User entry and exit detection method based on image acquisition

Abstract

The present invention relates to the field of image detection, and in particular to a user access detection method based on image acquisition. The method comprises the following steps: inputting facial information and associated user information corresponding to a number of targets, capturing real-time facial information and environmental information of the target to be monitored, generating pre-learning data corresponding to the real-time facial information, learning the facial pre-learning data to generate a facial comparison result, comparing the result with a similarity threshold, judging whether the facial comparison result passes, learning environmental information to generate an environmental comparison result, judging whether the environmental comparison result passes according to a similarity threshold, and judging whether to open access control according to a comprehensive comparison result. By utilizing machine learning, facial recognition is performed on the target to be monitored, effectively avoiding loss or misentry of the target, and realizing automatic recognition of the target and the associated user and access control. This method not only improves the security of user access detection, but also provides a more convenient and personalized service for the target owner.

Description

Technical Field

[0001] The present invention relates to the field of image detection, and in particular to a method for detecting user entry and exit based on image acquisition. Background Art

[0002] With technological advancements, automated monitoring and management of user access is becoming increasingly important. Image-based user access detection methods primarily rely on computer vision technology. These methods capture images or video of a target through a camera, and then analyze and identify the images using image processing algorithms and machine learning models, enabling automated monitoring of user access.

[0003] With the continuous development of deep learning technology, by combining multiple modal information such as images and sounds, more accurate and efficient machine learning models can be trained, which can achieve more comprehensive and accurate identification and management of users, and further improve the accuracy and efficiency of user entry and exit detection.

[0004] Chinese Patent Authorization Publication No. CN109784208B discloses an image-based target behavior detection method, which includes the following steps: first, a camera module captures multiple image samples of each target posture, labels and categorizes the target posture in each image sample to obtain an image dataset containing each target posture; then, a detection model is established, with a judgment error rate and a number of iterative training cycles set; then, the image dataset is imported into the detection model, and the detection model is trained according to a predetermined number of iterative training cycles in S2 to obtain a detection model for detecting target behavior; and when detecting the target behavior, the camera module captures a scene image of the target, which is then input into the detection model. The detection model extracts information from the scene image to determine the probability that the scene image belongs to a particular type of target posture. This invention provides an image-based target behavior detection method that helps users understand their target's status at all times.

[0005] Chinese patent application publication number: CN111507211A discloses a target monitoring method, device, equipment and storage medium for detecting whether a target active in a community is under the supervision of an owner and whether it is active within a specified time period, thereby facilitating target monitoring. The method includes: acquiring image data within a preset area; when it is determined that the acquisition time of the image data is within a first preset time period, performing target detection and portrait detection based on the image data; when it is determined that a target image is detected in the image data and the target image exists in a pre-established target database, if it is determined that the portrait detected in the image data does not include the target portrait, triggering an alarm message to be reported, wherein the target portrait is a portrait corresponding to the target image in the target database, and the target database stores the corresponding relationship between the target and the portrait of the target owner.

[0006] However, the above method has the following problems: the access control device cannot monitor the whereabouts of the target, thereby failing to reduce the risk of the target getting lost. Summary of the Invention

[0007] To this end, the present invention provides a user entry and exit detection method based on image acquisition to overcome the problem that the access control equipment in the prior art cannot monitor the whereabouts of the target, thereby failing to reduce the risk of the target getting lost.

[0008] To achieve the above objectives, the present invention provides a method for detecting user entry and exit based on image acquisition, comprising:

[0009] Enter facial information and associated user information corresponding to several targets;

[0010] Capture real-time facial information and environmental information of the target to be monitored;

[0011] Pre-learning the real-time facial information and generating corresponding pre-learning data;

[0012] The facial learning model learns the pre-learning data to generate a facial comparison result, compares the facial comparison result with a similarity threshold, and determines whether the facial comparison result passes based on the comparison result;

[0013] Learning the environmental information to generate an environmental comparison result, and determining whether the environmental comparison result is passed according to the similarity threshold;

[0014] In response to a passing facial comparison result or a passing environmental comparison result, a corresponding comprehensive comparison result is generated, and based on the comprehensive comparison result, it is determined whether to open the door access control;

[0015] Wherein, when the real-time facial information is pre-learned, the real-time facial information is also filtered and a number of facial pre-learning data with a learning rate of a standard learning rate are generated;

[0016] The standard learning rate is a learning rate that can be recognized by the facial learning model, which is inversely proportional to the resolution of the facial information. The higher the resolution, the lower the corresponding standard learning rate. For single learning, the corresponding standard learning rate is a single learning rate;

[0017] The pre-learning data is standardized data that can be recognized by the facial learning model;

[0018] The facial learning model is trained and generated based on a training set formed by a plurality of facial information.

