Fine measurement system for automobile cleaning

Abstract

The invention discloses a refined measurement system for automobile cleaning, which is characterized in that a laser scanning measurement module is introduced to carry out high-precision three-dimensional scanning on an automobile to obtain complete point cloud data, a data processing control module reconstructs a three-dimensional digital model reflecting the difference of automobile types based on the point cloud data and extracts key contour features, and the key contour features of the three-dimensional digital model are obtained; a cleaning control instruction completely matched with the features is generated; the actuating mechanism driving module plans and controls the movement track, posture and working parameters of the cleaning mechanism in real time according to the cleaning control instruction, so that a cleaning part can be tightly attached to unique curved surfaces and concave-convex structures of different vehicle types for operation, and the problem that the contour difference of different vehicle types cannot be recognized in the traditional technology is solved; the technical problem that cleaning blind areas are prone to being caused is solved, and no-dead-corner and consistent cleaning is achieved.

Description

Technical Field

[0001] This invention relates to the field of laser measurement technology, and more specifically to a precision measurement system for car washing. Background Technology

[0002] Traditional tunnel or gantry-style automatic car washes typically use preset programs for cleaning, with fixed brush and spray arm movement trajectories, or relying solely on simple photoelectric switches or ultrasonic sensors for coarse position detection. However, this traditional technology cannot recognize the differences in the shape and contours of different car models, easily creating blind spots in the cleaning process. Summary of the Invention

[0003] The purpose of this invention is to provide a precision measurement system for car washing, in order to solve the technical problem that traditional technologies cannot identify the differences in the shape contours of different car models, which easily leads to blind spots in cleaning.

[0004] The technical solution of this invention is implemented as follows:

[0005] A precision measurement system for car washing includes:

[0006] The laser scanning measurement module is used to acquire three-dimensional point cloud data of the vehicle to be cleaned;

[0007] The data processing and control module is used to process the three-dimensional point cloud data to construct a three-dimensional model of the vehicle and identify its external contour features, and generate cleaning control instructions based on the identification results.

[0008] An actuator drive module is used to receive the cleaning control command and clean the vehicle based on the cleaning control command;

[0009] The cleaning control command controls at least the movement trajectory of the actuator drive module so that it matches the identified vehicle outline features.

[0010] A further technical solution is that the laser scanning measurement module includes:

[0011] The laser emitting submodule is used to emit laser beams onto the vehicle surface;

[0012] The scanning drive submodule is used to drive the laser emission submodule to move according to a predetermined scanning mode, so that the laser beam covers the area of ​​the vehicle to be scanned.

[0013] A signal receiving submodule is used to receive laser signals reflected back from the surface of the vehicle;

[0014] The data acquisition submodule is used to calculate distance information based on the laser signal and simultaneously combine it with the scanning angle information to generate three-dimensional point cloud data of the vehicle.

[0015] A further technical solution is that the data acquisition submodule includes:

[0016] The signal processing unit is used to filter and amplify the laser electrical signal output by the signal receiving submodule;

[0017] The distance calculation unit calculates the straight-line distance between the laser reflection point and the scanning origin based on the processed electrical signal and according to the preset distance measurement principle.

[0018] The data fusion unit is connected to the distance calculation unit and the scanning drive submodule, respectively, and is used to receive the straight-line distance and real-time scanning angle information, and convert the two into three-dimensional coordinates in the same coordinate system;

[0019] The point cloud generation unit, connected to the data fusion unit, is used to serialize and encapsulate the three-dimensional coordinates to generate a standard three-dimensional point cloud data stream.

[0020] A further technical solution is that the data processing control module includes:

[0021] The point cloud registration submodule is used to denoise, filter, and process the coordinate system of the input raw 3D point cloud data to form a registered integrated point cloud.

[0022] The 3D model reconstruction submodule reconstructs the 3D surface model of the vehicle based on the integrated point cloud through surface fitting or meshing algorithms.

[0023] The feature recognition and extraction submodule is used to analyze the three-dimensional surface model and identify and extract the vehicle's external contour feature parameters.

[0024] The instruction generation submodule is used to generate the cleaning control instruction based on the extracted shape contour feature parameters.

[0025] A further technical solution is that the 3D model reconstruction submodule includes:

[0026] The cloud optimization unit is used to resample and remove outliers from the integrated point cloud to obtain an optimized point cloud.

[0027] A geometric structure building unit is used to perform surface fitting or triangular meshing calculations on the optimized point cloud to generate an initial three-dimensional geometric model.

[0028] The model repair unit is used to repair holes and smooth the surface of the initial three-dimensional geometric model, and output a complete three-dimensional surface model of the vehicle.

[0029] A further technical solution is that the feature recognition and extraction submodule includes:

[0030] The model region segmentation unit, based on geometric features and curvature analysis, divides the three-dimensional surface model into model partitions corresponding to different parts of the vehicle;

[0031] The feature geometry extraction unit is used to extract the geometric elements of the outline from each of the model partitions;

[0032] The parameter calculation unit calculates the shape contour feature parameters based on the geometric elements;

[0033] The parameter output unit is used to encapsulate and output the shape contour feature parameters according to a predefined format.

[0034] A further technical solution is that the specific operation steps of the feature geometry extraction unit include:

[0035] S1. For each model partition, based on the type of vehicle component to which it belongs, select a feature detection algorithm to initially obtain a set of discrete feature points of the outer contour;

[0036] S2. Based on the discrete feature point set, use a geometric prototype that matches the vehicle structure to perform fitting calculations to generate continuous geometric elements that describe the main body contour.

