Obstacle laser detection method and system for unmanned mine card

Abstract

The invention relates to the technical field of mine truck obstacle detection, and discloses an obstacle laser detection method and system for an unmanned mine truck, and the method comprises the steps: obtaining the signal-to-noise ratio of all detection echo signals, determining the adjustment strategy of the corresponding detection echo signals based on the signal-to-noise ratio, and determining the adjustment strategy of the corresponding detection echo signals; the method comprises the steps of constructing a point cloud topology network based on point cloud data, performing clustering to determine a spatial feature cluster, extracting spectrum features containing spectrum data based on the spatial feature cluster, determining a spatial spectrum feature cluster based on synchronous change of the spectrum features, determining an effective echo spectrum according to a spectrum space and spectrum time of the spatial spectrum feature cluster, and outputting the effective echo spectrum. And comparing the effective echo frequency spectrum with a historical frequency spectrum database to determine whether obstacle information of the effective echo frequency spectrum exists, determining the obstacle information based on an obstacle identification model and the effective echo frequency spectrum, and determining an effective obstacle of the driving route according to the obstacle information. According to the invention, the reliability and the recognition stability of obstacle detection are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of mining truck obstacle detection, and in particular to a laser obstacle detection method and system for unmanned mining trucks. Background Art

[0002] With the acceleration of intelligent mining, unmanned mining trucks have become a vital tool in open-pit mining due to their advantages in improving operational efficiency, reducing labor costs, and minimizing safety accidents. However, complex obstacles such as rockfall, intersecting vehicles, and temporary accumulations in mining environments place higher demands on the precision and accuracy of unmanned mining trucks' obstacle detection.

[0003] Chinese Patent No. CN111537994B discloses a method for detecting obstacles in unmanned mining trucks. The method converts obstacle data acquired by laser radar and millimeter-wave radar into corresponding vehicle coordinate systems. A 0 / 1 binary map of "ground-elevated points" is drawn using grid map height differences combined with neighborhood difference ground detection. A multi-parameter model is used to cluster elevated points. Based on the vehicle's trajectory, the method determines whether the clustered result represents an obstacle that could affect the vehicle's normal operation. The method detects whether the vehicle is within a drivable area. The method matches obstacle data acquired by the millimeter-wave radar with obstacle data acquired by the laser radar, and outputs a final result. The present invention is targeted at the actual application environment of mining dump trucks. It can be seen that using grid maps or simple density clustering makes it difficult to distinguish the spatial characteristics of ground protrusions from actual obstacles. In particular, irregular rock piles and low obstacles can be over- or under-clustered, leading to false detections or missed detections. Furthermore, the method does not identify obstacles through the dynamic changes in the laser radar spectrum, making it difficult to accurately distinguish between obstacles and the environment, resulting in limitations in obstacle matching. Therefore, it is necessary to design an obstacle laser detection method and system for unmanned mining trucks to solve the problems existing in the current technology. Summary of the Invention

[0004] In view of this, the present invention proposes a laser obstacle detection method and system for unmanned mining trucks, aiming to solve the problems existing in the current technology.

[0005] In one aspect, the present invention provides a laser obstacle detection method for an unmanned mining truck, comprising: transmitting a radar detection signal to the driving route based on a laser radar and determining all detection echo signals, obtaining a signal-to-noise ratio of all detection echo signals, determining an adjustment strategy for corresponding detection echo signals based on the signal-to-noise ratio, and determining echo time, spectrum data, and point cloud data based on all the adjusted detection echo signals; determining a constant driving strategy or a suspected obstacle strategy for the driving route based on the echo time; if the suspected obstacle strategy is determined, constructing a point cloud topology network based on the point cloud data and performing clustering to determine a spatial feature cluster; extracting spectral features including spectral data based on the spatial feature cluster; and determining a spatial spectral feature cluster based on synchronous changes in the spectral features; determining a valid echo spectrum based on the spectrum space and spectrum time of the spatial spectrum feature cluster, comparing the valid echo spectrum with a historical spectrum database to determine whether obstacle information in the valid echo spectrum exists, and when obstacle information in the valid echo spectrum does not exist, obtaining an obstacle sample set and constructing an obstacle recognition model; Obstacle information is determined based on the obstacle recognition model and the effective echo spectrum, and effective obstacles on the driving route are determined according to the obstacle information.

[0006] Furthermore, when determining an adjustment strategy for a corresponding detection echo signal based on the signal-to-noise ratio, the method includes: When the signal-to-noise ratio is greater than or equal to the signal-to-noise ratio threshold, phase compensation is performed based on the Doppler frequency; When the signal-to-noise ratio is less than the signal-to-noise ratio threshold, the complex signal of the short-time window of each detection echo signal is extracted, the conjugate product with the previous detection echo signal is determined, and the phase difference is extracted to determine the phase difference sequence, phase unwrapping is performed based on the phase difference sequence and accumulated to determine the accumulated phase, and phase compensation is performed based on the accumulated phase.

[0007] Furthermore, when determining the constant driving strategy or the suspected obstacle strategy of the driving route according to the echo time, it includes: Obtaining a standard echo time interval, and comparing the standard echo time interval with the echo time; When there is an echo time greater than the left boundary of the standard echo time interval and less than the right boundary of the standard echo time interval, the driving route is determined to be a constant driving strategy; otherwise, the driving route is determined to be a suspected obstacle strategy.

[0008] Furthermore, when constructing a point cloud topology network based on the point cloud data and performing clustering to determine spatial feature clusters, the method includes: removing isolated noise points in the point cloud data based on a bilateral filtering algorithm to determine valid point cloud data, and determining a neighborhood radius with each point of the valid point cloud data as the center, wherein the neighborhood radius is determined based on the point cloud density; For each point, search all other points within its neighborhood radius to determine whether they are bidirectionally adjacent. If so, establish undirected edges between the point and the bidirectionally adjacent points, and determine the edge attributes of each undirected edge. Otherwise, do not establish undirected edges. Constructing the point cloud topology network from all points and their corresponding undirected edges, determining the local connection density of each point, determining an initial clustering region based on the local connection density, traversing all undirected edges in each initial clustering region to determine a reflection intensity gradient, and determining a plurality of reflection sub-regions with the same reflection intensity based on a relationship between the reflection intensity gradient and the average gradient; The normal vector angles of all points in each reflection sub-region are determined, and the reflection sub-region is split based on the normal vector angles, and the split results are determined as the spatial feature clusters.

[0009] Furthermore, when extracting spectrum features including spectrum data based on the spatial feature cluster and determining the spatial spectrum feature cluster based on synchronous changes of the spectrum features, the method includes: Extracting segments of spectrum data corresponding to all points in the spatial feature cluster, comparing the acquisition time of each segment with the acquisition time of the corresponding point, eliminating segments with mismatched times, and determining all main frequency offsets, high-frequency energy, and main frequency change rates based on the remaining segments; The synchronization in the spatial coordinate diagram, spatial distribution diagram and spatial position diagram is determined respectively according to all main frequency offsets, high-frequency energy and main frequency change rate, the fragments with the highest synchronization are extracted, and the spatial feature cluster is split according to whether the corresponding points form a continuous area in space, and the split result is determined as the spatial spectrum feature cluster.