[0019] Furthermore, the step of entering facial information and associated user information corresponding to a plurality of targets includes:

[0020] The terminal platform sends summary instructions to several sub-platforms;

[0021] The sub-platform captures corresponding facial information and collects corresponding associated user information;

[0022] The sub-platform sends the facial information and the associated user information to the terminal platform;

[0023] The associated user information includes user name, contact information and / or corresponding target associated information.

[0024] Furthermore, the step of photographing the real-time facial information and environmental information of the target to be monitored includes:

[0025] Several image acquisition devices are set up at the access control point;

[0026] When the monitor detects the target to be monitored, the transmitter sends a shooting instruction to the image acquisition device;

[0027] The image acquisition device photographs the target to be monitored to obtain corresponding real-time facial information and environmental information;

[0028] Transmitting the real-time facial information and the environmental information into a pre-learner;

[0029] Wherein, the image acquisition device is provided with a camera with adjustable angle.

[0030] Furthermore, the step of pre-learning the real-time facial information includes:

[0031] The pre-learner performs pixel filtering on the real-time facial information and forms corresponding filtering information;

[0032] Selecting several indicator features of the filtering information;

[0033] Standardizing the filtered information and generating corresponding pre-learning data;

[0034] Wherein, a pixel threshold is set in the pre-learner;

[0035] The indicator features include the pixel size, target breed and / or fur color of the filtered information;

[0036] The standardization process is to segment the filter information according to a standard learning rate.

[0037] Furthermore, the step of pixel filtering includes:

[0038] The pre-learner obtains pixel information of the real-time facial information;

[0039] comparing the pixel information with the pixel threshold;

[0040] When the pixel information is greater than the pixel threshold, the real-time facial information is retained; when the pixel information is less than the pixel threshold, the real-time facial information is filtered.

[0041] Furthermore, the step of the face learning model learning the pre-learning data includes:

[0042] adjusting the learning rate of the face learning model to the standard learning rate;

[0043] Passing the pre-learning data into the face learning model;

[0044] The facial learning model learns the pre-learning data and obtains corresponding facial comparison results.

[0045] Furthermore, the facial comparison result is compared with the similarity threshold. When the facial comparison result is less than the similarity threshold, the transmitter sends an access control instruction, and the access control device closes the access control in response to the access control instruction.

[0046] Furthermore, when the facial comparison result is greater than the similarity threshold, the environmental information is learned to generate a corresponding environmental comparison result; when the environmental comparison result is less than the similarity threshold, the transmitter sends an access control instruction, and the access control device closes the access control in response to the access control instruction.

[0047] Furthermore, when the environment comparison result is greater than the similarity threshold, the corresponding associated user information is matched according to the facial comparison result, and a confirmation message for going out is sent to the corresponding associated user. When the associated user confirms to go out, the door control is opened in response to the going out instruction; when the associated user confirms to close the door, the door control is closed in response to the closing instruction;

[0048] Wherein, the exit instruction is an instruction sent by the transmitter when the confirmation information of the associated user to open the door is obtained;

[0049] The closing instruction is an instruction sent by the transmitter when the confirmation closing information of the associated user is obtained.

[0050] Furthermore, when the monitor detects more than one target to be monitored, the transmitter sends the access control instruction, and the access control device closes the access control in response to the access control instruction.

[0051] Compared with the prior art, the present invention records the facial information and associated user information corresponding to several targets, captures the real-time facial information and environmental information of the target to be monitored, and generates pre-learning data corresponding to the real-time facial information. The facial pre-learning data is learned to generate a facial comparison result, which is compared with the similarity threshold to determine whether the facial comparison result passes. The environmental information is learned to generate an environmental comparison result, and whether the environmental comparison result passes according to the similarity threshold is determined. The access control is determined based on the comprehensive comparison result. Machine learning is used to perform facial recognition on the target to be monitored, which effectively avoids the loss and misentry of the target, realizes automatic recognition of the target and the associated user, and access control, which not only improves the security of user entry and exit detection, but also provides more convenient and personalized services for the target owner.

