Non-motor vehicle parking space monitoring and evaluation method and system

Abstract

The invention relates to the technical field of intelligent traffic management, particularly discloses a non-motor vehicle parking space monitoring and evaluation method and system, and aims to solve the problems of low manual inspection efficiency, insufficient parking form recognition precision and lack of dynamic resource management capability in the prior art. The method comprises the following steps: deploying a high-definition camera in a parking area to collect video data in real time; constructing a parking space digital model based on a GIS; a YOLO deep learning model is combined with shielding scene data generated by a GAN, and the number and the parking form of non-motor vehicles are recognized; evaluating a parking state by calculating a supply-demand rate and an order index; and finally generating a visual report and triggering an optimization decision. The system is composed of a sensing layer, a transmission layer and a platform layer, integrates hardware equipment, an algorithm model and a GIS platform, and realizes all-weather monitoring, accurate identification and dynamic management. According to the invention, a decision basis is provided for urban non-motor vehicle parking management decision.

Description

Technical Field

[0001] The present invention belongs to the technical field of image analysis, and in particular relates to a non-motor vehicle parking space monitoring and evaluation method and system. Background Art

[0002] With the acceleration of urbanization, non-motorized vehicles (such as bicycles and electric bicycles) have become an important means of short-distance travel due to their convenience and environmental friendliness. However, the management of non-motorized vehicle parking is becoming increasingly problematic, especially in densely populated urban areas. The imbalance between supply and demand for roadside non-motorized vehicle parking spaces and the chaotic parking order have become major challenges for urban traffic management.

[0003] Currently, the monitoring of non-motor vehicle parking spaces mainly relies on manual patrols, with management personnel regularly inspecting and recording parking conditions. This approach has significant drawbacks: Inefficiency: Manual inspections require a lot of manpower and are difficult to achieve all-weather, full-coverage, real-time monitoring. Data lag: Manually collected parking data is discrete and infrequently updated, making it impossible to provide timely evidence for dynamic resource allocation. Subjective bias: The evaluation of parking order relies on manual judgment and lacks unified standards, which can easily lead to unscientific management decisions.

[0004] In recent years, intelligent monitoring technologies (such as video analysis and IoT perception) have been gradually applied in the field of traffic management, but technical bottlenecks still exist in non-motor vehicle parking scenarios: Existing camera systems are mostly designed for motor vehicles, and their recognition accuracy for non-motor vehicle parking configurations (such as vehicle head direction and parking offset angle) is insufficient; The lack of dynamic modeling capabilities integrated with geographic information systems (GIS) makes it difficult to achieve global optimization of parking resources; Traditional algorithms have poor adaptability to occlusion scenarios (such as mutual occlusion when vehicles are densely parked), resulting in insufficient reliability of detection results.

[0005] Therefore, there is an urgent need for a technical solution that can monitor the usage status of non-motor vehicle parking spaces in real time and accurately, and realize dynamic management of parking resources based on data-driven. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to address the deficiencies in the above-mentioned existing technologies and provide a non-motor vehicle parking space monitoring and evaluation method and system. By integrating high-definition video monitoring, deep learning algorithms and GIS digital modeling, it aims to solve the above-mentioned technical pain points and provide efficient and scientific decision-making support for urban traffic management.

[0007] The first aspect of the present invention discloses a method for monitoring and evaluating non-motor vehicle parking spaces, comprising the following steps: S1: Install a high-definition network camera on the side of the non-motor vehicle parking area. The camera resolution must be no less than 1080p, the frame rate must be no less than 30fps, and the installation angle must be no less than 45 degrees to cover the parking area and mark the parking area. S2: Constructing a digital model of non-motor vehicle parking based on a GIS environment, the model includes parking area points, ranges, and parking space configuration attributes, the attributes at least including number, number of parking spaces, number of parking requirements, and parking order; S3: Matching the digital model with the surveillance camera, and associating the parking space number with the camera coverage area; S4: Collecting video data through the camera, using the YOLO target detection model to identify the number of non-motor vehicles in the parking area and the positions of the vehicle heads and parking spaces, and marking the parking form of each non-motor vehicle; S5: Generate a parking demand change curve based on the number of continuously identified non-motor vehicles, calculate the supply-demand ratio (supply-demand ratio = parking demand / number of parking spaces × 100%), and calculate the absolute order index and relative order index based on the parking offset angle. The weighted summation results in the overall order index. S6: Evaluate the plans for adding or removing parking spaces based on the supply-demand ratio and the overall order index, and generate parking space adjustment recommendations based on the evaluation results.