[0010] Furthermore, when determining the effective echo spectrum according to the spectrum space and spectrum time of the spatial spectrum feature cluster, the method includes: Based on the point cloud topology network of the spatial spectrum feature cluster, the connection relationship of all points in each connected component is determined, segments of the spectrum data synchronously collected for all points in the spatial spectrum feature cluster are extracted, and the spectrum characteristics of each segment in the connected component in which it is located are verified to satisfy the transitive association; All segments that satisfy the transitive association are extracted to construct a spectrum segment set, the reflection intensity and high-frequency energy proportion of each point corresponding to the connected components of the spectrum segment set are extracted according to the corresponding connection relationship, and the correlation direction of the reflection intensity and high-frequency energy proportion of each point is counted, the main frequency change trend is determined based on the correlation direction, and the segments that are consistent with the main frequency change trend and the point cloud movement direction are retained to construct a valid spectrum set, and the effective echo spectrum is determined based on the valid spectrum set.

[0011] Furthermore, when comparing the effective echo spectrum with the historical spectrum database to determine whether obstacle information of the effective echo spectrum exists, the method includes: The historical spectrum database includes a plurality of historical valid echo spectrums and a plurality of historical obstacle information, and each historical valid echo spectrum corresponds to a piece of historical obstacle information; When a historical valid echo spectrum identical to the valid echo spectrum exists in the historical spectrum database, determining the historical obstacle information corresponding to the historical valid echo spectrum as the obstacle information, and determining that the obstacle information of the valid echo spectrum exists; When there is no historical valid echo spectrum identical to the valid echo spectrum in the historical spectrum database, it is determined that there is no obstacle information of the valid echo spectrum.

[0012] Furthermore, when there is no obstacle information in the valid echo spectrum, obtaining an obstacle sample set and building an obstacle recognition model includes: Dividing the obstacle sample set into a training set and a test set, and preselecting a convolutional neural network model; Using the training set to train the convolutional neural network model, and substituting the test set into the trained convolutional neural network model to determine the prediction accuracy; When the prediction accuracy is greater than or equal to the prediction accuracy threshold, the trained convolutional neural network model is determined as the obstacle recognition model.

[0013] Furthermore, when determining obstacle information based on the obstacle recognition model and the effective echo spectrum, and determining effective obstacles on the driving route according to the obstacle information, the method includes: Substituting the effective echo spectrum into the obstacle recognition model to determine obstacle information; The effective obstacles on the driving route are determined according to the relationship between the obstacle information and the size of the unmanned mining truck.

[0014] Compared with the prior art, the beneficial effects of the present invention are: by adjusting the detection echo signal through the signal-to-noise ratio, the interference of noise is effectively filtered out, laying a data foundation for subsequent analysis and processing; combining point cloud topology network clustering and spectral feature analysis, it accurately distinguishes ground protrusions from real obstacles, solves the risk of over-clustering or under-clustering of irregular rock piles and low obstacles, thereby reducing false detection and missed detection rates, and ensuring the stability of effective identification of complex obstacles in mining areas; introducing spectral data for dynamic analysis, and using the synchronous changes of spectral features to construct spatial spectral feature clusters; by comparing with the historical spectral database, it accurately captures the differences between obstacles and the environment, and copes with dynamically changing obstacle scenes such as rock rolling and temporary accumulations, as well as complex working conditions of multiple vehicles crossing; determines obstacle information through effective echo spectrum, and accurately identifies effective obstacles, ensuring the driving safety of unmanned mining trucks in complex environments, while reducing unnecessary deceleration or detours due to misjudgment.

[0015] On the other hand, the present application also provides an obstacle laser detection system for an unmanned mining truck, which is used to apply the above-mentioned obstacle laser detection method for an unmanned mining truck, including: an acquisition and adjustment module configured to transmit a radar detection signal to the driving route based on the laser radar and determine all detection echo signals, obtain a signal-to-noise ratio of all detection echo signals, determine an adjustment strategy for the corresponding detection echo signals based on the signal-to-noise ratio, and determine echo time, spectrum data, and point cloud data based on all the adjusted detection echo signals; a spatial spectrum module configured to determine a constant driving strategy or a suspected obstacle strategy for the driving route based on the echo time; when the suspected obstacle strategy is determined, construct a point cloud topology network based on the point cloud data and perform clustering to determine a spatial feature cluster; extract spectral features containing spectral data based on the spatial feature cluster; and determine a spatial spectrum feature cluster based on synchronous changes in the spectral features; an obstacle analysis module configured to determine a valid echo spectrum based on the spectrum space and spectrum time of the spatial spectrum feature cluster, compare the valid echo spectrum with a historical spectrum database to determine whether obstacle information exists in the valid echo spectrum, and when obstacle information does not exist in the valid echo spectrum, obtain an obstacle sample set and construct an obstacle recognition model; The obstacle determination module is configured to determine obstacle information based on the obstacle recognition model and the effective echo spectrum, and determine effective obstacles on the driving route according to the obstacle information.

[0016] It is understandable that the above-mentioned obstacle laser detection method and system for unmanned mining trucks have the same beneficial effects and will not be described in detail here. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings: Figure 1 A flowchart of a laser obstacle detection method for an unmanned mining truck provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of an obstacle laser detection system for an unmanned mining truck provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, unless there is a conflict, the embodiments of the present disclosure and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0019] See Figure 1 As shown, in some embodiments of the present application, a laser obstacle detection method for an unmanned mining truck includes: S100: Based on the laser radar, a radar detection signal is transmitted to the driving route and all detection echo signals are determined, the signal-to-noise ratio of all detection echo signals is obtained, an adjustment strategy for the corresponding detection echo signal is determined based on the signal-to-noise ratio, and the echo time, spectrum data, and point cloud data are determined based on all the adjusted detection echo signals.

[0020] S200: Determine a constant driving strategy or a suspected obstacle strategy for the driving route based on the echo time. When it is determined to be a suspected obstacle strategy, construct a point cloud topology network based on the point cloud data and perform clustering to determine a spatial feature cluster. Extract spectral features containing spectral data based on the spatial feature cluster, and determine the spatial spectral feature cluster based on the synchronous changes of the spectral features.

[0021] S300: Determine the effective echo spectrum based on the spectrum space and spectrum time of the spatial spectrum feature cluster, compare the effective echo spectrum with the historical spectrum database to determine whether there is obstacle information in the effective echo spectrum. If there is no obstacle information in the effective echo spectrum, obtain an obstacle sample set and build an obstacle recognition model.