[0052] Furthermore, by setting up a terminal platform and sending summary instructions to several sub-platforms, the sub-platforms send the collected facial information and associated user information to the terminal platform, which facilitates the collection of the target and the corresponding associated user information, and facilitates the analysis and processing of the facial information and associated user information.

[0053] Furthermore, by setting a camera with an adjustable angle on the image acquisition device to shoot the real-time facial information of the monitored target, it can be ensured that the real-time facial image at the best angle is captured. Through this process, real-time facial information monitoring of the target can be achieved, which improves the accuracy of subsequent machine learning in identifying the real-time facial image of the monitored target.

[0054] Furthermore, by performing pixel filtering on real-time facial information and selecting several indicator features to standardize the filtered information, it is convenient to remove noise and unimportant details, ensure the consistency and comparability of the data, and thus facilitate the extraction of clearer and more representative facial features.

[0055] Furthermore, by setting up a facial learning model, learning the pre-learning data and generating corresponding facial comparison results, it is possible to learn from the pre-learning data and improve its recognition ability, and ultimately achieve accurate recognition and comparison of the real-time facial information of the monitored target.

[0056] Furthermore, by setting a similarity threshold and comparing the facial comparison result with the similarity threshold, it ensures that only authorized monitored targets can pass through the access control, increasing security and convenience, and providing a safe, convenient and automated target access management detection method.

[0057] By entering the target's facial information and associated user information, capturing real-time facial information, performing pre-learning and facial learning, and controlling access control based on facial comparison results, a complete user access detection method based on image acquisition is constructed. From information entry, real-time facial information acquisition, pre-learning, facial learning, to access control, a closed-loop management method is formed, which improves the security and convenience of target access management, reduces manual intervention, and improves the efficiency of user access detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 Flowchart of the user entry and exit detection method based on image acquisition of the present invention;

[0059] Figure 2 This is a flow chart of the present invention for entering facial information corresponding to several targets and associated user information;

[0060] Figure 3 This is a flow chart of the present invention for photographing real-time facial information of a monitored target;

[0061] Figure 4 This is a flow chart of the present invention for pre-learning real-time facial information. DETAILED DESCRIPTION

[0062] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0063] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0064] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0065] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0066] See also Figure 1 As shown, it is a flow chart of the user entry and exit detection method based on image acquisition of the present invention, including:

[0067] Step S1, input facial information and associated user information corresponding to several targets;

[0068] Step S2, capturing real-time facial information and environmental information of the target to be monitored;

[0069] Step S3, pre-learning the real-time facial information and generating corresponding pre-learning data;

[0070] Step S4: the facial learning model learns the pre-learning data to generate a facial comparison result, compares the facial comparison result with a similarity threshold, and determines whether the facial comparison result passes based on the comparison result;

[0071] Step S5: learning the environmental information to generate an environmental comparison result, and judging whether the environmental comparison result is passed based on a similarity threshold;

[0072] Step S6, in response to the passing of the facial comparison result or the passing of the environmental comparison result, a corresponding comprehensive comparison result is generated, and whether to open the door control is determined according to the comprehensive comparison result;

[0073] During pre-learning, real-time facial information is filtered and a number of pre-learning data with a standard learning rate are generated;

[0074] The standard learning rate is the learning rate that the facial learning model can recognize. It is inversely proportional to the resolution of the facial information. The higher the resolution, the lower the corresponding standard learning rate. For single-shot learning, the corresponding standard learning rate is a single learning rate.

[0075] The pre-learning data is standardized data that can be recognized by the facial learning model;

[0076] The facial learning model is trained and generated based on a training set formed by several facial information.

[0077] In the specific implementation, when the learning rate is 150 / s-170 / s, the facial learning model has the best learning effect on the pre-learning data and the best facial comparison result generation effect. When the resolution is 100ppi, the preferred standard learning rate is 160 / s.

