Vehicle overrun type detection method and system based on wagon balance vibration and point cloud data

Abstract

The invention provides a vehicle over-limit type detection method and system based on wagon balance vibration and point cloud data, relates to the technical field of intelligent traffic detection, and aims to solve the problems that in the prior art, misclassification is caused by dependence on single data, the over-limit types of a whole loading vehicle and a bulk vehicle are difficult to effectively distinguish, an over-limit management strategy is lack of pertinence and effectiveness, and the detection accuracy is poor. And real vibration characteristics and the like generated in the lockage process of the vehicle are difficult to accurately separate. According to the method, a wagon balance vibration signal and original vehicle point cloud data are obtained, processing and feature extraction are carried out, vibration features and point cloud geometric features are fused through an SVM classifier, a vehicle type and a loading state are obtained, and judgment is carried out based on a result. According to the method, the problems existing in the prior art are solved, misclassification caused by traditional single data detection is effectively avoided, the overrun types of the whole loading vehicle and the bulk vehicle can be effectively distinguished, high robustness can be still kept in an extreme environment, and misjudgment and missing detection are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent traffic detection, and in particular relates to a method and system for detecting vehicle overload types based on weighbridge vibration and point cloud data. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Oversized transport vehicles pose a significant threat to road safety and the structural integrity of infrastructure. Oversized vehicles can easily lead to traffic accidents, resulting in casualties. Due to their excessive size, they can obstruct other vehicles' visibility, leading to accidents and even multi-vehicle chain accidents. Furthermore, oversized vehicles can cause structural damage to infrastructure such as roads and bridges, increasing maintenance costs and posing safety risks to public facilities.

[0004] However, traditional vehicle overload detection methods typically rely on single sensor technologies such as scales or lidar. Scales can only capture the static weight of a vehicle and are unable to detect vehicle dimensions, cargo distribution, or loading status, making them prone to misjudgment due to center of gravity shift. LiDAR and cameras are susceptible to interference from environmental factors such as rain, fog, and dust, leading to deviations in vehicle contour extraction, impacting measurement accuracy and detection reliability. Furthermore, multimodal pattern recognition still faces challenges such as large training sample requirements, high computational resource consumption, and difficulty determining load type. Furthermore, existing technologies struggle to effectively distinguish between overload types in fully loaded and unloaded vehicles, resulting in a lack of targeted and effective overload management strategies.

[0005] In addition, although vibration sensing technology has significant advantages such as anti-interference and high sensitivity, existing vibration signal processing methods, such as fixed-threshold wavelet decomposition, can often only extract relatively rough vibration features, making it difficult to accurately separate the true vibration features generated by the vehicle during the gate passing process. Moreover, since it is difficult to extract vehicle size features, it cannot be directly and independently applied to the detection of overload types of large transport vehicles. Summary of the Invention

[0006] To overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for detecting vehicle overload types based on scale vibration and point cloud data. By integrating vibration energy characteristics with point cloud geometric features and using a weighted multimodal classifier, the low-frequency energy proportion of the vibration signal, time domain entropy, and the surface area-to-volume ratio and surface smoothness of the point cloud are organically combined. This system can not only accurately distinguish between pre-loaded and unloaded vehicles, but also identify the vehicle's load status, effectively avoiding the misclassification problem caused by the one-sided information of traditional single sensors, thereby providing a more fine-grained decision-making basis for overload management.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention discloses a method for detecting vehicle overload types based on weighbridge vibration and point cloud data, comprising: Obtaining scale vibration signals and original vehicle point cloud data; The dynamic characteristic frequency bands of the vibration information of the scale are separated by improving the wavelet packet transform; The energy entropy threshold denoising algorithm is used to filter low entropy subbands from the separated characteristic frequency bands to obtain the denoised vibration signal. Perform feature extraction on the noise-reduced vibration signal to obtain the actual vibration characteristics of the vehicle; The original vehicle point cloud data is processed to remove background and noise, and the processed moving vehicle point cloud is clustered to obtain vehicle point cloud clusters; Extracting point cloud geometric features from vehicle point cloud clusters; The actual vibration characteristics and point cloud geometric features of the vehicle are transmitted to the weighted classifier, and weighted fusion is performed through the weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

[0008] As a further technical solution, the calculation formula of the vibration signal after noise reduction is: ; Constraints ; in, is the vibration signal after noise reduction, is the wavelet coefficient of the kth subband of the jth layer, is the adaptive wavelet basis function, The energy entropy of the j-th layer subband complexity.

