Road parking detection system and method with sentinel mode

Abstract

The present invention discloses a road parking detection system and method with a sentinel mode, which relates to the field of intelligent transportation technology. The system includes: a detection module, which is used to detect whether there is an object approaching or moving away from the detection device, and wake up the remaining modules after the object is detected; a camera module, which is used to capture and detect objects, and identify vehicle information and determine the vehicle's behavior diagram when the detection result is a vehicle; a network module, which is used to upload vehicle information and alarm information after the camera is blocked; a near-field communication module, which is used to transmit data when the detection device cannot connect to the network. According to the technical solution of the present application, the tracing function of the problem of malicious blocking or damage of the detection equipment in the parking fee scene can be realized, which has a high application value.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a road parking detection system and method with a sentinel mode. Background Art

[0002] In recent years, with the increasing number of cars, cars have become an essential means of transportation for people's daily commutes. The difficulty of finding parking in cities has become increasingly serious, especially in busy commercial areas and near public transportation hubs. To address this issue, various regions have implemented intelligent parking management systems, aiming to improve parking space turnover through efficient parking management. One such solution is a battery-powered, wire-free intelligent parking management solution with a built-in communication module. This solution requires no road construction and is quick and easy to install. However, it utilizes traditional embedded technology, resulting in a startup time of more than 5 seconds for the video acquisition and data processing modules. It only captures images of vehicles parked, exiting, and in available parking spaces. The lack of a complete parking evidence chain prevents accurate parking duration detection. Furthermore, in actual operation, some drivers obstruct or damage the detection equipment to avoid paying parking fees, resulting in direct financial losses for parking operators. Therefore, the development of a simple and efficient method to help intelligent parking management solutions maintain a complete evidence chain is urgently needed.

[0003] Currently available methods for intelligent parking management solutions to maintain a complete chain of evidence often suffer from shortcomings. For example, methods based on video inspection equipment regularly collect parking vehicle information and manually supplement the chain of evidence. This method can be susceptible to poor recognition rates and efficiency on busy roads, and can also lead to missing vehicle information. Therefore, to address this challenge, this application proposes a roadside parking detection system and method with a sentinel mode. Summary of the Invention

[0004] Technical Purpose

[0005] To address the aforementioned issues, the present invention aims to provide a road parking detection system and method with sentinel mode. This system and method not only addresses the lack of evidence in traditional intelligent parking management solutions, but also enables traceability of malicious obstruction or damage to detection equipment. This system and method have a wide range of application scenarios, are not easily affected by external factors, simplify and automate parking detection, and improve the accuracy of parking space detection.

[0006] Technical Solution

[0007] To achieve the above objectives, the present invention provides a road parking detection system and method with sentinel mode. By reducing the wake-up time of the camera module to capture the behavior of the vehicle entering the parking space, thereby forming a complete parking evidence chain, and by uploading and locally storing video data within a period of time after the vehicle enters the parking space, it can realize the traceability function of the problem of malicious obstruction or damage of the detection equipment.

[0008] In a first aspect, the present invention provides a road parking detection system with a sentinel mode, comprising:

[0009] The detection module is used to detect whether an object is approaching or moving away from the detection device and wake up the other modules after detecting an object;

[0010] A camera module is used to capture and detect objects, identify vehicle information when the detection result is a vehicle, and determine the vehicle's behavior map;

[0011] Network module, used to upload vehicle information and alarm information when the camera is blocked;

[0012] Near field communication module, used for data transmission when the detection device cannot connect to the network.

[0013] Furthermore, the system also includes a storage module for storing video data within a period of time after the object is detected.

[0014] Furthermore, the system also includes a fill light module for supplementing light for the detection equipment.

[0015] Furthermore, the detection module includes at least one of radar detection, infrared sensor detection, and machine vision detection.