[0022] S400: Determine obstacle information based on the obstacle recognition model and the effective echo spectrum, and determine effective obstacles on the driving route according to the obstacle information.

[0023] Specifically, FMCW and other lidars are deployed on unmanned mining trucks. After the lidar transmits a radar detection signal to the driving route of the unmanned mining truck, it will receive all the detection echo signals reflected from the front. At this time, the signal quality is evaluated by determining the signal-to-noise ratio of the detection echo signal. The signal-to-noise ratio reflects the intensity ratio of the effective signal to the noise. For example, in the dusty environment of the mining area, the echo of distant rocks may cause the signal-to-noise ratio to be lower than 10dB due to scattering, while the signal-to-noise ratio of close-range vehicles can reach more than 30dB. Based on the difference in signal-to-noise ratio, the adjustment strategy is determined to suppress the noise of the detection echo signal with low signal-to-noise ratio, and fine-tune the detection echo signal with high signal-to-noise ratio to avoid signal saturation. Finally, the adjusted detection echo signal is used to obtain the signal-to-noise ratio. Three key data types are extracted from the wave signal: echo time, spectrum data, and point cloud data. The echo time reflects the distance between the unmanned mining truck and the detection target. For example, for every 1μs increase in echo time, the corresponding distance increases by 0.15m. The spectrum data reflects the material and other properties of the detection target. For example, the main frequency of the spectrum of metal equipment is concentrated in 5-10kHz, while that of rock is in 10-20kHz. The point cloud data reflects the spatial form of the detection target. Each point contains three-dimensional coordinate information. By adjusting the detection echo signal, the risk of signal quality degradation caused by dust and strong light in the mining area is avoided, thereby ensuring the reliability of subsequent analysis data and preventing noise or distorted signals from interfering with the detection results.The driving strategy is judged based on the echo time. If the echo time of multiple consecutive frames is within the allowable braking time range, it indicates that there may be an obstacle in front of the unmanned mining truck. For example, 50m in front of the unmanned mining truck is flat ground, and the echo time of the flat ground is stable at 333μs, corresponding to a detection distance of 50m in front. If the echo time of a certain frame suddenly drops to 100μs, the distance corresponds to 15m, and the echo time is relatively reduced by 233μs, indicating that the distance of the unmanned mining truck has dropped sharply by 35m. It may be reflected by obstacles such as rock piles 15m in front. When it is determined that there may be obstacles, the suspected obstacle strategy is executed. At this time, the point cloud data is combined to form a topological network reflecting the spatial connection relationship, and then the spatial feature cluster is obtained by clustering. Compared with traditional grid maps or density clustering, spatial feature clusters have topological networks. It can more accurately distinguish between continuous targets (such as rock piles) and discrete points (such as ground protrusions). For example, the point cloud of a rock pile is clustered into a continuous spatial feature cluster through a topological network, while the isolated points of the ground protrusion are clustered separately to avoid under-clustering or over-clustering. The spectral characteristics of the spatial feature cluster are then extracted, and the spatial spectral feature cluster is determined by analyzing the synchronous changes in the spectrum. Targets with the same spatial form but different materials are further distinguished. The spatiotemporal binding relationship between point cloud data and spectral data effectively solves the problem of distinguishing between objects with similar spatial features but different materials. For example, the spatial characteristics of a rock pile and a soil pile may be similar, but the high-frequency energy of the rock accounts for a higher proportion than that of the soil, and the synchronous change trend of the main frequency is more stable. Spectral analysis can accurately distinguish them, avoiding misjudging the soil pile as an obstacle, thereby improving the reliability and stability of obstacle detection.

[0024] Specifically, the effective echo spectrum is screened out according to the spectrum space and spectrum time of the spatial spectrum feature cluster. The effective echo spectrum is a spectrum feature set of the detection echo signal detected by the lidar that can truly reflect the physical characteristics of the obstacle (such as material, surface state, movement trend) and has been screened and verified. It is the core spectrum information retained after removing noise, interference and abnormal signals from the original spectrum data. The effective echo spectrum is compared with the historical spectrum database. If the match is successful, the obstacle information (category, size, etc.) is directly determined based on the historical spectrum database. If the match cannot be made, it indicates that there is no obstacle information with a valid echo spectrum in the historical spectrum database. Then the obstacle sample set is obtained. The obstacle sample set contains at least 1,000 groups of sample data, covering effective echoes under different dust conditions. The spectrum and corresponding obstacle information are complex due to the variety of obstacle types in mining environments (rocks, metal equipment, soil piles, pipes, fallen rocks, etc.), and new targets that are not included in the sample set (such as temporarily stacked new building materials) may appear. Although the obstacle sample set contains at least 1,000 groups of sample data to cover common obstacles, there are still morphological differences among obstacles of the same type (such as broken rocks and intact rocks), making it difficult for the obstacle sample set to cover all morphologies. Therefore, by learning the common characteristics of obstacles of the same type through the model, generalized recognition of new targets or targets with changed morphologies can be performed. For example, even if there is no valid echo spectrum of a certain type of pipeline in the sample set, the model can still determine it as a metal obstacle based on its main frequency, high-frequency energy ratio and other characteristics, thereby effectively determining the obstacle information.

[0025] It is understandable that through topological network clustering and spectrum synchronization analysis, the risk of false detection and missed detection of irregular and low obstacles is avoided. The dynamic changes in the spectrum of the lidar are used to achieve dual recognition of spatial features and material features, ensuring the reliability of distinguishing between rocks and soil, metals and non-metals. Therefore, based on the obstacle information, it is judged whether it is a passable obstacle or an avoidable obstacle. For example, low stones can be driven directly past, while raised stone piles on the ground are avoided. In addition, obstacle information is determined based on the obstacle recognition model and the effective echo spectrum, realizing adaptive recognition of unknown obstacles, meeting the real-time obstacle detection requirements in the complex environment of the mining area, and providing reliable protection for the safe driving of unmanned mining trucks.

[0026] In some embodiments of the present application, when determining the adjustment strategy of the corresponding detection echo signal based on the signal-to-noise ratio, it includes: when the signal-to-noise ratio is greater than or equal to the signal-to-noise ratio threshold, performing phase compensation based on the Doppler frequency; when the signal-to-noise ratio is less than the signal-to-noise ratio threshold, extracting the complex signal of the short time window of each detection echo signal, determining the conjugate product with the previous detection echo signal, and extracting the phase difference to determine the phase difference sequence, performing phase expansion based on the phase difference sequence and accumulating to determine the accumulated phase, and performing phase compensation based on the accumulated phase.

