Methods, devices, equipment, and media for controlling ventilation fans based on tunnel environmental monitoring.
By identifying vehicle models and monitoring exhaust emissions, the operating status of tunnel fans can be controlled, solving the problem of the inability to personalize ventilation in long-distance tunnels and reducing energy consumption and operating costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHONGQING GUOXIANG NEW MATERIAL
- Filing Date
- 2023-02-08
- Publication Date
- 2026-05-26
AI Technical Summary
The ventilation and air exchange methods in long tunnels cannot be customized according to the tunnel environment, resulting in high energy consumption and high maintenance and operating costs.
By acquiring vehicle images and spectral imaging images, convolutional processing and deep residual networks are used to identify vehicle models, and exhaust emissions are monitored based on vehicle exhaust models to control the working status of tunnel ventilation fans.
It enables personalized control of the ventilation system based on the tunnel environment, reducing energy consumption and operating costs.
Smart Images
Figure CN116044800B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automated energy-saving control technology, and in particular to a method, apparatus, equipment and medium for controlling fans based on tunnel environment monitoring. Background Technology
[0002] With the continuous expansion of transportation infrastructure construction, tunnel construction has advanced rapidly, bringing tremendous convenience to travel. During this rapid development, long-distance tunnels require better ventilation equipment and more energy-efficient ventilation control. Currently, ventilation systems in long-distance tunnels require the entire system to be activated regardless of whether ventilation is needed, resulting in high energy consumption and maintenance and operating costs. Furthermore, these ventilation methods cannot provide individualized treatment for heavily polluted areas, cannot reduce ventilation volume for slightly polluted areas, and perform ineffective ventilation in unpolluted areas, leading to energy waste and increased costs. Therefore, a more reasonable technical solution is needed to address the problems of existing technologies. Summary of the Invention
[0003] To address the problems of high energy consumption and high maintenance and operating costs associated with existing ventilation methods for long tunnels, which cannot be customized to the tunnel environment.
[0004] In a first aspect, embodiments of the present invention propose a method for controlling a ventilation fan based on tunnel environment monitoring, the method comprising:
[0005] Acquire vehicle model image set and vehicle image and spectral imaging image of the vehicle corresponding to the vehicle entering the tunnel. The vehicle model image set includes sample images of several different vehicles.
[0006] Each vehicle image is convolved to obtain the first local feature vector of the vehicle model. Each vehicle model sample image in the model image set is also convolved to obtain the second local feature vector of the vehicle model. The intersection of the second local feature vectors of each vehicle model sample image is calculated and used as the common feature vector of the vehicles. The difference between the first local feature vector of the vehicle model and the common feature vector of the vehicles is calculated and used as the vehicle dissimilar feature vector of the vehicle image. The vehicle dissimilar feature vector is then input into the vehicle model recognition model to identify the corresponding vehicle model.
[0007] The spectral imaging image of each vehicle is used as input. The vehicle exhaust model is used to monitor exhaust emissions according to the vehicle model to obtain exhaust emission monitoring values. The working status of the tunnel ventilation fan is then controlled based on the exhaust emission monitoring values.
[0008] Preferably, each vehicle image is convolved to obtain a first vehicle model local feature vector corresponding to the vehicle image, including:
[0009] Obtain the convolution kernel function and determine its scale and orientation;
[0010] Based on the scale and orientation of the convolution kernel function, several feature extraction kernel functions are obtained by combination;
[0011] Each feature extraction kernel function is convolved with the vehicle image to obtain the first local feature sub-vector of the vehicle corresponding to each feature extraction kernel function;
[0012] The dimensionality of each local feature vector of the first vehicle model is reduced, and the reduced local feature vectors of the first vehicle model are used to form a column vector, which is then used as the local feature vector of the first vehicle model.
[0013] Preferably, the step of acquiring a vehicle image includes:
[0014] Obtain a first training dataset, wherein the first training dataset includes multiple first acquired images for training and a first label corresponding to each first acquired image for training;
[0015] The first label is used to indicate whether a vehicle exists in the first acquired image used for training; an initial vehicle recognition model based on a deep residual network is constructed; the initial vehicle recognition model is trained using the first training dataset as input to obtain the vehicle recognition model.
[0016] Preferably, before using the spectral imaging image of each vehicle as input and performing exhaust emission monitoring based on the vehicle's corresponding model using a vehicle exhaust emission model, the method further includes training a vehicle exhaust emission model corresponding to the vehicle model, including:
[0017] Obtain a second training dataset, which includes multiple training spectral imaging images containing vehicle models and a second label corresponding to each training spectral imaging image;
[0018] The second label is the vehicle exhaust corresponding to the spectral imaging image used for training;
[0019] Construct an initial vehicle exhaust gas model based on a deep residual network;
[0020] The initial vehicle exhaust model is trained using the second training dataset as input to obtain the vehicle exhaust model.
