An artificial intelligence assisted system for detecting abnormal size and weight vehicles at high speed

Abstract

The invention relates to a detection system (1) for real-time detection of abnormally sized and weighted vehicles (A) travelling at high speed on motorways. The detection system (1) subject to the invention comprises the following in its most basic form: a plurality of display units; a LIDAR sensor unit (3); a license plate detection unit (4); an axle number determination unit (5); at least one loop detector (8); at least one quartz weight measuring unit (9); at least one data processing unit (11) that calculates the weight and total weight of the vehicles (A) per axle group; a control unit (12) that creates a 3D profile of the vehicles (A) and creates a class data for each vehicle (A) and transition data containing the data produced for the vehicles by matching the data produced by the units in question with the transition images of the vehicles (A); a central server system (13) that produces oversize violation information and weight violation information according to the transit data and violation limit rules, and at least one competent authority system (18) that directly imposes penal sanctions based on the violation information produced.

Description

[0001] AN ARTIFICIAL INTELLIGENCE ASSISTED SYSTEM FOR DETECTING ABNORMAL SIZE AND WEIGHT VEHICLES AT HIGH SPEED

[0002] Technical Field

[0003] The invention relates to a system for real-time detection of abnormally sized and weighted vehicles travelling at high speed on motorways for direct enforcement purposes.

[0004] Prior Art

[0005] One of the preferred methods for transporting cargo is the use of land vehicles such as trailer trucks, lorries and pickup trucks. However, when transporting cargo with such vehicles, these vehicles can often be overloaded. Oversized and loaded vehicles on the move are not only dangerous to road users, but are also responsible for much of the damage to roads and bridges. Therefore, various methods are applied to measure the weight of heavy tonnage vehicles on highways. One of these methods is the use of static weighing stations. However, static weighing stations cause traffic flow to slow down. Another method for measuring the weight of heavy tonnage vehicles is the use of weight measurement systems in motion, and some of these systems are described below.

[0006] In a document entitled ‘Piezo Quartz Crystal HSWIM Product Specification’ in the prior art, a system for determining the weight of vehicles travelling at high speed is described. In its most general form, the system consists of piezo quartz crystal sensors placed under the asphalt; loop detectors placed in the lanes; a large number of cameras, some of which are arranged to zoom in on the licence plate images of vehicles and whose operation is triggered by the activation of the loop detectors; and a field computer. Thanks to this system, information such as the direction (straight / reverse) in which the vehicles are travelling in the lane, vehicle class (according to axle range, cycle information), chassis height code, vehicle length, vehicle speed, asphalt temperature, number of axles, axle range, axle and / or wheel weight, vehicle type, violation status can be recorded. In addition, said system can be controlled remotely by connecting to the field computer via a remote connection.

[0007] In the prior art United States patent document with publication number US20220196459A1 , a real-time overload detection method based on Convolutional Neural Network (CNN) is described. In the mentioned invention, the CNN method and (YOLO)-V3 detection algorithm are used to determine the number of axles, number of wheels, number of axles and wheelbase of vehicles moving on the motorway, the maximum load of the vehicle is determined in the national vehicle load standard with the determined wheelbase, and the maximum load determined is compared with the actual load measured by the piezoelectric sensor positioned under the vehicle and overload detection is performed.

[0008] In the Chinese utility model document with publication number CN208140032U in the prior art, a detection system for automatic overload detection without stopping the vehicles is disclosed. The detection system comprises a vehicle gauge measuring system, a vehicle weight detection system, a control system and laser radar detection devices (LIDAR). In this detection system, all processes such as LIDAR sensors, axle group vehicle weight and vehicle model detection can be carried out automatically without any manual intervention. In the mentioned invention, the values measured by the first LIDAR and second LIDAR sensors are transferred to the 3D coordinate system and thus the length, width and height information of the vehicle is obtained. As the vehicles pass through the weight measurement platform, a weight measurement signal is generated and the control system calculates the number of axles, axle type, axle weight and vehicle speed based on the generated weight measurement signal. In the mentioned invention, the external dimensions of the vehicles are detected by the LIDAR sensors in the vehicle size measurement system, and the weight of the vehicle is detected by the weight sensors while the vehicle passes through the weight detection system, and the detected vehicle external dimensions and weight are transmitted to a control system. Thus, overload detection is carried out automatically without stopping the vehicles.

