Anti-collision monitoring and early warning method and system for gate machine

Abstract

The present invention relates to the field of vehicle monitoring, and specifically to an anti-collision monitoring and early warning method and system for gates. An anti-collision monitoring and early warning system for gates, comprising: a vehicle detection module, a microwave radar module, a facial area image acquisition module, a facial expression recognition module and a monitoring and early warning module. The present invention integrates a variety of monitoring equipment such as photoelectric sensors, microwave radars and cameras, and can monitor the status of vehicles and their drivers in real time from different angles and dimensions; this multi-dimensional monitoring method not only improves the monitoring accuracy of the system, but also enhances the ability to identify potential risks, thereby more effectively preventing the occurrence of collision accidents. At the same time, the monitoring time interval is optimized through the particle swarm optimization algorithm, which further reduces the misjudgment rate and maximizes the monitoring efficiency, while minimizing the monitoring frequency while maintaining the detection accuracy.

Description

Technical Field

[0001] The present invention relates to the field of vehicle monitoring, and in particular to an anti-collision monitoring and early warning method and system for a gate. Background Art

[0002] In modern traffic management, turnstiles, as important traffic control equipment, are widely used in highway toll stations, parking lot entrances and exits, community access control and other occasions; their main function is to ensure traffic order and safety by controlling the passage of vehicles; however, with the increase in the number of vehicles and the increase in traffic flow, turnstiles will face some safety challenges in actual use; especially during the passage of vehicles, due to factors such as driver's emotional fluctuations, operational errors or vehicle failures, collision accidents may occur, causing the gate pole to bend and deform, and even causing the turnstile to be scrapped.

[0003] Traditional gate systems mainly rely on physical isolation and manual management, and have disadvantages such as slow response speed, limited monitoring range, and high misjudgment rate. In order to improve the safety and intelligence level of the gate system, modern scientific and technological means such as sensor technology, image recognition technology, and machine learning algorithms can be used to achieve real-time monitoring and early warning of vehicle and driver status. Based on these modern scientific and technological means, the present invention proposes a new anti-collision monitoring and early warning method and system for gates. This method realizes real-time monitoring and early warning of vehicle and driver status by comprehensively utilizing technical means such as photoelectric sensors, microwave radars, cameras, and facial expression recognition models, thereby effectively reducing the risk of collision accidents. Summary of the invention

[0004] The present invention integrates multiple monitoring devices such as photoelectric sensors, microwave radars and cameras, and can monitor the status of vehicles and their drivers in real time from different angles and dimensions. This multi-dimensional monitoring method not only improves the monitoring accuracy of the system, but also enhances the ability to identify potential risks, thereby more effectively preventing the occurrence of collision accidents. At the same time, the monitoring time interval is optimized through the particle swarm optimization algorithm, which further reduces the misjudgment rate and maximizes the monitoring efficiency, while minimizing the monitoring frequency and maintaining the detection accuracy.

[0005] An anti-collision monitoring and early warning method for a gate machine, comprising:

[0006] Use photoelectric sensors to detect whether there is a moving vehicle. When a moving vehicle is detected, start the microwave radar to obtain the vehicle's speed v and distance s from the gate;

[0007] Based on the current speed v of the vehicle, calculate and obtain the safe distance of the vehicle at the current speed; determine whether the current distance s between the vehicle and the gate is greater than the currently calculated safe distance. If so, no operation is performed; if not, execute alarm response measures.

[0008] Preferably, the calculation method of the safety distance is as follows:

[0009] Using the formula Calculate the safe distance s of the vehicle at the current speed safe , where s buffer is the preset buffer distance; a is the average deceleration of the vehicle; t is the average reaction time of the driver; k is the preset safety factor.

[0010] Preferably, the calculation method of the vehicle average deceleration a and the driver average reaction time t is as follows:

[0011] Obtain several vehicle braking time data of different vehicle models, each vehicle braking time data includes the initial speed of a vehicle and the time required for braking from the initial speed; for any vehicle braking time data, divide the initial speed by the time required for braking to calculate the deceleration of the current vehicle; calculate the average of all the obtained decelerations to obtain the vehicle average deceleration a;

[0012] A driver reaction time data set is obtained, which contains the reaction time required for several drivers from receiving warning information to taking braking measures; the average of all reaction times is calculated to obtain the average driver reaction time t.