[0021] Secondly, embodiments of the present invention propose a device for controlling a ventilation fan based on tunnel environment monitoring, the device comprising:
[0022] The image acquisition module is used to acquire vehicle image sets and vehicle images and spectral imaging images of vehicles entering the tunnel. The vehicle image set includes sample images of several different vehicles.
[0023] The vehicle model identification module performs convolution processing on each vehicle image to obtain a first local feature vector of the vehicle model corresponding to the vehicle image, and performs convolution processing on each vehicle model sample image in the vehicle model image set to obtain a second local feature vector of the vehicle model corresponding to each vehicle model sample image; it calculates the intersection of the second local feature vectors of the vehicle models corresponding to each vehicle model sample image, and uses the intersection as the common feature vector of the vehicles; it calculates the difference between the first local feature vector of the vehicle model and the common feature vector of the vehicles, and uses the difference as the vehicle dissimilar feature vector of the vehicle image; it inputs the vehicle dissimilar feature vector into the vehicle model recognition model to identify the vehicle model corresponding to the vehicle.
[0024] The infrared monitoring module takes the spectral image of each vehicle as input, monitors the exhaust emissions according to the vehicle model, obtains the exhaust emission monitoring value, and controls the working status of the tunnel ventilation fan based on the exhaust emission monitoring value.
[0025] Preferably, the vehicle model determination module includes: a processing unit, the processing unit being used for
[0026] Obtain the convolution kernel function and determine its scale and orientation;
[0027] Based on the scale and orientation of the convolution kernel function, several feature extraction kernel functions are obtained by combination;
[0028] Each feature extraction kernel function is convolved with the vehicle image to obtain the first local feature sub-vector of the vehicle corresponding to each feature extraction kernel function;
[0029] The dimensionality of each local feature vector of the first vehicle model is reduced, and the reduced local feature vectors of the first vehicle model are used to form a column vector, which is then used as the local feature vector of the first vehicle model.
[0030] Preferably, the image acquisition module includes: a vehicle model image unit, which is used for...
[0031] Obtain a first training dataset, wherein the first training dataset includes multiple first acquired images for training and a first label corresponding to each first acquired image for training;
[0032] The first label is used to indicate whether a vehicle exists in the first acquired image used for training; an initial vehicle recognition model based on a deep residual network is constructed; the initial vehicle recognition model is trained using the first training dataset as input to obtain the vehicle recognition model.
[0033] Preferably, it further includes a vehicle exhaust model module; wherein, the vehicle exhaust model module is used for
[0034] Obtain a second training dataset, which includes multiple training spectral imaging images containing vehicle models and a second label corresponding to each training spectral imaging image;
[0035] The second label is the vehicle exhaust corresponding to the spectral imaging image used for training;
[0036] Construct an initial vehicle exhaust gas model based on a deep residual network;
[0037] The initial vehicle exhaust model is trained using the second training dataset as input to obtain the vehicle exhaust model.
[0038] Thirdly, embodiments of the present invention provide a computer device including a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the method for controlling a fan based on tunnel environment monitoring as proposed in the above embodiments.
[0039] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing instructions that, when executed on a computer, perform the method for controlling a fan based on tunnel environment monitoring as proposed in the above embodiments.
[0040] Beneficial Effects: This technical solution acquires vehicle images of vehicles entering tunnels, their corresponding spectral imaging images, and a pre-stored set of vehicle model images. It extracts first local features of the vehicle model from the vehicle images, where these first local features are equivalent to the layout characteristics and shape information of the vehicle body. It also extracts second local features from each sample image of a vehicle model in the image set, using these extracted features to identify the vehicle model. The working principle of this invention for vehicle model identification is that different vehicle models necessarily contain both common and dissimilar features, and model information is contained within these dissimilar features. Therefore, in specific implementation, the intersection of all second local features of the vehicle model is taken as the common feature of the vehicle. Then, during model identification, the difference between the first local features of the vehicle model and the common features is taken as the vehicle's dissimilar feature. Finally, the vehicle dissimilar feature is input into the model identification model to obtain the vehicle model of the vehicle entering the tunnel. After obtaining the vehicle model information of the vehicles entering the tunnel, it is also necessary to obtain the corresponding exhaust emission data. By acquiring spectral imaging images, the vehicle's exhaust emission amount can be obtained, allowing for monitoring of the vehicle's exhaust emissions. Based on the monitoring data, exhaust emission monitoring values are obtained, and then the ventilation fans in the tunnel are controlled according to these values. By setting up multiple monitoring points to monitor the exhaust emission data of the corresponding vehicle model, the amount of polluting gases in the tunnel can be obtained more accurately. This allows for the control of the operation of the ventilation fans set at different monitoring points based on the exhaust emission monitoring data. This solves the problems of existing ventilation methods for long-distance tunnels, which cannot be customized according to the tunnel environment, resulting in high energy consumption and high maintenance and operating costs. Attached Figure Description
[0041] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:
[0042] Figure 1 This is a flowchart of a method for controlling a ventilation fan based on tunnel environmental monitoring;
[0043] Figure 2 This is a schematic diagram of the functional modules of a device for controlling a ventilation fan based on tunnel environmental monitoring.