[0016] Furthermore, the system primarily powers the detection module during operation, reducing energy consumption, except when an object approaches or moves away from the detection device, or when the device is in low light. This reduces the camera module's wake-up time to ensure the proper acquisition of vehicle entry behavior patterns. By using the sentry mode system, power is limited to a small number of modules when not needed, significantly reducing system energy consumption.

[0017] Furthermore, methods for reducing the camera module's wake-up time include: reducing the connection distance between the detection module and the camera module on the hardware circuit PCB; limiting the camera module's functions upon startup to only enable the image capture function; and increasing the detection module's sensitivity to wake up the camera module as quickly as possible, with the camera module wake-up speed reaching as fast as 200ms. This method enables the camera module to quickly activate and capture the vehicle's entry behavior when a vehicle enters a parking space, improving the system's real-time performance.

[0018] Furthermore, the detection module wakes up the camera module to capture and detect the object after detecting that an object is approaching the detection device. When the detection result is a vehicle, the camera module further detects whether the image data contains a license plate, and further identifies the license plate if it contains a license plate.

[0019] Furthermore, the vehicle's behavior graph is determined by comprehensively comparing the parking behavior confidence and license plate confidence of multiple frames captured by the camera module. This graph includes the vehicle's entry behavior graph, the vehicle's stable parking behavior graph, and the vehicle's exit behavior graph. This behavior graph, combined with the parking space availability graph after the vehicle exits, forms a complete parking evidence chain. By capturing the vehicle's entry behavior graph, the system fully records the vehicle's parking process, forming an effective parking evidence chain that helps resolve parking disputes.

[0020] Furthermore, the parking behavior confidence and license plate confidence are determined by comprehensively judging conditions such as the clarity of the image, the completeness of the license plate, the change in the size of the license plate, and the movement direction of the license plate.

[0021] Furthermore, the vehicle's entry information and the vehicle's status information after stopping are uploaded to the device management platform through the network module at one time; when uploading the alarm information after the camera is blocked, you can manually choose whether to upload the video data after the camera is blocked.

[0022] Furthermore, the storage module automatically stores the video data captured by the camera module and marks unobstructed video data during storage. When the required video data exceeds the storage module's capacity, the module prioritizes overwriting the newly obstructed video data over the marked video data. If no marked video data exists, the module prioritizes overwriting the oldest recorded video data. Each video data entry is approximately 10 to 20 MB in size. By uploading captured video data online and storing it locally, the system facilitates quick and efficient tracing of the vehicle's parking process in the event of a parking dispute. By recording video data in order of priority, data storage is optimized and storage efficiency is improved.

[0023] Furthermore, the near-field communication module is used to connect the detection device to an external device contactlessly when the detection device cannot connect to the network, thereby outputting video data stored locally in the detection device, or debugging the detection device and handling device failure problems.

[0024] Furthermore, the fill light module includes a photoresistor and a fill light. The photoresistor is used to measure the light conditions of the environment in which the detection device is located, and sends an instruction to enable the fill light when the light is lower than a certain threshold.

[0025] Furthermore, the system is provided with a parking space indicator light to display the current vacancy status of the parking space and the working status of the detection equipment.

[0026] Furthermore, the system uses an intelligent urban transportation network solution to display the location distribution of urban parking spaces on a map, providing real-time data for road traffic monitoring and urban transportation planning. By monitoring on-street parking availability in real time, it can quickly respond to parking issues, effectively allocate and manage parking resources, and provide drivers with fast and accurate parking information, reducing the time drivers spend searching for parking spaces.

[0027] Furthermore, the system uses a deep residual network solution to extract vehicle and parking space features and provide more accurate high-level feature representations. The residual learning mechanism captures spatiotemporal features within image sequences, enabling the system to extract abstract features in complex environments. By learning deep image features, it can more accurately identify vehicle and parking space status information, thereby improving parking detection accuracy. By capturing spatiotemporal features within image sequences, the system's ability to process complex data is enhanced, reducing the probability of false positives and false negatives.