[0037] S3. Analyze the spatial relationships between the continuous geometric elements corresponding to different model partitions, and calculate and record the intersecting, parallel, perpendicular or coplanar relationships between the continuous geometric elements;

[0038] S4. Integrate the continuous geometric elements and their relationships in the geometric topology network to generate a set of geometric elements for the vehicle's external outline.

[0039] A further technical solution is that S1 specifically includes:

[0040] S101. Obtain the semantic identifier of each model partition to determine the specific vehicle component type to which it belongs;

[0041] S102. Based on a preset vehicle component type mapping database, match at least one feature detection algorithm corresponding to the current component type;

[0042] S103. Initialize and configure the feature detection algorithm, and execute the algorithm to initially obtain the discrete feature point set of the outer contour.

[0043] A further technical solution is that S103 specifically includes:

[0044] S1031. Based on the geometric feature data of the current model partition, calculate and generate the initialization parameter set of the feature detection algorithm;

[0045] S1032. Perform coordinate transformation and resampling on the 3D point cloud or mesh data of the current model partition to generate data blocks that meet the input requirements of the feature detection algorithm;

[0046] S1033. Load the data block and the initialization parameter set into the feature detection algorithm kernel for operation, and output the initial feature position set;

[0047] S1034. Perform false detection elimination and neighbor point aggregation operations based on geometric constraints on the initial feature position set to generate the discrete feature point set of the outer contour.

[0048] A further technical solution is that the actuator drive module includes:

[0049] The instruction parsing and allocation submodule is used to receive and parse the cleaning control instructions, and decompose them into independent motion and action parameters corresponding to different cleaning actuators;

[0050] The multi-axis coordinated motion control submodule is used to generate multi-axis coordinated control signals to drive the servo motors in each cleaning actuator based on the motion and action parameters.

[0051] The safety monitoring submodule is used to monitor the position and stress status of each cleaning actuator in real time, and trigger emergency intervention when it exceeds the preset safe working boundary or contact force threshold.

[0052] The dynamic adjustment submodule is used to collect real-time status data of the actuator and compare it with the target value of the instruction in order to adjust the cleaning process.

[0053] The beneficial effects of this invention are as follows:

[0054] By introducing a laser scanning measurement module to perform high-precision 3D scanning of vehicles and obtain complete point cloud data, the data processing and control module reconstructs a 3D digital model reflecting the differences in vehicle models based on the point cloud data and extracts key contour features, thereby generating cleaning control commands that perfectly match the features. The actuator drive module plans and controls the motion trajectory, posture, and working parameters of the cleaning mechanism in real time according to the cleaning control commands, so that the cleaning components can closely fit the unique curved surfaces and concave and convex structures of different vehicle models to perform operations. This solves the technical problem that traditional technologies cannot identify the differences in the shape contours of different vehicle models, which easily leads to cleaning blind spots, and achieves cleaning without dead angles and with consistency. Attached Figure Description

[0055] Figure 1 A block diagram of a precision measurement system for car washing provided by the present invention;

[0056] Figure 2 A block diagram of the laser scanning measurement module provided by the present invention;

[0057] Figure 3 A block diagram of the data processing control module provided by the present invention;

[0058] Figure 4 A block diagram of the actuator driving module provided by the present invention. Detailed Implementation

[0059] To better understand the technical content of this invention, specific embodiments are provided below, and the invention will be further described in conjunction with the accompanying drawings.

[0060] See Figures 1 to 4 This invention provides a refined measurement system for car washing, comprising: a laser scanning measurement module for acquiring three-dimensional point cloud data of the vehicle to be washed; a data processing and control module for processing the three-dimensional point cloud data of the model repair unit to construct a three-dimensional model of the vehicle and identify its external contour features, and generating washing control commands based on the identification results; and an actuator drive module for receiving the model repair unit washing control commands and washing the vehicle based on the model repair unit washing control commands; wherein the model repair unit washing control commands at least control the motion trajectory of the model repair unit actuator drive module to match the identified vehicle external contour features.

[0061] It should be noted that 3D point cloud data refers to a collection of massive spatial point coordinates that represent the surface shape of an object, obtained through 3D measurement equipment. It is a discretized digital representation of the object's 3D shape.

[0062] Vehicle shape profile features refer to the set of parameters extracted from the vehicle's 3D model that describe its overall shape and local key structural geometric properties, and are the direct basis for control decisions.

[0063] Cleaning control commands refer to the set of operation commands output by the data processing control module that drive the actuators to perform specific cleaning operations. Their content is generated based on the real-time identified vehicle contour features.

[0064] Specifically, when a vehicle to be cleaned enters the system's scanning area, the laser scanning measurement module starts working. One or more 3D laser scanners emit laser beams onto the vehicle surface and receive the reflected signals. By calculating the laser's time of flight or phase difference, the module obtains the spatial coordinates of a large number of measurement points on the vehicle surface. This massive collection of spatial points constitutes the 3D point cloud data. This data provides a raw digital outline of the vehicle's exterior.

[0065] The data processing control module processes the received raw 3D point cloud data. This includes denoising and registration preprocessing, unifying point clouds from multiple scanners into a single coordinate system to form a complete and clean vehicle point cloud model. These discrete points are then connected and fitted to construct a continuous and complete 3D vehicle surface model.

[0066] The actuator drive module receives and parses the cleaning control commands. These commands drive servo motors, hydraulic cylinders, and other actuators, precisely controlling the cleaning actuators to move along a pre-planned trajectory that matches the vehicle's contours. For example, the top brush adaptively oscillates along the extracted roofline, while the side brushes automatically avoid or adjust pressure when approaching the rearview mirror, thus achieving refined cleaning for different vehicle models.

