directly fused. For example, for each 3D location point in the fused point cloud, its attributes (such as spatial coordinates and reflection intensity) can be obtained by multiplying the attributes of the corresponding point in each wave system at that location by their respective fusion weights and then adding them together (or by performing other weight-based combinations, such as maximum value selection).

[0109] Through a concise and effective weighted fusion operation, the advantages of multi-source information are combined. For example, in foggy conditions, the high-weighted mid-infrared data ensures the accurate recovery of the basic outline of objects; while even with lower weights, near-infrared data can still contribute its detailed features that are not completely obscured by noise. The final output enhanced point cloud has significantly improved point cloud density, target outline integrity, and geometric accuracy compared to point clouds of any single band, providing a reliable environmental perception foundation for downstream applications such as autonomous driving.

[0110] In this embodiment, the provided laser point cloud data processing method firstly utilizes a multispectral sensor combined with a support vector machine model for environmental perception, achieving accurate and quantitative identification of weather type and intensity level. This provides a reliable decision-making basis for subsequent adaptive control, overcoming the shortcomings of traditional threshold methods or insufficient accuracy of single sensors. Secondly, based on weather state parameters, it adaptively and finely configures multiple parameters of the lidar, such as multi-band transmission power, scanning frequency, receiving polarization direction, and detector gain, optimizing the signal transmission and reception link at the hardware level to address specific weather challenges. Furthermore, a conditional judgment mechanism based on data missing rate is introduced, ensuring that the computationally intensive point cloud completion operation is triggered only when data quality is severely degraded. This achieves an optimal balance between processing efficiency and performance assurance, avoiding unnecessary computational consumption. Thirdly, in the completion stage, weather state parameters and original point cloud data are jointly input into a deep learning model, guiding the completion process with physical environment information. This effectively repairs point cloud defects caused by extreme attenuation, significantly improving the rationality and accuracy of data reconstruction. Finally, in the fusion stage, the fusion weights of point clouds in different bands are adaptively allocated according to the severity of the weather, dynamically combining the advantages of different bands to ensure that enhanced point clouds with complete outlines and rich details can be output under any weather conditions. The entire process forms a complete intelligent closed loop of "environmental perception - adaptive parameter configuration - data acquisition - quality assessment - intelligent completion - weighted fusion", which significantly improves the data integrity, target perception reliability and overall system robustness of LiDAR under complex and severe weather conditions.

[0111] Figure 4This is a flowchart illustrating another laser point cloud data processing method provided in this application. Compared to the previous embodiments, this embodiment focuses on a dual-band fusion scheme with relatively fixed hardware configuration but efficient processing flow, which is particularly suitable for application scenarios with specific requirements for system real-time performance and cost. This method can be executed by a laser point cloud data processing device, such as... Figure 4 As shown, the method may include the following steps:

[0112] Step 410: Detect the target object using a dual-band lidar to obtain the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data based on near-infrared and point cloud data based on mid-infrared.

[0113] For example, a dual-band lidar refers to a radar system that integrates two specific wavelength laser transmission and reception channels. This embodiment preferably employs a combination of near-infrared (e.g., 905nm or other wavelengths) and mid-infrared (e.g., 1550nm or other wavelengths) light. The interaction between these two wavelengths and common atmospheric precipitation particles (raindrops, fog droplets) differs significantly: near-infrared light has a shorter wavelength and is easily scattered by small particles, but typically has higher detector sensitivity and resolution; mid-infrared light has a longer wavelength, is less affected by Rayleigh scattering, and has better penetration capabilities through fog, haze, etc.

[0114] For example, referring to the methods for acquiring raw point cloud data of a target object as shown in S1-S4, a dual-band lidar can operate in a time-division or frequency-division multiplexing manner. However, through coaxial optical path design, it ensures that the two laser beams scan and detect the same spatial area almost synchronously, thereby obtaining two sets of point cloud data that are highly aligned in time and space. Specifically, a dual-band lidar can simultaneously emit laser pulses of two wavelengths through its transmitting module, which are then projected onto the area where the target object is located via a scanning module (such as a MEMS mirror). The echo reflected by the target object is captured by the receiving module and received by detectors of the corresponding bands. The three-dimensional coordinates of each detection point are calculated using principles such as Time-of-Flight (ToF), thus forming two independent sets of raw point cloud data from two different physical characteristics and observation perspectives.

[0115] This dual-band parallel acquisition method inherently provides data redundancy and complementarity. Even when severe weather causes serious degradation of data in one band, the data in the other band may still retain valuable information, providing indispensable raw material for subsequent information enhancement and recovery through algorithms.

