Method and system for monitoring traffic flow situation in snowfall weather based on deep learning

Abstract

The invention relates to the technical field of traffic detection, and discloses a deep learning-based traffic flow situation monitoring method and system in snowfall weather, and the method comprises the steps: obtaining vehicle operation video data on a road in real time, and carrying out the data processing of the vehicle operation video data; constructing an improved even-number complex wavelet transform network model, and removing snow particles in the video image; removing snow grains in the video image through an image prior algorithm; vehicle detection is carried out, and vehicles in the video image are identified through a yov5 algorithm to obtain a vehicle identification result; and detecting the snow-removed video image through a detection line algorithm, and calculating to obtain a traffic flow parameter. The road operation state is evaluated through the traffic flow parameters, accurate perception of the traffic flow situation in the ice and snow weather is achieved, and the adaptability and practicability of the monitoring system in the complex environment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic detection, and particularly to a method and system for monitoring traffic flow situation under snowfall weather based on deep learning. Background Art

[0002] In the modern traffic system, traffic flow situation monitoring plays a crucial role in ensuring road safety and improving traffic efficiency. Especially under complex meteorological conditions, accurate monitoring is even more important. Snowfall weather, as a common and highly influential adverse meteorological condition, brings many challenges to traffic flow situation monitoring. Currently, the video data of vehicle operations on the road is one of the important sources for obtaining traffic information. However, under snowfall weather, video images are severely disturbed. On the one hand, the presence of snow grains makes the image blurry, obscuring some road and vehicle information. Traditional video processing technologies are difficult to effectively extract the features of key targets such as vehicles in the face of a large amount of snow grain interference, resulting in a significant decrease in the accuracy of subsequent vehicle detection algorithms. For example, in some existing traffic monitoring systems, once encountering moderate to heavy snow weather, the false alarm rate of vehicle detection based on ordinary image recognition algorithms may soar from less than 5% under normal weather to 30% or even higher. On the other hand, the snow pattern phenomenon also has a negative impact on video images. Snow patterns present irregular textures and lines in the image, which are easily confused with road signs, vehicle contours, etc., further increasing the difficulty of image analysis. Some existing image enhancement or denoising algorithms, when dealing with snow pattern interference, either cannot completely remove the snow patterns, resulting in impaired vehicle recognition accuracy; or over-smooth the image while removing the snow patterns, causing the loss of vehicle detail information, which is also not conducive to accurate vehicle detection and traffic flow parameter calculation. In addition, under snowfall weather, traffic flow situation monitoring also faces the problem of algorithm adaptability. Advanced vehicle recognition algorithms such as yolov5 can achieve high-precision vehicle recognition on clear images under normal weather. However, when faced with low-quality images disturbed by snowfall, their performance will be greatly reduced. This is because the training data of these algorithms is mostly based on images under normal weather conditions and lacks sufficient learning and adaptation ability for special interference situations caused by snowfall. Moreover, traditional detection line algorithms rely on clear and accurate vehicle detection results when calculating traffic flow parameters. In the case where vehicle detection is inaccurate due to snowfall, the calculated traffic flow parameters, such as traffic volume, vehicle speed, etc., deviate greatly from the actual traffic conditions and cannot provide reliable data support for traffic management departments, seriously affecting the scientificity and timeliness of traffic decision-making. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a traffic flow situation monitoring method based on deep learning under snowy weather to solve the above problems.

[0004] To solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a traffic flow situation monitoring method based on deep learning under snowy weather, including: obtaining real-time vehicle operation video data on the road and performing data processing on the vehicle operation video data; Based on the vehicle operation video data, constructing an improved dual complex wavelet transform network model to remove snow grains in the video image; Based on the video data after removing snow grains, removing snow patterns in the video image through an image prior algorithm; Performing vehicle detection based on the video data after removing snow patterns, identifying vehicles in the video image through the yolov5 algorithm, and obtaining vehicle recognition results; Based on the vehicle recognition results, detecting the video image after snow removal through a detection line algorithm, and calculating traffic flow parameters.