[0067] In this embodiment of the invention, a laser scanning measurement module is introduced to perform high-precision three-dimensional scanning of the vehicle to obtain complete point cloud data. The data processing and control module reconstructs a three-dimensional digital model reflecting the differences in vehicle models based on the point cloud data and extracts key contour features, thereby generating cleaning control commands that perfectly match the features. The actuator drive module plans and controls the motion trajectory, posture, and working parameters of the cleaning mechanism in real time according to the cleaning control commands, so that the cleaning components can closely fit the unique curved surfaces and concave and convex structures of different vehicle models to perform operations. This solves the technical problem that traditional technologies cannot identify the differences in the shape contours of different vehicle models, which easily leads to blind spots in cleaning, and achieves cleaning without dead angles and with consistency.

[0068] Preferably, the laser scanning measurement module of the model repair unit includes: a laser emitting submodule for emitting a laser beam onto the vehicle surface; a scanning driving submodule for driving the laser emitting submodule of the model repair unit to move according to a predetermined scanning mode, so that the laser beam of the model repair unit covers the area of ​​the vehicle to be scanned; a signal receiving submodule for receiving the laser signal reflected back from the surface of the model repair unit vehicle; and a data acquisition submodule for calculating distance information based on the laser signal of the model repair unit, and simultaneously combining it with the scanning angle information to generate three-dimensional point cloud data of the model repair unit vehicle.

[0069] It should be noted that the predetermined scanning mode refers to the control program set in advance to ensure that the laser beam covers the target area in an orderly manner. It specifies the path, speed and density of the laser beam deflection and is the basis for data acquisition.

[0070] Scanning angle information refers to the angular parameters fed back by the scanning drive mechanism at the moment of laser ranging, which are used to define the spatial direction of the laser beam. These parameters typically include the horizontal azimuth and vertical pitch angles.

[0071] Three-dimensional point cloud data refers to a set of a large number of three-dimensional spatial coordinate points obtained through the above measurement process. It is a discrete digital representation of the geometric shape of the vehicle surface.

[0072] Specifically, the laser emission submodule is activated, and its core is typically a laser diode. It generates a laser beam with a specific wavelength and power, and then collimates it before directing it toward the vehicle surface. This is the physical signal source for the entire measurement process.

[0073] Next, the scanning drive submodule begins to operate. It contains motion mechanisms, such as high-speed galvanometers or rotary motors. Following a pre-set scanning pattern, either grid-like or linear, it drives the optical components in the laser emission submodule to perform rapid and orderly two-dimensional scanning motion on the vehicle surface. This ensures that the laser beam covers all key surfaces of the area to be scanned, acquiring contour information without omission.

[0074] When a laser beam strikes the vehicle's surface, a portion of the beam is reflected. The signal receiving submodule is responsible for capturing these weak laser signals reflected back from the vehicle's surface and converting them into corresponding electrical signals.

[0075] The data acquisition submodule processes the electrical signals. Internally, it includes a high-speed timing circuit or a phase detection circuit. Based on the time difference or phase difference between the emitted and received laser signals, and according to the principle of the constancy of the speed of light, it calculates the precise distance between the laser reflection point and the scanner origin. Simultaneously, it can receive feedback signals from the scanning drive submodule in real time to obtain the azimuth and elevation angles of the current laser beam, i.e., the scanning angle information.

[0076] The data acquisition submodule fuses the distance information measured at each instant with the corresponding scanning angle information. Through the built-in coordinate transformation model, it calculates the absolute coordinates of each measurement point in three-dimensional space with the scanner as the origin. As the scanning continues, a massive number of three-dimensional spatial coordinate points are generated, arranged, and output in real time, ultimately forming a complete three-dimensional point cloud data representing the vehicle's outline.

[0077] Furthermore, the model repair unit data acquisition submodule includes: a signal processing unit, used to filter and amplify the laser electrical signal output by the model repair unit signal receiving submodule; a distance calculation unit, which calculates the straight-line distance between the laser reflection point and the scanning origin based on the processed electrical signal and according to the preset ranging principle; a data fusion unit, connected to the model repair unit distance calculation unit and the model repair unit scanning drive submodule respectively, used to receive the straight-line distance and real-time scanning angle information of the model repair unit, and convert the two into three-dimensional coordinates in the same coordinate system; and a point cloud generation unit, connected to the model repair unit data fusion unit, used to serialize and encapsulate the three-dimensional coordinates of the model repair unit to generate a standard three-dimensional point cloud data stream.

[0078] It should be noted that the preset ranging principle refers to the physical ranging method that is determined and fixed in the hardware or software during the system design. It is the mathematical and physical basis for distance calculation, such as the time-of-flight method, which calculates distance by measuring the round-trip time of a light pulse.

[0079] Specifically, the signal processing unit begins operation. It receives the raw laser electrical signal from the signal receiving submodule. This signal is typically weak and contains interference such as ambient light noise and circuit thermal noise. This unit filters out noise outside a specific frequency band using analog or digital filtering circuits, and then amplifies the effective signal using operational amplifier circuits to improve the signal-to-noise ratio.

[0080] The processed electrical signal is sent to the distance calculation unit, which internally incorporates a preset ranging principle. If the time-of-flight method is used, a high-precision timer within the unit measures the time difference between laser emission and reception, calculating the distance based on the speed of light constant. If the phase difference method is used, the distance is calculated by detecting the phase shift between the emitted and received laser beams. Based on the chosen principle, the unit calculates the straight-line distance between the laser reflection point and the scanning origin for each laser pulse from the processed electrical signal.