[0009] As a further technical solution, feature extraction is performed on the noise-reduced vibration signal to obtain the actual vibration characteristics of the vehicle. The specific process is as follows: Perform short-time Fourier transform on the denoised vibration signal to obtain the time-frequency energy density: Based on the time-frequency energy density and the denoised vibration signal, the actual vibration characteristics of the vehicle are obtained, including the low-frequency energy ratio and time domain entropy.

[0010] As a further technical solution, the original vehicle point cloud data is processed to remove background and noise. The specific process is as follows: Dynamic background modeling is performed on the original vehicle point cloud data, and the background model is updated through exponential weighting to separate the moving vehicle point cloud in real time; Radius filtering is used to remove floating noise from the separated moving vehicle point cloud.

[0011] As a further technical solution, the processed moving vehicle point cloud is clustered to obtain vehicle point cloud clusters. Specifically, the processed moving vehicle point cloud is merged with adjacent small clusters through the improved DBSCAN algorithm to obtain vehicle point cloud clusters.

[0012] As a further technical solution, the point cloud geometric features include the surface area to volume ratio of the vehicle point cloud and the surface smoothness of the vehicle point cloud, which are calculated as follows: ; ; in, is the surface area to volume ratio of the vehicle point cloud, is the surface smoothness of the vehicle point cloud, is the surface area of ​​the clustered vehicle V, is the volume of the clustered vehicle V, is the ideal normal vector of a plane of vehicle V after clustering, The average value of the actual normal vectors of the matching planes of the vehicle V after clustering, and M is the number of planes.

[0013] As a further technical solution, weighted fusion is performed through weighted kernel function to obtain vehicle type and load status. The specific process is as follows: The vehicle type decision value is obtained by weighted fusion of the actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud through the weighted kernel function; The vehicle type and load status are determined according to the vehicle type decision value.

[0014] The second aspect of the present invention discloses a vehicle overload type detection system based on weighbridge vibration and point cloud data, comprising: Acquisition module, used to obtain the vibration signal of the scale and the original vehicle point cloud data; The vibration feature extraction module is used to separate the characteristic frequency bands of the scale vibration information through improved wavelet packet transform; an energy entropy threshold noise reduction algorithm is used to filter low-entropy sub-bands from the separated characteristic frequency bands to obtain the noise-reduced vibration signal; and feature extraction is performed on the noise-reduced vibration signal to obtain the actual vibration characteristics of the vehicle; The point cloud geometric feature extraction module is used to filter the background and noise of the original vehicle point cloud data, cluster the processed moving vehicle point cloud to obtain vehicle point cloud clusters, and extract point cloud geometric features from the vehicle point cloud clusters; The over-limit detection and type determination module is used to transmit the actual vibration characteristics and point cloud geometric features of the vehicle to the weighted classifier, and perform weighted fusion through the weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

[0015] The third aspect of the present invention discloses a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the program.

[0016] The fourth aspect of the present invention aims to provide a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method described in the first aspect of the present invention.

