Intelligent street light control method and system based on fuzzy video images
By defuzzing and feature extraction of blurred video images, identifying abnormal road conditions, and adjusting street light brightness using deep learning models, the problem of the inability to adjust street light brightness in the existing technology is solved, and timely response to special situations and energy savings are achieved.
Patent Information
- Application Number
- CN202211407262.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-11-10
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2042-11-10
AI Technical Summary
The prior art cannot adjust the brightness of street lights according to abnormal road conditions, cannot ensure traffic safety in special circumstances, and there is a problem of waste of resources.
By defuzzing and feature extraction of blurred video images, road features are identified, such as traffic accidents, pedestrian falls, fires, severe water accumulation and manhole cover loss, and the street light brightness is adjusted using deep learning models.
It has achieved timely response to abnormal road conditions, ensured traffic safety, saved energy, and avoided waste of resources.
Smart Images

Figure CN115715044B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of street lamp control, and in particular relates to an intelligent street lamp control method and system based on fuzzy video images. Background Art
[0002] Energy conservation and emission reduction are key supports for achieving carbon peak and carbon neutrality. Road lighting consumes a lot of energy. Intelligent analysis of street light monitoring videos can be used to intelligently control the street light system, adjust the lighting in the middle of the night with low traffic flow, and reduce operating power, which is an important means of energy saving in road lighting. In addition, road lighting should be based on the lighting level of the road, the real-time lighting level of the road surface at night, and the changes in traffic flow, vehicle speed, and ambient brightness in different time periods. The lighting level to be achieved in the corresponding period should be determined, and the road illumination or brightness should be adjusted through intelligent control. However, the average illumination of expressways, trunk roads, and secondary trunk roads after adjustment shall not be less than 10lx, and the average illumination of branch roads shall not be less than 8lx. "
[0003] There are two main lighting solutions in the prior art. One is to preset the dimming time node through the built-in controller of the street lamp. After the street lamp is turned on, the street lamp light output is automatically reduced after running for a period of time to achieve the purpose of energy saving. For example: First, the street lamp is set to 100% light output after turning on, and the timing is started at the same time. After running for 4 hours (entering the late night period), the light output is automatically reduced to 50%. Second, when the built-in controller of the street lamp has communication and clock calibration functions, the street lamp is set to run in sections, with 100% light output after turning on the light, reduced to 50% light output after 22:00, reduced to 30% light output after 24:00, and restored to 100% light output at 5 am (if it is not time to turn off the light). Second, a camera device and / or a motion sensor is used to detect whether there is a target object in motion within the preset range of the road. Patent ZL201611131825.2 discloses that if there is a target object in motion within a preset range of a road, then according to the lighting level information of the road, the street lights within the preset range of the road are controlled to increase their brightness and last for a preset period of time. If there is no target object in motion within the preset range of the road, then the street lights within the preset range of the road are controlled to remain unchanged and / or reduce their brightness.
[0004] Although both of the above two solutions can adjust the lighting period and brightness of street lamps, the adjustments of the above solutions are aimed at lighting control under normal road conditions. They cannot adjust the brightness of street lamps according to abnormal road conditions, and cannot ensure road and traffic safety under special circumstances to avoid causing greater waste of resources. Summary of the invention
[0005] In order to solve the problems in the prior art, the present invention provides an intelligent street lamp control method and system based on blurred video images, which can identify and extract road features in blurred videos and adjust the brightness of street lamps according to the identification results, thereby ensuring traffic safety and saving resources.
[0006] In order to achieve the above object, the present invention provides the following technical solutions:
[0007] An intelligent street light control method based on fuzzy video images comprises the following steps:
[0008] S1: Deblurring the blurred video image to obtain a deblurred video image;
[0009] S2: Identify road features in the deblurred video image, and perform feature extraction on different features on the road surface to obtain an image after feature extraction;
[0010] S3: Obtain an image sub-image of the image after feature extraction, compare and calibrate the image sub-image according to preset road surface features, and input the compared and calibrated image sub-image into an existing deep learning training model for learning and training, and adjust the brightness of the road street lamp according to the detection results output by the model.
[0011] Furthermore, the road surface characteristics include traffic accidents, pedestrian falls, fires, severe water accumulation and missing manhole covers.
[0012] Furthermore, according to the detection results output by the model, if any one of the above-mentioned road surface characteristics is detected on the road surface, the control system controls the brightness of the street lamp to increase; if one of the above-mentioned road surface characteristics is not detected on the road surface, the brightness of the street lamp is normal.
[0013] Furthermore, the deblurring process is performed using an image deblurring algorithm.