[0081] Next, the data fusion unit synchronously receives data from two sources: one is the straight-line distance value provided by the distance calculation unit, and the other is the scanning angle information describing the current spatial direction of the laser beam, namely the horizontal azimuth and vertical elevation angles, which are fed back in real time by the scanning drive submodule. Based on the transformation relationship from spherical coordinates to Cartesian coordinates, each set of distance, azimuth, and elevation angles is converted in real time into three-dimensional coordinates in a unified spatial coordinate system with the optical center of the scanner as the origin.

[0082] Finally, the point cloud generation unit organizes and manages the discrete 3D coordinate points continuously output by the data fusion unit. It adds timestamps and other information to the 3D coordinate points generated in chronological order, and serializes and encapsulates them according to a predetermined data packet structure including frame header, point coordinates, intensity values, and frame tail, ultimately forming a continuous and ordered standard 3D point cloud data stream, which is then output to the data processing and control module, thus completing all the underlying data acquisition tasks from analog laser signals to digital spatial models.

[0083] Preferably, the model repair unit data processing control module includes: a point cloud registration submodule, used to denoise, filter, and process the input raw 3D point cloud data to form a registered integrated point cloud; a 3D model reconstruction submodule, which reconstructs the 3D surface model of the vehicle based on the integrated point cloud of the model repair unit through surface fitting or meshing algorithms; a feature recognition and extraction submodule, used to analyze the 3D surface model of the model repair unit, identify and extract the vehicle's outline feature parameters; and an instruction generation submodule, used to generate cleaning control instructions for the model repair unit based on the extracted outline feature parameters of the model repair unit.

[0084] It should be noted that point cloud registration refers to the process of aligning multiple 3D point cloud datasets to the same coordinate system through spatial transformation;

[0085] Integrated point cloud refers to a set of three-dimensional point data that is unified in a single coordinate system and completely covers the target object after denoising, filtering and registration processing.

[0086] A three-dimensional surface model refers to a digital model generated from point clouds to represent the continuous geometry of an object's surface. Common forms include triangular mesh models or parametric surface models.

[0087] Surface fitting / meshing algorithms refer to two main types of computational methods for generating continuous surfaces from discrete point clouds. Surface fitting aims to obtain a smooth mathematical surface representation; meshing generates a surface composed of polygonal patches.

[0088] External contour feature parameters refer to the set of data extracted from the 3D model that can describe the overall shape of the vehicle and the geometric and positional attributes of its key local structures.

[0089] Specifically, the point cloud registration submodule begins receiving raw 3D point cloud data from the data acquisition submodule. This data typically contains noise caused by environmental dust and stray light, and is acquired by multiple scanners from different perspectives, resulting in inconsistencies in coordinates. Noise removal and filtering are required to eliminate invalid data points. Subsequently, crucial coordinate system processing is performed, namely, using an iterative nearest-point algorithm to precisely align point cloud data from different perspectives and time points to the same global coordinate system.

[0090] The 3D model reconstruction submodule transforms discrete point clouds into continuous surface representations. If a surface fitting algorithm is used, it generates a smooth surface model by fitting the point cloud and its normal vectors using implicit functions; if a meshing algorithm is used, it connects adjacent points into a network of triangular facets.

[0091] The feature recognition and extraction submodule uses algorithms such as geometric analysis and pattern recognition to understand the vehicle structure represented by the model. It automatically identifies key component areas such as the roof, windows, doors, bumpers, and rearview mirrors, and extracts quantifiable shape contour feature parameters from them.

[0092] The instruction generation submodule combines input feature parameters, such as the rearview mirror position, with preset strategies, such as "the side brush sweeps past the edge of the rearview mirror with a 5cm gap," to generate specific, executable cleaning control instructions through calculation. These instructions are detailed action sequences that clearly define the movement trajectory, speed, start and stop time, and interaction force or distance with the vehicle body surface of each cleaning actuator.

[0093] Furthermore, the 3D model reconstruction submodule of the model repair unit includes: a cloud optimization unit, which is used to resample and remove outliers from the integrated point cloud of the model repair unit to obtain an optimized point cloud; a geometric structure construction unit, which is used to perform surface fitting or triangular meshing calculations on the optimized point cloud of the model repair unit to generate an initial 3D geometric model; and a model repair unit, which is used to repair holes and smooth the surface of the initial 3D geometric model of the model repair unit to output a complete 3D surface model of the vehicle.

[0094] It should be noted that resampling refers to the process of changing the density and distribution of point cloud data through specific sampling rules, aiming to reduce data redundancy and homogenize density in order to improve the efficiency and stability of subsequent processing.

[0095] Outliers are isolated data points in 3D point cloud data that deviate significantly from the main point set on the real surface of an object in terms of spatial distance. They are usually caused by measurement errors or environmental interference.

[0096] Optimized point cloud refers to point cloud data obtained after resampling and outlier removal. It has the characteristics of uniform density and fewer noise points, making it more suitable for stable 3D reconstruction.

[0097] The initial three-dimensional geometric model refers to the three-dimensional digital model initially generated from geometric structural building blocks. It expresses the basic shape of the vehicle, but may contain defects such as holes and rough surfaces.

[0098] Hole repair refers to the automatic identification and filling of surface damage areas caused by missing data on a 3D mesh model.

[0099] Specifically, the cloud optimization unit receives the integrated point cloud from upstream. Although this point cloud has been registered, its data point distribution may be uneven, and it still contains a few outliers caused by measurement errors or residual noise. The cloud optimization unit first performs a resampling operation, typically using a voxel rasterization method to divide the three-dimensional space into uniform small cubes, retaining only the centroid point or a random point within each voxel, thereby homogenizing the point cloud density and preserving shape characteristics while reducing the amount of data. Next, outlier removal is performed, for example using statistical filtering, calculating the average distance and standard deviation of each point with its nearest neighbors, and removing points whose distance exceeds a certain threshold as noise.