[0013] Preferably, the safe distance of the vehicle at the current speed can be obtained by calculating the driver's facial expression recognition:

[0014] When a moving vehicle is detected, the camera is started to obtain the vehicle image at the current monitoring time point, and the vehicle image is recognized using a cascade classifier based on Haar features to obtain a rectangular frame containing the facial area; the driver's facial area image is extracted from the vehicle image based on the facial area rectangular frame;

[0015] Input the driver's facial area image obtained at the current monitoring time point into the facial expression recognition model, and output the facial expression recognition result; based on the facial expression recognition result, determine whether the driver is in a normal emotional state;

[0016] If the driver is in a normal emotional state, based on the current speed v of the vehicle, the calculation formula of the safety distance is used, and the safety factor k is set to 1 to calculate the first safety distance; determine whether the current distance s between the vehicle and the gate is greater than the first safety distance. If so, no operation is performed; if not, the alarm response measures are executed;

[0017] If the driver is in an abnormal emotional state, the second safe distance is calculated based on the vehicle's current speed v, using the safety distance calculation formula, and the safety factor k is taken as 1.5; it is determined whether the current distance s between the vehicle and the gate is greater than the second safe distance. If so, no operation is performed; if not, alarm response measures are executed.

[0018] Preferably, the facial expression recognition model is established based on a CNN model, including an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; the input layer is used to input a driver's facial area image; the convolution layer is used to extract features of the driver's facial area image; the pooling layer is used to reduce the spatial dimension of the features; the fully connected layer is used to further learn feature representation; the output layer is used to output the facial expression recognition result;

[0019] For the training of facial expression recognition model, the specific operations are:

[0020] Acquire several facial expression recognition training samples, each of which contains an image of the driver's facial region with annotated facial expression recognition results; divide all acquired facial expression recognition training samples into an expression recognition training set and an expression recognition test set; train the facial expression recognition model using the expression recognition training set, and then input the expression recognition test set into the facial expression recognition model to obtain test results; set training conditions, and if the test results meet the training conditions, output the trained facial expression recognition model; otherwise, continue to train the facial expression recognition model using the expression recognition training set.

[0021] Preferably, the specific operation of using the cascade classifier based on Haar features to identify the vehicle image is as follows:

[0022] The input vehicle image is grayed to obtain a gray image; an integrated graph is generated based on the gray image; the generated integrated graph is used to calculate the Haar eigenvalue of each candidate area in the gray image; each Haar eigenvalue is judged by a pre-trained weak classifier to screen out the candidate facial area; the candidate facial area is screened layer by layer through a cascade classifier, each level contains several weak classifiers; if a certain level is passed, it enters the next level, otherwise it is judged as a non-target area; the candidate facial area judged by all cascade classifiers is the detected facial area.

[0023] Preferably, the time interval between every two monitoring time points is adjusted by a particle swarm optimization algorithm with the goal of minimizing the misjudgment rate and maximizing the time interval duration. The specific operation is as follows:

[0024] A1: Randomly generate a particle swarm in the range of 0-1 second, and the position of each particle represents a candidate time interval;

[0025] A2: For any particle, initialize the velocity of the particle and use the current position as the individual optimal position of the particle; use the formula Calculate the fitness value F of the particle, where P represents the misjudgment rate corresponding to the current candidate time interval, and L represents the duration of the current candidate time interval; and They are and the weight of L;

[0026] A3: Traverse all particles and take the position corresponding to the particle with the highest fitness value F as the global optimal position;

[0027] A4: For any particle, update the speed and position of the particle, and recalculate the fitness value; if the updated fitness value is higher than the fitness value of the individual best position of the particle, update the individual best position of the particle to the current position;

[0028] A5: Traverse all particles. If the fitness value of the individual best position of any particle is higher than the fitness value of the global best position, the global best position is updated to the individual best position of the particle.