[0014] Furthermore, the deblurring algorithm includes deblurring and super-resolution processing on the same blurred video image, wherein the deblurring module is used to remove noise in the image and improve the clarity and pixels of the image, and the super-resolution module is used to improve the resolution in the blurred image.
[0015] Furthermore, the image sub-image is an enlarged image after the video image is deblurred.
[0016] Furthermore, the magnification factor of the enlarged image is at least a multiple of the area of the image feature in the enlarged image accounting for half of the area of the entire image.
[0017] Furthermore, when extracting features of different features on the road surface, a feature extraction algorithm is used to extract the preset road surface features in the video image as feature points.
[0018] Furthermore, the method for calibrating the road surface characteristics is:
[0019] When comparing and calibrating the road surface characteristics, historical data including the road surface characteristics are obtained in advance;
[0020] The original image is input into the pre-trained neural network model for training to obtain a feature vector with a length and width that is one quarter of the original video image;
[0021] Upsample the feature vector to obtain an upsampled image;
[0022] The target detection algorithm is used to detect and select features of the upsampled image as the target area;
[0023] A multi-classification algorithm is used on the image sub-image to obtain the probability of each of the above road features and compare it with the pre-acquired probability value. If the probability value exceeds the threshold, the light brightness is enhanced; if it is lower than or equal to the threshold, the light brightness remains unchanged or decreases.
[0024] An intelligent street light control system based on fuzzy video images, comprising:
[0025] A deblurring module, used for deblurring a blurred video image to obtain a deblurred video image;
[0026] The feature recognition and feature extraction module is used to identify the road surface features in the deblurred video image and extract the features of different features on the road surface to obtain an image after feature extraction;
[0027] The brightness adjustment module is used to obtain the image sub-image of the image after feature extraction, calibrate the image sub-image according to the preset road surface features, and input the calibrated image sub-image into the existing deep learning training model for learning and training, and adjust the brightness of the road street lamp according to the detection results output by the model.
[0028] Compared with the prior art, the advantages of the present invention are:
[0029] The present invention discloses an intelligent street lamp control method based on fuzzy video images, which acquires fuzzy video images monitored by street lamps, deblurs the acquired fuzzy video images, extracts features from the deblurred video images, and compares and calibrates the image sub-images after feature extraction to adjust the brightness of street lamps containing special road features, so that the brightness of street lamps changes with the road conditions. On the one hand, it provides timely and sufficient lighting for special conditions on the road surface to ensure the safety of vehicles and pedestrians on the road surface. On the other hand, it can provide feedback to street lamp maintenance personnel through the increase in street lamp brightness, and can timely handle special road conditions such as traffic accidents, pedestrian falls, fires, serious water accumulation, and missing manhole covers. At the same time, through the adjustment of street lamps, energy is reasonably distributed to avoid energy waste. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:
[0031] Figure 1 A schematic diagram of a flow chart of an intelligent street light control method based on fuzzy video images according to the present invention;
[0032] Figure 2 The present invention is a flowchart of a deblurring algorithm for an intelligent street light control method based on blurred video images. DETAILED DESCRIPTION
[0033] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.
[0034] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.
[0035] Example 1
[0036] The present invention provides an intelligent street light control method based on fuzzy video images, such as Figure 1 As shown, the following steps are included:
[0037] S1: Deblurring the blurred video image to obtain a deblurred video image;
[0038] S2: using video detection technology to identify road features in the deblurred image, and extracting features of different features on the road surface to obtain a feature-extracted image;
[0039] S3: Obtain an image sub-image of the image after feature extraction, compare and calibrate the image sub-image according to preset road surface features, and input the compared and calibrated image sub-image into an existing deep learning training model for learning and training, and adjust the brightness of the road street lamp according to the detection results output by the model.
[0040] The blurred video image of the present invention also refers to image blur, which is a very important factor in image degradation. There are many factors that may cause image blur during the image acquisition, transmission, and processing, including blur in the video image caused by the movement of elements in the image, compression blur, blur caused by transmission, and other blurs. For example, when acquiring an image, incorrect focusing will produce out-of-focus blur, the relative movement of the scene and the camera will cause motion blur, and blur caused by high-frequency loss after image compression. Blurring including but not limited to the above-mentioned methods reduces the clarity of the image, seriously affects the image quality, and causes difficulties or even failures in image analysis, processing, and reception. All blurs are referred to in the present invention, not specifically referring to a certain type or category of blur.