[0009] A dynamic vehicle weight measurement system is disclosed in the Chinese patent document numbered CN1 16242465A, which is in the prior art. In the weight measurement areas in the said document, images of vehicles passing through the said lanes are taken by means of a large number of imaging devices positioned on the motorway lanes, and the number plates and appearance of the vehicles (e.g. vehicle model, colour, make, etc.) are determined by using deep learning algorithms such as CNN. The vehicle appearance and licence plate identification results are then combined to determine whether the vehicle is speeding, the condition of the vehicle, whether it is overloaded and whether it is in breach of regulations.

[0010] The vast majority of the heavy tonnage vehicle in motion violation detection systems in the state of the art is gross weight, weight per axle and weight per axle group violation detection systems. There are no direct sanctions with real-time violation detection in most of these systems, and sanctions have to be applied manually. In addition, in a significant part of the systems that detect the weights of vehicles in motion, the category of vehicles subject to violation is determined according to the number of axles, axle groups and axle distances obtained with the WIM (Weigh in Motion) sensor placed on the road or determined using a 3D LIDAR profile. However, there are vehicle classes and types that are not covered at a high rate in vehicle categorisation using these methods. Again, since the existing systems use a nondynamic categorisation system, the rate of unclassified or underclassified vehicles is high. In addition, existing systems often use strain gauge sensors for weight detection in high-speed motion. However, strain gauge sensors have various limitations. Some of these limitations include less acceptable accuracy and inconsistency across a wide range of vehicle types, sensitivity to temperature and humidity in measurements, sensitivity to overload, and the inability to perform bidirectional measurements. In addition, the fact that weather conditions are not taken into account when determining the violation limit tolerances for weight violations in existing systems may cause false violations to be generated. Therefore, there is a need for a detection system that eliminates these problems in the state of the art, includes abnormally sized vehicle violation detection together with weight violation detection, uses a more comprehensive and dynamically updated vehicle classification system compared to existing systems, provides direct enforcement with real-time violation detection without the need for manual intervention, and eliminates the negative effects of weather conditions on weight measurement performance. Brief Description of the Invention

[0011] The purpose of the invention is to realise a detection system that enables real-time detection and direct enforcement of abnormally sized and weighted vehicles travelling at high speed on motorways.