[0029] A6: Repeat steps A4 to A5 until the preset maximum number of iterations is reached, and the obtained global best position is the optimal time interval.

[0030] Preferably, an anti-collision monitoring and early warning system for a gate includes:

[0031] The vehicle detection module is used to detect the presence of a moving vehicle using a photoelectric sensor, and activate the microwave radar and camera when a moving vehicle is detected;

[0032] Microwave radar module, used to obtain the vehicle's speed v and distance s from the gate;

[0033] The facial region image acquisition module is used to use the camera to acquire the vehicle image at the current monitoring time point, and use the cascade classifier based on Haar features to identify the vehicle image and obtain a rectangular frame containing the facial region; and extract the driver's facial region image from the vehicle image according to the facial region rectangular frame;

[0034] The facial expression recognition module is used to input the driver's facial area image obtained at the current monitoring time point into the facial expression recognition model and output the facial expression recognition result; based on the facial expression recognition result, it is determined whether the driver is in a normal emotional state;

[0035] The monitoring and early warning module is used to perform different operations according to the driver's emotional state. If the driver is in a normal emotional state, the first safe distance is calculated based on the current speed v of the vehicle; whether the current distance s between the vehicle and the gate is greater than the first safe distance, if so, no operation is performed; if not, alarm response measures are performed; if the driver is in an abnormal emotional state, the second safe distance is calculated based on the current speed v of the vehicle; whether the current distance s between the vehicle and the gate is greater than the second safe distance, if so, no operation is performed; if not, alarm response measures are performed.

[0036] The present invention has the following advantages:

[0037] 1. The present invention integrates multiple monitoring devices such as photoelectric sensors, microwave radars and cameras to monitor the status of vehicles and their drivers in real time from different angles and dimensions. This multi-dimensional monitoring method not only improves the monitoring accuracy of the system, but also enhances the ability to identify potential risks, thereby more effectively preventing the occurrence of collision accidents. At the same time, the monitoring time interval is optimized through the particle swarm optimization algorithm, which further reduces the misjudgment rate and maximizes the monitoring efficiency, while minimizing the monitoring frequency and maintaining the detection accuracy.

[0038] 2. The present invention recognizes vehicle images through a cascade classifier based on Haar features, obtains the driver's facial area image, and then uses a CNN-based facial expression recognition model to determine whether the driver is in a normal emotional state; according to the driver's emotional state, the system will dynamically adjust the safety distance, calculate the first safety distance and the second safety distance respectively, and judge whether the vehicle is within the safety distance accordingly, so that the system can more flexibly respond to safety needs in different situations, thereby improving the accuracy of early warning and the response speed of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a structural schematic diagram of the anti-collision monitoring and early warning system for gates adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0041] Embodiment 1, a method for anti-collision monitoring and early warning for a gate, comprising:

[0042] Photoelectric sensors are used to detect whether there is a moving vehicle. Photoelectric sensors are sensors that can detect changes in light intensity and are often used to detect the existence and movement of objects. In this system, when the sensor detects a change in light intensity, it indicates that a vehicle has passed, thereby triggering the subsequent monitoring process; when a moving vehicle is detected, the microwave radar is started. The microwave radar can emit microwave signals at high frequency, and use the reflection principle of microwave signals to obtain the vehicle's speed v and distance s from the gate in real time; at the same time, the camera is started to obtain the vehicle image at the current monitoring time point, and the vehicle image is recognized using a cascade classifier based on Haar features to obtain a rectangular frame containing the facial area; based on the facial area rectangular frame, the system extracts the driver's facial area image from these areas from the vehicle image, focusing on the driver's facial image to provide accurate input for the facial expression recognition model;