[0078] By entering the facial information and associated user information corresponding to several targets, capturing the real-time facial information and environmental information of the target to be monitored, and generating pre-learning data corresponding to the real-time facial information, the facial pre-learning data is learned to generate a facial comparison result, which is compared with the similarity threshold to determine whether the facial comparison result passes. The environmental information is learned to generate an environmental comparison result, and the environmental comparison result is determined to pass based on the similarity threshold. The access control is determined based on the comprehensive comparison result. Machine learning is used to perform facial recognition on the target to be monitored, effectively avoiding the loss and misentry of the target, and realizing automatic recognition of the target and the associated user and access control. This not only improves the security of user entry and exit detection, but also provides more convenient and personalized services for the target owner.

[0079] See also Figure 2 As shown, it is a flow chart of the present invention for entering facial information corresponding to several targets and associated user information, including:

[0080] Step S11: The terminal platform sends a summary instruction to several sub-platforms;

[0081] Step S12: The sub-platform captures the corresponding facial information and collects the corresponding associated user information;

[0082] Step S13: The sub-platform sends the facial information and associated user information to the terminal platform;

[0083] The associated user information includes user name, contact information and / or corresponding target associated information.

[0084] In practice, after receiving the summary command, each sub-platform begins capturing the target's facial information and simultaneously collects associated user information related to these targets. When collecting associated user information, it obtains the user's name, contact information, and other information through user input or database query.

[0085] By setting up a terminal platform and sending summary instructions to several sub-platforms, the sub-platforms will send the collected facial information and related user information to the terminal platform, which facilitates the collection of the target and the corresponding related user information, and facilitates the analysis and processing of the facial information and related user information.

[0086] See also Figure 3 As shown, it is a flow chart of the present invention for photographing real-time facial information of a monitored target, including:

[0087] Step S21: several image acquisition devices are set up at the access control area, specifically, they can be set up at the door handle position to facilitate shooting;

[0088] Step S22: When the monitor detects the target to be monitored, the transmitter sends a shooting instruction to the image acquisition device;

[0089] Step S23: The image acquisition device captures the target to be monitored and obtains corresponding real-time facial information and environmental information;

[0090] Step S24, transferring the real-time facial information and environmental information to the pre-learner;

[0091] Wherein, the image acquisition device is provided with a camera with adjustable angle.

[0092] In a specific implementation, the target to be monitored can be a human or an animal, and the monitor can be a motion sensor or other types of sensors. When the target is detected approaching the access control area, the transmitter is automatically triggered to send a shooting instruction to the image acquisition device. By automatically responding to the appearance of the target, it is ensured that the target's facial information can be captured in time when the target reaches the access control area.

[0093] The camera in the image acquisition device can adjust its angle to adapt to different shooting environments and target heights to ensure that the best facial image can be captured. During the shooting process, the image quality should be ensured to meet the requirements of facial recognition, such as sufficient clarity and contrast.

[0094] In a specific implementation, the image acquisition device includes an error handling mechanism. When shooting fails or the image quality is poor, the device retakes the image or notifies maintenance personnel to check the equipment.

[0095] By setting up a camera with an adjustable angle on the image acquisition device to shoot the real-time facial information of the monitored target, it can be ensured that the real-time facial image at the best angle is captured. Through this process, real-time facial information monitoring of the target can be achieved, which improves the accuracy of subsequent machine learning in identifying the real-time facial image of the monitored target.

[0096] See also Figure 4 As shown in FIG, it is a flow chart of the present invention for pre-learning real-time facial information, including:

[0097] Step St1: the pre-learner performs pixel filtering on the real-time facial information and generates corresponding filtered information;

[0098] Step St2, selecting several indicator features of the filtering information;

[0099] Step St3, standardize the filtered information and generate corresponding pre-learning data;

[0100] Among them, a pixel threshold is set in the pre-learner;

[0101] Indicator characteristics include pixel size of filtered information, target breed and / or coat color;

[0102] The normalization process is to divide the filtered information according to the standard learning rate.

[0103] Image filtering techniques (such as mean filtering, Gaussian filtering, etc.) are applied to smooth the image and reduce noise. Edge detection algorithms (such as Canny edge detection) are used to highlight facial contours and feature edges. According to the needs of facial recognition, a series of key facial features are selected, such as the position and shape of the eyes, nose, and mouth, the contour and symmetry of the face, and skin texture. These features should have sufficient discrimination to represent different individuals.