[0044] Figure 3 This is a schematic diagram of the functional modules of another device for controlling fans based on tunnel environmental monitoring;
[0045] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation
[0046] To more clearly illustrate the technical solutions in the embodiments of the invention or the prior art, the invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the invention, but does not constitute a limitation on the invention.
[0047] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.
[0048] For the first aspect, please refer to... Figure 1This invention proposes a method for controlling ventilation fans based on tunnel environment monitoring, applied to a server connected to multiple sets of ventilation fans. The server can be, but is not limited to, a computer device with certain computing resources, such as a personal computer (PC, referring to a multi-purpose computer of size, price, and performance suitable for personal use; desktop computers, laptops, mini-laptops, tablets, and ultrabooks all fall under the category of personal computers), a smartphone, or a personal digital assistant (PDA). Electronic devices such as assistants, PADs, or platform servers are used to acquire vehicle image sets and vehicle images and spectral imaging images of vehicles entering the tunnel. The vehicle image set includes sample images of several different vehicle models. Each vehicle image is convolved to obtain a first local feature vector of the vehicle model. Similarly, each sample image in the vehicle image set is convolved to obtain a second local feature vector of the vehicle model. The intersection of the second local feature vectors of each sample image is calculated and used as the common feature vector of the vehicles. The difference between the first local feature vector and the common feature vector is calculated and used as the vehicle's distinct feature vector. This distinct feature vector is input into a vehicle model to identify the corresponding vehicle model. The spectral imaging image of each vehicle is used as input, and a vehicle exhaust model monitors exhaust emissions based on the vehicle model to obtain exhaust emission values. The operating status of the tunnel ventilation fan is then controlled based on these values. This solves the problems of existing ventilation methods for long-distance tunnels, which cannot be customized according to the tunnel environment, resulting in high energy consumption and high maintenance and operating costs.
[0049] It is understood that the aforementioned execution entity does not constitute a limitation on the embodiments of this application. Accordingly, the operation steps of this method may be, but are not limited to, the steps S11 to S14 below.
[0050] Step S11: Obtain the vehicle model image set and the vehicle image and spectral imaging image of the vehicle corresponding to the vehicle entering the tunnel. The vehicle model image set includes sample images of several different vehicles.
[0051] In practice, multiple monitoring points are set up inside the tunnel, each with a corresponding monitoring area. Real-time vehicle images and spectral imaging images are acquired as vehicles pass through the monitoring area after entering the tunnel. The vehicle images are used to identify the vehicle model, and the spectral imaging images are used to obtain the real-time exhaust emissions of that model after identification. To monitor whether a vehicle has entered the monitoring area, vehicle sensors can be installed. These sensors include, but are not limited to, vehicle microwave sensors, vehicle infrared sensors, or image recognition sensors to acquire vehicle sensing signals, thereby enabling the acquisition of vehicle images and spectral imaging images as soon as a vehicle enters the monitoring area.
[0052] Step S12: Perform convolution processing on each vehicle image to obtain the first local feature vector of the vehicle model corresponding to the vehicle image; perform convolution processing on each vehicle model sample image in the vehicle model image set to obtain the second local feature vector of the vehicle model corresponding to each vehicle model sample image; calculate the intersection between the second local feature vectors of the vehicle models corresponding to each vehicle model sample image, and use the intersection as the common feature vector of the vehicles; calculate the difference between the first local feature vector of the vehicle model and the common feature vector of the vehicles, and use the difference as the vehicle dissimilar feature vector corresponding to the vehicle image; input the vehicle dissimilar feature vector into the vehicle model recognition model to identify the vehicle model corresponding to the vehicle.
[0053] Once a vehicle is detected entering the monitoring area, its image and spectral imaging are acquired to identify the vehicle model. After confirming the model, the vehicle's real-time exhaust emissions are monitored. In this embodiment, model identification involves first extracting a first local feature vector from the model image and a second local feature vector from each model sample image in the sample image set. The extracted local feature vectors are then used for model identification. The model image set includes sample images of several different models. Vehicle images can be, but are not limited to, captured by a camera. This camera is installed near the monitoring area to acquire vehicle images within that area. After acquiring the model image set and the vehicle images, local vehicle features are extracted to enable model identification.
[0054] Step S13: Take the spectral imaging image of each vehicle as input, and use the vehicle exhaust model to monitor exhaust emissions according to the vehicle model.
[0055] After obtaining the vehicle type information of the vehicles entering the tunnel, the spectral imaging image of each vehicle is used as input. The vehicle exhaust model is used to monitor exhaust emissions according to the vehicle type to obtain the exhaust emission data of the corresponding vehicle. By obtaining the spectral imaging image, the exhaust emission amount of the vehicle can be obtained, and then the exhaust emission amount of the vehicle can be monitored. The exhaust monitoring value is obtained based on the monitoring data, and then the ventilation fan in the tunnel is controlled based on the monitoring value.