[0027] Specifically, the phase of the detection echo signal will produce linear offsets due to obstacles and the movement of unmanned mining trucks, and will also cause random disturbances due to dust and clutter in the mining area. Phase distortion will directly affect the extraction accuracy of spectral features such as the main frequency and the proportion of high-frequency energy. The signal-to-noise ratio threshold is set to 15dB. The specific signal-to-noise ratio threshold can be dynamically set according to the scale of the mining area and the degree of environmental pollution. In the mining environment, when the signal-to-noise ratio is greater than or equal to 15dB, the effective component of the detection echo signal accounts for approximately 97%, while noise and other interference (dust scattering, clutter) accounts for approximately 3%. The Doppler frequency (a key parameter reflecting target movement) can be stably extracted. When the signal-to-noise ratio is less than 15dB, the noise ratio will exceed 3%, and the Doppler frequency extraction error will increase, which cannot be directly used for compensation. At this time, the short-time window length is determined according to the sampling period of the lidar. , and intercept the complex signal of the short time window of the detection echo signal of each frame (retaining the phase information), so as to calculate the conjugate product with the previous frame signal to offset the common noise component, and then obtain the phase difference sequence, and then through phase expansion and accumulation, the phase winding problem is solved, so that the phase change trend is restored in the noise, and the phase jitter caused by random noise is offset. The phase distortion caused by target motion and noise interference is eliminated through differentiated phase compensation, ensuring the accuracy of the spectrum data. In addition, the differentiated phase compensation takes into account the compensation accuracy under different signal-to-noise ratios. It is used to improve the compensation efficiency at high signal-to-noise ratio and enhance the anti-noise ability at low signal-to-noise ratio, ensuring that the spectrum data after phase compensation can truly reflect the obstacle, providing a reliable foundation for the subsequent spatial spectrum feature cluster construction and obstacle recognition, and improving the robustness of laser detection in complex environments in mining areas.

[0028] In some embodiments of the present application, when determining the constant driving strategy or suspected obstacle strategy of a driving route based on the echo time, it includes: obtaining the standard echo time interval, and comparing the standard echo time interval with the echo time. When there is an echo time that is greater than the left boundary of the standard echo time interval and less than the right boundary of the standard echo time interval, the driving route is determined to be a constant driving strategy; otherwise, the driving route is determined to be a suspected obstacle strategy.

[0029] Specifically, the standard echo time interval is determined based on the braking distance of an unmanned mining truck on a flat, obstacle-free road. If there are no obstacles ahead, the time between the lidar's detection signal and the vehicle's return is primarily determined by ground reflection. The resulting echo time is the minimum or typical echo delay during normal driving, reflecting the detection distance when the vehicle is unobstructed. In actual operation, unmanned mining trucks must maintain a minimum safe stopping distance to avoid obstacles, which is determined by vehicle speed, braking performance, and road conditions. If the distance corresponding to the echo time is less than this braking distance, even emergency braking may not be enough to avoid the obstacle. Therefore, the braking distance is used as a lower limit, forming a standard echo time interval. The left boundary is determined by the echo time converted from the braking distance, and the right boundary is determined by the detection delay on a flat, obstacle-free road. Although mining roads are naturally undulating, the echo time changes continuously and slightly under normal circumstances. However, the presence of obstacles causes a step-like change in the echo time, and the interval boundaries capture this sudden change. By determining the standard echo time interval, it not only avoids overreaction to small road surface fluctuations, but also accurately identifies anomalies caused by obstacles. When the driving route is determined to be a constant driving strategy, it can continue driving. When the driving route is determined to be a suspected obstacle strategy, it indicates that there may be suspected obstacles on the road, and further analysis of point cloud data is required, ensuring the reliability and stability of obstacle detection.

[0030] In some embodiments of the present application, when constructing a point cloud topology network based on point cloud data and clustering to determine spatial feature clusters, it includes: removing isolated noise points in the point cloud data based on a bilateral filtering algorithm to determine valid point cloud data, and determining a neighborhood radius with each point of the valid point cloud data as the center, the neighborhood radius is determined based on the point cloud density, searching all other points within its neighborhood radius for each point to determine whether they are bidirectionally adjacent, if so, establishing an undirected edge between the point and the bidirectionally adjacent points, and determining the edge attributes of each undirected edge, otherwise, no undirected edge is established, all points and their corresponding undirected edges are constructed into a point cloud topology network, and the local connection density of each point is determined, the initial clustering area is determined based on the local connection density, and all undirected edges are traversed in each initial clustering area to determine the reflection intensity gradient, and several reflection sub-areas with the same reflection intensity are determined based on the relationship between the reflection intensity gradient and the average gradient, the normal vector angle of all points in each reflection sub-area is determined, and the reflection sub-area is split based on the normal vector angle, and the split result is determined as a spatial feature cluster.

[0031] Specifically, in a mining environment, the point cloud data of the lidar may contain isolated noise points caused by dust scattering and sensor noise. These isolated noise points are not only isolated in space, but also have characteristics that are irrelevant to the surrounding points. The range kernel of the bilateral filter can capture this difference and remove it, ensuring the reliability and stability of the valid point cloud data. The neighborhood radius is determined with each point of the valid point cloud data as the center. The neighborhood radius is the radius of an area centered on a certain point, which is used to search for neighboring points. The neighborhood radius is dynamically adjusted with the point cloud density. If the point cloud density is higher, the neighborhood radius is smaller to avoid including irrelevant points, and if the point cloud density is lower, the neighborhood radius is larger to ensure that enough neighboring points are captured. The point cloud density refers to the number of points per unit volume. It is an important attribute feature of the point cloud, reflecting the spatial distribution and density of the point cloud. Generally speaking, for areas with higher point cloud density (100 points per cubic meter), the field radius The radius of the domain is 0.08m, and there are about 8 points in the domain radius. For areas with low point cloud density (50 points per cubic meter), the domain radius is 0.11m, and there are about 9 points in the neighborhood radius. The dynamic domain radius can be used to adaptively search for neighboring points to ensure the accuracy of the search. For each point, all other points in its neighborhood radius are searched to determine whether they are two-way neighbors. If point A is in the neighborhood of point B, and point B is also in the neighborhood of point A, it means that the spatial correlation between the two points is strong and they are not accidental neighbors. In this case, they are judged to be two-way neighbors and an undirected edge is established between the two points. The edge attributes include reflection intensity and normal vector angle, etc., which lay the data foundation for subsequent analysis. One-way proximity may be accidental (such as a point happens to fall on the edge of the point cloud of another point), while two-way proximity can ensure the stability of the connection. The point cloud topology network intuitively reflects the spatial correlation through the point-edge relationship, providing a structured basis for clustering.