[0041] Specifically, blur and clarity are two relative but interrelated concepts that describe the clarity (blur) of an image. The clearer the image, the higher the quality, the greater the clarity and the smaller the blur; the less clear (blurrier) the image, the lower the quality, the smaller the clarity and the greater the blur. Therefore, when describing the clarity of an image, both clarity and blur can be used, but the two indicators are inversely proportional in value; the present invention uses the indicator of image blur as the basis for analyzing and explaining the technical solution of the invention.
[0042] Specifically, road surface characteristics include traffic accidents, pedestrian falls, fires, severe water accumulation and missing manhole covers. Traffic accidents include but are not limited to accidents between vehicles, as well as accidents between vehicles and people. Whether a traffic accident has occurred is determined by the length of time a moving vehicle or pedestrian stays on the road. A traffic accident is determined when a moving vehicle or pedestrian suddenly stops or the speed changes dramatically. The way pedestrians fall is consistent with the method for judging traffic accidents, which means detecting a person in motion. If the speed of a person in motion is 0 within a certain period of time, it is determined that the pedestrian has fallen. If a fire occurs, the color change of the road surface and the maximum value of the concentration of carbon dioxide and other gases in the air on the road surface are used as the basis for judgment to determine whether a fire has occurred on the road surface. Severe water accumulation refers to when the water on the road surface occupies half or more of the road surface width, it is determined to be road water accumulation. The missing manhole cover includes but is not limited to the complete loss of the manhole cover and the manhole cover is not completely set on the ground. The loss of the manhole cover is determined based on the brightness of the ground at night. The road surface is detected using an intelligent digital color sensor or a high-precision digital color sensor. If a fixed area on the road surface is detected to have a dark or too dark color with an obvious black shadow caused by the loss of the manhole cover, it is determined that the manhole cover is lost; specifically, when the color shape on the road surface is a fixed shape or area of a black shadow or pattern such as a circle or an ellipse, it is determined that the manhole cover is lost. If a partial black shadow is detected, such as a black shadow of a semicircular or semi-elliptical structure, it is determined that the manhole cover is not completely set on the ground. Specifically, the optional sensor model is the CL1-N3S1 (NPN) / CL1-P3S1 (PNP) intelligent digital color sensor, or the CL2-N3A1 (NPN) / CL2-P3A1 (PNP) high-precision digital color sensor.
[0043] Specifically, according to the detection results output by the model, if any of the above road surface features is detected on the road surface, the control system controls the brightness of the street lamp to increase; if one of the above road surface features is not detected on the road surface, the brightness of the street lamp is normal. Specifically, when low traffic flow is detected late at night, the lighting is adjusted to reduce the operating power to further save energy. Low traffic in this embodiment means that the vehicle flow does not exceed a preset number or the pedestrian flow does not exceed a preset number within a preset time or a certain period of time, which is low traffic, such as the vehicle flow does not exceed 10 vehicles or the pedestrian flow does not exceed 10 people within an hour.
[0044] Specifically, the control system includes a monitoring system, and the monitoring facilities of the monitoring system are arranged on the traffic light nearest to the street light, such as a monitor, etc.; the picture display device and the image detection and analysis device of the monitoring equipment are arranged in the monitoring room, and the monitoring facilities and the picture display device of the monitoring system are connected wirelessly or by wire, and the monitoring pictures of the monitoring facilities are transmitted in real time through the network to the analysis system or instrument of the picture display device and the picture monitoring equipment for analysis.
[0045] Specifically, the deblurring process is performed using an image deblurring algorithm, including deblurring and super-resolution processing on the same blurred video image, wherein the deblurring module is used to remove noise in the image and improve the clarity and pixels of the image, and the super-resolution module is used to improve the resolution of the blurred image.
[0046] In the initial state, the street lights are in the condition of reducing the operating power in the middle of the night, and the collected video images have the characteristics of blur and low resolution. Directly using video detection technology on such videos will cause large errors and low recognition accuracy, so the present invention adopts a surveillance video deblurring algorithm based on multi-task learning. Among them, multi-task learning refers to multiple tasks sharing the underlying neural network model, which includes the deblurring and super-resolution processing of the same blurred video image in the present invention. The two tasks can be performed simultaneously or in sequence in an arrangement and combination, but the two tasks are based on the same video image for clarity processing, so the use of a multi-task learning mechanism can effectively reduce the number of model parameters. The video deblurring and super-resolution algorithm under the entire low-light condition includes video image feature extraction, super-resolution network branch, and deblurring network branch. Specifically, super-resolution and deblurring share the underlying feature extraction part, and the overall algorithm is: the image to be deblurred is input into the constructed neural network model, and feature extraction is performed on the input image in a preset manner so that the image after feature extraction is one-fourth of the length and width of the original input image, followed by deblurring and super-resolution processing. When performing deblurring, the image after feature extraction of the input neural network is transformed to twice the length and width of the image after feature extraction, that is, transformed to one-half of the original input image, and then the transformed image is deconvolved to restore the details of the image to obtain a deblurred image; when performing super-resolution processing, the neural network model is connected in a manner of multiple residual block length leaps to improve the details and image quality of the image to be super-resolution processed to obtain a super-resolution image; the deblurred image is obtained by combining the super-resolution and deblurred images or analyzing them separately.
