A monitoring and early warning method for an electric vehicle entering an elevator

Abstract

The application provides a kind of electric vehicle into elevator monitoring early warning method, belong to monitoring safety technical field, solve the technical problems of lower accuracy of monitoring result and higher monitoring cost in the prior art in the monitoring of electric vehicle into elevator.A kind of electric vehicle into elevator monitoring early warning method is applied to SS chip, YOLOX_Tiny detection model and Darknet_Tiny classification model are configured in the SS chip, the method comprises: obtaining detection image;Receive preset detection area;Detection image is detected by YOLOX_Tiny detection model, and detection result is obtained;According to the detection result, judge whether there is no stretcher category, and there is electric vehicle category, and reach detection score threshold;If so, the target frame region in current detection image is cut out, and target frame image is obtained, and the target frame image is classified and identified by Darknet_Tiny classification model, and classification result is obtained;According to the classification result, judge whether it is electric vehicle category, and reach classification score threshold.

Description

Technical Field

[0001] This invention relates to the field of monitoring and security technology, and in particular to a method for monitoring and early warning of electric vehicles entering elevators. Background Technology

[0002] In recent years, electric vehicles have become a popular choice for short- and medium-distance transportation, leading to a year-on-year increase in electric vehicle fires and battery explosions. The disorderly parking of electric vehicles in residential areas poses a serious safety hazard, with potentially severe consequences in the event of a fire. Fire department statistics show that 80% of electric vehicle fires occur during charging, primarily due to battery malfunctions, followed by short circuits and power line failures. Furthermore, 90% of injuries and fatalities from electric vehicle fires are caused by vehicles being parked in lobbies, hallways, or passageways. Addressing the issue of illegally parked electric vehicles—such as using elevators to access residential buildings and then charging them in hallways or homes—poses significant safety risks, yet many communities have been unable to completely resolve this problem.

[0003] Currently, most video surveillance equipment used in elevators in residential communities across the country only has the single function of saving video recordings. It lacks functions such as identifying and issuing warnings for abnormal electric vehicle behavior, requiring 24-hour manual monitoring, which is a waste of manpower. Furthermore, manual monitoring is prone to fatigue and may result in missed detections. Alternatively, devices with detection algorithms based on SVM or YOLOv3 can be used for preliminary electric vehicle classification and monitoring in elevator scenes, but the accuracy of the monitoring results cannot be guaranteed. Moreover, current devices that can perform target detection in video scenes based on detection algorithms require local servers or cloud servers to support the high computing power required by the algorithms, which further increases the operating costs of community property management.

[0004] Therefore, existing technologies for monitoring electric vehicles entering elevators suffer from low accuracy and high monitoring costs. Summary of the Invention

[0005] The purpose of this invention is to provide a monitoring and early warning method for electric vehicles entering elevators. Existing technologies have technical problems such as low accuracy of monitoring results and high monitoring costs in monitoring electric vehicles entering elevators.

[0006] In a first aspect, the present invention provides a monitoring and early warning method for electric vehicles entering elevators, applied to an SS chip, wherein the SS chip is configured with a YOLOX_Tiny detection model and a Darknet_Tiny classification model, and the method includes:

[0007] Acquire the detection image;

[0008] Receive from the preset detection area;

[0009] The detection image is detected and identified using the YOLOX_Tiny detection model, and the detection results are obtained.

[0010] Based on the test results, determine whether there is no stretcher category, but there is an electric vehicle category, and the test score threshold is reached;

[0011] If so, the target bounding box region in the current detection image is cut out to obtain the target bounding box image. The target bounding box image is then classified and recognized using the Darknet_Tiny classification model to obtain the classification result.

[0012] Based on the classification results, determine whether it belongs to the electric vehicle category and whether it reaches the classification score threshold;

[0013] If so, determine whether the target bounding box of the electric vehicle category in the current detection image is located within the detection area;

[0014] If it is within the detection area, the cumulative count of the alarm category will begin based on the current detected image;

[0015] When the cumulative count of alarm categories reaches the preset number of times within a preset frame, an alarm command and an elevator braking command are sent.