[0008] The above-mentioned non-motor vehicle parking space monitoring and evaluation method, said S4 also includes: GAN is used to generate a training dataset under occlusion scenarios. Combined with bicycle category data from the COCO dataset, the YOLO object detection model is trained to improve the detection accuracy under occlusion conditions.

[0009] In the above-mentioned non-motor vehicle parking space monitoring and evaluation method, in S5: The calculation formula of the absolute order index is:

[0010] The calculation formula of the relative order index is:

[0011] The overall order index is the sum of the absolute order index and the relative order index weighted at 50%.

[0012] In the above-mentioned non-motor vehicle parking space monitoring and evaluation method, when constructing the digital model in S2, ArcGIS or QGIS software is used, with a satellite image or a topographic map as the background layer, and the closed space of the parking area is marked by a polygon drawing tool.

[0013] In the above-mentioned non-motor vehicle parking space monitoring and evaluation method, multiple cameras are configured in the special parking area in S1 to ensure coverage without blind spots.

[0014] The aforementioned non-motor vehicle parking space monitoring and evaluation method, and the logic for generating adjustment suggestions include: (1) When the supply-demand ratio is ≥95% for three consecutive days, a capacity expansion recommendation is triggered; (2) When the overall order index is less than 60 points for 5 consecutive days, a line modification recommendation is triggered; (3) Dynamically adjust parking space allocation based on historical data predictions.

[0015] A second aspect of the present invention discloses a non-motor vehicle parking space monitoring and evaluation system, comprising: Monitoring module: including a high-definition network camera with a resolution of not less than 1080p, a frame rate of not less than 30fps, and an installation angle of not less than 45 degrees, for real-time video data collection of the parking area; Modeling module: Build a digital model of non-motor vehicle parking based on the GIS environment and match it with the camera; Recognition module: equipped with the YOLO target detection model to identify the number of non-motor vehicles and parking patterns; Analysis module: calculates the supply-demand ratio, absolute order index and relative order index, and generates parking resource assessment reports; Optimization module: Propose parking space management plan based on the evaluation results.

[0016] In the above-mentioned non-motor vehicle parking space monitoring and evaluation system, the recognition module further includes an occlusion scene training data set generated by GAN.

[0017] In the above-mentioned non-motor vehicle parking space monitoring and evaluation system, the analysis module supports dynamic threshold setting, and when the supply-demand ratio is ≥95%, it is determined to be a parking saturation state.

[0018] Compared with the prior art, the present invention has the following advantages: 1. This invention proposes installing high-definition network cameras (resolution ≥ 1080p, frame rate ≥ 30fps) on the sides of parking areas. This system captures real-time video coverage of the parking area, directly replacing manual inspections and enabling 24 / 7, uninterrupted monitoring. This addresses the issues of high labor costs and delayed data updates. Camera parameters (resolution, frame rate, and mounting angle) ensure clear and comprehensive surveillance images, preventing data loss due to obstructions or viewing angle limitations.

[0019] 2. The present invention uses the YOLO target detection model (S4) in combination with GAN-generated occlusion scene training data to identify the front and parking positions of non-motor vehicles and annotate the parking patterns. This deep learning model improves detection accuracy in complex scenarios (such as vehicle occlusion), overcoming the shortcomings of traditional algorithms in insufficient recognition of non-motor vehicle parking patterns. The combination of the high real-time performance of the YOLO target detection model and GAN data enhancement significantly enhances the model's robustness in occlusion and dense parking scenarios.