[0012] The detection system defined in the first claim and the claims related to this claim to achieve the purpose of this invention comprises the following: a plurality of imaging units for capturing images of vehicles; at least one LIDAR (Laser Imaging Detection and Ranging) sensor system that enables the size of vehicles to be determined and point cloud data to be created; at least one license plate detection unit configured to detect license plates of vehicles by processing images received from the at least one imaging unit; at least one axle number detection unit configured to detect the number of axles on the vehicles by processing images received from at least one imaging unit; at least one loop detector adapted to generate a passing signal of vehicle passage when a vehicle passes over it; at least one quartz weight measurement unit containing quartz crystal sensors that enables the detection of weights per axle, axle distances, axle groups and wheelbase of vehicles; at least one data processing unit configured to determine the weights and total weight of the vehicles per axle group based on measurements obtained by the quartz weight measurement unit; a control unit configured to receive pass images, point cloud data, weights per axle, weights per axle group, total weights, axle distances, axle groups and wheelbase information for the vehicles, to create a three-dimensional profile of the vehicles according to the said point cloud data, and to match the said data with the images taken by the imaging units to create a pass data containing a class data and matched data for each vehicle; a central server system configured to determine the detailed vehicle category for each vehicle according to the 3D profile, class and axle distribution information contained in the pass data of the vehicles and to generate gauge violation information according to the predetermined violation limit rules of the category, class and axle distribution information of the passing vehicles and weight violation information according to the predetermined additional weight carrying permit limits and at least one competent authority system structured to directly impose penal sanctions based on the gauge and weight infringement information produced. Unlike the existing systems, the detection system of the invention not only detects gross weight, weight per axle and weight per axle group violations, but also detects abnormally sized vehicle violations together with these weight violations. In the inventive detection system, quartz piezoelectric sensors are used instead of strain gauge sensors for weight detection. Quartz piezoelectric sensors are more compact than strain gauges, easier to calibrate and place on the motorway, more durable and resistant to mechanical damage, more accurate at higher speeds and more consistent sensitivity. The detection system of the invention uses a more inclusive vehicle categorisation method than similar systems, in other words, it uses a categorisation method that covers vehicles at a very high rate by using the number of axles, axle distances and axle groups together with the classification obtained with the three-dimensional profile. In addition, the vehicle classification system can be dynamically updated by statistical learning with historical data and corrected vehicle classifications obtained by the detection system. This minimises the proportion of vehicle types that are not covered because they have not yet been detected. Unlike existing similar systems, the detection system of the invention dynamically adapts the weight limit error tolerances according to weather conditions. Thus, weight limit tolerances can be dynamically updated according to the capacity of temperatures to support heavy loads on the road surface. In the detection system of the invention, violation warning can be generated by determining the vehicle's gauge, number of axles, axle distribution parameters and weight anomalies. The inventive system can provide direct sanctions with real-time violation detection without the need for manual intervention. By effectively enforcing load and size limits, the inventive detection system helps to protect the infrastructure and improves traffic safety.

[0013] Detailed Description of the Invention

[0014] The detection system implemented to achieve the purpose of this invention is shown in the attached figure;

[0015] Figure 1 is the schematic view of an application of the detection system that is the subject of the invention.

[0016] The parts in the figure are enumerated one by one and the parts correspond to these numbers are given in the following.