[0043] The driver's facial region image acquired at the current monitoring time point is input into the facial expression recognition model, and the facial expression recognition results are output. These facial expression recognition results include the types of facial expressions, such as calmness, tension, anger, and fatigue; based on the facial expression recognition results, it is judged whether the driver is in a normal emotional state. If the driver's facial expression is calm or there is no obvious negative emotion, the system will artificially determine that the driver is in a normal emotional state; if the driver's facial expression is tense, angry or tired, it means that the driver has abnormal emotions. Under abnormal emotions, the driver's attention and reaction may be affected. Therefore, different monitoring conditions need to be formulated for drivers with abnormal emotions;

[0044] If the driver is in a normal emotional state, the first safe distance is calculated based on the current speed v of the vehicle; it is determined whether the current distance s between the vehicle and the gate is greater than the first safe distance, which is the minimum safe distance that the vehicle needs to maintain to avoid collision when the driver is in a normal emotional state; if so, the system considers the vehicle to be in a safe state and no further action is required; if not, the system considers that there is a risk of collision and executes alarm response measures;

[0045] If the driver is in an abnormal emotional state, the second safety distance is calculated based on the vehicle's current speed v. The second safety distance will be longer than the first safety distance because when the driver is in an abnormal emotion, more time and distance may be required to brake the vehicle. Determine whether the current distance s between the vehicle and the gate is greater than the second safety distance. If so, no operation is performed; if not, alarm response measures are executed.

[0046] The facial expression recognition model is built on the basis of the CNN model. The convolutional neural network, or CNN, is a deep learning model that is particularly suitable for processing image data. It can automatically and effectively extract features from images by simulating the working principle of the human visual system. The facial expression recognition model includes an input layer, a convolution layer, a pooling layer, a fully connected layer, and an output layer. The input layer is used to input the driver's facial area image. The convolution layer extracts the local features of the driver's facial area image through convolution operations. The convolution operation uses a set of learnable filters that slide on the image to calculate the weighted sum of the local area and generate a feature map. These feature maps capture features such as edges and textures in the image. The pooling layer is used to reduce the spatial dimension of the features, reduce the amount of calculation, and increase the invariance to image displacement. The fully connected layer is used to further learn feature representation. In the fully connected layer, each neuron is connected to all neurons in the previous layer for nonlinear transformation and feature combination. This layer helps the model learn more complex feature representations in preparation for the final classification task. The output layer is used to output the facial expression recognition results.

[0047] For the training of facial expression recognition model, the specific operations are:

[0048] A number of facial expression recognition training samples are obtained, each of which contains an image of the driver's facial area with annotated facial expression recognition results; all the acquired facial expression recognition training samples are divided into an expression recognition training set and an expression recognition test set, and the model gradually adjusts its internal parameters by learning the images and their annotations in the expression recognition training set to improve the recognition accuracy; the expression recognition test set is used to evaluate the model performance; the test set does not participate in the training process, and is used to test the recognition effect of the model after the training is completed to ensure the generalization ability of the model; the facial expression recognition model is trained using the expression recognition training set, in this step, the model gradually optimizes its parameters by repeatedly learning the images and annotations in the training set; the training process involves forward propagation and back propagation, the model generates prediction results through forward propagation, and then adjusts the parameters through back propagation to reduce the prediction error; the expression recognition test set is input into the facial expression recognition model to obtain the test results; the training conditions are set, and if the test results meet the training conditions, the trained facial expression recognition model is output; otherwise, the facial expression recognition model is continued to be trained using the expression recognition training set.

[0049] The specific operation of using the Haar feature-based cascade classifier to identify vehicle images is as follows:

[0050] The input vehicle image is grayed to obtain a grayscale image; converting a color image to a grayscale image can reduce the amount of data and simplify the processing process; a grayscale image only contains brightness information, not color information, which helps to improve the efficiency of image processing; an integrated graph is generated based on the grayscale image. The integrated graph is also called a feature graph or feature map, which is an image used to calculate Haar eigenvalues; the integrated graph is calculated from the local area of ​​the grayscale image, and usually contains positive and negative features, indicating the brightness contrast of different areas in the image; the generated integrated graph is used to calculate the Haar eigenvalues ​​of each candidate area in the grayscale image, and these eigenvalues ​​are It is obtained by calculating the pixel intensity difference in a specific area of ​​the image; a pre-trained weak classifier is used to judge each Haar feature value to screen out candidate facial regions. A weak classifier is a simple classifier, usually based on a decision tree or a simple rule; each weak classifier judges the Haar feature value to decide whether the candidate region may be a facial region; the candidate facial regions are screened layer by layer through cascade classifiers, each level contains several weak classifiers; if a certain level passes, it enters the next level, otherwise it is judged as a non-target region; the candidate facial region judged by all cascade classifiers is the detected facial region.

[0051] Most of the cascade classifiers based on Haar features are ready-made, have been trained and optimized with a large amount of data, and can be directly applied to tasks such as face detection; they are widely integrated in many computer vision libraries, such as OpenCV, which provides pre-trained models that can be directly called for face detection or other image recognition tasks; these pre-trained models have been extensively tested and verified and can work stably in a variety of scenarios and conditions, so you only need to load the model and apply it to your own image data without additional training steps. This can save a lot of time and resources, while also ensuring the accuracy and efficiency of recognition.

[0052] The calculation method of the first safety distance is as follows:

[0053] Obtain several vehicle braking time data of different vehicle models. This is a set of data obtained through experiments or collection. The data collection involves multiple vehicle models to ensure the universality and accuracy of the results. Each vehicle braking time data contains the initial speed of a vehicle and the time required for braking from the initial speed. For any vehicle braking time data, the initial speed is divided by the time required for braking to calculate the deceleration of the current vehicle. Calculate the average of all the decelerations obtained to obtain the average vehicle deceleration a. This average can be used as a reference value when calculating the safety distance to help predict how much distance the vehicle needs to stop completely in an emergency.

[0054] Obtain a driver reaction time data set, which contains the reaction time required for several drivers from receiving warning information to taking braking measures. This data can be obtained through experiments or simulation tests; calculate the average of all reaction times to obtain the driver's average reaction time t, which represents the average reaction time of general drivers in emergency situations;

[0055] Using the formula Calculate the first safe distance s1 of the vehicle at the current speed, where s buffer This is the preset buffer distance. Since the vehicle does not necessarily decelerate uniformly when parking, the above average deceleration only represents the value under ideal conditions. In actual situations, the distance required for parking is generally greater than the ideal distance. Therefore, a buffer distance needs to be set to prevent the ideal distance from being insufficient to completely stop the vehicle.

[0056] The calculation method of the second safety distance is as follows:

[0057] Using the formula Calculate and obtain the second safe distance s2 of the vehicle at the current speed, where k is the preset safety factor. When the driver has abnormal emotions, such as tension and fatigue, the driver may not be able to take braking measures in time, so more distance is needed for the driver to stop safely. Since the degree of emotional influence may vary from person to person, the safety factor k cannot be specifically calculated here, and can be set directly, such as k=1.5.

[0058] The time interval between every two monitoring time points is adjusted by the particle swarm optimization algorithm with the goal of minimizing the misjudgment rate and maximizing the time interval. The specific operations are as follows:

[0059] A1: Randomly generate a particle swarm in the range of 0-1 second, and the position of each particle represents a candidate time interval;

[0060] A2: For any particle, initialize the velocity of the particle and use the current position as the individual optimal position of the particle; use the formula Calculate the fitness value F of the particle, where P represents the misjudgment rate corresponding to the current candidate time interval, and L represents the duration of the current candidate time interval; and They are and the weight of L;

[0061] A3: Traverse all particles and take the position corresponding to the particle with the highest fitness value F as the global optimal position;

[0062] A4: For any particle, update the speed and position of the particle, and recalculate the fitness value; if the updated fitness value is higher than the fitness value of the individual best position of the particle, update the individual best position of the particle to the current position;

[0063] A5: Traverse all particles. If the fitness value of the individual best position of any particle is higher than the fitness value of the global best position, the global best position is updated to the individual best position of the particle.