[0028] In a second aspect, the present invention further provides a road parking detection method with a sentinel mode, the method being based on the system described in the first aspect, comprising:

[0029] The detection module detects whether an object is approaching or moving away from the detection device and wakes up the other modules after detecting an object;

[0030] Capture and detect objects, identify vehicle information and determine the vehicle's behavior if the detection result is a vehicle;

[0031] Upload vehicle information and alarm information when the camera is blocked;

[0032] Selectively upload and locally store video data within a period of time after an object is detected.

[0033] In a third aspect, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements the aforementioned road parking detection method with a sentinel mode.

[0034] The present invention reduces the wake-up time of the camera module to capture the behavior diagram of the vehicle entering the parking space, thereby forming a complete parking evidence chain; by uploading the video data within a period of time after the vehicle enters the parking space to the network and storing it locally, it realizes the traceability function of the problem of malicious obstruction or damage of the detection equipment; through the urban intelligent transportation network solution, the location distribution of urban parking spaces is displayed in the form of a map, providing real-time data for road traffic monitoring and urban traffic planning; through the deep residual network solution, the vehicle and parking space features are extracted and a more accurate high-level feature representation is provided. The residual learning mechanism captures the spatiotemporal features in the image sequence, enabling the system to extract abstract features in complex environments. This system and method not only solves the problem of the lack of evidence chain in traditional intelligent parking management solutions, but also realizes the traceability function of the problem of malicious obstruction or damage of the detection equipment. The system and method have a wide range of application scenarios and are not easily interfered with by external factors. It realizes the simplification and automation of parking detection and improves the accuracy of parking space detection.

[0035] Beneficial effects

[0036] By implementing the above-mentioned road parking detection system and method with sentinel mode provided by the present invention, the following technical effects are achieved:

[0037] (1) The present invention reduces the wake-up time of the camera module to capture the behavior diagram of the vehicle entering the parking space, thereby forming a complete parking evidence chain; it improves the real-time performance of the system by quickly starting the camera module to capture the behavior diagram of the vehicle entering the parking space when the vehicle enters the parking space; by capturing the behavior diagram of the vehicle entering the parking space, the system completely records the parking process of the vehicle, forming an effective parking evidence chain, which helps to resolve parking disputes; through the sentinel mode system, only a small number of modules are powered at non-essential times, thereby greatly reducing the energy consumption of the system.

[0038] (2) The present invention achieves the function of tracing the problem of malicious obstruction or damage of detection equipment by uploading video data within a period of time after the vehicle enters the parking space to the Internet and storing it locally; it is convenient to quickly and effectively trace the parking process of the vehicle when a parking dispute occurs by uploading the captured video data to the Internet and storing it locally; it optimizes data storage and improves storage efficiency by recording video data in priority order.

[0039] (3) The present invention displays the location distribution of urban parking spaces in the form of a map through an urban intelligent transportation network solution, providing real-time data for road traffic monitoring and urban traffic planning; it quickly responds to parking problems by real-time monitoring of the status of on-street parking spaces, can effectively allocate and manage parking space resources, provide drivers with fast and accurate parking space information, and reduce the time drivers spend looking for parking spaces.

[0040] (4) The present invention extracts vehicle and parking space features through a deep residual network solution and provides a more accurate high-level feature representation. It captures the spatiotemporal features in the image sequence through the residual learning mechanism, enabling the system to extract abstract features in complex environments. By learning the deep features of the image, it can more accurately identify the status information of the vehicle and parking space, thereby improving the accuracy of parking detection. By capturing the spatiotemporal features in the image sequence, the system's ability to process complex data is improved, and the probability of false alarms and missed alarms is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To make the above-mentioned on-road parking detection system and method with sentinel mode more clearly understood, the following briefly introduces the drawings required for use in the specific embodiments of the present invention. Obviously, the drawings described below are merely some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0042] Figure 1 A schematic diagram showing a road parking detection method with a sentry mode;

[0043] Figure 2 Schematic diagram showing the effect of the urban intelligent transportation network solution. DETAILED DESCRIPTION