[0017] 1. Detection system

[0018] 2. Console post 3. LIDAR sensor unit

[0019] 4. License plate detection unit

[0020] 5. Axle number detection unit

[0021] 6. Event camera

[0022] 7. Temperature measurement unit

[0023] 8. Loop detector

[0024] 9. Quartz weight measurement unit

[0025] 10. Field cabin

[0026] 11. Data processing unit

[0027] 12. Control unit

[0028] 13. Central server system

[0029] 14. Database

[0030] 15. Network video recording unit

[0031] 16. Operator interface unit

[0032] 17.Continuous learning and improvement unit

[0033] 18. Competent authority system

[0034] A. Vehicle

[0035] S. Highway station

[0036] A detection system (1 ) for real-time detection of abnormally sized and weighted vehicles (A) travelling at high speed on motorways, comprises the following; a plurality of imaging units adapted to take transit images of vehicles (A); at least one LIDAR sensor unit (3) adapted to detect the dimensions (width, length, height) of vehicles (A) and to generate point cloud data; at least one license plate detection unit (4) configured to detect the license plates of the vehicles (A) by processing their passing images; at least one axle number detection unit (5) configured to detect the axle number of the vehicles (A) by processing the passing images of the vehicles (A); at least one loop detector (8) adapted to generate a passing signal when a vehicle (A) passes over it. In one embodiment of the invention, a console post (2), preferably having a ‘LT shape, is provided on a highway station (S) on or close to at least one lane in which moving vehicles (A) subject to abnormal size and weight detection are to pass without stopping. However, the cantilever post (2) is not limited to this, but may also have an T form, an ‘L’ form or other forms. There are imaging units on the console post (2), such as a camera, that take video images of the vehicles (A) and / or imaging units, such as a camera, that allow taking photographs of the vehicles (A), some of the said imaging units are positioned on the same console post (2) or on a different console pole (2) in such a way that they capture the front view of vehicles (A) approaching the console post (2) and the rear view of vehicles (A) travelling away from the console post (2). Thus, bidirectional measurement and violation detection is possible in both flow directions of traffic in each lane. At least part of the LIDAR sensor unit (3), license plate detection unit (4) and axle number detection unit (5) are arranged on the same console post (2) or on different console posts (2). When a vehicle (A) moving in the direction of traffic flow on a motorway reaches the area of interest of the LIDAR sensor unit (3) comprising a plurality of LIDAR sensors arranged on the console post (2) and a processor unit, the LIDAR sensor unit (3) performs a measurement during the passage of the vehicle (A) through the area of interest in question and transmits a signal to one or more of the imaging units for taking photographs or video footage of the passage of the vehicle (A) in question. Thus, during the passage of the vehicle (A), front and rear images of the vehicle (A) are taken, the images are processed by the number plate detection unit (4) electrically connected to the imaging units, and the number plate of the vehicle (A) is identified from the processed images. While the vehicle (A) is in the region of interest, the images taken from the imaging units are also processed by the axle count detection unit (5) to determine the total number of axles that the vehicle (A) has. At least one loop detector (8) is positioned on at least one lane of the motorway, close to the console post (2), when the tires of a moving vehicle (A) land on the loop detector (8), a signal regarding the presence of the vehicle (A) is created by the loop detector (8) and sent to a control unit (12) over a cable or via a wireless data transmission unit (e.g. Wi-Fi, Bluetooth, GSM module, GPRS module, 4G, 4.5G, 5G module, etc.).

[0037] The detection system (1 ) subject to the invention comprises the following: at least one quartz weight measurement unit (9) comprising quartz crystal sensors to detect the weights per axle, axle distances, axle groups and wheelbase of the vehicles (A); at least one data processing unit (1 1 ) configured to calculate the weights and total weight of the vehicles (A) per axle group according to the data obtained by the quartz weight measurement unit (9); a control unit (12) configured to generate a 3D profile of the vehicles (A) according to said point cloud data and to match the data generated by said units with the passage images of the vehicles (A) to generate a class data for each vehicle (A) and a passage data containing the matched data; a central server system (13) configured to determine the detailed categories of vehicles (A) according to the 3D profile, class and axle distribution information contained in the passage data of vehicles (A) and to generate gauge violation information and weight violation information by comparing the category, class and axle distribution information and weight information of the passing vehicles (A) with the predetermined violation limit rules; at least one database (14) adapted to record the data obtained on vehicles (A) and at least one competent authority system (18) configured to directly impose penal sanctions based on the violation information generated. The quartz weight measurement unit (9) in the detection system (1 ) of the invention is used embedded in the highway ground. As the axles of the vehicle (A) pass over the quartz weight measurement unit (9) embedded in the motorway ground, the high-precision quartz crystal sensors in the quartz weight measurement unit (9) detect the weights per axle, axle distances, axle groups and wheelbase of the vehicle (A) and send the detected data to the data processing unit (1 1 ) via at least one cable or wireless data transmission unit. The data processing unit (11 ) calculates the weights per axle group and total weight of the vehicle (A) according to the data sent by the quartz weight measurement unit (9) and sends the weights per axle group and total weight information together with the weights per axle group, axle distances, axle groups (in other words, the number of axles in axle groups) and wheelbase information sent by the quartz weight measurement unit (9) to the control unit (12). The data processing unit (11 ) and the control unit (12) are preferably installed in a field cabin (10), so that they can operate in close proximity to field sensors and imaging units without being affected by outdoor conditions. The control unit (12) receives data on the passage of vehicles (A), i.e. photographs and / or video images of the passage of vehicles (A), the detected number plate information from the number plate detection unit (4), the axle information of vehicles (A) from the axle number detection unit (5), the point cloud data sent from the LIDAR sensor unit (3), and the axle information from the quartz weight measurement unit (9). Then, the control unit (12) creates the 3D profile of the vehicle (A) by using the point cloud data sent from the LIDAR sensor unit (3). In addition, the control unit (12) compares the axle numbers in the axle groups coming from the quartz weight measurement unit (9) with the axle numbers coming from the axle number determination unit (5). In addition, the control unit (12) fuses the data provided by these field sensors, i.e. the number plate data from the number plate detection unit (4), the axle information from the axle number detection unit (5) and the images provided by the display units, i.e. combines and correlates these data. The control unit (12) processes all the data it collects, generates transition data and class data containing the collected data for each vehicle (A), and creates a transition document containing these data. The control unit (12) sends the processed pass data, the pass document, the pass images and the 3D profile files, preferably in a single file, to the central server system (13). The central server system (13) opens the file from the control unit (12), processes the pass data and stores it in a database (14). During the processing of the pass data, the central server system (13) determines the detailed category information of the vehicles (A) from the vehicle (A) class and axle distribution information received from the control unit (12). The central server system (13) queries the violation rules configuration already stored in the database (14) using the category, class and axle distribution information of the passing vehicle (A). The central server system (13) also queries the additional weight transport authorisation limits already stored in the competent authority system (18) for the passing vehicle (A). As a result of these queries, the central server system (13) generates gauge (width, length, height) and weight (total weight, weight per axle, weight per axle group) violation information from the pass data according to the violation configuration limits for vehicle (A) and records the generated violation information in the database (14). The central server system (13) then sends the violation information on the vehicles (A) detected for direct enforcement, together with the photographs taken and / or a short video of the passage, via a wired data communication unit (e.g. Ethernet module) or a wireless data communication unit (GSM module, GPRS module, Wi-Fi module, etc.) to the competent authority system (18) via a data communication network such as the Internet or intranet. Thus, direct sanction can be provided by the competent authority system (18) after real-time gauge and weight violation detection without requiring manual intervention.