[0064] A6: Repeat steps A4 to A5 until the preset maximum number of iterations is reached, and the obtained global best position is the optimal time interval.

[0065] Embodiment 2, a collision prevention monitoring and early warning system for a gate machine, such as Figure 1 As shown, including:

[0066] The vehicle detection module is used to detect the presence of a moving vehicle using a photoelectric sensor, and activate the microwave radar and camera when a moving vehicle is detected;

[0067] Microwave radar module, used to obtain the vehicle's speed v and distance s from the gate;

[0068] The facial region image acquisition module is used to use the camera to acquire the vehicle image at the current monitoring time point, and use the cascade classifier based on Haar features to identify the vehicle image and obtain a rectangular frame containing the facial region; and extract the driver's facial region image from the vehicle image according to the facial region rectangular frame;

[0069] The facial expression recognition module is used to input the driver's facial area image obtained at the current monitoring time point into the facial expression recognition model and output the facial expression recognition result; based on the facial expression recognition result, it is determined whether the driver is in a normal emotional state;

[0070] The monitoring and early warning module is used to perform different operations according to the driver's emotional state. If the driver is in a normal emotional state, the first safe distance is calculated based on the current speed v of the vehicle; whether the current distance s between the vehicle and the gate is greater than the first safe distance, if so, no operation is performed; if not, alarm response measures are performed; if the driver is in an abnormal emotional state, the second safe distance is calculated based on the current speed v of the vehicle; whether the current distance s between the vehicle and the gate is greater than the second safe distance, if so, no operation is performed; if not, alarm response measures are performed.