[0005] As a preferred solution of the traffic flow situation monitoring method based on deep learning under snowy weather according to the present invention, wherein: obtaining real-time vehicle operation video data on the road includes: Obtaining the original video image containing vehicles under actual road snowy weather, processing the video of the actual road, and cutting the video by frame skipping to obtain pictures actually containing vehicles; Performing data annotation on the vehicle pictures, including target categories and target bounding box information, generating a data set, wherein the target categories are vehicles and pedestrians, and the target bounding box is the minimum circumscribed rectangle of the target in the picture; Mixing the processed data set and dividing it into a training set and a validation set according to a preset ratio.

[0006] As a preferred solution of the traffic flow situation monitoring method based on deep learning under snowy weather according to the present invention, wherein: constructing an improved dual complex wavelet transform network model includes: Improving the backbone network part of the dual complex wavelet transform algorithm and introducing the Res2Net network; The Res2Net network consists of a feature extraction unit composed of a convolutional layer, a BN layer, and a ReLu activation function, Adopting the shortcut connection method, and the shortcut connection method is to perform convolution on the output of the first layer network and the final output through the shortcut connection method.

[0007] As a preferred solution of the traffic flow situation monitoring method under snowfall weather based on deep learning according to the present invention, wherein: removing snow grains in the video image includes: Training an improved dual-complex wavelet transform network model through the training set; Normalizing the width and height of the bounding boxes of all targets in the training set, using the normalized width and height as the input of the K-means algorithm, setting the value of K, clustering to obtain the center points, and using them as the prior information of the candidate box size for anchor box generation; In the model training stage, using the annealing algorithm to dynamically adjust the loss function.

[0008] As a preferred solution of the traffic flow situation monitoring method under snowfall weather based on deep learning according to the present invention, wherein: removing snow streaks in the video image through the image prior algorithm includes: Improving the ResNet network, adding a guiding filter in the network structure, separating the original picture into a detail layer and a low-frequency layer, and constructing an improved snow streak removal network model; Training the improved snow streak removal network model through the training set to obtain a convolutional network for snow streak removal, and removing snow streaks from the image data.

[0009] Wherein, the improved snow streak removal network model divides the picture into a detail layer and a low-frequency layer through the guiding filtering algorithm, and uses the ResNet network and negative mapping as the basic structure of the model. The network model is divided into two layers. In the first layer, the input picture is decomposed into a high-frequency detail layer and a low-frequency layer, and the detail layer picture is used as the input of the second layer network. In the second layer, the input detail layer picture passes through the negative mapping network to learn the characteristics of snow streaks in the picture.

[0010] As a preferred solution of the traffic flow situation monitoring method under snowfall weather based on deep learning according to the present invention, wherein: identifying vehicles in the video image through the yolov5 algorithm includes: Setting the detection line position according to the angle and height of the actual traffic video, inputting the actual traffic video data, and identifying vehicles in the video image through the yolov5 algorithm; Wherein, the detection line position is adjusted manually or automatically in combination with the road direction, the camera view range and the vehicle driving trajectory.

[0011] As a preferred solution of the traffic flow situation monitoring method under snowfall weather based on deep learning according to the present invention, wherein: calculating traffic flow parameters includes: Calculating traffic flow parameters according to the vehicle trajectory information output by the vehicle recognition result and the detection line area, and the traffic flow parameters include traffic volume, average vehicle speed, and headway; Among them, the traffic flow volume is calculated by the number of vehicles passing through the detection line per unit time, the average vehicle speed is obtained by dividing the vehicle displacement in consecutive frames by the time interval, and the headway is calculated by the time difference between consecutive vehicles.

[0012] In a second aspect, the present invention provides a traffic flow situation monitoring system under snowy weather based on deep learning, including: A roadside detection module for real-time acquiring vehicle operation video data on a road and performing data processing on the vehicle operation video data; A snow particle removal module for constructing an improved dual complex wavelet transform network model based on the vehicle operation video data to remove snow particles in the video image; A snow pattern removal module for removing snow patterns in the video image through an image prior algorithm based on the video data after snow particle removal; A vehicle detection module for performing vehicle detection based on the video data after snow pattern removal, identifying vehicles in the video image through the yolov5 algorithm, and obtaining vehicle identification results; A traffic flow parameter acquisition module for detecting the video image after snow removal through a detection line algorithm based on the vehicle identification results and calculating traffic flow parameters.