[0038] In one embodiment of the invention, the detection system (1 ) also comprises at least one operator interface unit (16) that allows users to enter data to define violation limit rules regarding vehicle (A) class, type, axle and all violations committed. The operator interface unit (16) works on an electronic device such as a computer that is connected to the central server system (13) for data transmission and provides monitoring and management of the detection system (1 ). The operator interface unit (16) is a user interface and shows a summary of the data obtained in the detection system (1 ) with an indicator panel displayed in this interface, provides the display and management of highway stations (S) on the map, lists all vehicle (A) passes and violations through highway stations (S), and contains statistics and graphs regarding such crossings and violations, enables the user to define the vehicle (A) class, type, axle and limit configurations for all detected violations manually through a user interface using a data entry unit such as a keyboard, mouse, touch screen, etc. Erroneous data (e.g. erroneous vehicle classification data) corrected by the user via the operator interface unit (16) are stored in the database (14) by the operator interface unit (16).

[0039] In one embodiment of the invention, the detection system (1 ) further comprises a continuous learning and improvement unit (17) configured to update the vehicle classification configuration and to identify unusual measurements and anomalies by statistical learning based on data relating to vehicles (A) stored in the database (14). In this embodiment of the invention, the user operator rearranges the incorrect vehicle classifications by listing the violations on the operator interface unit (16), in other words, the user operator corrects the incorrect vehicle classifications and enters the correct vehicle classification via the user interface and the corrected data is recorded in the database (14). The continuous learning and improvement unit (17) uses this historical data stored in the database (14) to update the vehicle classification configuration by statistically learning from the parameters (width, length, height, axle configuration, 3D profile and transition photographs) of the relevant vehicles (A).