[0100] The geometric structure building unit takes the optimized point cloud as input and outputs an initial three-dimensional geometric model that can express the basic shape of the vehicle shell.

[0101] Finally, the model repair unit performs fine-tuning post-processing on the initial model, primarily addressing two issues: First, hole repair. For data-deficient areas on the model surface caused by scanning blind spots, such as the edge of the car's undercarriage or behind the rearview mirror, new triangular facets are automatically generated based on the geometric trends around the holes to fill the gaps and form a closed model. Second, surface smoothing. For non-physical bumps and depressions on the model surface caused by noise or triangulation, a smoothing filtering algorithm is used to fine-tune the vertex positions while maintaining the overall geometric features, eliminating irregular undulations and achieving a smooth surface visual effect.

[0102] Furthermore, the model repair unit feature recognition and extraction submodule includes: a model region segmentation unit, which, based on geometric features and curvature analysis, segments the three-dimensional surface model of the model repair unit into model partitions corresponding to different parts of the vehicle; a feature geometry extraction unit, which extracts the geometric elements of the outline from each model repair unit model partition; a parameter calculation unit, which calculates the outline feature parameters based on the geometric elements of the model repair unit; and a parameter output unit, which encapsulates and outputs the outline feature parameters of the model repair unit according to a predefined format.

[0103] It should be noted that geometric feature and curvature analysis refers to the method of calculating and analyzing the local geometric properties of the surface of a 3D model, such as the degree of curvature of the surface and the direction of the normal vector. It is the basis for distinguishing between planar, cylindrical, and spherical regions.

[0104] Model partitioning refers to the process of dividing a single, continuous 3D surface model representing the entire vehicle into multiple geometrically relatively uniform sub-region models that semantically correspond to specific vehicle components, using a segmentation algorithm.

[0105] Geometric elements refer to the basic geometric entities extracted from model partitions that are used to describe their outlines and shapes; they are an intermediate representation between the original mesh data and the final numerical parameters.

[0106] External contour feature parameters refer to a set of specific values ​​that describe the overall size, shape, position, orientation, and other attributes of a vehicle and its components, obtained by calculating the extracted geometric elements.

[0107] Specifically, the model region segmentation unit can calculate geometric features such as curvature and normal vector changes for each region of the model surface. By identifying abrupt boundary changes in features, i.e., the transition from a high-curvature body edge to a flat window, and using region growing algorithms or clustering segmentation algorithms, the model surface is automatically segmented into multiple continuous, semantically independent model partitions. Each partition corresponds to a specific vehicle component, such as the "left front door partition," "windshield partition," and "right rearview mirror partition," thereby decomposing the overall model into component-level logical units.

[0108] The feature geometry extraction unit operates on each identified model partition. For example, for the "roof partition," this unit extracts its outer boundary ring and the main ridge line representing the top bulge; for the "window partition," it extracts the planar quadrilateral boundary surrounding the window glass; and for the "rearview mirror partition," it extracts the envelope contour of its protruding shell portion. These geometric elements are higher-level mathematical abstractions than the original mesh, typically represented by basic geometric primitives such as points, lines, curves, planes, or simple surfaces.

[0109] The parameter calculation unit performs calculations based on the extracted geometric elements. According to predefined parameter definition rules, the abstract geometric elements are transformed into specific numerical shape contour feature parameters. For example, the coordinates of the highest point and the longitudinal curvature are calculated based on the main ridge line of the roof; the tilt angle relative to the horizontal plane is calculated based on the window plane; and the centroid coordinates and the maximum protrusion distance relative to the door panel are calculated based on the envelope contour of the rearview mirror.

[0110] The parameter output unit formats and prepares all calculated feature parameters for transmission. For example, it uses structured data formats such as JSON and XML, or custom binary data frames, to associate and encapsulate each parameter with its semantic tags, forming a complete and ordered data packet.

[0111] Furthermore, the specific operational steps of the model repair unit's feature geometry extraction unit include:

[0112] S1. For each model repair unit model partition, according to the type of vehicle component to which it belongs, select a feature detection algorithm to initially obtain the discrete feature point set of the outer contour;

[0113] S2. Based on the discrete feature point set of the model repair unit, use the geometric prototype that matches the vehicle structure to perform fitting calculations and generate continuous geometric elements that describe the main body contour.

[0114] S3. Analyze the spatial relationships between continuous geometric elements of the model repair unit corresponding to different model repair unit partitions, and calculate and record the intersecting, parallel, perpendicular or coplanar relationships between continuous geometric elements of the model repair unit.

[0115] S4. Integrate the continuous geometric elements of the model repair unit and their relationships in the geometric topology network to generate a set of geometric elements for the vehicle's outline.

[0116] It should be noted that vehicle component type refers to the functional classification of the automotive components represented by the model partition, such as hood, trunk lid, fender, etc. This classification drives the selection of subsequent processing strategies.

[0117] Feature detection algorithms refer to computer algorithms used to automatically identify and locate edges, corners, and ridges from three-dimensional data;

[0118] A geometric prototype refers to a predefined parametric basic geometric model for fitting common shapes of vehicle parts;

[0119] Continuous geometric elements refer to geometric entities described by mathematical formulas, obtained through fitting calculations.

[0120] Spatial relationship refers to the relative position and orientation of two or more geometric elements in three-dimensional space;

[0121] A geometric topology network refers to a graphical model that uses geometric elements as nodes and spatial relationships as edges to define the connectivity and constraints of the overall geometric structure of a vehicle.

[0122] The set of geometric elements refers to the total structured data of all geometric elements and their complete topological relationships.

