Abnormal street lamp detection system and method based on YOLOv8 model

Abstract

The invention provides an abnormal street lamp detection system and method based on a YOLOv8 model, the system comprises an image acquisition module, a target detection module and an abnormal street lamp identification module, the image acquisition module is used for acquiring a monitoring image on a road in real time, the target detection module adopts the YOLOv8 model to carry out target detection on image data, and the abnormal street lamp identification module is used for identifying the abnormal street lamp. Identifying a street lamp area in the image, positioning a street lamp in the image and marking the position of the street lamp; and the abnormal street lamp identification module analyzes the detected street lamp state by using image features, performs multiple times of abnormal detection on multiple street lamp images collected in the same time period by adopting a multiple-time identification strategy, and judges a final result through a voting mechanism, thereby identifying the abnormal street lamp. According to the invention, automatic detection and state analysis are carried out on the street lamps by using the YOLOv8 model, so that real-time monitoring and identification of abnormal street lamps are realized, and the intelligent management level of an urban road illumination system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of street lamp detection, and in particular to an abnormal street lamp detection system and method based on a YOLOv8 model. Background Art

[0002] At present, street lamp anomaly detection mainly relies on manual inspection or traditional sensor detection methods, which have many problems.

[0003] In terms of process, traditional street light inspection usually requires manual inspection, relying on staff to check street lights one by one within a specific time, which is not only time-consuming and labor-intensive, but also prone to missed detection and misjudgment. Some detection methods use sensors to detect the status of street lights, but these sensors need to be installed in large numbers on each street light, increasing the complexity of system deployment and maintenance costs. In addition, sensors are greatly affected by the environment, especially in severe weather conditions, and may cause false alarms or failures.

[0004] In terms of real-time performance, traditional methods cannot achieve real-time monitoring due to their reliance on manual inspections, especially on wide-area urban roads, where it is difficult to quickly discover and handle street lamp failures. Although sensor monitoring can achieve a certain degree of automation, it is limited by the number and distribution of sensors, and the monitoring range is limited, making it difficult to cover the real-time status of all street lamps.

[0005] In terms of cost, manual inspections require a lot of manpower, and long inspections also increase maintenance costs. At the same time, the initial installation cost of the sensor system is high, and the equipment is easily damaged and requires regular maintenance and replacement, which further increases operating costs. In addition, the sensor coverage is limited and cannot be deployed over a large area, which increases the difficulty of system expansion.

[0006] Therefore, the existing abnormal street light detection methods have problems such as complex procedures, poor real-time performance, and high cost. There is an urgent need for a more efficient, real-time, and low-cost street light abnormality detection method and system. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention proposes an abnormal street light detection system and method based on the YOLOv8 model. The present invention uses computer vision and deep learning technology to utilize the YOLOv8 model to automatically detect and analyze the status of street lights, aiming to achieve real-time monitoring and identification of abnormal street lights and improve the intelligent management level of urban road lighting systems.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An abnormal street lamp detection system based on a YOLOv8 model comprises an image acquisition module, a target detection module and an abnormal street lamp recognition module. The image acquisition module is used to acquire monitoring images on the road in real time. The target detection module uses the YOLOv8 model to perform target detection on image data, recognizes the street lamp area in the image, locates the street lamp in the image and marks its position; the abnormal street lamp recognition module uses image features to analyze the detected street lamp status, adopts multiple recognition strategies, performs multiple abnormality detections on multiple street lamp images collected in the same period, judges the final result through a voting mechanism, and thus recognizes the abnormal street lamp.

[0010] Preferably, the target detection module uses the YOLOv8 model to perform target detection on image data, and during the model training process, the channel attention mechanism and the spatial attention mechanism are combined to enhance the network's attention to key features and improve the accuracy of target detection.

[0011] Preferably, for the channel attention module: first, perform global average pooling and maximum pooling on the input feature map to generate two different feature vectors, which respectively represent the importance of each channel; secondly, pass these two vectors through two fully connected layers, and use ReLU activation function and Sigmoid activation function to generate attention weights for each channel; finally, apply the weights to the input feature map to adjust the influence of each channel.

[0012] Preferably, for the spatial attention module: first, the channel dimension is compressed to obtain a two-dimensional feature map, which represents the importance of each spatial position; secondly, a convolution operation is applied to the compressed feature map to generate a spatial attention map; finally, the spatial attention map is multiplied with the input feature map to enhance the features of important areas and weaken the influence of unimportant areas.