[0013] In a third aspect, the present invention provides a computer device, including: A memory and a processor; The memory is used for storing computer-executable instructions, and the processor is used for executing the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the traffic flow situation monitoring method under snowy weather based on deep learning are implemented.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, and when the computer-executable instructions are executed by a processor, the steps of the traffic flow situation monitoring method under snowy weather based on deep learning are implemented.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: By constructing an improved dual complex wavelet transform network model to remove snow grains and then using an image prior algorithm to remove snow patterns, the present invention can effectively improve the quality of video images in snowy weather, making subsequent vehicle detection and traffic flow parameter calculation more accurate; Based on the high-quality video data after removing snow grains and snow patterns, the yolov5 algorithm is used for vehicle recognition, which can accurately detect vehicles in video images, improve the accuracy of vehicle detection, and reduce misdetection and missed detection; The detection line algorithm is used to detect and calculate traffic flow parameters for the snow-removed video images, and the video images have been effectively de-snowed in the early stage, so that the calculated traffic flow parameters can more truly reflect the actual traffic conditions and provide accurate data support for traffic management and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of the overall process of a traffic flow situation monitoring method in snowy weather based on deep learning according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0019] Refer to Figure 1 , an embodiment of the present invention provides a traffic flow situation monitoring method in snowy weather based on deep learning, including: S101, real-time obtain vehicle operation video data on the road and perform data processing on the vehicle operation video data; S102, based on the vehicle operation video data, construct an improved dual complex wavelet transform network model to remove snow grains in the video image; S103, based on the video data after removing snow grains, remove snow patterns in the video image through an image prior algorithm; S104, perform vehicle detection based on the video data after removing snow patterns, identify the vehicles in the video images through the YOLOv5 algorithm, and obtain the vehicle recognition results; S105, based on the vehicle recognition results, detect the video images after snow removal through the detection line algorithm, and calculate the traffic flow parameters.

[0020] Preferably, the real-time acquisition of vehicle operation video data on the road includes: Obtain the original video images containing vehicles under the actual road snowfall weather, process the video of the actual road, and cut the video frame by frame into pictures actually containing vehicles; Perform data annotation on the vehicle pictures, including the target category and the target bounding box information, and generate a data set. Among them, the target categories are vehicles and pedestrians, and the target bounding box is the minimum circumscribed rectangle of the target in the picture; Mix the processed data set and divide it into a training set and a validation set according to a preset ratio.

[0021] Optionally, the video processing of the actual road can use the Ffmpeg tool to cut the video frame by frame into pictures actually containing vehicles, perform annotation and data augmentation preprocessing on the original images, and divide the data into a training set and a validation set. Among them, in this embodiment, the preset ratio is 7:3, and the validation set does not participate in the training of the model, but only serves as the verification of the model.

[0022] Preferably, constructing an improved Dual-Tree Complex Wavelet Transform (DTCWT) network model includes: Improve the backbone network part of the Dual-Tree Complex Wavelet Transform algorithm and introduce the Res2Net network; The Res2Net network consists of a feature extraction unit composed of a convolutional layer, a BN layer, and a ReLu activation function, which is used to improve the iterative optimization speed of the model; Adopt the shortcut connection method. The shortcut connection method is to perform convolution on the output of the first layer network and the final output through the shortcut connection method, which is used to solve the problem of gradient degradation under multi-layer networks.