[0020] 3. The present invention constructs a digital parking model based on GIS (S2) and matches it with cameras (S3). The parking demand change curve and order index (S5) are combined to assess the supply and demand balance. The GIS model integrates spatial data and real-time monitoring information to support the global dynamic management of parking resources, compensating for the lack of systematic analysis in traditional methods. The linkage between model attributes (number, number of parking spaces, parking order) and camera data provides a quantitative basis for adding or removing parking spaces (S6).

[0021] 4. This invention defines the supply-demand ratio (parking demand / number of parking spaces × 100%) and the order index (weighted calculation of the absolute order index and the relative order index). Through mathematical modeling, it eliminates subjective evaluation bias and provides an objective parking status assessment standard. The dual analysis of the supply-demand ratio threshold (e.g., ≥95% for saturation) and the order index provides actionable data support for management decisions.

[0022] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a specific implementation flow chart of the non-motor vehicle parking space monitoring and evaluation method.

[0024] Figure 2 This is an example diagram of non-motor vehicle identification and parking form calculation.

[0025] Figure 3 This is the architecture diagram of the non-motor vehicle parking space monitoring and evaluation system. DETAILED DESCRIPTION

[0026] like Figure 1 As shown, a method for monitoring and evaluating non-motor vehicle parking spaces includes the following steps: Step 1: Hardware deployment and data collection; Install a high-definition network camera on the side of the target non-motor vehicle parking area, and the camera meets the following parameters: resolution not less than 1080p, frame rate not less than 30 frames per second (fps); The installation angle should be no less than 45 degrees to ensure that the camera covers the parking area from a bird's-eye view. For special areas with obstructions or curved roads, multiple cameras are deployed to eliminate monitoring blind spots.

[0027] In the camera monitoring screen, use image annotation tools (such as LabelImg) to mark the boundary range of the parking area to clearly define the parking area.

[0028] Step 2: GIS digital model construction; Use geographic information system software (ArcGIS or QGIS) to build a digital model of non-motorized vehicle parking spaces. The specific steps are as follows: import satellite images or high-precision terrain maps as background layers; Use the polygon drawing tool to mark the closed space of the non-motor vehicle parking area on the background layer; Define attribute fields for each parking area, including: Number: unique identifier (such as area code); Parking Space: The maximum number of non-motor vehicles that can be accommodated in the area; Parking demand number: the actual number of parked vehicles updated dynamically; Parking order: The order score generated by the subsequent algorithm.

[0029] Step 3: spatial matching between the camera and the digital model; Bind the camera's installation location (latitude and longitude coordinates) to the corresponding parking area number in the GIS model to form a "camera-parking area" mapping relationship table; The video stream data collected by the camera and the attribute information of the GIS model are synchronized in real time through the API interface to ensure data consistency.

[0030] Step 4: Non-motor vehicle identification and parking form analysis, such as Figure 2 As shown; Data preparation: Use public datasets (such as the bicycle category in the COCO dataset) as the base training set; use generative adversarial networks (GANs) to synthesize images of non-motor vehicles in occluded scenarios (such as partially obscured vehicles) to expand the training data; Model training: YOLOv5 was used as the object detection model, with an input image size of 640 × 640 pixels and a training period of 300 epochs. The CIoU loss function was used, the optimizer was stochastic gradient descent (SGD), the momentum parameter was set to 0.937, and the initial learning rate was 0.01. Real-time detection and annotation: The camera video stream is processed frame by frame to identify the location of non-motorized vehicles and annotate the bounding boxes of the front and rear of the vehicles (in x, y, w, h format). Combined with the parking area boundary lines in the GIS model, the parking offset angle of each vehicle is calculated (defined as the angle between the vehicle's central axis and the perpendicular direction of the parking area).