[0013] Preferably, the system further comprises an alarm module, which outputs the detection results after multiple identifications and voting. If an abnormal street lamp is detected, its location is marked and relevant data is recorded, and the alarm module generates alarm information.

[0014] Preferably, the system further comprises a tracking module, and the tracking module tracks the abnormal street lights detected by the detection model through an improved DeepSORT algorithm.

[0015] Furthermore, the present invention also provides an abnormal street light detection method based on the YOLOv8 model, comprising the following steps:

[0016] Collect images on the road in real time, use the YOLOv8 model to detect targets in the images, and identify and mark street light areas;

[0017] Perform multi-dimensional feature extraction on the labeled street light area images, and use the attention mechanism to optimize the feature extraction effect and identify possible abnormal street lights;

[0018] Perform multiple anomaly detections on images collected within the same time period, determine the final result through a voting mechanism, identify abnormal street lights, and mark their location information.

[0019] Preferably, the step of performing multi-dimensional feature extraction on the labeled street lamp area image, optimizing the feature extraction effect by using the attention mechanism, and identifying possible abnormal street lamps specifically includes the following steps:

[0020] Use a pre-trained convolutional neural network to extract multi-dimensional features from the cropped streetlight area image;

[0021] An attention mechanism is introduced to dynamically weight the importance of features, highlight key features related to anomaly detection, and suppress irrelevant or redundant features.

[0022] Preferably, the step of performing multiple anomaly detections on images collected within the same time period, judging the final result through a voting mechanism, identifying abnormal street lights, and marking the location information of the abnormal street lights specifically includes the following steps:

[0023] Perform independent anomaly detection on multiple images collected in the same time period;

[0024] The final abnormal result is determined through a voting mechanism;

[0025] According to the final result of the voting mechanism, the status of abnormal street lights is marked.

[0026] Preferably, the method further comprises the following steps: using an improved DeepSORT algorithm to track abnormal street lights detected by the detection model.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. In terms of process, the present invention makes full use of the existing road monitoring system or other image acquisition equipment, avoids additional hardware installation or equipment debugging, and reduces the complexity of project implementation. The YOLOv8 model is used to automatically detect and label the street light area, which can accurately identify abnormal street lights, and adopts multiple detection and voting mechanisms to further improve the reliability and accuracy of detection. By adding the Attention mechanism to optimize the effect of feature extraction, the model's recognition accuracy of abnormal street lights is improved. The entire process has a high degree of automation, reduces manual intervention, simplifies the detection operation process, and can operate effectively in complex environments, significantly improving detection efficiency and data processing accuracy.

[0029] 2. In terms of real-time performance, the present invention can obtain images in real time through the road monitoring system and use the YOLOv8 model to instantly detect the status of street lamps. The introduction of the Attention mechanism further improves the real-time response capability of the model. Through regular image acquisition, multiple detections, and a voting mechanism where the minority obeys the majority, the present invention can reflect the operating status of street lamps in real time while ensuring detection accuracy. There is no need to wait for a specific detection time or close the road. The system can monitor the status of street lamps at any time under normal traffic conditions to ensure the safety and stability of road lighting.

[0030] 3. In terms of reducing false alarms and missed alarms, the present invention effectively reduces the risk of false alarms or missed alarms in a single detection through multiple detections and voting mechanisms. In complex environments, such as light changes or vehicles passing by, single detection errors that may be caused can be automatically corrected through the voting mechanism, thereby improving the reliability and accuracy of the overall system. The decision-making rule of the minority obeys the majority ensures the robustness of street light anomaly detection and significantly improves the fault tolerance of the system.

[0031] 4. In terms of cost, the present invention makes full use of the existing monitoring system and image acquisition equipment, without the need to purchase expensive hardware equipment. The street light anomaly detection is achieved through software algorithms, reducing the dependence on physical equipment. The Attention mechanism combined with the YOLOv8 model not only improves the detection effect, but also further reduces the long-term use cost of the system through continuous software optimization and model updates. There is no need to make physical changes to the road or set up additional markings, avoiding high construction and maintenance costs, and has good economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0033] Figure 1 This is a structural block diagram of the abnormal street light detection system based on the YOLOv8 model of the present invention;

[0034] Figure 2 This is a schematic diagram of the voting mechanism for electing abnormal street lamps in an embodiment of the present invention;

[0035] Figure 3 This is a flowchart of the abnormal street light detection method based on the YOLOv8 model of the present invention. DETAILED DESCRIPTION

[0036] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0037] Specifically, the present invention provides an abnormal street light detection system based on the YOLOv8 model, such as Figure 1 As shown, the system includes an image acquisition module 1, a target detection module 2 and an abnormal street light recognition module 3.