[0044] Example 1:

[0045] A road parking detection system and method with a sentry mode is provided, wherein the system includes: a detection module, specifically an ultrasonic radar detection module, which is used to detect whether an object is approaching or moving away from the detection device and wake up the other modules after detecting an object; a camera module, which is used to capture and detect objects, and identify vehicle information and determine the vehicle's behavior diagram when the detection result is a vehicle; a network module, which is used to upload vehicle information and alarm information when the camera is blocked; and a near-field communication module, which is used to transmit data when the detection device cannot connect to the network. Figure 1 shown.

[0046] The road parking detection system with sentinel mode provided by the present invention specifically includes:

[0047] The system further includes a storage module for storing video data within a period of time after the object is detected.

[0048] The system also includes a supplementary light module for supplementing light for the detection equipment.

[0049] The detection module includes at least one of radar detection, infrared sensor detection, and machine vision detection.

[0050] During operation, the system only supplies power to the detection module most of the time, except when an object approaches or moves away from the detection device or the detection device is located in an environment with insufficient light. The system also ensures the normal acquisition of the vehicle entry behavior map by reducing the wake-up time of the camera module.

[0051] Methods for reducing the camera module wake-up time include: reducing the connection distance between the detection module and the camera module on the hardware circuit PCB board; limiting the functions of the camera module at startup, and only pre-enabling the image capture function; and increasing the sensitivity of the detection module to wake up the camera module as early as possible. The camera module wake-up speed can be as fast as within 200ms.

[0052] Furthermore, the detection module wakes up the camera module to capture and detect the object after detecting that an object is approaching the detection device. When the detection result is a vehicle, the camera module further detects whether the image data contains a license plate, and further identifies the license plate if it contains a license plate.

[0053] The vehicle behavior graph is determined by comprehensively comparing the parking behavior confidence and license plate confidence of multiple frames of vehicle images obtained by the camera module, which includes the vehicle's entry behavior graph, the vehicle's parking behavior graph, and the vehicle's exit behavior graph; the vehicle's behavior graph and the parking space vacancy graph after the vehicle exits are combined to form a complete parking evidence chain.

[0054] The parking behavior confidence and license plate confidence are determined by comprehensively judging conditions such as the clarity of the image, the completeness of the license plate, the change in the size of the license plate, and the movement direction of the license plate.

[0055] The vehicle's entry information and the vehicle's status information after stopping are uploaded to the device management platform through the network module at one time; when uploading the alarm information after the camera is blocked, you can manually choose whether to upload the video data after the camera is blocked.

[0056] Furthermore, the storage module automatically stores the video data acquired by the camera module and marks the video data that is not obstructed by the camera during storage; when the video data that needs to be stored exceeds the storage capacity of the storage module, the storage module gives priority to overwriting the new obstructed video data on the marked video data. If the marked video data no longer exists, the storage module gives priority to overwriting the video data with the earliest time record; the size of each video data is approximately 10M to 20M.

[0057] Furthermore, the near-field communication module is used to connect the detection device to an external device contactlessly when the detection device cannot connect to the network, thereby outputting video data stored locally in the detection device, or debugging the detection device and handling device failure problems.

[0058] The fill light module includes a photoresistor and a fill light. The photoresistor is used to measure the light conditions in the environment where the detection device is located, and sends an instruction to enable the fill light when the light is lower than a certain threshold.

[0059] The system is provided with a parking space indicator light to display the current vacancy status of the parking space and the working status of the detection equipment.

[0060] The system uses the city intelligent transportation network solution to display the location distribution of urban parking spaces in the form of a map, providing real-time data for road traffic monitoring and urban traffic planning, including: matching the data identified by the road parking detection system with the geographic information in the city intelligent transportation network solution database, which contains the geographic coordinates, belonging area, type and other information of the parking spaces; combining the real-time collected data with the city intelligent transportation network solution, and performing spatial analysis and processing, using the city intelligent transportation network solution to monitor and manage vehicle locations and parking status in real time; and displaying parking status, vehicle location and related statistical information in the form of a map. The effect of the city intelligent transportation network solution is as follows: Figure 2 shown.