[0104] The selected features are standardized to ensure their consistency and comparability across different images. Standardization includes scaling feature values ​​to a specific range (such as 0-1 or -1 to 1), or performing normalization (such as converting feature values ​​to relative values), and dividing the filtered information at a standard learning rate. Through standardization, the scale and distribution differences between different images can be eliminated, improving the accuracy of subsequent feature extraction, thereby improving the accuracy and efficiency of tasks such as facial recognition.

[0105] By performing pixel filtering on real-time facial information and selecting several indicator features to standardize the filtered information, it is easy to remove noise and unimportant details, ensuring the consistency and comparability of the data, thereby facilitating the extraction of clearer and more representative facial features.

[0106] Specifically, the steps of pixel filtering include:

[0107] The pre-learner obtains pixel information of real-time facial information;

[0108] comparing pixel information to a pixel threshold;

[0109] When the pixel information is greater than the pixel threshold, the real-time facial information is retained; when the pixel information is less than the pixel threshold, the real-time facial information is filtered.

[0110] In practice, a pixel threshold is set to determine whether to retain facial information captured in real time. Specifically, when pixel information (such as brightness, grayscale value, or a specific color channel value) is greater than the set threshold, the pixel is considered valid facial information and is retained. When the pixel information is less than the threshold, the pixel is considered likely to be background, noise, or unimportant details and is filtered out.

[0111] This method is very useful in the pre-learning phase of facial recognition or facial feature extraction because it helps remove irrelevant information from the image, thereby highlighting facial features and improving the accuracy and efficiency of subsequent processing. However, setting an appropriate pixel threshold is a key issue. If the threshold is set too high, some important facial information may be mistakenly filtered out; if the threshold is set too low, it may not effectively remove noise and background information.

[0112] Determining the appropriate pixel threshold often requires some experimentation and tuning. This involves observing the effects of filtered images at different thresholds and evaluating the performance of these images in subsequent facial recognition or feature extraction tasks. Through trial and error, an optimal threshold can be found that effectively removes noise while preserving key facial information.

[0113] Furthermore, it's worth noting that pixel threshold-based filtering methods may not be suitable for all situations. For example, in certain scenes with poor lighting conditions or where facial features are not obvious, this method may fail. Therefore, in practical applications, it is necessary to combine other pre-learning techniques (such as image enhancement and denoising) with more advanced feature extraction and recognition algorithms to improve the accuracy and robustness of facial recognition.

[0114] Preferably, the pixel threshold is typically set to 1080px.

[0115] Among some possible implementations:

[0116] The pixel threshold is set to 1080px. If the pixel information of A's real-time facial information is 1000px, which is less than the pixel threshold, the pre-learner will filter A's real-time facial information.

[0117] Among some possible implementations:

[0118] The pixel threshold is set to 1080px. If the pixel information of B's ​​real-time facial information is 2000px, which is greater than the pixel threshold, the pre-learner will retain B's real-time facial information.

[0119] Among some possible implementations:

[0120] The pixel threshold is set to 900px. If the pixel information of C's real-time facial information is 1000px, which is greater than the pixel threshold, the pre-learner will retain C's real-time facial information.

[0121] Specifically, the steps for the face learning model to learn pre-learning data include:

[0122] Adjust the learning rate of the face learning model to the standard learning rate;

[0123] Pass the pre-learning data into the face learning model;

[0124] The facial learning model learns the pre-learning data and obtains the corresponding facial comparison results.

[0125] In practice, the learning rate is a crucial hyperparameter in model training, determining the step size used to update the model's weights in each iteration. An appropriate learning rate helps the model converge to the optimal solution more quickly, while an excessively high learning rate can cause the model to oscillate around the optimal solution or even diverge. An excessively low learning rate can significantly slow down the model training process.

[0126] Standard learning rate: Before training a face learning model, the learning rate is usually set to a preset "standard" value. This value may be derived from previous experiments or experience, or it may be estimated based on the complexity of the model and the size of the data.

[0127] Adjusting the learning rate: During actual training, the learning rate may be dynamically adjusted based on the model's training performance (such as the rate of decrease of the loss function). This can be achieved through techniques such as learning rate decay and learning rate scheduler.

[0128] Pre-learning data refers to pre-learned facial image data. This data will be used to train the facial learning model so that it can learn facial features and be used for subsequent facial recognition or classification tasks.

