Tunnel lamp state detection method, system and device and storage medium

Abstract

The invention discloses a tunnel lamp state detection method, system and device and a storage medium, and the method comprises the steps: calibrating the corresponding relation between a lamp and a camera in a target tunnel, and determining a capture equipment set; detecting a vehicle position through a millimeter wave radar, triggering a target tunnel to automatically light according to the vehicle position, and obtaining a target image of a corresponding lamp through a camera; inputting the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result; and determining a lamp abnormal area according to the lamp state detection result, and notifying corresponding personnel to perform fixed-point maintenance according to the lamp abnormal area. The method improves the efficiency and accuracy of tunnel lamp state detection, reduces the cost and potential risk of manual inspection, improves the intelligent level of tunnel safety monitoring, and can be applied to the technical field of machine vision.

Description

Technical Field

[0001] The present invention relates to the field of machine vision technology, and in particular to a tunnel lamp status detection method, system, device and storage medium. Background Art

[0002] Tunnels are key transportation facilities, and their safe operation is extremely important for protecting the safety of personnel and vehicles. The lighting inside the tunnel directly affects driving safety, so regular inspection of lighting equipment is indispensable. At present, tunnel maintenance mainly relies on manual labor, and very few areas have introduced tunnel robots for automatic inspection. Manual inspection of tunnels has problems such as high labor intensity, lack of unified standards, low efficiency, easy omissions and delays, which affect the efficiency and accuracy of tunnel lighting status detection; tunnel robot inspections mainly focus on lane events, tunnel structures and other contents, and lack the means to check the status of tunnel electromechanical equipment. Summary of the invention

[0003] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0004] Therefore, an object of an embodiment of the present invention is to provide a method for detecting the state of a tunnel lamp, which improves the efficiency and accuracy of detecting the state of a tunnel lamp.

[0005] Another object of an embodiment of the present invention is to provide a tunnel lamp status detection system.

[0006] In order to achieve the above technical objectives, the technical solutions adopted by the embodiments of the present invention include:

[0007] In a first aspect, an embodiment of the present invention provides a method for detecting a state of a tunnel lamp, comprising the following steps:

[0008] Calibrate the correspondence between the lamps and cameras in the target tunnel and determine the capture device set;

[0009] The vehicle position is detected by a millimeter wave radar, the target tunnel is triggered to automatically light up according to the vehicle position, and the corresponding target image of the lamp is acquired by the camera;

[0010] Inputting the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result;

[0011] The abnormal area of ​​the lamp is determined according to the lamp status detection result, and corresponding personnel are notified to perform fixed-point maintenance according to the abnormal area of ​​the lamp.

[0012] Furthermore, in one embodiment of the present invention, the calibration of the correspondence between the lamps and the cameras in the target tunnel and the determination of the capture device set specifically includes:

[0013] Numbering the lamps in the target tunnel according to their positions to obtain a complete set of lamps;

[0014] Determine a plurality of lamps corresponding to the shooting areas of each of the cameras in the target tunnel, and obtain a subset of lamps corresponding to each of the cameras;

[0015] When the union of the subsets of lamps is the full set of lamps, determining that the shooting area of ​​the camera covers all the lamps;

[0016] A first mapping relationship between each of the lamps and each of the cameras is calibrated according to the lamp subset, and the capture device set is determined according to the first mapping relationship.

[0017] Further, in one embodiment of the present invention, the detecting the vehicle position by the millimeter wave radar, triggering the automatic lighting of the target tunnel according to the vehicle position, and acquiring the corresponding target image of the lamp by the camera specifically includes:

[0018] The millimeter-wave radar disposed in the target tunnel detects the position of the vehicle, and determines whether a vehicle is located in the target tunnel and whether a vehicle is about to enter the target tunnel according to the position of the vehicle;

[0019] When a vehicle is located in the target tunnel and / or a vehicle is about to enter the target tunnel, the lamp in the target tunnel is triggered to light up continuously, and the target image of the corresponding shooting area is acquired through the camera;

[0020] When no vehicle is in the target tunnel and no vehicle is about to enter the target tunnel, the lamps in the target tunnel are triggered to turn off, and the camera is controlled to stop shooting.