[0032] It can be understood that the local connection density is the number of undirected edges of a point divided by the neighborhood volume, which reflects the degree of dense connection of the point in space. The higher the value, the stronger the correlation between it and the surrounding points, and the more likely it is to belong to the same obstacle entity. In the open-pit mine environment, the average local connection density of the rock pile is 12 per cubic meter, and the minimum is 8 per cubic meter. The average local connection density of isolated rocks is 6 per cubic meter, and the minimum is 4 per cubic meter (because the point cloud is sparse, but still higher than the 3 per cubic meter on the ground). The average local connection density of metal vehicles (trucks, excavators) is 14 per cubic meter because the metal surface is flat. Through experimental data and simulation verification of the mine, the local connection density of different obstacle types can be determined, and the area where the point is located can be divided according to the value of its local connection density, thereby forming an initial clustering area. The initial clustering area reflects that it may belong to the same obstacle entity. All undirected edges in the initial clustering area are traversed to confirm The reflection intensity gradient is determined. The reflection intensity gradient is the reflection intensity difference of an undirected edge divided by the distance between two points. It reflects the rate of change of reflection intensity along the undirected edge. In the initial clustering area, regions with a reflection intensity gradient less than or equal to twice the regional average gradient are divided, such as the interface between metal and rock. This results in reflection sub-regions with the same reflection intensity, ensuring material consistency within the reflection sub-region. Furthermore, within each reflection sub-region, the reflection sub-region is further split based on the normal vector angle of the point. The normal vector angle reflects the surface flatness. A larger angle indicates an uneven surface, such as rock, while a smaller angle indicates a flat surface, such as a metal plate. If the normal vector angle between two points is greater than 30°, the region is further split and the result is determined as a spatial feature cluster. Otherwise, it is directly determined as a spatial feature cluster. Spatial feature clusters can reflect the spatial position, reflection characteristics, and surface morphology of obstacles, laying the foundation for subsequent analysis combined with spectral data and ensuring the reliability of obstacle detection.

[0033] In some embodiments of the present application, when extracting spectral features containing spectral data based on spatial feature clusters and determining spatial spectral feature clusters based on synchronous changes of spectral features, it includes: extracting segments of spectral data corresponding to all points in the spatial feature cluster, and comparing the acquisition time of each segment with the acquisition time of the corresponding point, eliminating segments with mismatched time and determining all main frequency offsets, high-frequency energy and main frequency change rates based on the remaining segments, determining the synchronization in the spatial coordinate diagram, spatial distribution diagram and spatial position diagram based on all main frequency offsets, high-frequency energy and main frequency change rates, extracting the segments with the highest synchronization, and splitting the spatial feature cluster according to whether the corresponding points form a continuous area in space, and determining the splitting results as spatial spectral feature clusters.

[0034] Specifically, spectrum data contains the frequency characteristics of the radar echo signal (such as the main frequency). Because there may be a slight time difference between the acquisition of the lidar point cloud data and the spectrum data, by comparing the acquisition time of the segment with the acquisition time of the corresponding point, misaligned data can be eliminated (for example, if a spectrum segment actually corresponds to a point one second ago, it is irrelevant to the current point cloud), ensuring the temporal and spatial consistency of subsequent analysis. The main frequency offset refers to the difference between the main peak frequency extracted from the spectrum data at a certain moment and a determined reference frequency. The reference frequency is determined by real-time extraction of the stable component of the detection echo signal after adjusting the signal-to-noise ratio. The offset reflects the relative radial motion of the obstacle with respect to the sensor or the dynamic changes of the obstacle surface / structure. The main frequency offset of a stationary or slowly changing obstacle is close to zero or fluctuates slightly, while the offset is larger for obstacles with significant radial velocity or structural changes. High-frequency energy refers to the sum of energy in a pre-set high-frequency band within a certain segment. The high-frequency band is usually set to above 1kHz and can be adjusted according to the area and complexity of the mining area. The main frequency change rate reflects the speed of change of the main frequency over time. The change rate of a uniformly moving object is stable between ±2Hz / s, while the main frequency change rate of a variable-speed object fluctuates greatly. The main frequency offsets of all spectra within the spatial feature cluster are marked on the spatial coordinate diagram of the corresponding point cloud (with the x-axis as the horizontal coordinate of the point cloud and the y-axis as the main frequency offset), forming a scattered distribution of offsets and spatial positions. If the scattered points show a continuous unidirectional trend (such as gradually rising or falling from left to right without abrupt jumps), it is judged to be highly synchronized. If the scattered points are randomly distributed (irregular up and down fluctuations), the synchronization is low. The high-frequency energy is marked on the spatial distribution map of the corresponding point cloud (the depth of color represents the size of the high-frequency energy, usually red represents higher high-frequency energy and blue represents lower high-frequency energy). If the color presents a continuous gradient area (such as from red to blue from the center of the cluster to the edge, without sudden alternation of red and blue), the synchronization is high. If the color is randomly distributed (red and blue are mixed in the same small area), the synchronization is low.The main frequency change rate and the spatial position of the corresponding point cloud are plotted as trend lines (each point corresponds to a main frequency change rate and spatial position, connected by a broken line). If the trend line shows an orderly change pattern (for example, as the spatial position coordinate increases, the main frequency change rate continues to increase or remains stable), the synchronization is high. If the trend line fluctuates violently (suddenly rising and falling, with irregular changes), the synchronization is low. For the same spatial feature cluster, three types of highly synchronized segments are selected to observe whether their corresponding point clouds form a continuous region in space (for example, the point cloud coordinates are not broken, forming a complete block or strip region). If a continuous region can be formed, the spatial feature cluster is retained and recorded as a spatial spectrum feature cluster. Otherwise, the spatial feature cluster is split and the resulting spatial feature clusters are identified as spatial spectrum feature clusters. Based on the spectrum data, the misidentification of discrete entities as entities of the same obstacle is further avoided. Through spatiotemporal matching, multi-dimensional feature synchronization, and spatial continuity verification, a comprehensive clustering method is achieved from morphological clustering to a combination of morphology and spectrum, improving the precision and accuracy of obstacle recognition in complex mining scenarios.

[0035] In some embodiments of the present application, when determining the effective echo spectrum based on the spectral space and spectral time of the spatial spectrum feature cluster, it includes: determining the connection relationship of all points in each connected component based on the point cloud topology network of the spatial spectrum feature cluster, extracting fragments of the spectrum data synchronously collected for all points in the spatial spectrum feature cluster, and verifying whether the spectral characteristics of each fragment in its connected component satisfy the transitive association, extracting all fragments that satisfy the transitive association to construct a spectrum fragment set, extracting the reflection intensity and high-frequency energy proportion of each point according to the corresponding connection relationship of the connected components corresponding to the spectrum fragment set, and counting the correlation direction of the reflection intensity and high-frequency energy proportion of each point, determining the main frequency change trend based on the correlation direction, and retaining the fragments consistent with the main frequency change trend and the point cloud movement direction to construct a valid spectrum set, and determining the effective echo spectrum based on the valid spectrum set.