[0031] Step 5: Parking supply and demand balance and parking order assessment; Supply-demand ratio calculation: Count the actual number of parked vehicles (parking demand) and the number of parking spaces per hour, and calculate the supply-demand ratio using the following formula: Supply-demand ratio = parking demand / number of parking spaces × 100%; Set a threshold (e.g., supply-demand ratio ≥ 95%) to determine parking saturation; Order index calculation: Absolute order index: calculated based on the average deviation angle, the formula is:

[0032] Relative order index: Based on the sum of the absolute deviations between the vehicle offset angle and the average offset angle, the formula is:

[0033] Overall order index: The absolute order index and the relative order index are weighted together at 50% weight, and the formula is: overall order index = 0.5 × absolute order index + 0.5 × relative order index; Step 6: Dynamic optimization and decision support; Generate visual reports on the GIS platform, including supply and demand rate heat maps and order index distribution maps; Generate parking space adjustment suggestions based on the assessment results: (1) When the supply-demand ratio is ≥95% for three consecutive days, a capacity expansion recommendation is triggered; (2) When the overall order index is less than 60 points for 5 consecutive days, a line modification recommendation is triggered; (3) Dynamically adjust parking space allocation based on historical data predictions.

[0034] like Figure 3 As shown, a non-motor vehicle parking space monitoring and evaluation system includes: Perception layer: Deploy a cluster of high-definition network cameras (such as the Hikvision DS-2CD2043G0-I) to cover all target parking areas. Edge computing devices (such as the NVIDIA Jetson Xavier NX) are responsible for local pre-processing of video streams (including noise reduction and keyframe extraction). Transport layer: transmits video streams and metadata to cloud servers via 5G or fiber networks; Platform layer: Monitoring module, which manages camera status, video stream distribution, and storage; modeling module, which builds and maintains digital models of parking areas based on ArcGIS Pro 3.0; recognition module, which deploys the YOLOv5 model and uses GPU clusters (such as Alibaba Cloud ECS instances) for real-time target detection; analysis module, which uses the Apache Spark framework to batch calculate supply and demand rates and order indices; and optimization module, which generates parking space management recommendations using a decision tree algorithm and pushes them to the management terminal.

[0035] The module interaction process of the non-motor vehicle parking space monitoring and evaluation system is as follows: Data flow: The camera collects video data → edge device preprocessing → cloud storage; the recognition module calls the video stream → outputs non-motor vehicle location and shape data → writes to the GIS model; the analysis module reads the GIS model data → calculates the supply and demand rate and order index → ​​generates a visualization report; the optimization module analyzes the report results → triggers management instructions (such as SMS notifications or system alarms); Control flow: Users set monitoring parameters (such as supply and demand rate alarm thresholds) through the web interface; the system automatically adjusts the camera angle (such as through pan-tilt control) to optimize the monitoring field of view.

[0036] The specific embodiments of the present invention are as follows: Example 1: Parking management around subway stations; If the system detects that the supply-demand ratio at a subway station exit area is ≥98% for two consecutive hours, it will automatically send a text message to the management staff, suggesting that temporary parking spaces be added; Example 2: Optimizing parking order in commercial areas; Based on the distribution of the order index, the system pushes a prompt message in the user's APP that "it is recommended to park with the front of the car facing north" to guide users to park in a standardized manner.

[0037] Hardware configuration for actual implementation: Camera: Hikvision DS-2CD2043G0-I (resolution 1080p, frame rate 30fps); Edge device: NVIDIA Jetson Xavier NX (GPU 384 cores, 8GB memory); Server: Alibaba Cloud ECS instance (GPU-accelerated, 8-core CPU, 32GB memory); Software configuration for actual implementation: GIS platform: ArcGIS Pro 3.0; Deep learning framework: PyTorch 1.12 + Ultralytics YOLOv5; Data analysis tool: Apache Spark 3.3; Communication protocols: RTSP (video streaming transmission), MQTT (command issuance).