[0021] Furthermore, in one embodiment of the present invention, the tunnel lamp status detection model is trained by the following steps:

[0022] Acquire a plurality of tunnel lamp sample images, and determine a sample label of each of the tunnel lamp sample images;

[0023] Constructing a training data set according to the tunnel lamp sample images and the sample labels;

[0024] Inputting the training data set into a pre-built convolutional neural network to obtain a trained tunnel lamp state detection model;

[0025] The sample labels include all lights on, some lights on, and all lights off, and the convolutional neural network includes an initial convolutional layer, an inverted residual structure with a linear bottleneck, an average pooling layer, and a fully connected layer.

[0026] Furthermore, in one embodiment of the present invention, the step of inputting the training data set into a pre-built convolutional neural network to obtain the trained tunnel lamp state detection model specifically includes:

[0027] Inputting the tunnel lamp sample image into the convolutional neural network to obtain a lamp state recognition result;

[0028] A loss value is determined according to the lamp state recognition result and the sample label, and channel pruning is performed on the convolutional neural network according to the loss value and the parameters of the convolutional neural network are updated to obtain a trained tunnel lamp state detection model.

[0029] Furthermore, in one embodiment of the present invention, the loss value is determined according to the lamp state recognition result and the sample label, and the convolutional neural network is channel pruned and the parameters of the convolutional neural network are updated according to the loss value, which specifically includes:

[0030] Calculate a cross entropy loss value according to the lamp state recognition result and the sample label;

[0031] The penalty factor is determined according to the product of the preset penalty intensity coefficient and the L1 norm of the current weight parameter;

[0032] Determine the loss value according to the sum of the cross entropy loss value and the penalty factor;

[0033] Channel pruning is performed on the convolutional neural network according to the current weight parameters, the parameters of the convolutional neural network are updated according to the cross entropy loss value, and the sample image of the tunnel lamp is returned to the convolutional neural network until the loss value reaches a preset first threshold.

[0034] Further, in an embodiment of the present invention, determining the abnormal area of ​​the lamp according to the lamp state detection result specifically includes:

[0035] Determine that the target image whose lamp status detection result is partially lit or fully lit is an abnormal lamp image;

[0036] The target camera corresponding to the abnormal lamp image is determined, and multiple target lamps corresponding to the target camera are determined according to the corresponding relationship between the lamps and the cameras, and the area where the target lamps are located is used as the lamp abnormal area.

[0037] In a second aspect, an embodiment of the present invention provides a tunnel lamp status detection system, including:

[0038] The device calibration module is used to calibrate the correspondence between the lamps and cameras in the target tunnel and determine the capture device set;

[0039] An image acquisition module, used to detect the vehicle position through a millimeter wave radar, trigger the target tunnel to automatically light up according to the vehicle position, and acquire the corresponding target image of the lamp through the camera;

[0040] A lamp state detection module, used for inputting the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result;

[0041] The abnormal area determination module is used to determine the abnormal area of ​​the lamp according to the lamp status detection result, and notify corresponding personnel to perform fixed-point maintenance according to the abnormal area of ​​the lamp.

[0042] In a third aspect, an embodiment of the present invention provides a tunnel lamp state detection device, comprising:

[0043] at least one processor;

[0044] at least one memory for storing at least one program;

[0045] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned method for detecting the state of a tunnel lamp.

[0046] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to execute the above-mentioned method for detecting the state of a tunnel lamp.

[0047] The advantages and beneficial effects of the present invention will be partly given in the following description, partly become apparent from the following description, or be understood through the practice of the present invention:

[0048] The embodiment of the present invention calibrates the correspondence between the lamps and cameras in the target tunnel, determines the capture device set, detects the vehicle position through the millimeter wave radar, triggers the target tunnel to automatically light up according to the vehicle position, obtains the target image of the corresponding lamp through the camera, inputs the target image into the pre-trained tunnel lamp status detection model, obtains the lamp status detection result, determines the lamp abnormal area according to the lamp status detection result, and notifies the corresponding personnel to perform fixed-point maintenance according to the lamp abnormal area. The embodiment of the present invention improves the efficiency and accuracy of tunnel lamp status detection, reduces the cost and potential risks of manual inspection, and thus improves the intelligent level of tunnel safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solution in the embodiments of the present invention, the following introduction is made to the drawings required for use in the embodiments of the present invention. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solution of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 A flowchart of a method for detecting the state of a tunnel lamp provided by an embodiment of the present invention;

[0051] Figure 2 A structural block diagram of a tunnel lamp status detection system provided by an embodiment of the present invention;

[0052] Figure 3 This is a structural block diagram of a tunnel lamp status detection device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0054] In the description of the present invention, the meaning of "a plurality" is two or more than two. If there is a description of "a first" or "a second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used in this document have the same meaning as those commonly understood by those skilled in the art.