[0116] Step 420: Perform weighted fusion processing on the original point cloud data to determine the enhanced point cloud data of the target object.

[0117] For example, combining the advantages of dual-band data through intelligent algorithms can generate a final result of superior quality compared to either single-band data. Weighted fusion processing can involve assigning appropriate weights to the two band point cloud data based on their inherent quality or a pre-defined fusion strategy before synthesizing them.

[0118] For example, a preliminary analysis can be performed on two sets of point clouds. For instance, the quality of data in different spatial regions can be evaluated based on indicators such as signal-to-noise ratio, local point density, or intensity consistency. Typically, in areas heavily affected by weather (such as distant areas or foggy regions), mid-infrared point cloud data, due to its stronger penetration, tends to have higher quality (e.g., integrity) and is therefore assigned a higher fusion weight. Conversely, in areas with good weather and close proximity, near-infrared point clouds may have richer details and thus receive a higher weight. Weighting can be based on fixed rules or dynamic calculations based on simple statistical characteristics. After ensuring that the two sets of point clouds have achieved spatial alignment through hardware synchronization or software registration, a weighted calculation is performed for each spatial location (or point cloud feature). For example, for a fusion point on the same target surface, its final three-dimensional coordinates can preferentially use data from the high-weight point cloud to reduce errors, while its reflectance intensity attribute can be obtained by weighted averaging of the intensity values ​​of the two bands, thus comprehensively reflecting the reflectance characteristics of the target under different spectra.

[0119] In addition, data fusion processing methods from other embodiments can be referenced. For example, the original point cloud data can be first supplemented to improve its density. Then, the weights of the near-infrared and mid-infrared data in the point cloud data under different weather conditions can be adjusted (the basic rule can be: increase the weight of near-infrared point clouds (rich in detail) when weather conditions are good; and increase the weight of mid-infrared point clouds (stronger penetration and more stable) under severe weather conditions such as rain and fog. For example, under dense fog conditions, W_NIR=0.3 and W_MIR=0.7 may be set) to combine the details of near-infrared data with the contours of mid-infrared data for weighted fusion processing, thereby generating a complete three-dimensional point cloud of the target.

[0120] In this embodiment, based on the acquisition of dual-band raw data with complementary characteristics, enhanced point cloud data with improved quality is finally generated through adaptive weighted fusion. Compared with the raw point cloud of any single band, it is improved in terms of spatial coverage integrity (especially under adverse weather conditions) and target information richness, and can still effectively improve the perception effect, thereby significantly improving the overall perception performance of the lidar system in complex scenarios (including adverse weather conditions).

[0121] It should be noted that, Figure 4 The laser point cloud data processing method shown can also be combined with other embodiments, and the same or similar concepts or processes will not be described again here.

[0122] Figure 5 This is a schematic diagram of the structure of a laser point cloud data processing device provided in an embodiment of this application. Figure 5 As shown, the device 50 includes: an acquisition unit 501, a configuration unit 502, a detection unit 503, and a processing unit 504.

[0123] Acquisition unit 501 is used to acquire the weather state parameters of the current environment;

[0124] The configuration unit 502 is used to adaptively configure the detection parameters of the lidar according to the weather state parameters; wherein, the detection parameters include transmission parameters and reception parameters, the transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction;

[0125] The detection unit 503 is used to control the lidar to perform detection based on the configured detection parameters and acquire the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data corresponding to multiple bands respectively;

[0126] The processing unit 504 is used to perform data completion and fusion processing based on the original point cloud data and weather state parameters to determine the enhanced point cloud data of the target object.

[0127] In one alternative approach, obtaining unit 501 is specifically used for:

[0128] The environmental spectral signal of the current environment is acquired using a multispectral sensor;

[0129] The environmental spectral signal is input into the support vector machine model to analyze the position of the scattering peak and the ambient light intensity, and the weather state parameters of the current environment are output; the weather state parameters include at least the weather type and intensity level.

[0130] In one alternative embodiment, processing unit 504 is specifically used for:

[0131] The original point cloud data and weather state parameters are input together into a pre-trained point cloud completion model for data completion processing to obtain the completed point cloud data for each band.

[0132] Based on weather condition parameters, the completed point cloud data for each band is fused to determine the enhanced point cloud data of the target object.

[0133] In one alternative embodiment, processing unit 504 is specifically used for:

[0134] Based on weather condition parameters, different fusion weights are assigned to the completed point cloud data of different bands; among them, the stronger the penetration of the band, the higher the weight is assigned to the point cloud data under severe weather conditions.