[0056] Step S14: Obtain the exhaust gas monitoring value and control the working status of the tunnel ventilation fan based on the exhaust gas monitoring value.
[0057] In this technical solution, vehicle images of vehicles entering tunnels, their corresponding spectral imaging images, and a pre-stored set of vehicle model images are acquired. First local features of the vehicle model are extracted from the vehicle images; these first local features are equivalent to the layout characteristics and shape information of the vehicle body. Second local features of the vehicle model are also extracted from each sample image of the vehicle model in the image set. These extracted features are then used for vehicle model identification. The working principle of this invention for vehicle model identification is that different vehicle models necessarily contain both common and dissimilar features, and model information is contained within these dissimilar features. Therefore, in specific implementation, the intersection of all second local features of the vehicle model is taken as the common feature of the vehicle. Then, during model identification, the difference between the first local features of the vehicle model and the common features is taken as the vehicle's dissimilar feature. Finally, the vehicle dissimilar feature is input into the model identification model to obtain the vehicle model of the vehicle entering the tunnel. After obtaining the vehicle model information of the vehicles entering the tunnel, it is also necessary to obtain the corresponding exhaust emission data. By acquiring spectral imaging images, the vehicle's exhaust emission amount can be obtained, allowing for monitoring of the vehicle's exhaust emissions. Based on the monitoring data, exhaust emission monitoring values are obtained, and then the ventilation fans in the tunnel are controlled according to these values. By setting up multiple monitoring points to monitor the exhaust emission data of the corresponding vehicle model, the amount of polluting gases in the tunnel can be obtained more accurately. This allows for the control of the operation of the ventilation fans set at different monitoring points based on the exhaust emission monitoring data. This solves the problems of existing ventilation methods for long-distance tunnels, which cannot be customized according to the tunnel environment, resulting in high energy consumption and high maintenance and operating costs.
[0058] Preferably, each vehicle image is convolved to obtain a first vehicle model local feature vector corresponding to the vehicle image, including:
[0059] Obtain the convolution kernel function and determine its scale and orientation;
[0060] Based on the scale and orientation of the convolution kernel function, several feature extraction kernel functions are obtained by combination;
[0061] Each feature extraction kernel function is convolved with the vehicle image to obtain the first local feature sub-vector of the vehicle corresponding to each feature extraction kernel function;
[0062] The dimensionality of each local feature vector of the first vehicle model is reduced, and the reduced local feature vectors of the first vehicle model are used to form a column vector, which is then used as the local feature vector of the first vehicle model.
[0063] Based on the above-disclosed content, this invention discloses a specific process for feature extraction, namely, using convolution kernel functions of different scales and directions to convolve the vehicle image, thereby obtaining multiple local feature vectors of the first vehicle model. Then, each local feature vector of the first vehicle model is subjected to dimensionality reduction processing. Finally, the dimensionality-reduced local feature vectors of the first vehicle model are concatenated to obtain the local feature vector of the first vehicle model. In essence, it uses the dimensionality-reduced local feature vectors of the first vehicle model to form a column vector, and the column vector formed is the local feature vector of the first vehicle model. Correspondingly, each feature extraction kernel function is convolved with the vehicle image to obtain the local feature vector of the first vehicle model corresponding to each feature extraction kernel function.
[0064] Calculate the difference between the second local feature vector of each vehicle sample image and the common feature vector of the vehicle to obtain the vehicle dissimilar feature sample vector corresponding to each vehicle sample image;
[0065] Obtain the vehicle model corresponding to each vehicle sample image, and use the vehicle model of each vehicle sample image and the vehicle dissimilar feature sample vector corresponding to each vehicle sample image to form a training dataset, wherein each vehicle dissimilar feature sample vector in the training dataset corresponds to a vehicle model.
[0066] Using the training dataset as input and the predicted age corresponding to each vehicle dissimilar feature sample vector in the training dataset as output, an RBF neural network is trained. During the training process, the network weights of the RBF neural network are iteratively updated so that the vehicle recognition model is obtained when the iteration termination condition is met.
[0067] Preferably, the step of acquiring a vehicle image includes:
[0068] Obtain a first training dataset, wherein the first training dataset includes multiple first acquired images for training and a first label corresponding to each first acquired image for training;
[0069] The first label is used to indicate whether a vehicle exists in the first acquired image used for training; an initial vehicle recognition model based on a deep residual network is constructed; the initial vehicle recognition model is trained using the first training dataset as input to obtain the vehicle recognition model.