[0017] One or more of the above technical solutions have the following beneficial effects: In this embodiment, through the improved wavelet packet transform technology, it is possible to adaptively focus on the vibration frequency band unique to the vehicle, effectively stripping off the high-frequency noise in the environment, greatly improving the signal-to-noise ratio of the vibration signal, and thus more accurately reflecting the actual operating status of the vehicle. Combined with the energy entropy threshold denoising algorithm, it is possible to intelligently screen out low-entropy sub-bands that reflect real vehicle vibrations, avoiding the signal distortion problem caused by noise interference in traditional methods. In terms of lidar point cloud processing, dynamic background modeling technology separates moving vehicles from static environments in real time, and combines multi-scale statistical filtering with optimized clustering algorithms to significantly enhance the completeness of vehicle contour extraction, ensure high accuracy of dimensional measurement, and maintain stable detection performance even under complex and changeable weather conditions.

[0018] This embodiment utilizes a dynamic background update mechanism and an adaptive parameter adjustment strategy. For example, in rainy and snowy conditions, dynamic background modeling is used to rapidly update the background point cloud using an exponentially weighted model, effectively filtering out noise introduced by raindrops and snowflakes. In strong winds, an improved wavelet packet transform automatically adjusts the decomposition bandwidth to lock onto the dominant vehicle vibration frequency, preventing wind-induced noise from interfering with signal analysis. This adaptive processing approach enables the system to maintain high robustness even in extreme environments, reducing false positives and missed detections.

[0019] In this embodiment, a weighted multimodal SVM classifier is used to fuse vibration energy features with point cloud geometric features. The low-frequency energy proportion and time-domain entropy of the vibration signal are organically combined with the surface area-to-volume ratio and surface smoothness of the point cloud to form a multi-dimensional feature fusion strategy. This strategy not only accurately distinguishes between packaged and bulk vehicles, but also identifies the vehicle's load status (e.g., packaged solids versus bulk solids), providing a more granular decision-making basis for overload management. For example, for bulk vehicles, the system can determine the looseness of the cargo based on the time-domain entropy of the vibration signal, further verified by combining it with the surface smoothness of the point cloud. This effectively avoids the misclassification problem caused by the one-sided information of traditional single sensors.

[0020] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0022] Figure 1 This is a flow chart of a method for detecting vehicle overload types based on weighbridge vibration and point cloud data according to a first embodiment of the present invention. DETAILED DESCRIPTION

[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0024] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention.

[0025] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0026] Explanation of terms Wavelet basis function: The core function in wavelet transform, used to decompose and reconstruct signals. It meets certain mathematical conditions, such as orthogonality and normalization. By translating and scaling the wavelet basis function, a series of wavelet functions can be generated to analyze the characteristics of the signal at different scales and locations.

[0027] Dynamic characteristic frequency bands: Dynamic characteristic frequency bands refer to the frequency ranges in which signals are primarily concentrated during a specific dynamic process. In signal processing, different dynamic processes often exhibit significant characteristics within specific frequency ranges. By identifying and analyzing these dynamic characteristic frequency bands, useful information can be extracted from the signal for subsequent feature analysis and classification.

[0028] Energy entropy: Energy entropy is a measure of the complexity of a signal's energy distribution. It quantifies the distribution of signal energy across different subbands or frequency components, reflecting the degree of energy dispersion. A higher energy entropy indicates a more complex and dispersed signal energy distribution; a lower energy entropy indicates that the signal's energy is concentrated in a few subbands or frequencies.

[0029] Support Vector Machine (SVM) is a supervised learning algorithm based on statistical learning theory, primarily used for classification and regression analysis. Its basic idea is to find an optimal hyperplane in the feature space that maximizes the margin between different categories, thereby achieving data classification.

[0030] Example 1 This embodiment discloses a method for detecting vehicle overload types based on weighbridge vibration and point cloud data.