[0047] Specifically, the image input to the neural network model is a picture. When the monitored video screen is transmitted to the neural network, the video image to be input is automatically divided into several pictures according to the number of frames, etc., and the picture is input into the neural network in the form of a picture for deblurring. After all the pictures are deblurred, all the deblurred pictures are re-integrated into a complete video image in the order of division for analysis, or individual pictures are analyzed to obtain the final result.
[0048] Specifically, the video image feature extraction in this preferred embodiment takes the original input image size of n×32×100 as an example to illustrate the size change in the feature extraction process, where 32 is the length of the image and 100 is the width of the image. Specifically, the feature extraction module is based on the structure of the convolutional neural network and the residual block for extracting video image features. The feature extraction module is shown in the following table. The input selection adjusts the size of the original input video image to 32x100, and the output obtained by the feature extraction module is a feature vector of 512x8x25, that is, the length and width are converted to 1 / 4 of the original image.
[0049]
[0050] Specifically, in the table: in feature extraction, Kernel represents the convolution kernel size of the convolutional neural network, Stride represents the convolution kernel sampling step, and Padding represents the sampling border size.
[0051] Specifically, Figure 2 As shown in the figure, the depth of convolutional neural networks is very important for image super-resolution. However, image super-resolution networks with deeper channel dimensions are usually difficult to train. Low-resolution input features contain rich low-frequency information, which is treated equally across channels, thus hindering the representation ability of CNN. Therefore, a deep residual network based on the channel attention mechanism is adopted. First, a residual connection module is used to avoid overfitting through the residual mode and pay more attention to the detailed features in the original image. A channel attention mechanism is added to the algorithm, and finally upsampling technology is used to obtain the super-resolution video image.
[0052] Specifically, the deblurring process adopts a multi-scale convolutional neural network, which makes full use of the feature vector obtained by video image feature extraction, and adopts a convolutional network based on residual blocks to convert the feature vector scale into 32x100, that is, the deblurred video image of the same size as the original input. The pixel similarity between the deblurred image and the clear image at the same time of the acquired training set is calculated as the loss function to constrain the model training.
[0053] Specifically, the image sub-image is an enlarged image after the video image is deblurred. The enlargement factor of the enlarged image is at least a multiple of the area of the image feature in the enlarged image accounting for half of the area of the entire image.
[0054] Specifically, when extracting features of different features on the road surface, a feature extraction algorithm is used to extract preset road surface features in the video image as feature points.
[0055] Specifically, the method for calibrating the road surface features is as follows: when comparing and calibrating the road surface features, historical data containing the above-mentioned road surface features are obtained in advance; the original image is input into a pre-trained neural network model for training to obtain a feature vector whose length and width are one-fourth of the original video image; the feature vector is up-sampled to obtain an up-sampled image; a target detection algorithm is used to detect and select features from the up-sampled image as a target area; and a multi-classification algorithm is used for the image sub-image to obtain the probability of each of the above-mentioned road surface features and compare it with the pre-acquired probability value. If the probability value exceeds a threshold, the light brightness is enhanced; if it is lower than or equal to the threshold, the light brightness remains unchanged or decreases.
[0056] Specifically, the historical data is the historical data values of the road surface characteristics included in this application occurring on the detected road surface, including data values based on traffic accident statistics and data values based on weather condition statistics; the above threshold is the average value of the historical data values.
[0057] Specifically, according to common abnormal situations such as traffic accidents, pedestrian falls, fires, serious water accumulation, missing manhole covers, etc., image detection technology is first used to obtain the target sub-image in the video image. The current common image target detection algorithm can complete the detection and selection of the target area, and then the sub-image is used with a multi-classification algorithm to obtain the probability of each abnormal situation, and the threshold is set according to the probability value under historical abnormal situations. When the probability predicted by the multi-classification model exceeds the threshold, it is considered that an abnormal situation has occurred on the scene, and the lighting control mechanism is triggered to automatically brighten the scene lights, and a message is sent to the on-duty personnel for the first time to remind them. When the lighting control mechanism is triggered after passing through the abnormal detection module, the brightness of the on-site street lights is increased, and the clarity of the video acquisition image is increased. The image detection technology and multi-classification module in the abnormal detection module are called again for the higher-definition video image to obtain the probability of the abnormal event. If it exceeds the threshold again, the on-duty personnel are reminded to pay attention to the abnormal event again. The brightness of the street lights continues until the probability of the abnormal event is less than the threshold, and then the brightness of the lights is automatically reduced after a period of time.