[0055] Manual inspections in tunnels are labor-intensive, lack uniform standards, are inefficient, and prone to omissions and delays, all of which can lead to safety risks; robot inspections in tunnels mainly focus on lane events, tunnel structures, and other content, and lack the means to check the status of tunnel electromechanical equipment. At the same time, in remote areas with less traffic and long tunnels, automatic dimming systems based on millimeter-wave radars and high-definition cameras are usually installed. These devices can automatically sense vehicles and adjust lighting, while automatically recording and storing images, providing conditions for the automatic inspection of tunnel lighting equipment.

[0056] Based on this, the present invention proposes a method for detecting the status of tunnel lamps, which can perform all-weather intelligent real-time monitoring of the status of lamps in the tunnel, significantly improve the efficiency and accuracy of inspections, and enhance the operating efficiency and safety of the tunnel through continuous monitoring and intelligent analysis, thereby effectively preventing the occurrence of accidents.

[0057] Reference Figure 1 The embodiment of the present invention provides a method for detecting the state of a tunnel lamp, which specifically includes the following steps:

[0058] S101, calibrating the correspondence between lamps and cameras in the target tunnel, and determining a capture device set;

[0059] S102, detecting the vehicle position by using a millimeter wave radar, triggering the target tunnel to automatically light up according to the vehicle position, and acquiring a target image of the corresponding lamp by using a camera;

[0060] S103, inputting the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result;

[0061] S104: Determine the abnormal area of ​​the lamp according to the lamp status detection result, and notify the corresponding personnel to perform fixed-point maintenance according to the abnormal area of ​​the lamp.

[0062] The embodiments of the present invention improve the efficiency and accuracy of tunnel lamp status detection, reduce the cost and potential risks of manual inspection, and thus improve the intelligent level of tunnel safety monitoring.

[0063] As an optional implementation, the correspondence between the lamps and the cameras in the target tunnel is calibrated to determine the capture device set, which specifically includes:

[0064] Number the lamps in the target tunnel according to their positions to obtain the complete set of lamps;

[0065] Determine a plurality of lamps corresponding to the shooting areas of each camera in the target tunnel, and obtain a subset of lamps corresponding to each camera;

[0066] When the union of the subsets of lamps is the complete set of lamps, it is determined that the shooting area of ​​the camera covers all lamps;

[0067] A first mapping relationship between each lamp and each camera is calibrated according to the lamp subset, and a capture device set is determined according to the first mapping relationship.

[0068] Specifically, the lamps in the tunnel are numbered, starting from the tunnel entrance, to obtain the lamp set X{x1,x2...x n}; Select the fixed camera installed in the tunnel. For a camera C1, the subsequence of the lamp collection captured by it is X1{x1,x2...x i}, record the camera C1 and the light fixture collection subsequence X1{x1,x2...x i} is a pair of mapping relationships; select the next camera C2, whose shooting picture is adjacent to the previous camera C1, and the set of lamps in the picture is X2{x k ,x k+1 ...x k+n}, where X1 and X2 have an intersection, and the subsequence {x k ...x k+j} and the subsequence {x i-j ...x i Repeat the above steps to calibrate the camera Ci and the light fixture collection subsequence X i The mapping relationship until X i The union is equal to the lamp set X. Confirm that the camera image capture range covers all lamps, and get the capture device C{C1,C2...C n}collection.

[0069] As an optional implementation, the vehicle position is detected by a millimeter wave radar, the target tunnel is automatically lit according to the vehicle position, and the target image of the corresponding lamp is obtained by a camera, which specifically includes:

[0070] The millimeter-wave radar installed in the target tunnel detects the vehicle position, and determines whether there is a vehicle in the target tunnel and whether there is a vehicle about to enter the target tunnel based on the vehicle position;

[0071] When a vehicle is in the target tunnel and / or is about to enter the target tunnel, the lamps in the target tunnel are triggered to light up continuously, and a target image of the corresponding shooting area is acquired through the camera;

[0072] When there is no vehicle in the target tunnel and no vehicle is about to enter the target tunnel, the lights in the target tunnel are triggered to turn off and the camera is controlled to stop shooting.