[0070] A first training dataset is obtained. This dataset includes multiple first spectral imaging images for training and a first label corresponding to each image. The first label indicates whether a vehicle exists in the first spectral imaging image. Specifically, in this embodiment, the first training dataset is a dataset of spectral imaging images containing vehicles; that is, a portion of the first training dataset contains spectral imaging images containing vehicles, while the remaining dataset does not contain vehicles. The method for generating the first training dataset is as follows: a spectral imaging module periodically acquires spectral imaging images containing vehicles and those not containing vehicles, while simultaneously recording whether a vehicle exists in the acquired spectral imaging image. The result of whether a vehicle exists is used as the first label corresponding to the spectral imaging image, and the first training dataset is generated based on the correspondence between the spectral imaging images and the first label. For example, the first training dataset contains two categories: the first category contains spectral imaging images containing vehicles, and the second category does not contain vehicles. The first training dataset is used to train a deep residual network model for predicting whether a vehicle is present in the acquired spectral imaging images, i.e., to train a vehicle recognition model.
[0071] Preferably, before using the spectral imaging image of each vehicle as input and performing exhaust emission monitoring based on the vehicle's corresponding model using a vehicle exhaust emission model, the method further includes training a vehicle exhaust emission model corresponding to the vehicle model, including:
[0072] Obtain a second training dataset, which includes multiple training spectral imaging images containing vehicle models and a second label corresponding to each training spectral imaging image;
[0073] The second label is the vehicle exhaust corresponding to the spectral imaging image used for training;
[0074] Construct an initial vehicle exhaust gas model based on a deep residual network;
[0075] The initial vehicle exhaust model is trained using the second training dataset as input to obtain the vehicle exhaust model.
[0076] The second training dataset is a dataset of spectral imaging images of vehicle models under different vehicle exhaust emission conditions. This dataset is used to train a deep residual network model for predicting different vehicle exhaust emissions, i.e., to train a vehicle exhaust emission prediction model. In this embodiment, vehicle exhaust emissions can include low emissions, high emissions, and normal emissions. Therefore, the second training dataset is used to train a deep residual network model to distinguish between these three categories of vehicle exhaust emissions. The method for generating the second training dataset is as follows: A spectral imaging module periodically acquires spectral imaging images containing vehicle models, and simultaneously records the subjective vehicle exhaust emission feedback during the acquisition of the spectral imaging images. The vehicle exhaust emission is used as the category label corresponding to the spectral imaging image, i.e., as the second label corresponding to the spectral imaging image. The second training dataset is generated based on the correspondence between the spectral imaging image and the second label.
[0077] For the second aspect, please refer to... Figures 2 to 3 The present invention provides a device 100 for controlling a ventilation fan based on tunnel environment monitoring, comprising:
[0078] The image acquisition module 110 is used to acquire a vehicle model image set and vehicle images and spectral imaging images of the vehicles entering the tunnel. The vehicle model image set includes sample images of several different vehicles.
[0079] Multiple monitoring points are set up inside the tunnel, each with a corresponding monitoring area. Real-time vehicle images and spectral imaging images are acquired as vehicles pass through the monitoring area after entering the tunnel. The vehicle images are used to identify the vehicle model, and the spectral imaging images are used to obtain the real-time exhaust emissions of that model after identification. To monitor whether a vehicle has entered the monitoring area, vehicle sensors can be installed. These sensors include, but are not limited to, vehicle microwave sensors, vehicle infrared sensors, or image recognition sensors to acquire vehicle sensing signals, thus enabling the acquisition of vehicle images and spectral imaging images as soon as a vehicle enters the monitoring area.
[0080] The vehicle model identification module 120 is used to perform convolution processing on each vehicle image to obtain the first local feature vector of the vehicle model corresponding to the vehicle image, and to perform convolution processing on each vehicle model sample image in the vehicle model image set to obtain the second local feature vector of the vehicle model corresponding to each vehicle model sample image; to calculate the intersection between the second local feature vectors of each vehicle model sample image, and to use the intersection as the common feature vector of the vehicle; to calculate the difference between the first local feature vector of the vehicle model and the common feature vector of the vehicle, and to use the difference as the vehicle dissimilar feature vector corresponding to the vehicle image; and to input the vehicle dissimilar feature vector into the vehicle model recognition model to identify the vehicle model corresponding to the vehicle.
[0081] Once a vehicle is detected entering the monitoring area, its image and spectral imaging are acquired to identify the vehicle model. After confirming the model, the vehicle's real-time exhaust emissions are monitored. In this embodiment, model identification involves first extracting a first local feature vector from the model image and a second local feature vector from each model sample image in the sample image set. The extracted local feature vectors are then used for model identification. The model image set includes sample images of several different models. Vehicle images can be, but are not limited to, captured by a camera. This camera is installed near the monitoring area to acquire vehicle images within that area. After acquiring the model image set and the vehicle images, local vehicle features are extracted to enable model identification.
[0082] The spectral monitoring module 130 is used to take the spectral imaging image of each vehicle as input and perform exhaust gas monitoring according to the vehicle model based on the corresponding vehicle type.