[0036] Specifically, in a point cloud topology network, a connected component is the maximum contiguous set of points connected by undirected edges (bidirectional proximity). A maximum contiguous point set is an independent region of interconnected points in the point cloud. Each connected component corresponds to a spatially continuous entity, such as a complete rock or the body of a mine cart. For example, in a point cloud topology network, if points ABC are connected by an undirected edge, and points DE are connected by an undirected edge, and A and D are unconnected, then {A, B, C} and {D, E} are two independent connected components, corresponding to two spatially separated entities. Transitive associations reflect the indirect correlation of spectral features within connected components. For example, if the spectral features of points A and B are similar (e.g., the difference in the main frequency offset is less than 5 Hz), and the spectral features of points B and C are similar, then the spectral features of points A and C must also be similar (transitive association through B). This transitivity reflects the physical law that the spectral characteristics of the same entity are continuously distributed in space (the spectral characteristics of different parts of the same object will not suddenly jump). For the spectral fragment within a connected component, the verification process is as follows: select any point P in the connected component as a reference, record its spectral characteristics, and traverse all directly adjacent points of the checkpoint P (points connected by undirected edges). If the difference between the spectral characteristics of the adjacent point Q and the spectral characteristics of point P is within the threshold range (such as the difference in the main frequency offset is less than 5Hz), then the marked point Q is directly associated with point P. If all points in the connected component satisfy the requirement that any two points maintain consistency through transitive association, then the spectral characteristics of the fragment satisfy the transitive association. If the transitive difference between a point and the reference point exceeds the threshold range, it is judged as not satisfied. By verifying the transitive association, false fragments with similar local spectra but global contradictions can be eliminated (for example, the accidental spectrum of dust is similar to that of rock, but cannot meet the global transitivity), further ensuring that the fragments come from the same entity. The high-frequency energy proportion is expressed as the ratio of the energy in the high-frequency band (above 1kHz) in the spectrum to the total energy, reflecting the surface roughness of the obstacle (rock is rougher than metal, and rough surfaces scatter more high-frequency signals). According to the undirected edge connection relationship of the connected component, the reflection intensity and high-frequency energy proportion of each point are extracted in turn to form a correspondence between points, reflection intensity, and high-frequency energy, ensuring that the data comes from continuous points of the same connected component.

[0037] Specifically, the correlation direction refers to the relationship between the changing trends of the two. There are two typical cases: positive correlation: when the reflection intensity increases, the proportion of high-frequency energy also increases (such as the rough metal surface, the reflection intensity is high and the high-frequency scattering is more), negative correlation: when the reflection intensity increases, the proportion of high-frequency energy decreases (such as the smooth metal surface, the reflection intensity is high but the high-frequency scattering is less). For the points in the connected component, the correlation coefficient between the reflection intensity and the high-frequency energy proportion is determined based on the Pearson correlation algorithm. When the correlation coefficient is greater than 0.6, it is judged to be a positive correlation direction, otherwise it is judged to be a negative correlation direction. The main frequency change trend is the overall change direction of the main frequency in the connected component over time (increase / decrease / stable). Deduced from the correlation direction, if it is positively correlated, the main frequency of the rough surface is more significantly affected by movement. If the mine card is close to an obstacle, the main frequency offset will increase with time. If it is negatively correlated, the main frequency of the smooth surface is relatively stable, but during movement, if the mine card is away from the obstacle, the main frequency offset will decrease with time. The truck's direction of motion is determined by calculating the difference between point cloud frames (e.g., the coordinate changes between two consecutive point cloud frames). If the main frequency change trend matches the motion direction (e.g., when the truck moves forward, the main frequency offset increases over time), then the spectrum fragment is from a truly moving entity. If it does not match (e.g., when the truck moves forward, but the main frequency offset decreases), it may be noise or a false signal. This constructs a valid spectrum set, which retains all fragments with consistent trends. Its comprehensive characteristics (the mean of the spectral characteristics) are the effective echo spectrum, which can truly reflect the physical properties and motion state of the obstacle.

[0038] It can be understood that by eliminating noise fragments with consistent local spectra but globally contradictory, it is ensured that the corresponding spectra come from the same obstacle entity, so that the effective echo spectrum can accurately reflect the material, roughness and motion state of the obstacle, reducing the risk of obstacle identification errors in complex mining environments.

[0039] In some embodiments of the present application, when comparing an effective echo spectrum with a historical spectrum database to determine whether obstacle information of the effective echo spectrum exists, it includes: the historical spectrum database includes several historical effective echo spectra and several historical obstacle information, and each historical effective echo spectrum corresponds to a historical obstacle information; when a historical effective echo spectrum identical to the effective echo spectrum exists in the historical spectrum database, the historical obstacle information corresponding to the historical effective echo spectrum is determined as the obstacle information, and it is determined that the obstacle information of the effective echo spectrum exists; when a historical effective echo spectrum identical to the effective echo spectrum does not exist in the historical spectrum database, it is determined that the obstacle information of the effective echo spectrum does not exist.

[0040] Specifically, the historical spectrum database is an archive of spectral characteristics of common obstacles in mining areas (such as rocks, mine carts, metal equipment, earth piles, etc.). Each historical valid echo spectrum corresponds to unique historical obstacle information (such as type, size, material, etc.). The valid echo spectrum contains the core characteristics of the current obstacle (main frequency offset, high-frequency energy ratio, main frequency change trend, etc.). These characteristics are unique in that one object has one spectrum. The spectral characteristics of different obstacles (such as rocks and mine carts) are different, and the spectral characteristics of the same type of obstacles are highly similar. Therefore, if there is a record identical to the current valid echo spectrum in the historical spectrum database, it means that the unmanned mining truck has encountered a known obstacle and the historical information can be directly reused. If it does not exist, it is a new and unknown obstacle and needs to be processed by subsequent models. By comparing the historical spectrum database, there is no need to perform complex model reconstruction for each obstacle, thereby adapting to the real-time requirements of unmanned mining trucks in the mining area. In addition, the historical spectrum database has been verified for a long time, avoiding the risk of repeated identification errors of similar obstacles. When an unknown obstacle appears, its valid echo spectrum and confirmed obstacle information are stored in the historical spectrum database. The content of the historical spectrum database is continuously enriched in a data-driven manner, which realizes the rapid identification of obstacles and ensures the accuracy and reliability of detection.

[0041] In some embodiments of the present application, when there is no obstacle information in the effective echo spectrum, obtaining an obstacle sample set and constructing an obstacle recognition model includes: dividing the obstacle sample set into a training set and a test set, pre-selecting a convolutional neural network model, using the training set to train the convolutional neural network model, and substituting the test set into the trained convolutional neural network model to determine the prediction accuracy; when the prediction accuracy is greater than or equal to the prediction accuracy threshold, the trained convolutional neural network model is determined as the obstacle recognition model.