[0073] Specifically, several millimeter-wave radars are installed in the tunnel to detect vehicles entering or exiting the tunnel, as well as whether there are vehicles in the tunnel; based on the detection results of the millimeter-wave radar, when a vehicle enters the tunnel or there is a vehicle in the tunnel, all lamps in the tunnel are triggered to turn on automatically, and the cameras in the capture device set start shooting; based on the detection results of the millimeter-wave radar, when a vehicle exits the tunnel and there are no vehicles driving in the tunnel, all lamps in the tunnel are triggered to turn off automatically, and the cameras in the capture device set stop shooting; the captured image is scaled and normalized to obtain the target image.

[0074] As an optional implementation, the tunnel lamp status detection model is trained by the following steps:

[0075] Acquire multiple tunnel lamp sample images, and determine a sample label of each tunnel lamp sample image;

[0076] Construct a training dataset based on tunnel lamp sample images and sample labels;

[0077] Input the training data set into the pre-built convolutional neural network to obtain a trained tunnel lamp status detection model;

[0078] Among them, the sample labels include all lights on, some lights on, and all lights off. The convolutional neural network includes an initial convolutional layer, an inverted residual structure with a linear bottleneck, an average pooling layer, and a fully connected layer.

[0079] Specifically, for the cameras in the capture device set, take images of normal lights and abnormal lights (including partial lights and lights off), take the normal lights images as positive samples and the abnormal lights images as negative samples to form the original image dataset. The original dataset is scaled and normalized to obtain the preprocessed dataset; the preprocessed dataset is divided into a training set and a validation set.

[0080] In the embodiment of the present invention, a 4-megapixel high-definition camera is used to shoot color images, and the original image resolution is 2560×1440; the original data set is uniformly scaled to 224×224 using an interpolation method, and then the pixel values ​​are scaled to a range using a normalization method to obtain a preprocessed data set; the preprocessed data set is divided into a training set and a test set at a ratio of 7:3.

[0081] The convolutional neural network model adopts a MobileNetV3-like structure, including an initial convolutional layer, an inverted residual structure (Bottleneck) with a linear bottleneck, an average pooling layer and a fully connected layer. The convolutional neural network model is trained using the constructed enhanced dataset. During the training process, the convolutional neural network model is optimized using a channel pruning method based on the BN (Batch Normalization) layer. The tunnel lighting status detection model is obtained by iteratively calculating the gradient of the loss function and updating the model parameters.

[0082] As an optional implementation, the training data set is input into a pre-built convolutional neural network to obtain a trained tunnel lamp state detection model, which specifically includes:

[0083] Input the sample images of tunnel lamps into the convolutional neural network to obtain the lamp status recognition results;

[0084] The loss value is determined according to the lamp state recognition results and sample labels. The convolutional neural network is channel pruned and the parameters of the convolutional neural network are updated according to the loss value to obtain a trained tunnel lamp state detection model.

[0085] As an optional implementation, a loss value is determined according to the lamp state recognition result and the sample label, and the convolutional neural network is channel pruned and the parameters of the convolutional neural network are updated according to the loss value, which specifically includes:

[0086] Calculate the cross entropy loss value based on the lamp state recognition result and sample label;

[0087] The penalty factor is determined according to the product of the preset penalty intensity coefficient and the L1 norm of the current weight parameter;

[0088] Determine the loss value based on the sum of the cross entropy loss value and the penalty factor;

[0089] The convolutional neural network is channel pruned according to the current weight parameters, the parameters of the convolutional neural network are updated according to the cross entropy loss value, and the sample image of the tunnel lamp is input into the convolutional neural network again until the loss value reaches a preset first threshold.

[0090] Specifically, the inverted residual structure (Bottleneck) with a linear bottleneck includes an expansion layer, a depthwise separable convolution layer, a SE (Squeeze-and-Excitation) module, and a projection layer: First, the standard convolution layer of the expansion layer is used for dimensionality increase and then sent to the depthwise separable convolution layer. In the depthwise separable convolution layer, firstly, a depthwise convolution is performed, and the convolution kernel size is 3×3×M or 5×5×M, where M is the number of input channels; the attention mechanism SE module is introduced, and global average pooling is applied to the feature map of each channel, and a new weight coefficient is assigned to each channel after the ReLU activation function and the Hard-Sigmoid activation function; finally, point-by-point convolution is performed, and the convolution kernel size is 1×1×N, where N is the number of output channels. It is sent to the projection layer and the standard convolution layer is used for dimensionality reduction.