[0040] In one embodiment of the invention, the continuous learning and improvement unit (17) is configured to update the vehicle classification configuration by weighting the output of the Random Forest learning algorithm to process the numerical data related to the vehicles (A), and the outputs of the Pointnet and Convolutional Neural Networks deep learning algorithms to process the image data at predetermined magnitudes. The updated vehicle classification configuration is recorded in the database (14) by the continuous learning and improvement unit (17). Thus, by using artificial intelligence during the classification of vehicles (A) by the detection system (1 ), errors that may arise regarding the classification of vehicles (A) are prevented. In one embodiment of the invention, the continuous learning and improvement unit (17) is configured to use the Mahalanobis Distance and Isolation Forest algorithms to identify unusual measurements and anomalies related to the gauge (width, length, height), number of axles, axle distribution parameters and weights of vehicles (A). In this embodiment of the invention, the data regarding the dimensions, number of axles and axle distributions of the vehicle (A) determined by the continuous learning and improvement unit (17) are taken into consideration and the Mahalanobis distance of the said data is calculated and possible unusual weight measurements are determined. An Isolation Forest model is also trained by the continuous learning and improvement unit (17) using weight, dimensions, number of axles and axle distributions, and among the weight, dimensions, number of axles and axle distribution information obtained for the isolated samples, i.e. for each vehicle (A), those that are contrary to the data in the trained Isolation Forest model are considered as abnormal measurements. The continuous learning and improvement unit (17) updates the outputs of both algorithms to the database (14) as unusual measurement records. Thus, unusual measurements and anomalies are detected with high accuracy using artificial intelligence support.

[0041] In one embodiment of the invention, the continuous learning and improvement unit (17) is configured to dynamically update weight limit error tolerances for use by the central server system (13) during the generation of violation information based on temperature data measured by a temperature measurement unit (7) using a Support Vector Regression algorithm. In one embodiment, the temperature measurement unit (7) comprises at least one temperature sensor. The continuous learning and improvement unit (17) uses a Support Vector Regression algorithm to create a dynamic weight limit tolerances model based on the temperature data obtained in the past, and the dynamic weight limit tolerances model updates the weight limit error tolerances to the database (14) based on real-time changes with the temperature information obtained by the temperature measurement unit (7). Thus, changes in the support capacity of heavy loads due to softening or tightening of the road surface due to temperature are taken into account, and therefore weight limit error tolerances are dynamically updated according to different weather conditions.

[0042] In one embodiment of the invention, the detection system (1 ) further comprises at least one event camera (6) for capturing video images of vehicles (A) and at least one network video recording unit (15) adapted to record video images captured by the event camera (6). The central server system (13) uses the video images stored in the network video recording unit (15) to generate a short passage video of the vehicle (A) passing through a region of interest, and stores the images of the vehicle (A) passing and the passage video in the database (14). Thus, the created short pass video is transmitted to the competent authority system (18) and the said pass video is used by the competent authority system (18) to impose criminal sanctions.

[0043] The number plate detection unit (4), the axle number detection unit (5), the data processing unit (1 1 ) and the control unit (12) mentioned in the above-described embodiments of the invention may be a microprocessor, a microcontroller, a field- programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a digital signal processor (DSP), or a processing unit such as a microprocessor, a microcontroller, a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a digital signal processor (DSP) or an electronic card comprising such processing unit or an electronic device comprising such electronic card. Again, in mentioned the applications, the central server system (13) and the competent authority system (18) are systems that comprise one or more servers and are adapted to establish wired or wireless data communication with each other and / or with the other units mentioned herein over an internet network.