[0031] To more clearly illustrate this embodiment, the vehicle overload type detection process based on scale vibration and point cloud data can be specifically described as follows: like Figure 1 As shown, this embodiment provides a method for detecting vehicle overload types based on weighbridge vibration and point cloud data, including: S1. Obtaining the vibration signal of the weighbridge and the original vehicle point cloud data; S2, separating the characteristic frequency bands of the vibration information of the scale by improving the wavelet packet transform; An energy entropy threshold denoising algorithm is used to filter low-entropy subbands from the separated characteristic frequency bands to obtain a denoised vibration signal. Feature extraction is performed on the denoised vibration signal to obtain the actual vibration characteristics of the vehicle. S3. Perform background and noise filtering on the original vehicle point cloud data, and cluster the processed moving vehicle point cloud to obtain vehicle point cloud clusters; extract point cloud geometric features from the vehicle point cloud clusters; S4, transmitting the actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud to the weighted classifier, performing weighted fusion through the weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

[0032] like Figure 1 As shown, in step S1, the vibration signal of the weighing scale and the original vehicle point cloud data are obtained.

[0033] S1-1. Obtain vibration signal of the weighing scale.

[0034] In this embodiment, a piezoelectric high-sensitivity floor scale sensor is used, the vibration signal sampling frequency is 2 kHz, and the measuring range is 0-100 tons.

[0035] Specifically, when a vehicle enters the weighing scale, the sensor array continuously collects vibration signals at a sampling rate of 2kHz, and the raw data is cached in the edge computing unit.

[0036] S1-2. Obtain original vehicle point cloud data.

[0037] In this embodiment, a 16-line mechanical radar is used with a horizontal field of view of 120°, a vertical resolution of 0.1°, and a maximum detection range of 150m. It is equipped with a GPU accelerator (NVIDIA Jetson AGX Xavier) to enable parallel computing of signal processing and classification algorithms.

[0038] Specifically, the lidar outputs point cloud data at a frequency of 10 Hz.

[0039] After the above steps, the point cloud data obtained can accurately monitor the vehicle envelope, and the vibration data obtained can accurately monitor whether the vehicle is fully loaded.

[0040] like Figure 1 As shown, in step S2, dynamic characteristic frequency band separation is performed on the vibration information of the weighing scale by improving the wavelet packet transform; The energy entropy threshold denoising algorithm is used to screen low entropy subbands from the separated characteristic frequency bands to obtain the denoised vibration signal; the feature extraction of the denoised vibration signal is performed to obtain the actual vibration characteristics of the vehicle.

[0041] S2-1. Dynamic characteristic frequency band separation of scale vibration information is performed through improved wavelet packet transform.

[0042] In this embodiment, when the vehicle is unloaded, the scale and the vehicle itself vibrate, and the traditional wavelet packet decomposition uses a fixed bandwidth. Therefore, the improved wavelet packet transform (MWPT) is adopted to dynamically adjust the wavelet basis bandwidth, focus on the characteristic frequency band of vehicle vibration, and suppress high-frequency noise. Specifically, the wavelet packet transform (MWPT) decomposition is improved. By dynamically adjusting the bandwidth of the wavelet basis function, the decomposition process is focused on the characteristic frequency band of vehicle vibration (usually 5-50 Hz), suppressing the high-frequency components of environmental noise (such as wind vibration and mechanical interference). The formula is: ; ; ; ; in, is the original vibration signal of the weighbridge (time domain), is the decomposition scale (level), is the subband position index, For the Tier The wavelet coefficients of the subbands, is the adaptive wavelet basis function, For the Adaptive bandwidth, is the bandwidth adjustment factor, The dominant frequency with the highest energy in the signal spectrum.

[0043] After the above steps, the wavelet packet transform (MWPT) dynamically adjusts the bandwidth according to the signal spectrum, so that the decomposition result is more consistent with the actual vibration characteristics of the vehicle, laying the foundation for noise reduction.

[0044] S2-2. Use the energy entropy threshold denoising algorithm to filter the low entropy sub-band from the separated characteristic frequency band to obtain the denoised vibration signal.

[0045] In this embodiment, traditional noise reduction methods (such as wavelet threshold denoising) have limited effects on complex noise. Therefore, energy entropy threshold denoising is introduced to screen low entropy subbands (reflecting the actual vibration of the vehicle) and improve the signal-to-noise ratio.