[0129] The process of a facial learning model learning from pre-learning data refers to the process of minimizing a loss function by iteratively optimizing its internal parameters (such as weights and biases). The loss function is usually used to measure the difference between the model's predictions and the actual results.

[0130] Forward propagation: In each iteration, the facial learning model receives a batch of pre-learning data as input and performs forward propagation through its internal network structure (such as convolutional layer, pooling layer, fully connected layer, etc.), and finally outputs the facial comparison result or classification result.

[0131] Calculate loss: Compare the output of the facial learning model with the true label and calculate the loss value.

[0132] Backpropagation: Based on the loss value, the gradient of each parameter is calculated by the chain rule, and the parameters of the facial learning model are updated in the opposite direction of the gradient to obtain the facial comparison result.

[0133] Facial comparison results refer to the results obtained by the model after extracting features and comparing the input facial images during or after training. These results can be used for tasks such as facial recognition and identity verification.

[0134] Feature extraction: The facial learning model extracts facial features of the input image through the learned facial feature representation.

[0135] Feature matching: The extracted facial features are compared with the features in the database to determine whether the input image matches an individual in the database.

[0136] By setting up a facial learning model, learning the pre-learning data and generating corresponding facial comparison results, it is possible to learn from the pre-learning data and improve its recognition ability, and ultimately achieve accurate recognition and comparison of the real-time facial information of the monitored target.

[0137] Specifically, the facial comparison result is compared with a similarity threshold. When the facial comparison result is less than the similarity threshold, the transmitter sends an access control instruction, and the access control device closes the access control in response to the access control instruction.

[0138] In a specific implementation, when the facial comparison result does not meet the similarity requirement (ie, is less than the similarity threshold), the transmitter (which may be a controller or computer of the access control system) will send an access control command.

[0139] This instruction is usually an electrical signal or network message used to instruct the access control device to perform a specific action.

[0140] Access control device response: After receiving the access control command, the access control device (such as electromagnetic lock, access gate, etc.) will perform corresponding actions according to the content of the command.

[0141] In this scenario, the access control device will respond to the access control command and close the access control, that is, prevent unauthorized monitored targets from entering.

[0142] It's important to note that in practical applications, the security of access control systems is paramount. Therefore, it's crucial to ensure the accuracy and robustness of facial learning models, as well as the rationality of similarity thresholds. Furthermore, privacy protection must be prioritized. When processing facial data, relevant privacy protection regulations and standards must be adhered to, ensuring the legitimacy and security of personal information. User experience is also crucial, and the design and implementation of access control systems must consider user experience, such as recognition speed, false alarm rate, and user feedback.

[0143] The facial comparison result indicates the degree of similarity between the real-time facial information and the preset facial information, while the similarity threshold is a preset standard used to determine whether the real-time facial image and the preset facial information are similar enough. If the facial comparison result is less than the similarity threshold, it means that the facial images are too different to be considered the same target. At this time, security measures need to be taken to close the access control.

[0144] Among some possible implementations:

[0145] Set the similarity threshold to 80%. If the facial comparison result is 70%, the facial comparison result is less than the similarity threshold, and the condition for closing the access control is met, the transmitter sends the access control command, and the access control device responds to the access control command to close the access control.

[0146] Specifically, when the facial comparison result is greater than the similarity threshold, the environmental information is learned to generate the corresponding environmental comparison result. When the environmental comparison result is less than the similarity threshold, the transmitter sends an access control command, and the access control device responds to the access control command to close the access control.

[0147] If the facial comparison result is greater than the similarity threshold, the system will automatically assume that the real-time facial information is similar enough to the preset facial information and may be the same target. If the environmental comparison result is less than the similarity threshold, the system will consider the environment unsafe and will not execute the access control command.

[0148] Among some possible implementations:

[0149] The similarity threshold is set to 80%. If the facial comparison result is 85%, the facial comparison result is greater than the similarity threshold, and the environmental information is learned to generate the corresponding environmental comparison result.

[0150] By setting a similarity threshold and comparing the facial comparison result with the similarity threshold, it ensures that only authorized monitored targets can pass through the access control, increasing security and convenience, and providing a safe, convenient and automated target access management detection method.