[0044] Thanks to the detection system (1 ) subject to the invention, abnormally sized and weighted vehicles (A) traveling at high speed on highways can be detected in real time, violation cases can be determined and thus direct sanctions can be applied. In addition, thanks to the detection system (1 ) of the invention, the size (gauge) and weight violations of the vehicles (A) travelling in the normal traffic flow on the motorway can be detected in real time with high reliability supported by artificial intelligence and a violation sanction can be created accordingly. Instead of static weighing stations for heavy tonnage vehicles (A) on motorways, such size and weight violation detection operations can be carried out at high speed without slowing down the traffic flow and without stopping the vehicles (A) to be detected. Thus, the cost for violation detection and direct enforcement is reduced, traffic jam is prevented by providing fast passage measurement, and detection efficiency is increased and revenue loss due to nondetection of violations can be reduced. During the determination of the violation limit tolerances for weight violations by the inventive detection system (1 ), weather conditions, which have a negative effect on the road surface and thus on the weight measurement performance, are also taken into account, so that the weight limit error tolerances are dynamically adapted according to the weather conditions, thus preventing the generation of false violations. In addition, in the detection system (1 ) subject to the invention, thanks to the effective application of loading and size limits, the infrastructure is protected, the road life is extended and traffic safety is increased.

Claims

CLAIMS1. A detection system (1 ) that comprising; a plurality of imaging units arranged in such a way as to receive images of the passage of vehicles (A), allowing real-time detection of abnormally sized and weighted vehicles (A) travelling at high speed on motorways; at least one LIDAR sensor unit (3) adapted to detect the dimensions of the vehicles (A) and generate point cloud data; at least one license plate detection unit (4) configured to detect the license plates of the vehicles (A) from the passing images; at least one axle number detection unit (5) configured to detect the axle number of vehicles (A) from passing images; at least one loop detector (8) adapted to generate a passing signal when a vehicle (A) passes over it, and characterized by comprising; at least one quartz weight measurement unit (9) comprising quartz crystal sensors to detect the weights per axle, axle distances, axle groups and wheelbase of the vehicles (A); at least one data processing unit (1 1 ) configured to calculate the weights and total weight of the vehicles (A) per axle group according to the data obtained by the quartz weight measurement unit (9); a control unit (12) configured to generate a 3D profile of the vehicles (A) according to said point cloud data, and to match the data generated by said units with the passage images of the vehicles (A) to generate a class data for each vehicle (A) and a passage data containing the matched data; a central server system configured to determine the categories of vehicles according to the 3D profile, class and axle distribution information contained in the pass data of vehicles (A) and to generate gauge violation information and weight violation information by comparing the category, class and axle distribution and weight information of the passing vehicles (A) with the predetermined violation limit rules (13); at least one database (14) adapted to record data obtained on vehicles (A) and at least one competent authority system (18) configured to directly impose penal sanctions based on the infringement information generated.

2. The detection system (1 ) according to claim 1 , characterized by comprising at least one operator interface unit (16) for enabling users to input data for defining violation limit rules for vehicle (A) class, type, axle and all violations committed.

3. The detection system (1 ) according to claim 1 or 2, characterized by comprising a continuous learning and improvement unit (17) configured to update vehicle classification rules and to identify unusual measurements and anomalies by statistical learning based on data relating to vehicles (A) recorded in the database (14).

4. The detection system (1 ) according to claim 3, characterized by comprising a continuous learning and improvement unit (17) configured to update the vehicle classification configuration by weighting, at predetermined magnitudes, the output of a Random Forest learning algorithm for processing numerical data relating to vehicles (A), and the outputs of the Pointnet and Convolutional Neural Network deep learning algorithms for processing image data.

5. The detection system (1 ) according to claim 3 or 4, characterized by comprising a continuous learning and improvement unit (17) configured to identify unusual measurements and anomalies related to the gauge, number of axles, axle distribution parameters and weights of vehicles (A) using Mahalanobis Distance and Isolation Forest algorithms.

6. The detection system (1 ) according to any one of claims 3 to 5, characterized by comprising a continuous learning and improvement unit (17) configured to dynamically update weight limit error tolerances for use by the central server system (13) during the generation of violation information, based on temperature data measured by a temperature measurement unit (7) using a Support Vector Regression algorithm.

7. The detection system (1 ) according to any one of the preceding claims, characterized by comprising at least one event camera (6) adapted to capture video images of vehicles (A) and at least one network video recording unit (15) adapted to record video images captured by the event camera (6).

Images