[0061] Example 2:

[0062] Based on the previous embodiment, a deep residual network solution is added to extract vehicle and parking space features and provide more accurate high-level feature representation. The residual learning mechanism captures the spatiotemporal features in the image sequence, enabling the system to extract abstract features in complex environments.

[0063] Among them, the video data needs to be preprocessed to meet the input requirements of the deep learning model. The normalization formula is as follows:

[0064]

[0065] in, Represents raw data, such as pixel values ​​of an image; represents the mean of the data; represents the standard deviation of the data; represents the normalized data; Represents the scaling factor, which is used to control the scaling degree of the normalized data.

[0066] ResNet is selected as the basic model, and its residual block is used to improve the efficiency of training deep networks. The design formula of the residual block is expressed as:

[0067]

[0068] in, Indicates the The input of the layer; Indicates the The output of the layer; Represents the residual function, which represents the operation inside the residual block; Indicates the The weight parameters of the layer; Represents the adjustment factor, which is used to control the influence of the residual function on the output.

[0069] The key features of the video are extracted through the trained ResNet model. The feature extraction process is as follows:

[0070]

[0071] in, Represents the feature map, which is the output of the activation function; Represents input data; Represents the residual block function with weights; represents the weight parameter; Represents a parameterized rectified linear unit activation function, used to increase nonlinear characteristics; Represents the parameters of the activation function, which are used to control the strength of the nonlinearity.

[0072] Parking behavior analysis is performed based on the extracted features and is implemented through the following decision function:

[0073]

[0074] in, Represents the predicted output, i.e., the category probability distribution; represents the weight of the fully connected layer; represents the bias of the fully connected layer; Represents the feature map extracted from ResNet; Represents the activation function, which is used in the output layer of multi-classification problems to convert the output into a probability distribution; Represents the weight decay term, used to adjust The confidence level of the output.

[0075] The output formula of the system is:

[0076]

[0077] in, Indicates the final action instruction, such as whether to send an alarm; represents the weight of the output layer; represents the bias of the output layer; Represents the output of the previous layer; express Activation function, used in the output layer of a binary classification problem, converts the output into a probability value between 0 and 1; Represents the threshold parameter.

[0078] For example, suppose there is an image data pixel value , the mean value of the pixel and standard deviation , selected :

[0079]

[0080] Assuming the residual function Through a simple operation ,and , selected To balance the residual and the original input:

[0081]

[0082] Selected As parameters of the parameterized activation function:

[0083]

[0084] Assumptions , , selected As a weight decay term:

[0085]

[0086] Assumptions , , selected As threshold parameter:

[0087]

[0088] The effect of the deep residual network solution is shown in Table 1:

[0089] Table 1. Summary of the effects of deep residual network solutions

[0090]

[0091] As shown in Table 1, the model optimized with the deep residual network solution further improved its accuracy, significantly reduced its false positive and false negative rates, and shortened its response time. This demonstrates that the deep residual network solution can significantly improve the accuracy of the sentinel-based on-road parking detection method.

[0092] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable non-transitory storage media containing computer-usable program code.

[0093] The present invention can provide computer program instructions to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the system.

[0094] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device that implements the functions of the system.

[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions of the described system.