[0083] After obtaining the vehicle type information of the vehicles entering the tunnel, the spectral imaging image of each vehicle is used as input. The vehicle exhaust model is used to monitor exhaust emissions according to the vehicle type to obtain the exhaust emission data of the corresponding vehicle. By obtaining the spectral imaging image, the exhaust emission amount of the vehicle can be obtained, and then the exhaust emission amount of the vehicle can be monitored. The exhaust monitoring value is obtained based on the monitoring data, and then the ventilation fan in the tunnel is controlled based on the monitoring value.
[0084] The ventilation control module 140 is used to acquire exhaust gas monitoring values and control the working status of the tunnel ventilation fan based on the exhaust gas monitoring values.
[0085] In this technical solution, vehicle images of vehicles entering tunnels, their corresponding spectral imaging images, and a pre-stored set of vehicle model images are acquired. First local features of the vehicle model are extracted from the vehicle images; these first local features are equivalent to the layout characteristics and shape information of the vehicle body. Second local features of the vehicle model are also extracted from each sample image of the vehicle model in the image set. These extracted features are then used for vehicle model identification. The working principle of this invention for vehicle model identification is that different vehicle models necessarily contain both common and dissimilar features, and model information is contained within these dissimilar features. Therefore, in specific implementation, the intersection of all second local features of the vehicle model is taken as the common feature of the vehicle. Then, during model identification, the difference between the first local features of the vehicle model and the common features is taken as the vehicle's dissimilar feature. Finally, the vehicle dissimilar feature is input into the model identification model to obtain the vehicle model of the vehicle entering the tunnel. After obtaining the vehicle model information of the vehicles entering the tunnel, it is also necessary to obtain the corresponding exhaust emission data. By acquiring spectral imaging images, the vehicle's exhaust emission amount can be obtained, allowing for monitoring of the vehicle's exhaust emissions. Based on the monitoring data, exhaust emission monitoring values are obtained, and then the ventilation fans in the tunnel are controlled according to these values. By setting up multiple monitoring points to monitor the exhaust emission data of the corresponding vehicle model, the amount of polluting gases in the tunnel can be obtained more accurately. This allows for the control of the operation of the ventilation fans set at different monitoring points based on the exhaust emission monitoring data. This solves the problems of existing ventilation methods for long-distance tunnels, which cannot be customized according to the tunnel environment, resulting in high energy consumption and high maintenance and operating costs.
[0086] Preferably, the vehicle model determination module 120 includes: a processing unit 121, which is used for...
[0087] Obtain the convolution kernel function and determine its scale and orientation;
[0088] Based on the scale and orientation of the convolution kernel function, several feature extraction kernel functions are obtained by combination;
[0089] Each feature extraction kernel function is convolved with the vehicle image to obtain the first local feature sub-vector of the vehicle corresponding to each feature extraction kernel function;
[0090] The dimensionality of each local feature vector of the first vehicle model is reduced, and the reduced local feature vectors of the first vehicle model are used to form a column vector, which is then used as the local feature vector of the first vehicle model.
[0091] Based on the above-disclosed content, this invention discloses a specific process for feature extraction, namely, using convolution kernel functions of different scales and directions to convolve vehicle images to obtain multiple local feature sub-vectors of the first vehicle model. Then, each local feature sub-vector of the first vehicle model is subjected to dimensionality reduction processing. Finally, the dimensionality-reduced local feature sub-vectors of the first vehicle model are concatenated to obtain the local feature vector of the first vehicle model. In essence, it uses the dimensionality-reduced local feature sub-vectors of the first vehicle model to form a column vector, and the column vector formed is the local feature vector of the first vehicle model.
[0092] Preferably, the image acquisition module 110 includes: a vehicle model image unit 111, which is used for...
[0093] Obtain a first training dataset, wherein the first training dataset includes multiple first acquired images for training and a first label corresponding to each first acquired image for training;
[0094] The first label is used to indicate whether a vehicle exists in the first acquired image used for training; an initial vehicle recognition model based on a deep residual network is constructed; the initial vehicle recognition model is trained using the first training dataset as input to obtain the vehicle recognition model.
[0095] A first training dataset is obtained. This dataset includes multiple first spectral imaging images for training and a first label corresponding to each image. The first label indicates whether a vehicle exists in the first spectral imaging image. Specifically, in this embodiment, the first training dataset is a dataset of spectral imaging images containing vehicles; that is, a portion of the first training dataset contains spectral imaging images containing vehicles, while the remaining dataset does not contain vehicles. The method for generating the first training dataset is as follows: a spectral imaging module periodically acquires spectral imaging images containing vehicles and those not containing vehicles, while simultaneously recording whether a vehicle exists in the acquired spectral imaging image. The result of whether a vehicle exists is used as the first label corresponding to the spectral imaging image, and the first training dataset is generated based on the correspondence between the spectral imaging images and the first label. For example, the first training dataset contains two categories: the first category contains spectral imaging images containing vehicles, and the second category does not contain vehicles. The first training dataset is used to train a deep residual network model for predicting whether a vehicle is present in the acquired spectral imaging images, i.e., to train a vehicle recognition model.