[0091] During the training process, L1 regularization is used to perform sparse training on the model, penalizing unimportant parameters, making the parameters in the filter sparse, and the loss function L = L CE +λγ1, where L CE The cross entropy loss function is the loss function of the convolutional neural network, λ is the penalty coefficient, and γ is the current weight parameter. The current weight parameter in the BN layer is used to judge the importance of the channel. When γ is close to 0, it means that the channel has little impact on the BN layer and can be removed. The pruning rate is set to 40%, and the initial λ = 10 -5 ,λ=10 -5 , the sparse training times are 150 times to obtain the pruned network model; fine-tuning is performed again according to the model performance requirements to obtain the final pruned network model.

[0092] As an optional implementation, determining the abnormal area of ​​the lamp according to the lamp state detection result specifically includes:

[0093] Determine that the target image whose lamp status detection result is partially lit or fully lit is an abnormal lamp image;

[0094] The target camera corresponding to the abnormal lamp image is determined, and multiple target lamps corresponding to the target camera are determined according to the corresponding relationship between the lamp and the camera, and the area where the target lamp is located is used as the lamp abnormal area.

[0095] Specifically, the target image is identified according to the intelligent detection model of the tunnel lamp status to obtain the target image classification result; when the target image classification result is an abnormal lighting image, the abnormal lamp status area is located through the correspondence between the lamp and the camera.

[0096] The method steps of the embodiment of the present invention are described above. It can be appreciated that the embodiment of the present invention improves the efficiency and accuracy of tunnel lamp status detection, reduces the cost and potential risks of manual inspection, and thus improves the intelligent level of tunnel safety monitoring.

[0097] Reference Figure 2 , an embodiment of the present invention provides a tunnel lamp status detection system, comprising:

[0098] The device calibration module is used to calibrate the correspondence between the lamps and cameras in the target tunnel and determine the capture device set;

[0099] An image acquisition module is used to detect the vehicle position through a millimeter-wave radar, trigger the target tunnel to automatically light up according to the vehicle position, and acquire the target image of the corresponding lamp through a camera;

[0100] A lamp state detection module is used to input the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result;

[0101] The abnormal area determination module is used to determine the abnormal area of ​​the lamp according to the lamp status detection result, and notify the corresponding personnel to perform fixed-point maintenance according to the abnormal area of ​​the lamp.

[0102] The contents of the above method embodiments are all applicable to the present system embodiments. The functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0103] Reference Figure 3 The embodiment of the present invention provides a tunnel lamp status detection device, comprising:

[0104] at least one processor;

[0105] at least one memory for storing at least one program;

[0106] When the at least one program is executed by the at least one processor, the at least one processor implements the tunnel lamp status detection method.

[0107] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0108] An embodiment of the present invention further provides a computer-readable storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to execute the above-mentioned method for detecting the state of a tunnel lamp.

[0109] A computer-readable storage medium according to an embodiment of the present invention can execute a tunnel lamp status detection method provided by an embodiment of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0110] The embodiment of the present invention also discloses a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1 The method shown.

[0111] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the above-mentioned boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0112] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified to the contrary, one or more of the above-mentioned functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0113] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0114] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0115] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the above-mentioned program is printed, since the above-mentioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or processing in other suitable ways as necessary, and then stored in a computer memory.

[0116] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0117] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0118] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0119] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for detecting the state of a tunnel lamp, characterized in that: The following steps are involved: Calibrate the correspondence between the lamps and cameras in the target tunnel and determine the capture device set; The vehicle position is detected by a millimeter wave radar, the target tunnel is triggered to automatically light up according to the vehicle position, and the corresponding target image of the lamp is acquired by the camera; Inputting the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result; The abnormal area of ​​the lamp is determined according to the lamp status detection result, and corresponding personnel are notified to perform fixed-point maintenance according to the abnormal area of ​​the lamp.