[0038] The image acquisition module 1 is used to acquire monitoring images on the road in real time. In this embodiment, the image acquisition module 1 can acquire real-time street lamp image videos through a drone device. The drone is equipped with a high-definition camera and can fly in the street lamp area according to a predetermined route to capture real-time video streams in the area. The module can dynamically adjust the camera angle and shooting range to ensure that the target street lamp area can be covered and provide clear image input for subsequent detection.

[0039] The target detection module 2 uses the YOLOv8 model to detect and identify street lamps, uses the YOLOv8 model to identify the street lamp area in the image, locates the street lamp in the image and marks its position.

[0040] The YOLOv8 model is used to perform preliminary target detection on image data, and the Attention mechanism is added during model training to focus on the brightness, color, and shape characteristics of street lamps, reduce the impact of environmental interference on detection results, and optimize feature extraction related to street lamp status. At the same time, the training data set is expanded through data enhancement technology, including image rotation, cropping, brightness adjustment, etc., to improve the generalization ability of the model.

[0041] The CBAM (Convolutional Block Attention Module) module is a lightweight attention mechanism module that can adaptively assign weights to important features in convolutional neural networks to improve the expressiveness and accuracy of the model. The attention mechanism is combined with the CBAM module. By adding channel and spatial attention modules after the feature extraction layer of YOLOv8, the system can enhance the perception of specific areas and ensure more accurate recognition of the location and status of street lights. The CBAM module improves the accuracy of target detection by combining channel attention and spatial attention mechanisms to increase the network's attention to key features. For the channel attention module: first, the input feature map is globally averaged and max-pooled to generate two different feature vectors, which represent the importance of each channel respectively; secondly, these two vectors are passed through two fully connected layers, and the ReLU activation function and the Sigmoid activation function are used to generate the attention weights of each channel; finally, the weights are applied to the input feature map to adjust the influence of each channel. For the spatial attention module: first, the channel dimension is compressed, usually using average pooling and maximum pooling, to obtain a two-dimensional feature map that represents the importance of each spatial position; secondly, a convolution operation is applied to the compressed feature map to generate a spatial attention map; finally, the spatial attention map is multiplied with the input feature map to enhance the features of important areas and weaken the influence of unimportant areas.

[0042] The abnormal street light recognition module 3 uses image features to analyze the detected street light status and recognize abnormal street lights, such as damaged or unlit street lights. Figure 2 As shown, the present invention adopts a multiple recognition strategy, that is, for multiple street lamp images collected in the same period, multiple abnormality detections are performed, and the final result is determined by a voting mechanism, and the minority obeys the majority principle is followed to reduce false positives and missed positives. If most of the recognition results show that the street lamp is abnormal, it is finally judged as an abnormal street lamp, otherwise it is regarded as a normal street lamp. This mechanism effectively reduces the error and contingency of a single detection and improves the reliability of overall recognition.

[0043] In an optional embodiment, the system also includes an alarm module, which outputs the detection results after multiple identifications and voting. If an abnormal street light is detected, its location is marked and relevant data is recorded. The alarm module generates alarm information, which includes: the location of the abnormal street light, the detection time and the type of abnormality. This information will be recorded and stored locally in a timely manner so that the management personnel can handle it in time.

[0044] In an optional embodiment, the system further includes a tracking module, which tracks the abnormal street lamp detected by the detection model through an improved DeepSORT algorithm, uses a Kalman filter to predict the position information of the target street lamp at the next moment, uses the Hungarian algorithm to perform cascade matching and IoU matching on the predicted trajectory and the detection result, comprehensively judges the degree of matching between the two, completes the tracking and matching of the target street lamp, uses the more reliable Mahalanobis distance and cosine distance weighted values ​​as the association metric during cascade matching, integrates the appearance and motion information into the matching strategy, and uses the χ 2 The distribution threshold is used to determine whether the match is successful, reducing the number of ID switches that occur when the target is blocked and then reappears.

[0045] Furthermore, the present invention also provides a method for detecting abnormal street lights based on the YOLOv8 model, such as Figure 3 As shown, the method comprises the following steps:

[0046] S1: Real-time acquisition of images on the road, use the YOLOv8 model to detect targets on the acquired images, and identify and mark street lamp areas;

[0047] Specifically, step S1 includes the following steps:

[0048] S11: Real-time video or images on the road are collected by drone equipment;

[0049] In order to ensure that all street lights on the road can be detected, the drone equipment should be set at a high position to cover the panoramic view of the road and street lights. One frame of image is extracted from the continuous video stream every second as the input of the system. The extracted image frame contains timestamp information to facilitate subsequent detection records and tracking.