[0046] Specifically, the energy entropy threshold denoising algorithm is used to retain the real vibration component of the vehicle by screening low entropy subbands (noise has high randomness and large entropy value). The formula is: ; ; ; Constraints: ; in, is the vibration signal after noise reduction, The energy entropy of the j-th layer subband complexity, is the energy proportion of the kth coefficient in the subband.

[0047] After the above steps, the energy entropy mechanism can distinguish regular vibrations from random noise, avoiding the dilemma of "signal distortion" and "noise residual".

[0048] S2-3. Extract features from the noise-reduced vibration signal to obtain actual vibration features of the vehicle.

[0049] In this embodiment, short-time Fourier transform is performed on the noise-reduced vibration signal to obtain time-frequency energy density.

[0050] Perform short-time Fourier transform on the noise-reduced vibration signal to extract the 5-20Hz low-frequency energy ratio ( ); Calculate the time domain entropy ( ), which quantifies the non-stationarity of the signal and reflects the looseness of the cargo.

[0051] Based on the time-frequency energy density and the denoised vibration signal, the actual vibration characteristics of the vehicle are obtained, including the low-frequency energy ratio and time domain entropy.

[0052] Specifically, vibration energy feature extraction is performed by obtaining time-frequency energy density through short-time Fourier transform, and extracting low-frequency energy ratio and time domain entropy as energy features. The formula is: ; ; ; in, is the time-frequency energy density, that is, the frequency Energy at time window τ; The time resolution is 0.5s, which can usually capture the typical vehicle vibration cycle; It is the proportion of low-frequency (usually 5-20Hz) energy, reflecting the looseness of the cargo; is the time domain entropy, which quantifies the non-stationarity of the vibration signal. Bulk trucks are more random due to the shaking of the cargo; is the window function; is the discrete vibration signal after noise reduction; N is the total number of samples of the discrete signal; is the total vibration energy of the denoised signal, i.e., the absolute value integral.

[0053] After the above steps, the feature extraction and noise reduction steps form a closed loop, preventing traditional methods from losing key information (such as the shaking characteristics of bulk vehicles) due to over-smoothing. The two features complement each other in terms of frequency domain distribution and time domain randomness, significantly improving the accuracy of subsequent classifiers in distinguishing between packaged and bulk vehicles, solving the problems of large computational complexity and high latency in traditional multimodal models (such as CNN fusion), and meeting the real-time detection needs of toll stations.

[0054] like Figure 1As shown, in step S3, the original vehicle point cloud data is subjected to background and noise filtering, and the processed moving vehicle point cloud is clustered to obtain a vehicle point cloud cluster; Extract point cloud geometric features from vehicle point cloud clusters.

[0055] In this embodiment, exponentially weighted dynamic background modeling is used to separate the moving vehicle point cloud from the static background in real time; multi-scale statistical filtering and improved DBSCAN clustering are combined to eliminate floating noise points and forcefully merge adjacent small clusters to ensure the integrity of the vehicle point cloud. The specific process is as follows: S3-1. Perform background and noise filtering on the original vehicle point cloud data.

[0056] (1) Dynamic background modeling is performed on the original vehicle point cloud data, and the background model is updated through exponential weighting to separate the moving vehicle point cloud in real time.

[0057] In this embodiment, a dynamic background model is constructed in real time, and the background model is updated by exponential weighting to separate the moving vehicle point cloud in real time. The formula is: ; Constraints: ; in, is the vehicle point cloud at time t; is the original point cloud at time t-1; is the original point cloud at time t; is the background update rate, the larger the value, the more stable the background; is any point at time t; It is the minimum motion threshold, which is determined flexibly according to the on-site speed limit.

[0058] After the above steps, the vehicle can be separated from the dynamic environment (rain, snow, wind and sand).

[0059] (2) Remove floating noise points from the separated moving vehicle point cloud by radius filtering.

[0060] In this embodiment, multi-scale statistical filtering is performed on the separated point cloud to remove isolated noise clusters with a radius less than 0.1m.