2. A tunnel lamp status detection method according to claim 1, characterized in that: The calibration of the correspondence between the lamps and the cameras in the target tunnel and determining the capture device set specifically includes: Numbering the lamps in the target tunnel according to their positions to obtain a complete set of lamps; Determine a plurality of lamps corresponding to the shooting areas of each of the cameras in the target tunnel, and obtain a subset of lamps corresponding to each of the cameras; When the union of the subsets of lamps is the full set of lamps, determining that the shooting area of ​​the camera covers all the lamps; A first mapping relationship between each of the lamps and each of the cameras is calibrated according to the lamp subset, and the capture device set is determined according to the first mapping relationship.

3. A tunnel lamp status detection method according to claim 1, characterized in that: The method of detecting the vehicle position by the millimeter wave radar, triggering the target tunnel to automatically light up according to the vehicle position, and acquiring the corresponding target image of the lamp by the camera specifically includes: The millimeter-wave radar disposed in the target tunnel detects the position of the vehicle, and determines whether a vehicle is located in the target tunnel and whether a vehicle is about to enter the target tunnel according to the position of the vehicle; When a vehicle is located in the target tunnel and / or a vehicle is about to enter the target tunnel, the lamp in the target tunnel is triggered to light up continuously, and the target image of the corresponding shooting area is acquired through the camera; When no vehicle is in the target tunnel and no vehicle is about to enter the target tunnel, the lamps in the target tunnel are triggered to turn off, and the camera is controlled to stop shooting.

4. A tunnel lamp status detection method according to claim 1, characterized in that: The tunnel lamp status detection model is trained by the following steps: Acquire a plurality of tunnel lamp sample images, and determine a sample label of each of the tunnel lamp sample images; Constructing a training data set according to the tunnel lamp sample images and the sample labels; Inputting the training data set into a pre-built convolutional neural network to obtain a trained tunnel lamp state detection model; The sample labels include all lights on, some lights on, and all lights off, and the convolutional neural network includes an initial convolutional layer, an inverted residual structure with a linear bottleneck, an average pooling layer, and a fully connected layer.

5. A tunnel lamp status detection method according to claim 4, characterized in that: The step of inputting the training data set into a pre-built convolutional neural network to obtain a trained tunnel lamp state detection model specifically includes: Inputting the tunnel lamp sample image into the convolutional neural network to obtain a lamp state recognition result; A loss value is determined according to the lamp state recognition result and the sample label, and channel pruning is performed on the convolutional neural network according to the loss value and the parameters of the convolutional neural network are updated to obtain a trained tunnel lamp state detection model.

6. A method for detecting the state of a tunnel lamp according to claim 5, characterized in that: The step of determining a loss value according to the lamp state recognition result and the sample label, performing channel pruning on the convolutional neural network according to the loss value, and updating the parameters of the convolutional neural network specifically includes: Calculate a cross entropy loss value according to the lamp state recognition result and the sample label; The penalty factor is determined according to the product of the preset penalty intensity coefficient and the L1 norm of the current weight parameter; Determine the loss value according to the sum of the cross entropy loss value and the penalty factor; Channel pruning is performed on the convolutional neural network according to the current weight parameters, the parameters of the convolutional neural network are updated according to the cross entropy loss value, and the sample image of the tunnel lamp is returned to the convolutional neural network until the loss value reaches a preset first threshold.

7. A tunnel lamp status detection method according to any one of claims 1 to 6, characterized in that: The step of determining the abnormal area of ​​the lamp according to the lamp state detection result specifically includes: Determine that the target image whose lamp status detection result is partially lit or fully lit is an abnormal lamp image; The target camera corresponding to the abnormal lamp image is determined, and multiple target lamps corresponding to the target camera are determined according to the corresponding relationship between the lamps and the cameras, and the area where the target lamps are located is used as the lamp abnormal area.

8. A tunnel lamp status detection system, characterized in that: include: The device calibration module is used to calibrate the correspondence between the lamps and cameras in the target tunnel and determine the capture device set; An image acquisition module, used to detect the vehicle position through a millimeter wave radar, trigger the target tunnel to automatically light up according to the vehicle position, and acquire the corresponding target image of the lamp through the camera; A lamp state detection module, used for inputting the target image into a pre-trained tunnel lamp state detection model to obtain a lamp state detection result; The abnormal area determination module is used to determine the abnormal area of ​​the lamp according to the lamp status detection result, and notify corresponding personnel to perform fixed-point maintenance according to the abnormal area of ​​the lamp.

9. A tunnel lamp status detection device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a tunnel lamp status detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to execute a tunnel lamp status detection method according to any one of claims 1 to 7 when executed by the processor.

Images