[0151] Specifically, when the environment comparison result is greater than the similarity threshold, the corresponding associated user information is matched according to the facial comparison result, and a confirmation message for going out is sent to the corresponding associated user. When the associated user confirms to go out, the access control is opened in response to the going out instruction. When the associated user confirms to close the door, the access control is closed in response to the closing instruction.

[0152] Among them, the exit command is a command sent by the transmitter when the confirmation information of the associated user is obtained;

[0153] The closing instruction is a instruction sent by the transmitter when the confirmation of closing the door is obtained from the associated user.

[0154] In specific implementations, when the environmental comparison result is greater than the similarity threshold, the system will search for associated user information that matches the facial feature vector and send a confirmation message for going out. The confirmation message may contain some basic information, such as the user's name, going out time, access control location, etc. It may also contain some additional information, such as the authorization status of going out (whether allowed), the destination prompt after going out, etc. The confirmation message can be sent to the associated user in a variety of ways, such as mobile phone text messages, emails, APP push notifications, etc. The choice of which sending method depends on the user's preference and system configuration.

[0155] After receiving the confirmation message, the user can take corresponding actions based on the content of the message, such as confirming to go out, canceling to go out, etc.

[0156] The system can also perform follow-up processing based on user feedback, such as updating access control status, recording exit logs, etc.

[0157] It should be noted that:

[0158] 1. Access control accuracy: Ensure the accuracy and robustness of the facial learning model to reduce false positives and false negatives.

[0159] 2. Real-time access control: The access control system needs to be able to quickly respond to facial comparison results and send a confirmation message within the user's waiting time.

[0160] 3. Access control security: protect users’ privacy information and ensure the security and confidentiality of messages during sending.

[0161] 4. User experience: Provide a friendly user interface and operation process to improve user satisfaction and convenience.

[0162] The associated user confirms closing the door through some means (such as a button on a mobile phone APP, touch sensing on the access control panel, voice command, etc.). Once the system receives the information that the associated user confirms closing the door, it will trigger an internal mechanism to generate a closing instruction. This closing instruction is an electrical signal or network message specifically used to instruct the access control device to perform the closing action.

[0163] After receiving a closing instruction, the access control device (such as an electromagnetic lock, access control gate, sliding door, etc.) will execute the closing action according to the instruction content.

[0164] This may involve the movement of mechanical structures, the locking of electronic locks, or the establishment of other forms of physical barriers to prevent unauthorized access to the monitored object.

[0165] After the access control is successfully closed, the system usually updates its internal status record and marks the access control as currently closed. This helps the system monitor and record the use of the access control and provides a basis for subsequent access control decisions.

[0166] In some cases, the system may send a message or prompt to the user to confirm that the access control is closed to enhance user trust and satisfaction. This can be achieved through mobile APP notification, SMS, email or other communication methods.

[0167] Specifically, when the monitor detects more than one target to be monitored, the transmitter sends an access control instruction, and the access control device closes the access control in response to the access control instruction.

[0168] In an access control system, sensors are configured to identify and count objects within a specific area, such as in front of a gated entrance.

[0169] The target to be monitored may be a target that has been pre-registered and authorized in the system, or any target may be considered as an object that needs to be monitored according to the settings.

[0170] The monitor uses a camera to monitor the number of objects in front of the access control entrance.

[0171] In specific implementations, the camera identifies targets through facial recognition, body feature recognition, weight sensing or infrared sensing, and counts the number of targets appearing in the monitoring area at the same time.

[0172] When the monitor detects more than one target to be monitored, an internal mechanism is triggered to determine whether an access control command needs to be sent.

[0173] If the system is configured to allow only a single target to pass or to require special authorization to allow multiple targets to pass, and the current situation does not meet these conditions, the monitor will send a door access command to the access control system's transmitter.

[0174] By entering the target's facial information and associated user information, capturing real-time facial information, performing pre-learning and facial learning, and controlling access control based on facial comparison results, a complete user access detection method based on image acquisition is constructed. From information entry, real-time facial information acquisition, pre-learning, facial learning, to access control, a closed-loop management method is formed, which improves the security and convenience of target access management, reduces manual intervention, and improves the efficiency of user access detection.