[0061] Multi-scale statistical filtering uses radius filtering to remove floating noise points (such as dust and raindrops). The formula is: ; in, is the set of valid point clouds, is any point after background filtering at time t, is the value after background filtering at time t Any point outside.

[0062] After the above steps, local noise (dust / raindrop reflections) can be removed.

[0063] S3-2. Cluster the processed moving vehicle point cloud to obtain vehicle point cloud clusters.

[0064] In this embodiment, the improved DBSCAN algorithm is used to cluster the remaining point clouds, and small clusters with a spacing less than 0.3m are forced to be merged to avoid vehicle contour breaks caused by occlusion.

[0065] Specifically, the improved DBSCAN algorithm forces the merging of adjacent small clusters to ensure the integrity of the vehicle point cloud. The objective function is: ; Constraints: ; in, For clustering The maximum point spacing, is the minimum distance between two clusters, It is a compact factor and needs to be determined flexibly according to the weather conditions on site. is an arbitrary point used for clustering.

[0066] After the above steps, a continuous and complete point cloud cluster can be output to ensure the reliability of subsequent geometric features and over-limit judgment.

[0067] S3-3. Extract point cloud geometric features from vehicle point cloud cluster.

[0068] In this embodiment, based on the clustered vehicle point cloud, the surface area volume ratio ( ), the ratio of packaged vehicles is low, and the ratio of bulk vehicles is high; the surface smoothness of the vehicle point cloud is calculated by surface normal vector analysis ( ), identifying surface irregularities in bulk truck cargo.

[0069] Through the geometric information after clustering, the surface volume ratio and surface smoothness are extracted as geometric features. Among them, the point cloud geometric features include the surface area volume ratio of the vehicle point cloud and the surface smoothness of the vehicle point cloud. The formula is: ; ; in, is the surface area to volume ratio of the vehicle point cloud, is the surface smoothness of the vehicle point cloud, is the surface area of ​​the clustered vehicle V, is the volume of the clustered vehicle V, is the ideal normal vector of a plane of vehicle V after clustering, The average value of the actual normal vectors of the matching planes of the vehicle V after clustering, and M is the number of planes.

[0070] After the above steps, it can be determined whether it is a fully loaded or bulk truck based on the roughness.

[0071] like Figure 1 As shown, in step S4, the actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud are transmitted to the weighted classifier, and weighted fusion is performed through the weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

[0072] S4-1. The actual vibration characteristics and point cloud geometric characteristics of the vehicle are transmitted to the weighted classifier, and weighted fusion is performed through the weighted kernel function to obtain the vehicle type and load status.

[0073] In this embodiment, the proportion of low-frequency vibration energy and time domain entropy are extracted to quantify the looseness of the cargo; the surface area-to-volume ratio and surface smoothness of the point cloud are extracted to distinguish between fully loaded and loose vehicles; and a weighted multimodal SVM classifier is designed to integrate vibration and geometric features to solve the problem of misjudgment by a single sensor.

[0074] Specifically, 、 、 and After normalization, the data is input into the SVM classifier. The specific process is as follows: (1) The actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud are weightedly fused through the weighted kernel function to obtain the vehicle type decision value.

[0075] In this embodiment, a weighted kernel function is used to fuse vibration and geometric features to solve the problem of single sensor misjudgment (for example, an empty vehicle may be misjudged as a fully loaded vehicle). The decision function formula is: ; in, , is the baseline vector in the no-load state (set by vehicle type), The kernel coefficient needs to be set according to the sensitivity requirements.

[0076] (2) Determine the vehicle type and load status based on the vehicle type decision value.

[0077] In this embodiment, according to The value of determines the vehicle type, fully loaded < fully loaded < loose fully loaded < loose loose loaded < fully loaded < loose loose loaded. The specific threshold value depends on the situation.

[0078] That is, the classifier outputs the vehicle type (fully loaded / bulk) and load status (empty / half loaded / fully loaded) according to the preset kernel function weights.