[0123] Specifically, the semantic labels of vehicle component types assigned to the model partitions are read. Based on these labels, a mapping database within the unit is activated, and a preset optimal feature detection algorithm is selected for different types of components. For example, for the "window" type, which is mainly composed of planes, an edge detection algorithm based on normal vector mutation is selected; for the "wheel arch" type, which contains complex curved surfaces, a key point extraction algorithm based on curvature changes is selected. After selecting the algorithm, it is applied to the 3D data of the current partition to initially obtain a set of discrete feature points of key positions of the component shape. This point set is still raw, unorganized low-level data.

[0124] Based on the component type of the current model partition, a matching geometric prototype is invoked. These geometric prototypes are predefined basic geometric models that conform to common morphologies in vehicle engineering, such as planes for fitting door panels, cylindrical surfaces or B-spline surfaces for fitting wheel arches, and spherical surfaces for fitting antenna bases. Next, the discrete feature point set obtained in S1 is fitted onto the selected geometric prototype using fitting calculations, and the plane equation coefficients, cylinder radius, and axis of the prototype are solved.

[0125] The system processes all continuous geometric elements generated by S2 from all model partitions in parallel. Through geometric calculations, it automatically analyzes the spatial relationships between these elements. For example, it calculates whether the "windshield plane" and the "roof surface" intersect on a spatial curve, i.e., the leading edge of the roof; determines whether the "left A-pillar edge" and the "right A-pillar edge" are approximately parallel; determines whether the "door panel plane" is perpendicular to the ground; or verifies whether the four "door frame planes" are coplanar.

[0126] All continuous geometric elements output by S2 are integrated with the relational information in the geometric topology network constructed by S3. This integration involves assigning a unique identifier to each geometric element and linking it to all relationships involving that element, such as "element A intersects with element B." This forms a complete and self-consistent geometric model representation that can be used for efficient computation, reasoning, and data exchange.

[0127] Furthermore, the model repair unit S1 specifically includes:

[0128] S101. Obtain the semantic identifier of each model partition to determine the specific vehicle component type to which it belongs;

[0129] S102. Based on the preset vehicle component type mapping relationship library, match at least one feature detection algorithm corresponding to the component type of the current model repair unit;

[0130] S103. Initialize and configure the feature detection algorithm for the model repair unit, and execute the algorithm to initially obtain the discrete feature point set of the outer contour.

[0131] Specifically, semantic identifiers assigned to each model partition are obtained from the metadata provided by the upstream model region segmentation unit. These identifiers are labels with clear engineering meanings, such as "front bumper," "left front fender," "roof," or "right rearview mirror." By reading these identifiers, the specific vehicle component type to which the model partition belongs is determined, thereby associating abstract geometric regions with specific automotive structural knowledge.

[0132] The strategy selection phase involves accessing a pre-built and stored expert knowledge base, specifically a pre-defined mapping database of vehicle component types and feature detection algorithms. This database defines one or more feature detection algorithms best suited for extracting the shape contour features of each vehicle component type. For example, the mapping might specify that for the "window glass" type, the optimal algorithm is an edge detection algorithm based on normal vector discontinuities; for the "door handle" type, it is mapped to a keypoint detection algorithm based on local curvature extrema. Based on the component type determined in S101, a query and match are performed in this database to match at least one corresponding feature detection algorithm for the current component type.

[0133] The initial configuration is performed, calculating and setting the optimal operating parameters for the algorithm based on the specific geometric feature data of the current model partition. For example, for a large, smooth door partition, the gradient threshold for edge detection is automatically increased to ignore minor textures; for a small, high-curvature rearview mirror partition, the curvature threshold for keypoint detection is decreased to improve sensitivity. Next, the feature detection algorithm matched in the previous step is loaded and executed using this initial configuration. This algorithm traverses and processes the 3D data of the current model partition, identifying and outputting all feature locations that meet the requirements based on its principles and set parameters.

[0134] Furthermore, the model repair unit S103 specifically includes:

[0135] S1031. Based on the geometric feature data of the current model partition, calculate and generate the initialization parameter set of the model repair unit feature detection algorithm;

[0136] S1032. Perform coordinate transformation and resampling on the 3D point cloud or mesh data of the current model partition to generate data blocks that meet the input requirements of the model repair unit feature detection algorithm.

[0137] S1033. Load the model patching unit data block and the model patching unit initialization parameter set into the model patching unit feature detection algorithm kernel for operation, and output the initial feature location set.

[0138] S1034. Perform geometric constraint-based false detection elimination and neighbor point aggregation operations on the initial feature position set of the model repair unit to generate a discrete feature point set of the model repair unit's outline.

[0139] Specifically, the geometric feature data of the current model partition is analyzed. This data includes the spatial distribution density of the point cloud, the mean and variance of the surface curvature, and the approximate size range of the partition in three-dimensional space. Based on this real-time analysis of feature data, an optimal set of initialization parameters is calculated and generated using a linear interpolation model based on point cloud density or a lookup table based on curvature range. For example, for regions with sparse point clouds, the generated parameters will reduce the gradient sensitivity of edge detection; for regions with drastic curvature changes, the neighborhood search radius for keypoint detection will be increased accordingly.

[0140] Coordinate transformation typically involves translating and rotating the data from its original global coordinate system to a local coordinate system with the geometric center or feature point of the partition as its origin. This simplifies calculations and eliminates the influence of irrelevant global position offsets. Subsequently, resampling is performed. If the data is a point cloud, voxel rasterization may be used for downsampling to achieve density uniformity; if the data is a grid, grid simplification may be performed.