[0038] The above implementation details the technical implementation path of this invention, covering hardware deployment, algorithm design, system architecture, and typical application scenarios, ensuring that those skilled in the art can reproduce this invention based on the disclosed content. Specific parameters (such as camera resolution and model training cycle) can be adjusted according to actual needs without departing from the core innovation of this invention.

Claims

1. A non-motor vehicle parking space monitoring and evaluation method, characterized in that: The following steps are involved: S1: Install a high-definition network camera on the side of the non-motor vehicle parking area, with a resolution of no less than 1080p, a frame rate of no less than 30fps, and an installation angle of no less than 45 degrees to cover the parking area and mark the parking range; S2: constructing a non-motor vehicle parking digital model based on the GIS environment, the model including the parking area points, range and parking space configuration attributes, the attributes at least including the number, number of parking spaces, number of parking requirements and parking order; S3: matching the digital model with the surveillance camera, and associating the parking space number with the camera coverage area; S4: Collecting video data through the camera, using the YOLO target detection model to identify the number of non-motor vehicles in the parking area and the positions of the vehicle heads and parking spaces, and marking the parking form of each non-motor vehicle; S5: Generate a parking demand change curve based on the number of continuously identified non-motor vehicles, calculate the supply-demand ratio, supply-demand ratio = parking demand / number of parking spaces × 100%, and calculate the absolute order index and relative order index based on the parking offset angle, and obtain the overall order index by weighted summation; S6: Evaluate the plans for adding or removing parking spaces based on the supply-demand ratio and the overall order index, and generate parking space adjustment recommendations based on the evaluation results.

2. A non-motor vehicle parking space monitoring and evaluation method according to claim 1, characterized in that: The S4 also includes: GAN is used to generate a training data set in an occlusion scenario, and the YOLO target detection model is trained with the bicycle category data of the COCO data set to improve the detection accuracy under occlusion conditions.

3. A non-motor vehicle parking space monitoring and evaluation method according to claim 1, characterized in that: In S5: The calculation formula of the absolute order index is: ； The calculation formula of the relative order index is: ； The overall order index is the sum of the absolute order index and the relative order index weighted at 50%.

4. A non-motor vehicle parking space monitoring and evaluation method according to claim 1, characterized in that: When constructing the digital model in S2, ArcGIS or QGIS software is used, with a satellite image or a topographic map as the background layer, and a polygon drawing tool is used to mark the closed space of the parking area.

5. A non-motor vehicle parking space monitoring and evaluation method according to claim 1, characterized in that: The special parking area in S1 is equipped with multiple cameras to ensure coverage without blind spots.

6. A non-motor vehicle parking space monitoring and evaluation method according to claim 1, characterized in that: Adjust the logic of generating suggestion solutions including: (1) When the supply-demand ratio is ≥ 95% for three consecutive days, a capacity expansion recommendation is triggered; (2) When the overall order index is less than 60 points for 5 consecutive days, a recommendation for line modification is triggered; (3) Dynamically allocate and adjust parking spaces based on historical data predictions.

7. A non-motor vehicle parking space monitoring and evaluation system, characterized in that: include: Monitoring module: including a high-definition network camera with a resolution of no less than 1080p, a frame rate of no less than 30fps, and an installation angle of no less than 45 degrees, for real-time collection of parking area video data; Modeling module: construct a non-motor vehicle parking digital model based on the GIS environment and match it with the camera; Identification module: equipped with the YOLO target detection model, used to identify the number of non-motor vehicles and parking patterns; Analysis module: calculates the supply-demand ratio, absolute order index and relative order index, and generates a parking resource evaluation report; Optimization module: Propose parking space management solutions based on the evaluation results.

8. The non-motor vehicle parking space monitoring and evaluation system according to claim 7 is characterized in that: The recognition module also includes an occlusion scene training data set generated by GAN.

9. The non-motor vehicle parking space monitoring and evaluation system according to claim 7, characterized in that: The analysis module supports dynamic threshold setting, and determines that the parking is saturated when the supply-demand ratio is ≥95%.

Images