[0096] Preferably, it further includes a vehicle exhaust emission model module 150; wherein, the vehicle exhaust emission model module 150 is used for
[0097] Obtain a second training dataset, which includes multiple training spectral imaging images containing vehicle models and a second label corresponding to each training spectral imaging image;
[0098] The second label is the vehicle exhaust corresponding to the spectral imaging image used for training;
[0099] Construct an initial vehicle exhaust gas model based on a deep residual network;
[0100] The initial vehicle exhaust model is trained using the second training dataset as input to obtain the vehicle exhaust model.
[0101] The second training dataset is a dataset of spectral imaging images of vehicle models under different vehicle exhaust emission conditions. This dataset is used to train a deep residual network model for predicting different vehicle exhaust emissions, i.e., to train a vehicle exhaust emission prediction model. In this embodiment, vehicle exhaust emissions can include low emissions, high emissions, and normal emissions. Therefore, the second training dataset is used to train a deep residual network model to distinguish between these three categories of vehicle exhaust emissions. The method for generating the second training dataset is as follows: A spectral imaging module periodically acquires spectral imaging images containing vehicle models, and simultaneously records the subjective vehicle exhaust emission feedback during the acquisition of the spectral imaging images. The vehicle exhaust emission is used as the category label corresponding to the spectral imaging image, i.e., as the second label corresponding to the spectral imaging image. The second training dataset is generated based on the correspondence between the spectral imaging image and the second label.
[0102] Please see Figure 4 The third aspect of this application provides a device for controlling a ventilation fan based on tunnel environment monitoring, including a memory and a processor connected in sequence. The memory stores a computer program, and the processor reads the computer program and executes the method for controlling a ventilation fan based on tunnel environment monitoring as described in the first aspect of the application. Specifically, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out memory (FIFO), and / or last-in-first-out memory (FILO), etc.; the processor may be, but is not limited to, a microprocessor of the STM32F105 series, an ARM (Advanced RISC Machines) processor, an x86 architecture processor, or a processor with integrated NPU (neural-network processing units).
[0103] The working process, working details and technical effects of the device provided in the third aspect of this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.
[0104] This fourth aspect of the embodiment provides a computer-readable storage medium storing instructions for the method of controlling a ventilation fan based on tunnel environment monitoring according to the first aspect of the embodiment. That is, the computer-readable storage medium stores instructions that, when executed on a computer, perform the method of controlling a ventilation fan based on tunnel environment monitoring as described in the first aspect. The computer-readable storage medium refers to a data storage medium, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
[0105] The working process, working details and technical effects of the computer-readable storage medium provided in the fourth aspect of this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.
[0106] The fifth aspect of this embodiment provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the method for controlling a fan based on tunnel environment monitoring as described in the first aspect of this embodiment. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
[0107] The embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and apply this without any creative effort.
[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, 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 ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a repository code merging device to execute the methods of various embodiments or some parts of the embodiments.
[0109] Finally, it should be noted that the above are merely preferred embodiments of the invention and are not intended to limit the scope of protection of the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of protection of the invention.
Claims
1. A method for controlling a ventilation fan based on tunnel environmental monitoring, characterized in that, The methods include: Acquire vehicle model image set and vehicle image and spectral imaging image of the vehicle corresponding to the vehicle entering the tunnel, wherein the vehicle model image set includes sample images of several different vehicles; Each vehicle image is convolved to obtain a first vehicle model local feature vector corresponding to the vehicle image. Each vehicle model sample image in the vehicle model image set is also convolved to obtain a second vehicle model local feature vector corresponding to each vehicle model sample image. The intersection of the second vehicle model local feature vectors corresponding to each vehicle model sample image is calculated, and this intersection is used as the vehicle common feature vector. The difference between the first vehicle model local feature vector and the vehicle common feature vector is calculated, and this difference is used as the vehicle dissimilar feature vector corresponding to the vehicle image. The vehicle dissimilar feature vector is input into the vehicle model recognition model to identify the vehicle model corresponding to the vehicle. The spectral imaging image of each vehicle is used as input, and exhaust gas monitoring is performed according to the vehicle model based on the corresponding vehicle type. Obtain exhaust gas monitoring values and control the working status of the tunnel ventilation fan based on these values.
2. The method for controlling a ventilation fan based on tunnel environment monitoring according to claim 1, characterized in that, Each of the vehicle images is convolved to obtain a first vehicle model local feature vector corresponding to the vehicle image, including: Obtain the convolution kernel function and determine the scale and direction of the convolution kernel function; Based on the scale and orientation of the convolution kernel function, several feature extraction kernel functions are obtained by combination; Each feature extraction kernel function is convolved with the vehicle image to obtain the first local feature sub-vector of the vehicle corresponding to each feature extraction kernel function; The dimensionality of each local feature subvector of the first vehicle model is reduced, and a column vector is formed by combining the dimensionality-reduced local feature subvectors of the first vehicle model. The column vector is then used as the local feature vector of the first vehicle model.