[0141] The data block generated in S1032, along with the initialization parameter set generated in S1031, is loaded into the computational kernel of the selected feature detection algorithm. Here, the algorithm kernel refers to the software module or hardware acceleration unit that implements the core mathematical model of the feature detection algorithm. The data block is traversed and computed based on the loaded threshold and scale parameters. For example, if the algorithm is 3D Canny edge detection, the gradient magnitude and direction of the data block are calculated, and non-maximum suppression is performed; if it is SIFT keypoint detection, a scale space is constructed and extreme points are detected. After the computation is complete, the kernel outputs an initial set of feature locations.

[0142] The initial feature location set output by S1033 undergoes post-processing. First, false detection removal is performed based on geometric constraints: constraint rules are set using the approximate continuous contour of the vehicle components. For example, the average distance between each feature point and its K nearest neighbors is calculated; if this distance is greater than the overall average, it is identified as an outlier and removed. Second, neighbor aggregation is performed: multiple feature points that are very close in spatial location are clustered into a representative point using a clustering algorithm. These two steps purify the data, generating a discrete feature point set for the outer contour, providing high-quality input for subsequent geometric fitting.

[0143] Preferably, the model repair unit actuator drive module includes:

[0144] The instruction parsing and allocation submodule is used to receive and parse the cleaning control instructions of the model repair unit, and decompose them into independent motion and action parameters corresponding to different cleaning actuators;

[0145] The multi-axis coordinated motion control submodule is used to generate multi-axis coordinated control signals to drive the servo motors in each cleaning actuator based on the motion and action parameters of the model repair unit.

[0146] The safety monitoring submodule is used to monitor the position and stress status of each cleaning actuator in real time, and trigger emergency intervention when it exceeds the preset safe working boundary or contact force threshold.

[0147] The dynamic adjustment submodule is used to collect real-time status data of the actuator and compare it with the target value of the instruction in order to adjust the cleaning process.

[0148] It should be noted that cleaning control commands refer to a structured set of commands generated by the data processing control module that describes the overall cleaning operation requirements. They typically include the expected sequence of actions, target trajectory, and target parameters of each actuator.

[0149] Motion and motion parameters refer to the specific control settings required to drive a single actuator after analytical decomposition, such as target position, target speed, target torque / force, and switching status.

[0150] Multi-axis collaborative control signals refer to the set of underlying drive electrical signals generated by the motion controller that are strictly corresponding in time in order to enable multiple servo motors to complete complex spatial trajectory movements synchronously and in coordination.

[0151] The safe working boundary refers to the limitation on the range of motion of the actuator in space set to protect the vehicle and equipment. It usually exists in the form of soft limit of position coordinates or dynamic safe distance envelope relative to the three-dimensional model of the vehicle.

[0152] The contact force threshold refers to the maximum permissible contact pressure value between the cleaning tool and the vehicle surface, set to ensure cleaning effectiveness without damaging the paint.

[0153] Emergency intervention refers to the highest-priority protective control action automatically triggered by a safety monitoring system when it detects a potential hazard, aimed at immediately stopping or correcting dangerous movements;

[0154] Real-time status data refers to data obtained by sensors in real time during the cleaning process, reflecting the actual operating status of the actuator, such as actual position, actual speed, and actual force.

[0155] The target value of an instruction refers to the state or output value that the system expects the actuator to achieve at a certain moment.

[0156] Specifically, the instruction parsing and allocation submodule receives cleaning control instructions from the data processing and control module. It parses the instructions to understand their overall intent and decomposes them into motion and action parameters corresponding to the actuators. For example, the instruction "clean the roof" is decomposed into the target motion trajectory of the large top brush, the target brush rotation speed, the target contact pressure, and the on / off timing and water pressure values ​​of the cooperating spray system.

[0157] The multi-axis coordinated motion control submodule receives motion and action parameters from upstream for each mechanism. Typically implemented by a multi-axis motion controller, its function is to perform multi-axis coordinated trajectory planning and real-time control. Based on the motion parameters of each mechanism, it calculates a smooth motion trajectory for multiple servo motors that is synchronized in time and coordinated in space. Subsequently, it generates corresponding multi-axis coordinated control signals to drive the servo motors in the cleaning actuators such as the large brush, side brush, and wheel brushes on the gantry, causing them to move the cleaning tools according to the planned trajectory and speed, ensuring that multiple mechanisms do not interfere with each other during complex cleaning operations.

[0158] Meanwhile, the safety monitoring submodule operates independently and in parallel. It uses high-precision encoders, six-dimensional force sensors, and other devices to monitor the position and stress state of each cleaning actuator in real time. It sets safe operating boundaries and contact force thresholds for different cleaning scenarios. If the monitored data exceeds these preset safety ranges—for example, if the side brush is about to collide with the rearview mirror or the pressure of the top brush increases abnormally—emergency intervention is immediately triggered, sending the highest priority stop, reversal, or decompression command to the motion control submodule to ensure the safety of the vehicle and equipment.

[0159] The dynamic adjustment submodule continuously collects real-time status data from each actuator, including the actual position, speed, and current of the motors, as well as the actual contact force of the force sensors. It compares this real-time data with the target values ​​of the commands issued by the command parsing and allocation submodule, calculating the position tracking error and pressure control error. Based on this deviation, it adjusts the control signals sent to the servo motors, thereby fine-tuning the cleaning process in real time. For example, when the actual contact force is detected to be less than the target value, the motor current is automatically increased to increase the pressure; when slight slippage of the track causes position tracking lag, a control signal is output in advance to compensate. This allows the entire cleaning process to resist interference, adapt to minor changes, and maintain stable operational performance.