[0050] S12: preprocessing the acquired video or image;

[0051] In order to meet the input requirements of the YOLOv8 model, the collected original image is resized to the specified size. The aspect ratio of the image is maintained during the resizing process to avoid image deformation and affect the detection accuracy of the model. The image pixel values ​​are normalized and scaled to the range of [0, 1] so that the model can process the input data more efficiently and reduce the impact of light intensity or brightness changes on detection. The image is converted from the original RGB color space to the color space commonly used by the YOLOv8 model (such as HSV or YCbCr), which can enhance the model's robustness to lighting and color changes.

[0052] S13: Input the preprocessed image into the YOLOv8 model to perform target detection, identify and mark the street light area.

[0053] During the detection process, YOLOv8 will perform multi-scale analysis on the image, identify objects in different areas of the image, and give the category and confidence score of each target. For each street light area detected by YOLOv8, the system will mark the corresponding bounding box on the original image. The bounding box is usually represented by four coordinates (x, y, width, height) and displayed in the image with a colored frame line. After the detection is completed, the annotated image frame is output to the visualization interface, and the operator can view the detection status and area annotation of the street light in real time.

[0054] S2: Perform multi-dimensional feature extraction on the labeled street lamp area image, and use the attention mechanism to optimize the feature extraction effect and identify possible abnormal street lamps;

[0055] Specifically, step S2 includes the following steps:

[0056] S21: Use the pre-trained convolutional neural network to extract multi-dimensional features from the cropped streetlight area image;

[0057] Through multiple layers of convolution and pooling operations, the convolutional neural network converts the spatial information of the image into a high-dimensional feature vector, capturing the texture, shape, lighting and other features of the street lamp. In the last layer of the convolutional neural network, global average pooling is used to compress the spatial information of each feature map into a single feature value. This operation further simplifies the feature dimension and helps the model focus on the global characteristics of the image.

[0058] S22: Introduce an attention mechanism to dynamically weight the importance of features, highlight key features related to anomaly detection, and suppress irrelevant or redundant features.

[0059] The Attention mechanism enables the model to decide how to allocate attention in the feature space based on the contextual information of each feature, thereby improving the accuracy and effect of feature extraction. The Attention mechanism calculates the correlation between features and assigns different weights to each feature. The more correlated the features, the greater the weight, and the model will pay more attention to these features. Based on the assigned weights, the feature vectors are recombined to highlight features with a greater possibility of abnormality (such as abnormal brightness of street lights, etc.).

[0060] S3: Perform multiple anomaly detection on the images collected in the same time period, determine the final result through a voting mechanism, identify abnormal street lights and mark their location information;

[0061] Specifically, step S3 includes the following steps:

[0062] S31: performing independent anomaly detection on multiple images collected in the same time period;

[0063] Each image will follow the process in step S2, undergoing operations such as feature extraction, attention mechanism optimization, and anomaly scoring to generate a corresponding anomaly score. The system will record the detection results of each image, including the anomaly score and whether it exceeds the set anomaly threshold. These data will be used for decision-making in the subsequent voting mechanism.

[0064] S32: Determine the final abnormal result through voting mechanism;

[0065] In order to avoid false positives or false negatives caused by individual image quality or temporary environmental changes (such as vehicles passing by, light changes), the system will use a voting mechanism to determine the final abnormal results. Among multiple detection results, the system will calculate how many detections are judged as abnormal (scores exceed the threshold) and how many detections are normal. According to pre-set rules (such as more than half of the votes are judged as abnormal), the system makes the final decision. The rule usually set is "the minority obeys the majority", that is, if most of the detection results show that the street light is abnormal, it is judged as abnormal; if most of the detection results show that it is normal, the individual abnormal results are ignored.

[0066] S33: Mark the status of abnormal street lights according to the final result of the voting mechanism.

[0067] If most tests show that the streetlight is in an abnormal state, the system will mark the streetlight as "abnormal" and output a warning or notification for maintenance. If most tests show that the streetlight is normal, the system will ignore individual abnormal detections and keep the streetlight in normal state. The test results after multiple identifications and voting will be output. If an abnormal streetlight is detected, the location of the abnormal streetlight will be marked and related data will be recorded.

[0068] In an optional embodiment, the method further comprises the following steps:

[0069] The improved DeepSORT algorithm is used to track abnormal street lights detected by the detection model.