[0079] Through the above steps, it can be determined whether the vehicle is not fully loaded but the envelope exceeds the limit.

[0080] S4-2. Based on the vehicle type and load status, the vehicle size is determined from the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If so, the vehicle is directly released. If not, it is determined whether there is room for adjustment. If so, adjustments are made and the judgment is repeated. If not, penalties are imposed in accordance with relevant regulations.

[0081] (1) Calculate the length, width, and height of the vehicle based on the point cloud outline and compare them with specific standard limits.

[0082] In this embodiment, the vehicle size is determined by the point cloud cluster outline, and compared with the specification to determine whether it exceeds the limit. The formula is: ; ; ; Where L, W, and H are the length, width, and height of the vehicle, respectively. Comparing these with the inspection specifications will determine whether the vehicle exceeds the limit.

[0083] (2) Determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If so, the vehicle will be allowed to pass directly. If not, determine whether there is room for adjustment. If so, make adjustments and determine again. If not, impose penalties in accordance with relevant regulations.

[0084] In this embodiment, vehicles that meet the requirements are directly released; vehicles that do not meet the requirements and have no adjustment space are punished according to relevant regulations; vehicles that do not meet the requirements but have adjustment space are guided to a safe area for adjustment and wait for retesting.

[0085] Specifically, according to the result of step S4-1, determine whether the vehicle is fully loaded with fully loaded goods at the same time. If so, handle it according to relevant specifications; determine whether the vehicle is fully loaded with loose goods at the same time. If so, manually check whether there is room for adjustment. Otherwise, handle it according to relevant specifications; if it does not meet the first two conditions, adjust and retest until the result is determined to be neither overloaded nor overloaded, and then release it.

[0086] Through the above steps, it can be determined whether the vehicle is not fully loaded but the envelope exceeds the limit.

[0087] Example 2 This embodiment discloses a vehicle overload type detection system based on weighbridge vibration and point cloud data, including: Acquisition module, used to obtain the vibration signal of the scale and the original vehicle point cloud data; The vibration feature extraction module is used to separate the characteristic frequency bands of the scale vibration information through improved wavelet packet transform; an energy entropy threshold noise reduction algorithm is used to filter low-entropy sub-bands from the separated characteristic frequency bands to obtain the noise-reduced vibration signal; and feature extraction is performed on the noise-reduced vibration signal to obtain the actual vibration characteristics of the vehicle; The point cloud geometric feature extraction module is used to filter the background and noise of the original vehicle point cloud data, cluster the processed moving vehicle point cloud to obtain vehicle point cloud clusters, and extract point cloud geometric features from the vehicle point cloud clusters; The over-limit detection and type determination module is used to transmit the actual vibration characteristics and point cloud geometric features of the vehicle to the weighted classifier, and perform weighted fusion through the weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

[0088] Based on providing a vehicle overload type detection system based on weighbridge vibration and point cloud data, the method steps in Example 1 are implemented. Example 3 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.

[0089] Example 4 The purpose of this embodiment is to provide a computer-readable storage medium having a program stored thereon, which implements the steps in the method described above when the program is executed by a processor.

[0090] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0091] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A vehicle overload detection method based on scale vibration and point cloud data, characterized in that: include: Obtaining scale vibration signals and original vehicle point cloud data; Separating the characteristic frequency bands of the vibration information of the scale by improving the wavelet packet transform; The energy entropy threshold denoising algorithm is used to filter low entropy subbands from the separated characteristic frequency bands to obtain the denoised vibration signal. Perform feature extraction on the noise-reduced vibration signal to obtain the actual vibration characteristics of the vehicle; Performing background and noise filtering on the original vehicle point cloud data, and clustering the processed moving vehicle point cloud to obtain vehicle point cloud clusters; Extracting point cloud geometric features from the vehicle point cloud cluster; The actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud are transmitted to a weighted classifier, and weighted fusion is performed through a weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

2. The vehicle overrun type detection method based on weighbridge vibration and point cloud data according to claim 1, characterized in that: The calculation formula of the vibration signal after noise reduction is: ； Constraints ； in, is the vibration signal after noise reduction, is the wavelet coefficient of the kth subband of the jth layer, is the adaptive wavelet basis function, The energy entropy of the j-th layer subband complexity.