[0058] Specifically, the method of the present invention can address the shortcomings of the prior art, namely, dimming control based only on normal road traffic behavior, lack of recognition, automatic judgment and intelligent dimming of abnormal situations, and inability to effectively detect and obtain abnormal situations in blurry videos under low-light conditions, and timely adjust the brightness of street lights corresponding to special road conditions. It can make timely responses and feedback to road conditions, save energy, and at the same time give reminders of abnormal road conditions, thereby ensuring traffic safety under special road conditions.
[0059] Example 2
[0060] An intelligent street light control system based on fuzzy video images, comprising:
[0061] A deblurring module, used for deblurring a blurred video image to obtain a deblurred video image;
[0062] The feature recognition and feature extraction module is used to identify the road surface features in the deblurred video image and extract the features of different features on the road surface to obtain an image after feature extraction;
[0063] The brightness adjustment module is used to obtain the image sub-image of the image after feature extraction, calibrate the image sub-image according to the preset road surface features, and input the calibrated image sub-image into the existing deep learning training model for learning and training, and adjust the brightness of the road street lamp according to the detection results output by the model.
[0064] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0065] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided 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 generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0066] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. An intelligent street light control method based on fuzzy video images, characterized in that: The following steps are included: S1: Deblurring the blurred video image to obtain a deblurred video image; S2: Identify road features in the deblurred video image, and perform feature extraction on different features on the road surface to obtain an image after feature extraction; S3: Obtain an image sub-image of the image after feature extraction, compare and calibrate the image sub-image according to the preset road surface features, and input the compared and calibrated image sub-image into the existing deep learning training model for learning and training, and adjust the brightness of the road street lamp according to the detection results output by the model; Among them, the road surface characteristics include traffic accidents, pedestrian falls, fires, severe water accumulation and missing manhole covers; According to the detection results output by the model, if any of the above road surface features is detected on the road surface, the control system controls the brightness of the street lamp to increase; if none of the above road surface features is detected on the road surface, the brightness of the street lamp is normal; The image sub-image is an enlarged image after the video image is deblurred.
2. According to claim 1, the intelligent street light control method based on fuzzy video images is characterized in that: The deblurring process is performed using an image deblurring algorithm.
3. The intelligent street light control method based on fuzzy video images according to claim 2 is characterized in that: The magnification factor of the enlarged image is at least a multiple of the image feature area in the enlarged image accounting for half of the overall image area.
4. The intelligent street light control method based on fuzzy video images according to claim 1 is characterized in that: When extracting features of different features on the road surface, a feature extraction algorithm is used to extract the preset road surface features in the video image as feature points.
5. The intelligent street light control method based on fuzzy video images according to claim 1 is characterized in that: The method for calibrating the road surface characteristics is: When comparing and calibrating the road surface characteristics, historical data including the road surface characteristics are obtained in advance; The original image is input into the pre-trained neural network model for training to obtain a feature vector with a length and width that is one quarter of the original video image; Upsample the feature vector to obtain an upsampled image; The target detection algorithm is used to detect and select features of the upsampled image as the target area; A multi-classification algorithm is used on the image sub-image to obtain the probability of each of the above road features and compare it with the pre-acquired probability value. If the probability value exceeds the threshold, the light brightness is enhanced; if it is lower than or equal to the threshold, the light brightness remains unchanged or decreases.
6. An intelligent street light control system based on fuzzy video images, characterized in that: include, A deblurring module, used for deblurring a blurred video image to obtain a deblurred video image; A feature recognition and feature extraction module is used to recognize road features in the deblurred video image and extract features of different features on the road surface to obtain an image after feature extraction; A brightness adjustment module is used to obtain an image sub-image of the image after feature extraction, calibrate the image sub-image according to preset road surface features, and input the calibrated image sub-image into an existing deep learning training model for learning and training, and adjust the brightness of the road street lamp according to the detection results output by the model; Among them, the road surface characteristics include traffic accidents, pedestrian falls, fires, severe water accumulation and missing manhole covers; According to the detection results output by the model, if any of the above road surface features is detected on the road surface, the control system controls the brightness of the street lamp to increase; if none of the above road surface features is detected on the road surface, the brightness of the street lamp is normal; The image sub-image is an enlarged image after the video image is deblurred.