[0023] Optionally, the improved Dual-Tree Complex Wavelet Transform (DTCWT) network model is specifically as follows: The improved DTCWT model includes feature extraction and feature fusion. The actual snowfall pictures are processed through feature extraction to output image feature information containing snow grains of different sizes. The features at different levels are fused, and finally the feature information output is obtained. Then, it enters the inverse processing stage, and the inverse wavelet transform is performed on the image for restoration to obtain the actual detection results; Feature extraction consists of a Res2Net network. The input image first passes through a pooling layer to reduce the input parameters, and then passes through the Res2Net network to obtain shallow features X1. The shallow features X1 sequentially pass through a pooling layer, a deconvolution layer, and a module composed of a GCN+BR layer in series to obtain further features X2. The features X2 are connected to the features X1 and sequentially pass through a deconvolution layer and a GCN+BR layer to obtain deep features X3. Feature fusion includes an upsampling layer and a splicing layer. Feature X3 contains 1 low-frequency subband and 6 high-frequency subbands output by wavelet transform. Inverse wavelet transform is performed on the 7 subbands. The convolution kernel sizes for convolution of the low-frequency subband consist of five types: 2*2, 3*3, 5*5, 7*7, and 9*9, and those for the high-frequency subbands consist of three types: 2*2, 3*3, and 5*5.

[0024] Preferably, removing snow grains in video images includes: Training an improved dual complex wavelet transform network model through a training set; Normalizing the width and height of the bounding boxes of all objects in the training set, using the normalized width and height as the input of the K-means algorithm, setting the K value, clustering to obtain the center points, and using them as the prior information of the candidate box size for anchor box generation; In the model training stage, the annealing algorithm is used to dynamically adjust the loss function.

[0025] Specifically, first normalize the width and height of the bounding boxes of all objects (vehicles, pedestrians) in the training set; then use the normalized width and height as the input of the K-means algorithm, set the K value to 9, cluster to obtain 9 center points, and use them as the prior information of the candidate box size for anchor box generation to improve the matching efficiency of the object detection model; in the model training stage, use the annealing algorithm to dynamically adjust the loss function to prevent falling into local optima and improve the global optimization ability of the model; the training parameters are set as follows: the number of training samples per batch batch = 64, the total number of training rounds is 300 rounds, the optimizer is Adam, and the initial learning rate is 0.001.

[0026] Preferably, removing snow streaks in video images through an image prior algorithm includes: Improve the ResNet network, add a guided filter to the network structure, separate the original picture into a detail layer and a low-frequency layer, and construct an improved snow streak removal network model to achieve the purpose of removing snow streaks; Train the improved snow streak removal network model through a training set to obtain a convolutional network for snow streak removal, and perform snow streak removal on the image data.

[0027] Among them, the improved snow pattern removal network model divides the picture into a detail layer and a low-frequency layer through the guided filtering algorithm, and uses the ResNet network and negative mapping as the basic structure of the model. The network model is divided into two layers. In the first layer, the input picture is decomposed into a high-frequency detail layer and a low-frequency layer, and the detail layer picture is used as the input of the second-layer network. In the second layer, the input detail layer picture passes through the negative mapping network to learn the characteristics of the snow patterns in the picture.

[0028] Preferably, the vehicle recognition in the video image by the yolov5 algorithm includes: According to the angle and height of the actual traffic video, set the detection line position, input the actual traffic video data, and use the yolov5 algorithm to recognize the vehicles in the video image to complete the traffic flow monitoring task; Among them, due to the differences in the installation height and angle of the intersection cameras in the actual traffic video, the detection line position is adjusted manually or automatically in combination with the road direction, camera viewing angle range and vehicle driving trajectory.

[0029] Specifically, for the setting of the detection line position, for example: for a straight lane, the detection line can be set in the vertical direction of the lane center line to count the passing vehicles; if the camera is installed diagonally, the diagonal detection line should be set at the position where the field of view is the most covered and the image distortion is the smallest. By analyzing the dense paths of the vehicle trajectories in the video, the main vehicle passing path can be automatically identified and the detection line can be assisted in setting, improving the accuracy of traffic flow monitoring to achieve the traffic flow monitoring task; removing the snow grains from the actual road video, passing through snow pattern removal and vehicle detection in sequence, and finally outputting the two-way traffic flow information, and displaying the processed information and saving the results.

[0030] Preferably, the calculated traffic flow parameters include: According to the vehicle trajectory information and detection line area output by the vehicle recognition result, calculate the traffic flow parameters, and the traffic flow parameters include traffic volume, average vehicle speed, and headway; Among them, the traffic volume is calculated by the number of vehicles passing through the detection line per unit time, the average vehicle speed is obtained by dividing the vehicle displacement in consecutive frames by the time interval, and the headway is calculated by the time difference between consecutive vehicles.