[0070] Specifically, the improved DeepSORT algorithm is used to track the abnormal street lights detected by the detection model, the Kalman filter is used to predict the next moment position information of the target street lights, the Hungarian algorithm is used to perform cascade matching and IoU matching on the predicted trajectory and the detection results, and the matching degree between the two is comprehensively judged to complete the tracking and matching of the target street lights. In the cascade matching, the weighted values ​​of the more reliable Mahalanobis distance and cosine distance are used as the association metric, and the appearance and motion information are also integrated into the matching strategy. The χ2 2 The distribution threshold is used to determine whether the match is successful, reducing the number of ID switches that occur when the target is blocked and then reappears.

[0071] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An abnormal street light detection system based on the YOLOv8 model, characterized in that: The system includes an image acquisition module, a target detection module and an abnormal street lamp recognition module. The image acquisition module is used to acquire monitoring images on the road in real time. The target detection module uses the YOLOv8 model to perform target detection on the image data, recognize the street lamp area in the image, locate the street lamp in the image and mark its position; The abnormal street light recognition module uses image features to analyze the detected street light status, adopts multiple recognition strategies, performs multiple anomaly detections on multiple street light images collected in the same period, and determines the final result through a voting mechanism, thereby identifying abnormal street lights.

2. According to the abnormal street light detection system based on the YOLOv8 model of claim 1, it is characterized in that: The target detection module uses the YOLOv8 model to perform target detection on image data. During the model training process, the channel attention mechanism and the spatial attention mechanism are combined to enhance the network's attention to key features and improve the accuracy of target detection.

3. The abnormal street light detection system based on the YOLOv8 model according to claim 2 is characterized in that: For the channel attention module: first, perform global average pooling and maximum pooling on the input feature map to generate two different feature vectors, which respectively represent the importance of each channel; secondly, pass these two vectors through two fully connected layers, and use ReLU activation function and Sigmoid activation function to generate the attention weight of each channel; Finally, the weights are applied to the input feature map to adjust the influence of each channel.

4. The abnormal street light detection system based on the YOLOv8 model according to claim 2 is characterized in that: For the spatial attention module: first, the channel dimension is compressed to obtain a two-dimensional feature map, which represents the importance of each spatial position; secondly, a convolution operation is applied to the compressed feature map to generate a spatial attention map; finally, the spatial attention map is multiplied with the input feature map to enhance the features of important areas and weaken the influence of unimportant areas.

5. The abnormal street light detection system based on the YOLOv8 model according to claim 1 is characterized in that: The system also includes an alarm module, which outputs the detection results after multiple identifications and voting. If an abnormal street lamp is detected, its location is marked and related data is recorded. The alarm module generates alarm information.

6. The abnormal street light detection system based on the YOLOv8 model according to claim 1, characterized in that: The system also includes a tracking module, which tracks the abnormal street lights detected by the detection model through an improved DeepSORT algorithm.

7. An abnormal street light detection method based on the YOLOv8 model, characterized in that: The method comprises the following steps: Collect images on the road in real time, use the YOLOv8 model to detect targets in the images, and identify and mark street light areas; Perform multi-dimensional feature extraction on the labeled street light area images, and use the attention mechanism to optimize the feature extraction effect and identify possible abnormal street lights; Perform multiple anomaly detections on images collected within the same time period, determine the final result through a voting mechanism, identify abnormal street lights, and mark their location information.

8. The abnormal street light detection method based on the YOLOv8 model according to claim 7, characterized in that: The step of performing multi-dimensional feature extraction on the labeled street lamp area image, optimizing the feature extraction effect by using the attention mechanism, and identifying possible abnormal street lamps specifically includes the following steps: Use a pre-trained convolutional neural network to extract multi-dimensional features from the cropped streetlight area image; An attention mechanism is introduced to dynamically weight the importance of features, highlight key features related to anomaly detection, and suppress irrelevant or redundant features.

9. The abnormal street light detection method based on the YOLOv8 model according to claim 7, characterized in that: The step of performing multiple anomaly detections on images collected within the same time period, judging the final result through a voting mechanism, identifying abnormal street lights, and marking the location information of the abnormal street lights specifically includes the following steps: Perform independent anomaly detection on multiple images collected in the same time period; The final abnormal result is determined through a voting mechanism; According to the final result of the voting mechanism, the status of abnormal street lights is marked.

10. The abnormal street light detection method based on the YOLOv8 model according to claim 7, characterized in that: The method also includes the following steps: using an improved DeepSORT algorithm to track abnormal street lights detected by the detection model.

Images