3. The vehicle overrun type detection method based on weighbridge vibration and point cloud data according to claim 1, characterized in that: Feature extraction is performed on the vibration signal after noise reduction to obtain the actual vibration characteristics of the vehicle. The specific process is as follows: Perform short-time Fourier transform on the noise-reduced vibration signal to obtain the time-frequency energy density: Based on the time-frequency energy density and the noise-reduced vibration signal, actual vibration characteristics of the vehicle are obtained, including a low-frequency energy ratio and a time-domain entropy.

4. The vehicle overrun type detection method based on weighbridge vibration and point cloud data according to claim 1, characterized in that: The original vehicle point cloud data is subjected to background and noise filtering, and the specific process is as follows: Dynamic background modeling is performed on the original vehicle point cloud data, and the background model is updated through exponential weighting to separate the moving vehicle point cloud in real time; Radius filtering is used to remove floating noise from the separated moving vehicle point cloud.

5. The vehicle overload type detection method based on weighbridge vibration and point cloud data according to claim 1, characterized in that: The processed moving vehicle point cloud is clustered to obtain vehicle point cloud clusters. Specifically, the processed moving vehicle point cloud is merged with adjacent small clusters through the improved DBSCAN algorithm to obtain vehicle point cloud clusters.

6. The vehicle overrun type detection method based on weighbridge vibration and point cloud data according to claim 1, characterized in that: The point cloud geometric features include the surface area to volume ratio of the vehicle point cloud and the surface smoothness of the vehicle point cloud, which are calculated as follows: ； ； in, is the surface area to volume ratio of the vehicle point cloud, is the surface smoothness of the vehicle point cloud, is the surface area of ​​the clustered vehicle V, is the volume of the clustered vehicle V, is the ideal normal vector of a plane of vehicle V after clustering, The average value of the actual normal vectors of the matching planes of the vehicle V after clustering, and M is the number of planes.

7. The vehicle overrun type detection method based on weighbridge vibration and point cloud data according to claim 6, characterized in that: The vehicle type and load status are obtained by weighted fusion through weighted kernel function. The specific process is as follows: Performing weighted fusion on the actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud through a weighted kernel function to obtain a vehicle type decision value; The vehicle type and load status are determined according to the vehicle type decision value.

8. The vehicle overload detection system based on scale vibration and point cloud data is characterized by: include: Acquisition module, used to obtain the vibration signal of the scale and the original vehicle point cloud data; A vibration feature extraction module, used for separating the characteristic frequency bands of the vibration information of the scale by improving the wavelet packet transform; An energy entropy threshold denoising algorithm is used to filter low-entropy subbands from the separated characteristic frequency bands to obtain a denoised vibration signal. Feature extraction is performed on the denoised vibration signal to obtain the actual vibration characteristics of the vehicle. a point cloud geometric feature extraction module, configured to perform background and noise filtering on the raw vehicle point cloud data, cluster the processed moving vehicle point clouds to obtain vehicle point cloud clusters, and extract point cloud geometric features from the vehicle point cloud clusters; An over-limit detection and type determination module is used to transmit the actual vibration characteristics of the vehicle and the geometric characteristics of the point cloud to a weighted classifier, and perform weighted fusion through a weighted kernel function to obtain the vehicle type and load status; Based on the vehicle type and load status, the vehicle size is determined according to the vehicle point cloud cluster and compared with the standard threshold to determine whether the vehicle meets the traffic requirements and whether there is room for adjustment. If it meets the requirements, it will be released directly. If not, it will be determined whether there is room for adjustment. If so, adjustments will be made and the judgment will be made again. If not, penalties will be imposed in accordance with relevant regulations.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

Images