[0160] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A precision measurement system for car washing, characterized in that, include: The laser scanning measurement module is used to acquire three-dimensional point cloud data of the vehicle to be cleaned; The data processing and control module is used to process the three-dimensional point cloud data to construct a three-dimensional model of the vehicle and identify its external contour features, and generate cleaning control instructions based on the identification results. An actuator drive module is used to receive the cleaning control command and clean the vehicle based on the cleaning control command; The cleaning control command controls at least the movement trajectory of the actuator drive module so that it matches the identified vehicle outline features.

2. The precision measurement system for car washing according to claim 1, characterized in that, The laser scanning measurement module includes: The laser emitting submodule is used to emit laser beams onto the vehicle surface; The scanning drive submodule is used to drive the laser emission submodule to move according to a predetermined scanning mode, so that the laser beam covers the area of ​​the vehicle to be scanned. A signal receiving submodule is used to receive laser signals reflected back from the surface of the vehicle; The data acquisition submodule is used to calculate distance information based on the laser signal and simultaneously combine it with the scanning angle information to generate three-dimensional point cloud data of the vehicle.

3. The precision measurement system for car washing according to claim 2, characterized in that, The data acquisition submodule includes: The signal processing unit is used to filter and amplify the laser electrical signal output by the signal receiving submodule; The distance calculation unit calculates the straight-line distance between the laser reflection point and the scanning origin based on the processed electrical signal and according to the preset distance measurement principle. The data fusion unit is connected to the distance calculation unit and the scanning drive submodule, respectively, and is used to receive the straight-line distance and real-time scanning angle information, and convert the two into three-dimensional coordinates in the same coordinate system; The point cloud generation unit, connected to the data fusion unit, is used to serialize and encapsulate the three-dimensional coordinates to generate a standard three-dimensional point cloud data stream.

4. The precision measurement system for car washing according to claim 1, characterized in that, The data processing control module includes: The point cloud registration submodule is used to denoise, filter, and process the coordinate system of the input raw 3D point cloud data to form a registered integrated point cloud. The 3D model reconstruction submodule reconstructs the 3D surface model of the vehicle based on the integrated point cloud through surface fitting or meshing algorithms. The feature recognition and extraction submodule is used to analyze the three-dimensional surface model and identify and extract the vehicle's external contour feature parameters. The instruction generation submodule is used to generate the cleaning control instruction based on the extracted shape contour feature parameters.

5. A precision measurement system for car washing according to claim 4, characterized in that, The 3D model reconstruction submodule includes: The cloud optimization unit is used to resample and remove outliers from the integrated point cloud to obtain an optimized point cloud. A geometric structure building unit is used to perform surface fitting or triangular meshing calculations on the optimized point cloud to generate an initial three-dimensional geometric model. The model repair unit is used to repair holes and smooth the surface of the initial three-dimensional geometric model, and output a complete three-dimensional surface model of the vehicle.

6. A precision measurement system for car washing according to claim 4, characterized in that, The feature recognition and extraction submodule includes: The model region segmentation unit, based on geometric features and curvature analysis, divides the three-dimensional surface model into model partitions corresponding to different parts of the vehicle; The feature geometry extraction unit is used to extract the geometric elements of the outline from each of the model partitions; The parameter calculation unit calculates the shape contour feature parameters based on the geometric elements; The parameter output unit is used to encapsulate and output the shape contour feature parameters according to a predefined format.

7. A precision measurement system for car washing according to claim 6, characterized in that, The specific operation steps of the feature geometry extraction unit include: S1. For each model partition, based on the type of vehicle component to which it belongs, select a feature detection algorithm to initially obtain a set of discrete feature points of the outer contour; S2. Based on the discrete feature point set, use a geometric prototype that matches the vehicle structure to perform fitting calculations to generate continuous geometric elements that describe the main body contour. S3. Analyze the spatial relationships between the continuous geometric elements corresponding to different model partitions, and calculate and record the intersecting, parallel, perpendicular or coplanar relationships between the continuous geometric elements; S4. Integrate the continuous geometric elements and their relationships in the geometric topology network to generate a set of geometric elements for the vehicle's external outline.

8. A precision measurement system for car washing according to claim 7, characterized in that, S1 specifically includes: S101. Obtain the semantic identifier of each model partition to determine the specific vehicle component type to which it belongs; S102. Based on a preset vehicle component type mapping database, match at least one feature detection algorithm corresponding to the current component type; S103. Initialize and configure the feature detection algorithm, and execute the algorithm to initially obtain the discrete feature point set of the outer contour.

9. A precision measurement system for car washing according to claim 8, characterized in that, S103 specifically includes: S1031. Based on the geometric feature data of the current model partition, calculate and generate the initialization parameter set of the feature detection algorithm; S1032. Perform coordinate transformation and resampling on the 3D point cloud or mesh data of the current model partition to generate data blocks that meet the input requirements of the feature detection algorithm; S1033. Load the data block and the initialization parameter set into the feature detection algorithm kernel for operation, and output the initial feature position set; S1034. Perform false detection elimination and neighbor point aggregation operations based on geometric constraints on the initial feature position set to generate the discrete feature point set of the outer contour.

10. A precision measurement system for car washing according to claim 9, characterized in that, The actuator drive module includes: The instruction parsing and allocation submodule is used to receive and parse the cleaning control instructions, and decompose them into independent motion and action parameters corresponding to different cleaning actuators; The multi-axis coordinated motion control submodule is used to generate multi-axis coordinated control signals to drive the servo motors in each cleaning actuator based on the motion and action parameters. The safety monitoring submodule is used to monitor the position and stress status of each cleaning actuator in real time, and trigger emergency intervention when it exceeds the preset safe working boundary or contact force threshold. The dynamic adjustment submodule is used to collect real-time status data of the actuator and compare it with the target value of the instruction in order to adjust the cleaning process.

Images