[0042] Specifically, when a valid echo spectrum has no matching historical valid echo spectrum in the historical spectrum database, an obstacle sample set is obtained. This set contains valid echo spectra of a large number of obstacles and their corresponding obstacle-related information (such as type, size, and material). This set is divided into a training set (70%-80%) and a test set (20%-30%). The training set allows the model to learn the correlation between spectral features and obstacle information, while the test set verifies the model's generalization capabilities. A convolutional neural network (CNN) was selected as the base model. CNN excels at processing data with local correlations (such as the frequency distribution and energy variation trends of the spectrum). It extracts spectral features through convolutional layers, compresses redundant information through pooling layers, and finally outputs obstacle predictions through fully connected layers. During the training process, the convolutional neural network model is iteratively trained using the training set. The model adjusts parameters (such as the convolution kernel size and the number of neurons) to minimize prediction errors. After each training session, the prediction accuracy is evaluated using the test set. The prediction accuracy threshold is preferably 0.9. When the prediction accuracy is greater than or equal to the preset prediction accuracy threshold, it indicates that the model has mastered stable recognition rules and is then identified as an obstacle recognition model. The model learns the spectral characteristics of obstacles, solving the problem of being unable to recognize obstacles without historical records. This covers new types of obstacles that suddenly appear in the mining area, such as rare rock forms and temporarily intrusive equipment. In addition, using the convolutional neural network model as the original model can automatically capture subtle features in the spectrum that are difficult to detect, such as small fluctuations in high-frequency energy, reducing the subjectivity of manual feature extraction and ensuring the stability of obstacle detection and recognition results.

[0043] In some embodiments of the present application, when determining obstacle information based on an obstacle recognition model and an effective echo spectrum, and determining effective obstacles on a driving route according to the obstacle information, it includes: substituting the effective echo spectrum into the obstacle recognition model to determine the obstacle information, and determining the effective obstacles on the driving route according to the relationship between the obstacle information and the size of the unmanned mining truck.

[0044] Specifically, after inputting the effective echo spectrum into the obstacle recognition model, the model outputs obstacle information, including obstacle type (rock, mine truck) and dimensions (length, width, height). This information serves as the basis for determining whether an obstacle affects driving. Constraints are established based on the inherent dimensions of the unmanned mining truck. If an obstacle is large enough to intrude into the truck's path or its height exceeds the truck's safe passage threshold (for example, a 4.2m high obstacle prevents the truck's chassis from passing over it), the obstacle is considered a valid obstacle. Conversely, if an obstacle is smaller (such as a 0.5m diameter rock) and located on the roadside, it will not affect the truck's passage and is considered an invalid obstacle. By determining valid obstacles along the driving route based on the relationship between obstacle information and the truck's dimensions, invalid obstacles (such as small rocks and low weeds) are filtered out. This allows for flexible and stable obstacle detection in complex environments such as mining areas, ensuring reliable and stable obstacle detection and recognition.

[0045] In summary, the beneficial effects of the present invention are: by adjusting the detection echo signal through the signal-to-noise ratio, the interference of noise is effectively filtered out, laying a data foundation for subsequent analysis and processing, combining point cloud topology network clustering and spectral feature analysis, accurately distinguishing ground protrusions from real obstacles, solving the risk of over-clustering or under-clustering of irregular rock piles and low obstacles, thereby reducing false detection and missed detection rates, and ensuring the stability of effective identification of complex obstacles in mining areas. Spectral data is introduced for dynamic analysis, and spatial spectrum feature clusters are constructed using the synchronous changes of spectral features. By comparing with the historical spectrum database, the differences between obstacles and the environment are accurately captured, and dynamically changing obstacle scenes such as rock rolling and temporary accumulations, as well as complex working conditions of multiple vehicles crossing, obstacle information is determined through the effective echo spectrum, and effective obstacles are accurately identified, ensuring the driving safety of unmanned mining trucks in complex environments, while reducing unnecessary deceleration or detours due to misjudgment.

[0046] In another preferred embodiment based on the above embodiment, refer to Figure 2 As shown, this embodiment provides an obstacle laser detection system for an unmanned mining truck, which is used to apply the above-mentioned obstacle laser detection method for an unmanned mining truck, including: The acquisition and adjustment module is configured to transmit a radar detection signal to the driving route based on the lidar and determine all detection echo signals, obtain the signal-to-noise ratio of all detection echo signals, determine the adjustment strategy of the corresponding detection echo signals based on the signal-to-noise ratio, and determine the echo time, spectrum data and point cloud data based on all the adjusted detection echo signals.

[0047] The spatial spectrum module is configured to determine a constant driving strategy or a suspected obstacle strategy for a driving route based on the echo time. When a suspected obstacle strategy is determined, a point cloud topology network is constructed based on the point cloud data and clustering is performed to determine a spatial feature cluster. Based on the spatial feature cluster, spectral features containing spectral data are extracted, and spatial spectrum feature clusters are determined based on the synchronous changes of the spectral features.

[0048] The obstacle analysis module is configured to determine the effective echo spectrum based on the spectrum space and spectrum time of the spatial spectrum feature cluster, compare the effective echo spectrum with the historical spectrum database to determine whether there is obstacle information in the effective echo spectrum, and when there is no obstacle information in the effective echo spectrum, obtain an obstacle sample set and build an obstacle recognition model.

[0049] The obstacle determination module is configured to determine obstacle information based on the obstacle recognition model and the effective echo spectrum, and determine effective obstacles on the driving route according to the obstacle information.

[0050] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0051] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0052] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A laser obstacle detection method for unmanned mining trucks, characterized in that: include: transmitting a radar detection signal to the driving route based on a laser radar and determining all detection echo signals, obtaining a signal-to-noise ratio of all detection echo signals, determining an adjustment strategy for corresponding detection echo signals based on the signal-to-noise ratio, and determining echo time, spectrum data, and point cloud data based on all the adjusted detection echo signals; determining a constant driving strategy or a suspected obstacle strategy for the driving route based on the echo time; if the suspected obstacle strategy is determined, constructing a point cloud topology network based on the point cloud data and performing clustering to determine a spatial feature cluster; extracting spectral features including spectral data based on the spatial feature cluster; and determining a spatial spectral feature cluster based on synchronous changes in the spectral features; determining a valid echo spectrum based on the spectrum space and spectrum time of the spatial spectrum feature cluster, comparing the valid echo spectrum with a historical spectrum database to determine whether obstacle information in the valid echo spectrum exists, and when obstacle information in the valid echo spectrum does not exist, obtaining an obstacle sample set and constructing an obstacle recognition model; Obstacle information is determined based on the obstacle recognition model and the effective echo spectrum, and effective obstacles on the driving route are determined according to the obstacle information.

2. The obstacle laser detection method for unmanned mining trucks according to claim 1, characterized in that: When determining an adjustment strategy for a corresponding detection echo signal based on the signal-to-noise ratio, the method includes: When the signal-to-noise ratio is greater than or equal to the signal-to-noise ratio threshold, phase compensation is performed based on the Doppler frequency; When the signal-to-noise ratio is less than the signal-to-noise ratio threshold, the complex signal of the short-time window of each detection echo signal is extracted, the conjugate product with the previous detection echo signal is determined, and the phase difference is extracted to determine the phase difference sequence, phase unwrapping is performed based on the phase difference sequence and accumulated to determine the accumulated phase, and phase compensation is performed based on the accumulated phase.