[0135] Based on the fusion weights, the completed point cloud data corresponding to each band are weighted and fused to obtain the enhanced point cloud data of the target object.

[0136] In an alternative approach, before inputting the raw point cloud data and weather state parameters into a pre-trained point cloud completion model for data completion processing, the processing unit 504 is further configured to:

[0137] Calculate the missing data rate of the original point cloud data; whereby the missing data rate is used to indicate the proportion of invalid echo points.

[0138] If the data missing rate is greater than or equal to the preset threshold, the step of inputting the original point cloud data and weather state parameters into the pre-trained point cloud completion model for data completion processing will be executed.

[0139] In one alternative embodiment, the transmission parameters further include a scanning frequency, and the reception parameters further include a detector gain; the configuration unit 502 is specifically used for:

[0140] The detection parameters of the lidar are adaptively configured based on the weather type and intensity level indicated by the weather state parameters.

[0141] In one alternative embodiment, the detection unit 503 is specifically used for:

[0142] The control module of the lidar emits lasers in multiple bands based on the configured laser emission power allocation parameters.

[0143] The scanning module of the control lidar performs coaxial scanning of multiple laser bands based on the configured scanning frequency;

[0144] The system receives the echo signal reflected by the target object, performs polarization filtering on the echo signal based on the configured polarizer direction, and amplifies and converts the filtered signal based on the configured detector gain to obtain the digital signal corresponding to each band.

[0145] Wavelet denoising is performed on the digital signals corresponding to each band, and soft thresholding is used for suppression to generate point cloud data corresponding to each band in the original point cloud data.

[0146] Figure 6 This is a schematic diagram of another laser point cloud data processing device provided in an embodiment of this application. Figure 6 As shown, the device 60 includes: a first processing unit 601 and a second processing unit 602.

[0147] The first processing unit 601 is used to detect the target object using a dual-band lidar and acquire the original point cloud data of the target object; wherein the original point cloud data includes near-infrared point cloud data and mid-infrared point cloud data.

[0148] The second processing unit 602 is used to perform weighted fusion processing on the original point cloud data to determine the enhanced point cloud data of the target object.

[0149] As can be seen from the above, the laser point cloud data processing device provided in this application optimizes the detection performance under different weather conditions from the source by sensing the weather conditions in real time and adaptively adjusting the multi-band transmission power and receiving polarization direction of the lidar accordingly. Furthermore, it uses weather state parameters to perform intelligent weighted fusion and completion of the multi-band original point cloud, generating a fused point cloud with significantly enhanced quality. This method forms a complete technical closed loop from environmental perception and hardware parameter adaptation to data post-processing, effectively solving the industry problem of severe point cloud quality degradation under adverse weather conditions such as rain, snow, and fog, and significantly improving the reliability and accuracy of all-weather environmental perception systems.

[0150] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The specific embodiments of this application do not limit the specific implementation of the electronic device.

[0151] like Figure 7 As shown, the electronic device may include: a processor 702, a communications interface 704, a memory 706, and a communications bus 708.

[0152] The processor 702, communication interface 704, and memory 706 communicate with each other via communication bus 708. Communication interface 704 is used to communicate with other network elements such as clients or other servers. The processor 702 executes program 710, specifically performing the relevant steps described in the embodiments of the laser point cloud data processing method.

[0153] Specifically, program 710 may include program code, which includes computer-executable instructions.

[0154] The processor 702 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The electronic device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs.

[0155] Memory 706 is used to store program 710. Memory 706 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0156] Specifically, program 710 can be called by processor 702 to cause the electronic device to execute the relevant steps described in the embodiment of the laser point cloud data processing method.

[0157] This application provides a computer-readable storage medium storing at least one executable instruction that, when executed on an electronic device / app, causes the electronic device / app to perform the laser point cloud data processing method in any of the above method embodiments.

[0158] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Furthermore, the embodiments in this application are not directed to any particular programming language.

[0159] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. Similarly, for the purpose of simplification and aiding understanding of one or more aspects of the invention, in the above description of exemplary embodiments of this application, various features of the embodiments are sometimes grouped together in a single embodiment, figure, or description thereof. The claims, which follow the detailed description, are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0160] Those skilled in the art will understand that the modules in the device of the embodiment can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiment can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components, except that at least some of such features and / or processes or units are mutually exclusive.