Claims

1. A road parking detection system with sentinel mode, characterized in that: include: The detection module is used to detect whether an object is approaching or moving away from the detection device, and wake up the other modules after detecting the object; A camera module, used to capture and detect objects, identify vehicle information and determine the vehicle's behavior map when the detection result is a vehicle; Network module, used to upload vehicle information and alarm information when the camera is blocked; Near field communication module, used for data transmission when the detection device cannot connect to the network; The system extracts vehicle and parking space features through a deep residual network solution and provides a more accurate high-level feature representation. The system captures the spatiotemporal features in the image sequence through the residual learning mechanism, enabling the system to extract abstract features in complex environments. The video data needs to be preprocessed to meet the input requirements of the deep learning model. The normalization formula is as follows: ,in, Represents the original data; represents the mean of the data; Indicates the standard deviation of the data; represents the normalized data; Represents the scaling factor, which is used to control the scaling degree of the normalized data; ResNet is selected as the basic model, and its residual block is used to improve the efficiency of training deep networks. The design formula of the residual block is expressed as: ,in, Indicates The input of the layer; Indicates The output of the layer; Represents the residual function, which represents the operation inside the residual block; Indicates The weight parameters of the layer; Represents the adjustment factor, which is used to control the influence of the residual function on the output; The key features in the video are extracted through the trained ResNet model. The feature extraction process is as follows: ,in, represents the feature map, which is the output of the activation function; Represents input data; represents the residual block function with weights; represents the weight parameter; represents the parameterized rectified linear unit activation function, which is used to increase nonlinear characteristics; Represents the parameters of the activation function, which is used to control the strength of nonlinearity; Parking behavior analysis is performed through the extracted features and implemented through the following decision function: ,in, Represents the predicted output, i.e., the category probability distribution; represents the weight of the fully connected layer; represents the bias of the fully connected layer; Represents the feature map extracted from ResNet; Represents the activation function, which is used in the output layer of multi-classification problems to convert the output into a probability distribution; Represents the weight decay term, which is used to adjust Confidence of the output; The output formula of the system is: ,in, Indicates the final action instruction; represents the weight of the output layer; represents the bias of the output layer; Represents the output of the previous layer; express The activation function is used in the output layer of the binary classification problem to convert the output into a probability value between 0 and 1; Represents the threshold parameter.

2. The road parking detection system with sentinel mode according to claim 1, characterized in that: During operation, the system only supplies power to the detection module most of the time, except when an object approaches or moves away from the detection device or the detection device is located in an environment with insufficient light. The system also ensures normal acquisition of the vehicle entry behavior map by reducing the wake-up time of the camera module.

3. The road parking detection system with sentinel mode according to claim 1 or 2, characterized in that: The vehicle behavior graph is determined by comprehensively comparing the parking behavior confidence and license plate confidence of multiple frames of the vehicle acquired by the camera module, and includes the vehicle's entry behavior graph, the vehicle's stopping behavior graph, and the vehicle's exit behavior graph; the vehicle's behavior graph and the parking space vacancy graph after the vehicle exits are combined to form a complete parking evidence chain.

4. The road parking detection system with sentinel mode according to claim 1, characterized in that: The vehicle's entry information and the vehicle's status information after it stops are uploaded to the device management platform through the network module at one time; when uploading the alarm information after the camera is blocked, it is manually selected whether to upload the video data after the camera is blocked.

5. The road parking detection system with sentinel mode according to claim 1, characterized in that: The system also includes a storage module for storing video data within a period of time after the object is detected. The storage module automatically stores the video data acquired by the camera module and marks the video data that is not obstructed by the camera during storage. When the video data that needs to be stored exceeds the storage capacity of the storage module, the storage module preferentially overwrites the new obstructed video data on the marked video data. If the marked video data no longer exists, the storage module preferentially overwrites the video data with the earliest time record.

6. The road parking detection system with sentinel mode according to claim 1 or 5, characterized in that: The system also includes a fill light module for supplementing light for the detection device. The fill light module includes a photoresistor and a fill light. The photoresistor is used to measure the light conditions of the environment in which the detection device is located, and sends an instruction to enable the fill light when the light is lower than a certain threshold.

7. The road parking detection system with sentinel mode according to claim 1, characterized in that: The system displays the location distribution of urban parking spaces in the form of a map through an urban intelligent transportation network solution, providing real-time data for road traffic monitoring and urban transportation planning.

Images