3. The obstacle laser detection method for unmanned mining trucks according to claim 2, characterized in that: When determining the constant driving strategy or the suspected obstacle strategy of the driving route according to the echo time, the method includes: Obtaining a standard echo time interval, and comparing the standard echo time interval with the echo time; When there is an echo time greater than the left boundary of the standard echo time interval and less than the right boundary of the standard echo time interval, the driving route is determined to be a constant driving strategy; otherwise, the driving route is determined to be a suspected obstacle strategy.

4. The obstacle laser detection method for unmanned mining trucks according to claim 3 is characterized in that: When constructing a point cloud topology network based on the point cloud data and performing clustering to determine spatial feature clusters, the method includes: removing isolated noise points in the point cloud data based on a bilateral filtering algorithm to determine valid point cloud data, and determining a neighborhood radius with each point of the valid point cloud data as the center, wherein the neighborhood radius is determined based on the point cloud density; For each point, search all other points within its neighborhood radius to determine whether they are bidirectionally adjacent. If so, establish undirected edges between the point and the bidirectionally adjacent points, and determine the edge attributes of each undirected edge. Otherwise, do not establish undirected edges. Constructing the point cloud topology network from all points and their corresponding undirected edges, determining the local connection density of each point, determining an initial clustering region based on the local connection density, traversing all undirected edges in each initial clustering region to determine a reflection intensity gradient, and determining a plurality of reflection sub-regions with the same reflection intensity based on a relationship between the reflection intensity gradient and the average gradient; The normal vector angles of all points in each reflection sub-region are determined, and the reflection sub-region is split based on the normal vector angles, and the split results are determined as the spatial feature clusters.

5. The obstacle laser detection method for unmanned mining trucks according to claim 4 is characterized in that: When extracting spectrum features including spectrum data based on the spatial feature cluster, and determining the spatial spectrum feature cluster based on synchronous changes of the spectrum features, the method includes: Extracting segments of spectrum data corresponding to all points in the spatial feature cluster, comparing the acquisition time of each segment with the acquisition time of the corresponding point, eliminating segments with mismatched times, and determining all main frequency offsets, high-frequency energy, and main frequency change rates based on the remaining segments; The synchronization in the spatial coordinate diagram, spatial distribution diagram and spatial position diagram is determined respectively according to all main frequency offsets, high-frequency energy and main frequency change rate, the fragments with the highest synchronization are extracted, and the spatial feature cluster is split according to whether the corresponding points form a continuous area in space, and the split result is determined as the spatial spectrum feature cluster.

6. The obstacle laser detection method for unmanned mining trucks according to claim 5, characterized in that: When determining the effective echo spectrum according to the spectrum space and spectrum time of the spatial spectrum feature cluster, the method includes: Based on the point cloud topology network of the spatial spectrum feature cluster, the connection relationship of all points in each connected component is determined, segments of the spectrum data synchronously collected for all points in the spatial spectrum feature cluster are extracted, and the spectrum characteristics of each segment in the connected component in which it is located are verified to satisfy the transitive association; All segments that satisfy the transitive association are extracted to construct a spectrum segment set, the reflection intensity and high-frequency energy proportion of each point corresponding to the connected components of the spectrum segment set are extracted according to the corresponding connection relationship, and the correlation direction of the reflection intensity and high-frequency energy proportion of each point is counted, the main frequency change trend is determined based on the correlation direction, and the segments that are consistent with the main frequency change trend and the point cloud movement direction are retained to construct a valid spectrum set, and the effective echo spectrum is determined based on the valid spectrum set.

7. The obstacle laser detection method for unmanned mining trucks according to claim 6, characterized in that: When comparing the effective echo spectrum with a historical spectrum database to determine whether obstacle information of the effective echo spectrum exists, the method includes: The historical spectrum database includes a plurality of historical valid echo spectrums and a plurality of historical obstacle information, and each historical valid echo spectrum corresponds to a piece of historical obstacle information; When a historical valid echo spectrum identical to the valid echo spectrum exists in the historical spectrum database, determining the historical obstacle information corresponding to the historical valid echo spectrum as the obstacle information, and determining that the obstacle information of the valid echo spectrum exists; When there is no historical valid echo spectrum identical to the valid echo spectrum in the historical spectrum database, it is determined that there is no obstacle information of the valid echo spectrum.

8. The obstacle laser detection method for unmanned mining trucks according to claim 7, characterized in that: When there is no obstacle information in the valid echo spectrum, obtaining an obstacle sample set and building an obstacle recognition model includes: Dividing the obstacle sample set into a training set and a test set, and preselecting a convolutional neural network model; Using the training set to train the convolutional neural network model, and substituting the test set into the trained convolutional neural network model to determine the prediction accuracy; When the prediction accuracy is greater than or equal to the prediction accuracy threshold, the trained convolutional neural network model is determined as the obstacle recognition model.

9. The obstacle laser detection method for unmanned mining trucks according to claim 8, characterized in that: When determining obstacle information based on the obstacle recognition model and the effective echo spectrum, and determining effective obstacles on the driving route according to the obstacle information, the method includes: Substituting the effective echo spectrum into the obstacle recognition model to determine obstacle information; The effective obstacles on the driving route are determined according to the relationship between the obstacle information and the size of the unmanned mining truck.

10. An obstacle laser detection system for an unmanned mining truck, used to apply the obstacle laser detection method for an unmanned mining truck according to any one of claims 1 to 9, characterized in that: include: an acquisition and adjustment module configured to transmit a radar detection signal to the driving route based on the laser radar and determine all detection echo signals, obtain a signal-to-noise ratio of all detection echo signals, determine an adjustment strategy for the corresponding detection echo signals based on the signal-to-noise ratio, and determine echo time, spectrum data, and point cloud data based on all the adjusted detection echo signals; a spatial spectrum module configured to determine a constant driving strategy or a suspected obstacle strategy for the driving route based on the echo time; when the suspected obstacle strategy is determined, construct a point cloud topology network based on the point cloud data and perform clustering to determine a spatial feature cluster; extract spectral features containing spectral data based on the spatial feature cluster; and determine a spatial spectrum feature cluster based on synchronous changes in the spectral features; an obstacle analysis module configured to determine a valid echo spectrum based on the spectrum space and spectrum time of the spatial spectrum feature cluster, compare the valid echo spectrum with a historical spectrum database to determine whether obstacle information exists in the valid echo spectrum, and when obstacle information does not exist in the valid echo spectrum, obtain an obstacle sample set and construct an obstacle recognition model; The obstacle determination module is configured to determine obstacle information based on the obstacle recognition model and the effective echo spectrum, and determine effective obstacles on the driving route according to the obstacle information.

Images