[0161] It should be noted that the above embodiments are illustrative of this application and not restrictive, and those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A laser point cloud data processing method, characterized in that, The method includes: Obtain the current weather status parameters; The detection parameters of the lidar are adaptively configured according to the weather state parameters; wherein, the detection parameters include transmission parameters and reception parameters, the transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction; The lidar is controlled to perform detection based on the configured detection parameters to acquire the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data corresponding to multiple wavebands respectively; Based on the original point cloud data and the weather state parameters, data completion and fusion processing are performed to determine the enhanced point cloud data of the target object.

2. The method according to claim 1, characterized in that, The acquisition of current environmental weather state parameters includes: The environmental spectral signal of the current environment is acquired using a multispectral sensor; The environmental spectral signal is input into a support vector machine model to analyze the position of the scattering peak and the ambient light intensity, and the weather state parameters of the current environment are output; wherein, the weather state parameters include at least the weather type and intensity level.

3. The method according to claim 1, characterized in that, The step of performing data completion and fusion processing based on the original point cloud data and the weather state parameters to determine the enhanced point cloud data of the target object includes: The original point cloud data and the weather state parameters are input together into a pre-trained point cloud completion model for data completion processing to obtain the completed point cloud data for each band. Based on the weather state parameters, the completed point cloud data corresponding to each band is subjected to data fusion processing to determine the enhanced point cloud data of the target object.

4. The method according to claim 3, characterized in that, The step of performing data fusion processing on the completed point cloud data corresponding to each band based on the weather state parameters to determine the enhanced point cloud data of the target object includes: Based on the weather state parameters, different fusion weights are assigned to the completed point cloud data of different bands; among them, the stronger the penetration of the band, the higher the weight is assigned to the point cloud data under severe weather conditions. Based on the fusion weights, the completed point cloud data corresponding to each band are weighted and fused to obtain the enhanced point cloud data of the target object.

5. The method according to claim 3, characterized in that, Before inputting the original point cloud data and the weather state parameters into a pre-trained point cloud completion model for data completion processing, the method further includes: Calculate the data missing rate of the original point cloud data; wherein, the data missing rate is used to indicate the proportion of invalid echo points; If the data missing rate is greater than or equal to a preset threshold, then the step of inputting the original point cloud data and the weather state parameters into a pre-trained point cloud completion model for data completion processing is performed.

6. The method according to any one of claims 1-5, characterized in that, The transmission parameters also include a scanning frequency, and the reception parameters also include a detector gain; the adaptive configuration of the lidar detection parameters based on the weather state parameters includes: The detection parameters of the lidar are adaptively configured based on the weather type and intensity level indicated by the weather state parameters.

7. The method according to claim 6, characterized in that, The control of the lidar to perform detection based on the configured detection parameters and acquire the raw point cloud data of the target object includes: The laser radar's emission module is controlled to emit lasers in multiple wavelengths based on the configured laser emission power allocation parameters; The scanning module of the control lidar performs coaxial scanning of multiple laser bands based on the configured scanning frequency; The system receives the echo signal reflected by the target object, performs polarization filtering on the echo signal based on the configured polarizer direction, and amplifies and converts the filtered signal based on the configured detector gain to obtain digital signals corresponding to each band. Wavelet denoising is performed on the digital signals corresponding to each band, and soft thresholding is used for suppression to generate point cloud data corresponding to each band in the original point cloud data.

8. A laser point cloud data processing method, characterized in that, The method includes: The target object is detected by a dual-band lidar to obtain the original point cloud data of the target object; wherein the original point cloud data includes point cloud data based on near-infrared and point cloud data based on mid-infrared. The original point cloud data is subjected to weighted fusion processing to determine the enhanced point cloud data of the target object.

9. A laser point cloud data processing device, characterized in that, The device includes: The acquisition unit is used to acquire the weather state parameters of the current environment; A configuration unit is used to adaptively configure the detection parameters of the lidar according to the weather state parameters; wherein the detection parameters include transmission parameters and reception parameters, the transmission parameters include at least multiple laser transmission power allocation parameters for bands with different weather penetration characteristics, and the reception parameters include at least the polarizer direction; The detection unit is used to control the lidar to perform detection based on the configured detection parameters and acquire the original point cloud data of the target object; wherein, the original point cloud data includes point cloud data corresponding to multiple wavebands respectively; The processing unit is used to perform data completion and fusion processing based on the original point cloud data and the weather state parameters to determine the enhanced point cloud data of the target object.

10. An electronic device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation of the laser point cloud data processing method as described in any one of claims 1-8.

Images