3. The method for controlling a ventilation fan based on tunnel environment monitoring according to claim 1, characterized in that, The steps for acquiring vehicle images include: Obtain a first training dataset, wherein the first training dataset includes multiple first acquisition images for training and a first label corresponding to each of the first acquisition images for training; The first label is used to indicate whether a vehicle exists in the first acquired image used for training; an initial vehicle recognition model based on a deep residual network is constructed; the initial vehicle recognition model is trained using the first training dataset as input to obtain a vehicle recognition model.
4. The method for controlling a ventilation fan based on tunnel environment monitoring according to claim 1, characterized in that, Before using the spectral imaging image of each vehicle as input and performing exhaust emission monitoring based on the vehicle's corresponding model using a vehicle exhaust emission model, the process further includes training a vehicle exhaust emission model corresponding to the vehicle model, including: Obtain a second training dataset, wherein the second training dataset includes multiple training spectral imaging images containing vehicle models and a second label corresponding to each training spectral imaging image; The second label is the vehicle exhaust gas corresponding to the training spectral imaging image; Construct an initial vehicle exhaust gas model based on a deep residual network; Using the second training dataset as input, the initial vehicle exhaust model is trained to obtain the vehicle exhaust model.
5. A device for controlling a ventilation fan based on tunnel environmental monitoring, characterized in that, The device includes: The image acquisition module is used to acquire a vehicle image set and vehicle images and spectral imaging images of the vehicles entering the tunnel, wherein the vehicle image set includes sample images of several different vehicles. The vehicle model identification module is used to perform convolution processing on each vehicle image to obtain a first local feature vector of the vehicle model corresponding to the vehicle image, and to perform convolution processing on each vehicle model sample image in the vehicle model image set to obtain a second local feature vector of the vehicle model corresponding to each vehicle model sample image; to calculate the intersection of the second local feature vectors of each vehicle model sample image, and to use the intersection as the common feature vector of the vehicle; to calculate the difference between the first local feature vector of the vehicle model and the common feature vector of the vehicle, and to use the difference as the vehicle dissimilar feature vector of the vehicle image; and to input the vehicle dissimilar feature vector into the vehicle model recognition model to identify the vehicle model corresponding to the vehicle. An infrared monitoring module is used to take the spectral imaging image of each vehicle as input and perform exhaust gas monitoring based on the vehicle model corresponding to the vehicle type. The ventilation control module is used to acquire exhaust gas monitoring values and control the working status of the tunnel ventilation fan based on these values.
6. The device for controlling a ventilation fan based on tunnel environment monitoring according to claim 5, characterized in that, The vehicle model determination module includes a processing unit, the processing unit being used for... Obtain the convolution kernel function and determine the scale and direction of the convolution kernel function; Based on the scale and orientation of the convolution kernel function, several feature extraction kernel functions are obtained by combination; Each feature extraction kernel function is convolved with the vehicle image to obtain the first local feature sub-vector of the vehicle corresponding to each feature extraction kernel function; The dimensionality of each local feature subvector of the first vehicle model is reduced, and a column vector is formed by combining the dimensionality-reduced local feature subvectors of the first vehicle model. The column vector is then used as the local feature vector of the first vehicle model.
7. The device for controlling a ventilation fan based on tunnel environmental monitoring according to claim 5, characterized in that, The image acquisition module includes: a vehicle model image unit, the vehicle model image unit being used for... Obtain a first training dataset, wherein the first training dataset includes multiple first acquisition images for training and a first label corresponding to each of the first acquisition images for training; The first label is used to indicate whether a vehicle exists in the first acquired image used for training; an initial vehicle recognition model based on a deep residual network is constructed; the initial vehicle recognition model is trained using the first training dataset as input to obtain a vehicle recognition model.
8. The device for controlling a ventilation fan based on tunnel environment monitoring according to claim 5, characterized in that, It also includes a vehicle exhaust emission model module; wherein, the vehicle exhaust emission model module is used for Obtain a second training dataset, wherein the second training dataset includes multiple training spectral imaging images containing vehicle models and a second label corresponding to each training spectral imaging image; The second label is the vehicle exhaust gas corresponding to the training spectral imaging image; Construct an initial vehicle exhaust gas model based on a deep residual network; Using the second training dataset as input, the initial vehicle exhaust model is trained to obtain the vehicle exhaust model.
9. A computer device, characterized in that, The device includes a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the method for controlling a fan based on tunnel environment monitoring as described in any one of claims 1 to 4.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a computer, perform the method for controlling a fan based on tunnel environment monitoring as described in any one of claims 1 to 4.