[0031] It should be noted that using these traffic flow parameters to evaluate the road operation state can achieve the accurate perception of the traffic flow situation in ice and snow weather, and improve the adaptability and practicability of the monitoring system in complex environments.

[0032] The present invention can effectively improve the quality of video images in snowy weather by constructing an improved dual complex wavelet transform network model to remove snow grains and then using an image prior algorithm to remove snow streaks, making subsequent vehicle detection and traffic flow parameter calculation more accurate. Based on the high-quality video data after removing snow grains and snow streaks, the yolov5 algorithm is used for vehicle recognition, which can accurately detect vehicles in video images, improve the accuracy of vehicle detection, and reduce misdetection and missed detection. The detection line algorithm is used to detect and calculate traffic flow parameters for the snow-removed video images. Moreover, since the video images have been effectively de-snowed in the early stage, the calculated traffic flow parameters can more truly reflect the actual traffic conditions, providing accurate data support for traffic management and decision-making.

[0033] The above is a schematic solution of a traffic flow situation monitoring method in snowy weather based on deep learning in this embodiment. It should be noted that the technical solution of the traffic flow situation monitoring system in snowy weather based on deep learning belongs to the same concept as the above-mentioned traffic flow situation monitoring method in snowy weather based on deep learning. For the details not described in detail in the technical solution of the traffic flow situation monitoring system in snowy weather based on deep learning in this embodiment, reference can be made to the description of the technical solution of the traffic flow situation monitoring method in snowy weather based on deep learning.

[0034] This embodiment also provides a traffic flow situation monitoring system in snowy weather based on deep learning, including: A roadside detection module for real-time acquisition of vehicle operation video data on the road and data processing of the vehicle operation video data; A snow grain removal module for constructing an improved dual complex wavelet transform network model based on the vehicle operation video data to remove snow grains in the video image; A snow streak removal module for removing snow streaks in the video image through an image prior algorithm based on the video data after removing snow grains; A vehicle detection module for vehicle detection based on the video data after removing snow streaks, and identifying vehicles in the video image through the yolov5 algorithm to obtain vehicle recognition results; A traffic flow parameter acquisition module for detecting the snow-removed video image through the detection line algorithm based on the vehicle recognition results and calculating traffic flow parameters.

[0035] This embodiment also provides a computer device applicable to the situation of traffic flow situation monitoring in snowy weather based on deep learning, including: A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the traffic flow situation monitoring method in snowy weather based on deep learning as proposed in the above embodiment.

[0036] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for monitoring traffic flow situation under snowfall weather based on deep learning as proposed in the above embodiment.

[0037] The storage medium proposed in this embodiment and the method for monitoring traffic flow situation under snowfall weather based on deep learning proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0038] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general-purpose hardware. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disc of a computer, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present invention.

[0039] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for monitoring traffic flow in snowy weather based on deep learning, characterized in that: include: Acquire vehicle operation video data on the road in real time, and perform data processing on the vehicle operation video data; Based on the vehicle operation video data, an improved dual complex wavelet transform network model is constructed to remove snow particles in the video image; Based on the video data after snow particles are removed, the snow patterns in the video images are removed by using an image prior algorithm; Vehicle detection is performed based on the video data after snow patterns are removed, and the vehicle in the video image is identified using the yolov5 algorithm to obtain the vehicle identification result; Based on the vehicle recognition result, the video image after snow removal is detected by a detection line algorithm to calculate the traffic flow parameters.

2. A method for monitoring traffic flow in snowy weather based on deep learning as claimed in claim 1, characterized in that: Real-time acquisition of vehicle operation video data on the road includes: Obtaining original video images containing vehicles on actual roads in snowy weather, processing the video of the actual road, and skipping frames in the video to cut the images that actually contain vehicles; Data annotation is performed on vehicle images, including target categories and target bounding box information, to generate a data set, where the target categories are vehicles and pedestrians, and the target bounding box is the minimum circumscribed rectangular box of the target in the image; The processed data sets are mixed and divided into training set and validation set according to a preset ratio.