[0175] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0176] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A user entry and exit detection method based on image acquisition, characterized in that: include: Enter facial information and associated user information corresponding to several targets; Capture real-time facial information and environmental information of the target to be monitored; Pre-learning the real-time facial information and generating corresponding pre-learning data; The facial learning model learns the pre-learning data to generate a facial comparison result, compares the facial comparison result with a similarity threshold, and determines whether the facial comparison result passes based on the comparison result; Learning the environmental information to generate an environmental comparison result, and determining whether the environmental comparison result is passed according to the similarity threshold; In response to a passing facial comparison result or a passing environmental comparison result, a corresponding comprehensive comparison result is generated, and based on the comprehensive comparison result, it is determined whether to open the door access control; Wherein, when the real-time facial information is pre-learned, the real-time facial information is also filtered and a number of facial pre-learning data with a learning rate of a standard learning rate are generated; The standard learning rate is a learning rate that can be recognized by the facial learning model, which is inversely proportional to the resolution of the facial information. The higher the resolution, the lower the corresponding standard learning rate. For single learning, the corresponding standard learning rate is a single learning rate; The pre-learning data is standardized data that can be recognized by the facial learning model; The facial learning model is trained and generated based on a training set formed by a plurality of facial information; Comparing the facial comparison result with the similarity threshold, when the facial comparison result is less than the similarity threshold, the transmitter sends an access control instruction, and the access control device closes the door in response to the access control instruction; When the facial comparison result is greater than the similarity threshold, the environmental information is learned to generate a corresponding environmental comparison result; when the environmental comparison result is less than the similarity threshold, the transmitter sends an access control instruction, and the access control device closes the door in response to the access control instruction; When the environment comparison result is greater than the similarity threshold, the corresponding associated user information is matched according to the facial comparison result, and a confirmation message for going out is sent to the corresponding associated user. When the associated user confirms to go out, the door control is opened in response to the going out instruction. When the associated user confirms to close the door, the door control is closed in response to the closing instruction. Wherein, the exit instruction is an instruction sent by the transmitter when the confirmation information of the associated user to open the door is obtained; The closing instruction is an instruction sent by the transmitter when the confirmation closing information of the associated user is obtained.

2. The user entry and exit detection method based on image acquisition according to claim 1, characterized in that: The steps of entering facial information and associated user information corresponding to a number of targets include: The terminal platform sends summary instructions to several sub-platforms; The sub-platform captures corresponding facial information and collects corresponding associated user information; The sub-platform sends the facial information and the associated user information to the terminal platform; The associated user information includes user name, contact information and / or corresponding target associated information.

3. The user entry and exit detection method based on image acquisition according to claim 2, characterized in that: The step of capturing the real-time facial information and environmental information of the target to be monitored includes: Several image acquisition devices are set up at the access control point; When the monitor detects the target to be monitored, the transmitter sends a shooting instruction to the image acquisition device; The image acquisition device photographs the target to be monitored to obtain corresponding real-time facial information and environmental information; Transmitting the real-time facial information and the environmental information into a pre-learner; Wherein, the image acquisition device is provided with a camera with adjustable angle.

4. The user entry and exit detection method based on image acquisition according to claim 3, characterized in that: The step of pre-learning the real-time facial information includes: Performing pixel filtering on the real-time facial information through a pre-learner and forming corresponding filtering information; Selecting several indicator features of the filtering information; Standardizing the filtered information and generating corresponding pre-learning data; Wherein, a pixel threshold is set in the pre-learner; The indicator features include the pixel size, target breed and / or fur color of the filtered information; The standardization process is to segment the filter information according to a standard learning rate.

5. The user entry and exit detection method based on image acquisition according to claim 4, characterized in that: The steps of pixel filtering include: The pre-learner obtains pixel information of the real-time facial information; comparing the pixel information with the pixel threshold; When the pixel information is greater than the pixel threshold, the real-time facial information is retained; when the pixel information is less than the pixel threshold, the real-time facial information is filtered.

6. The user entry and exit detection method based on image acquisition according to claim 5, characterized in that: The step of the face learning model learning the pre-learning data includes: adjusting the learning rate of the face learning model to the standard learning rate; Passing the pre-learning data into the face learning model; The facial learning model learns the pre-learning data and obtains corresponding facial comparison results.

7. The method for detecting user entry and exit based on image acquisition according to claim 6, characterized in that: When the monitor detects more than one target to be monitored, the transmitter sends the access control instruction, and the access control device closes the access control in response to the access control instruction.

Images