[0071] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for anti-collision monitoring and early warning for gates, characterized in that: include: Use photoelectric sensors to detect whether there is a moving vehicle. When a moving vehicle is detected, start the microwave radar to obtain the vehicle's speed v and distance s from the gate; Based on the current speed v of the vehicle, calculate and obtain the safe distance of the vehicle at the current speed; determine whether the current distance s between the vehicle and the gate is greater than the currently calculated safe distance. If so, no operation is performed; if not, an alarm response measure is executed; The calculation method of safety distance is as follows: Using the formula Calculate the safe distance s of the vehicle at the current speed safe , where s buffer is the preset buffer distance; a is the average deceleration of the vehicle; t is the average driver reaction time; k is the preset safety factor; The calculation method of the average vehicle deceleration a and the average driver reaction time t is as follows: Obtain several vehicle braking time data of different vehicle models, each vehicle braking time data includes the initial speed of a vehicle and the time required for braking from the initial speed; for any vehicle braking time data, divide the initial speed by the time required for braking to calculate the deceleration of the current vehicle; calculate the average of all the obtained decelerations to obtain the vehicle average deceleration a; Obtain a driver reaction time data set, which contains the reaction time required for several drivers from receiving warning information to taking braking measures; calculate the average of all reaction times to obtain the average driver reaction time t; The safe distance of the vehicle at the current speed can be calculated based on the driver's facial expression recognition: When a moving vehicle is detected, the camera is started to obtain the vehicle image at the current monitoring time point, and the vehicle image is recognized using a cascade classifier based on Haar features to obtain a rectangular frame containing the facial area; the driver's facial area image is extracted from the vehicle image based on the facial area rectangular frame; Input the driver's facial region image acquired at the current monitoring time point into the facial expression recognition model, and output the facial expression recognition result; Based on the facial expression recognition results, determine whether the driver is in a normal emotional state; If the driver is in a normal emotional state, based on the current speed v of the vehicle, the calculation formula of the safety distance is used, and the safety factor k is set to 1 to calculate the first safety distance; determine whether the current distance s between the vehicle and the gate is greater than the first safety distance. If so, no operation is performed; If not, the alarm response measures are executed; If the driver is in an abnormal emotional state, the second safety distance is calculated based on the current speed v of the vehicle, using the calculation formula of the safety distance, and taking the safety factor k as 1.5; it is determined whether the current distance s between the vehicle and the gate is greater than the second safety distance. If so, no operation is performed; if not, the alarm response measures are executed; The facial expression recognition model is built based on the CNN model, including input layer, convolution layer, pooling layer, fully connected layer and output layer; the input layer is used to input the driver's facial area image; the convolution layer is used to extract the features of the driver's facial area image; the pooling layer is used to reduce the spatial dimension of the features; the fully connected layer is used to further learn the feature representation; the output layer is used to output the facial expression recognition results; For the training of facial expression recognition model, the specific operations are: Acquire a number of facial expression recognition training samples, each of which includes an image of the driver's facial region with annotated facial expression recognition results; divide all acquired facial expression recognition training samples into an expression recognition training set and an expression recognition test set; train a facial expression recognition model using the expression recognition training set, and then input the expression recognition test set into the facial expression recognition model to obtain test results; Set the training conditions. If the test results meet the training conditions, output the trained facial expression recognition model; otherwise, continue to use the expression recognition training set to train the facial expression recognition model. The specific operation of using the Haar feature-based cascade classifier to identify vehicle images is as follows: The input vehicle image is grayed to obtain a gray image; an integrated graph is generated according to the gray image; and the Haar feature value of each candidate area in the gray image is calculated using the generated integrated graph; Use pre-trained weak classifiers to judge each Haar feature value and screen out candidate facial regions; screen candidate facial regions layer by layer through cascade classifiers, each level contains several weak classifiers; if a level passes, it enters the next level, otherwise it is judged as a non-target area; the candidate facial region judged by all cascade classifiers is the detected facial region; The time interval between every two monitoring time points is adjusted by the particle swarm optimization algorithm with the goal of minimizing the misjudgment rate and maximizing the time interval. The specific operations are as follows: A1: Randomly generate a particle swarm in the range of 0-1 second, and the position of each particle represents a candidate time interval; A2: For any particle, initialize the velocity of the particle and use the current position as the individual optimal position of the particle; use the formula Calculate the fitness value F of the particle, where P represents the misjudgment rate corresponding to the current candidate time interval, and L represents the duration of the current candidate time interval; and They are and the weight of L; A3: Traverse all particles and take the position corresponding to the particle with the highest fitness value F as the global optimal position; A4: For any particle, update the speed and position of the particle, and recalculate the fitness value; if the updated fitness value is higher than the fitness value of the individual best position of the particle, update the individual best position of the particle to the current position; A5: Traverse all particles. If the fitness value of the individual best position of any particle is higher than the fitness value of the global best position, the global best position is updated to the individual best position of the particle. A6: Repeat steps A4 to A5 until the preset maximum number of iterations is reached, and the obtained global best position is the optimal time interval.

2. An anti-collision monitoring and early warning system for gates, characterized in that: The system adopts the anti-collision monitoring and early warning method for gates as described in claim 1 above, comprising: The vehicle detection module is used to detect the presence of a moving vehicle using a photoelectric sensor, and activate the microwave radar and camera when a moving vehicle is detected; Microwave radar module, used to obtain the vehicle's speed v and distance s from the gate; The facial region image acquisition module is used to use the camera to acquire the vehicle image at the current monitoring time point, and use the cascade classifier based on Haar features to identify the vehicle image and obtain a rectangular frame containing the facial region; and extract the driver's facial region image from the vehicle image according to the facial region rectangular frame; The facial expression recognition module is used to input the driver's facial area image obtained at the current monitoring time point into the facial expression recognition model and output the facial expression recognition result; based on the facial expression recognition result, it is determined whether the driver is in a normal emotional state; The monitoring and early warning module is used to perform different operations according to the driver's emotional state. If the driver is in a normal emotional state, the first safe distance is calculated based on the current speed v of the vehicle; whether the current distance s between the vehicle and the gate is greater than the first safe distance, if so, no operation is performed; if not, alarm response measures are performed; if the driver is in an abnormal emotional state, the second safe distance is calculated based on the current speed v of the vehicle; whether the current distance s between the vehicle and the gate is greater than the second safe distance, if so, no operation is performed; if not, alarm response measures are performed.

Images