3. A method for monitoring traffic flow in snowy weather based on deep learning as claimed in claim 2, characterized in that: The construction of the improved dual complex wavelet transform network model includes: Improve the backbone network of the dual complex wavelet transform algorithm and introduce the Res2Net network; The Res2Net network is a feature extraction unit consisting of a convolutional layer, a BN layer and a ReLu activation function; A shortcut connection method is used, wherein the output of the first layer network is convolved with the final output through a shortcut connection.

4. The method for monitoring traffic flow in snowy weather based on deep learning as claimed in claim 3, characterized in that: Removing snow particles from video images includes: Using the training set, training an improved dual complex wavelet transform network model; Normalize the width and height of the bounding boxes of all objects in the training set, use the normalized width and height as the input of the K-means algorithm, set the K value, cluster to obtain the center point, and use it as the size prior information of the candidate box for anchor box generation; During the model training phase, the annealing algorithm is used to dynamically adjust the loss function.

5. A method for monitoring traffic flow situation in snowy weather based on deep learning as claimed in claim 2 or 4, characterized in that: Removing snow patterns from video images through image prior algorithms includes: Improve the ResNet network by adding a guided filter to the network structure, separate the detail layer and the low-frequency layer of the original image, and build an improved snow pattern removal network model; The snow pattern removal network model is improved by training the training set to obtain a convolutional network for snow pattern removal, and snow pattern removal is performed on image data; Among them, the improved snow pattern removal network model divides the image into a detail layer and a low-frequency layer through a guided filtering algorithm, and adopts a ResNet network and negative mapping as the basic structure of the model. The network model is divided into two layers. In the first layer, the input image is decomposed into a high-frequency detail layer and a low-frequency layer, and the detail layer image is used as the input of the second layer network. In the second layer, the input detail layer image is passed through a negative mapping network to learn the characteristics of snow patterns in the image.

6. The method for monitoring traffic flow situation in snowy weather based on deep learning as claimed in claim 1, characterized in that: The vehicle recognition in the video image by yolov5 algorithm includes: According to the angle and height of the actual traffic video, the detection line position is set, the actual traffic video data is input, and the vehicles in the video image are identified by the yolov5 algorithm; The position of the detection line is adjusted manually or automatically based on the road direction, camera viewing angle and vehicle driving trajectory.

7. A method for monitoring traffic flow in snowy weather based on deep learning as claimed in claim 6, characterized in that: The calculated traffic flow parameters include: Calculate traffic flow parameters according to the vehicle trajectory information and the detection line area outputted from the vehicle recognition result, wherein the traffic flow parameters include vehicle flow, average vehicle speed, and headway; The traffic volume is calculated by the number of vehicles passing through the detection line per unit time, the average vehicle speed is obtained by dividing the vehicle displacement in consecutive frames by the time interval, and the headway is calculated by the time difference between consecutive vehicles.

8. A system for monitoring traffic flow in snowy weather based on deep learning, using a method for monitoring traffic flow in snowy weather based on deep learning as claimed in any one of claims 1 to 7, characterized in that: include, A roadside detection module is used to obtain real-time video data of vehicle operation on the road and perform data processing on the vehicle operation video data; A snow particle removal module is used to construct an improved dual complex wavelet transform network model based on the vehicle operation video data to remove snow particles in the video image; A snow pattern removal module is used to remove snow patterns in video images by using an image prior algorithm based on the video data after snow particles are removed; The vehicle detection module is used to detect vehicles based on the video data after snow patterns are removed, and identify vehicles in the video image through the yolov5 algorithm to obtain vehicle recognition results; The traffic flow parameter acquisition module is used to detect the video image after snow removal through a detection line algorithm based on the vehicle recognition result, and calculate the traffic flow parameters.

9. A computer device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of a method for monitoring traffic flow situation in snowy weather based on deep learning as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of a method for monitoring traffic flow situation in snowy weather based on deep learning as described in any one of claims 1 to 7.

Images