[0016] Furthermore, the step of acquiring the detection image includes:

[0017] The system acquires detection images from the monitoring video stream at a rate of 8 frames per second.

[0018] Furthermore, after determining, based on the detection results, whether there is no stretcher category but there is an electric vehicle category, and whether the detection score threshold has been reached, the method further includes:

[0019] If not, return to the step of obtaining the detected image.

[0020] Furthermore, after the step of determining whether it belongs to the electric vehicle category based on the classification results and reaching the classification score threshold, the method further includes:

[0021] If not, return to the step of obtaining the detected image.

[0022] Furthermore, the preset frame is 24 frames, and the preset number of times is 16.

[0023] Furthermore, after the step of sending the alarm command and the electric braking command when the cumulative count of the alarm category reaches a preset number of times within a preset frame, the method further includes:

[0024] Determine whether the electric vehicle has left the detection area;

[0025] If so, then stop sending alarm commands and elevator braking commands;

[0026] If not, return to the step of determining whether the electric vehicle has left the detection area.

[0027] Furthermore, the step of sending an alarm command and an elevator braking command when the cumulative count of alarm categories reaches a preset number of times within a preset frame further includes:

[0028] Get the current image and save it.

[0029] Furthermore, the training method for the YOLOX_Tiny detection model is as follows:

[0030] A detection dataset containing both positive and negative samples is collected and data augmentation is performed. The categories of the detection dataset include electric vehicles, bicycles, people, other vehicles, and stretchers.

[0031] Model training is performed based on training images from the detection dataset;

[0032] The training image size is 576x320, the mini_batch is 32 per training session, the network depth parameter is set to self.depth = 0.33, and the network width parameter is set to self.width = 0.25.

[0033] Furthermore, the method for generating the Darknet_Tiny classification model is as follows:

[0034] The method for generating the Darknet_Tiny classification model includes an adjustment method and a training method;

[0035] The adjustment method for the Darknet_Tiny classification model is as follows:

[0036] Based on the original Darknet_Tiny classification model, when calculating the target value of logits of the input softmax layer, boundary processing is performed on the angle θ after adding the combined margin, and boundary protection is performed on the gradient during backpropagation (BP) of the error, resulting in a Darknet_Tiny classification model with a structure of 15 convolutional layers, 6 pooling layers, 1 InnerProduct fully connected layer, 1 softmax loss function layer, 1 Normalize layer, 1 Combined Margin layer and scale layer;

[0037] The training method for the Darknet_Tiny classification model is as follows:

[0038] Images in the detection dataset are cut out to obtain a classification dataset, which includes categories for electric vehicles and non-electric vehicles.

[0039] Model training is performed based on training images from a classification dataset.

[0040] The training image size is 128x128, and each training batch has a mini_batch size of 32.

[0041] In a second aspect, the present invention also provides a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to perform the method described in the first aspect.

[0042] This invention provides a monitoring and early warning method for electric vehicles entering elevators, applied to an SS chip. The SS chip is equipped with a YOLOX_Tiny detection model and a Darknet_Tiny classification model. The method includes: acquiring a detection image; receiving a preset detection area; performing detection and recognition on the detection image using the YOLOX_Tiny detection model to obtain a detection result; based on the detection result, determining whether there is no stretcher category and an electric vehicle category exists, and whether the detection score threshold is reached; if so, then performing image segmentation on the target box region in the current detection image to obtain a target box image, and performing classification and recognition on the target box image using the Darknet_Tiny classification model to obtain a classification result; based on the classification result, determining whether it is an electric vehicle category and whether the classification score threshold is reached; if so, determining whether the target box of the electric vehicle category in the current detection image is located within the detection area; if located within the detection area, then starting the cumulative count of alarm categories based on the current detection image; when the cumulative count of alarm categories reaches a preset number of times within a preset frame, sending an alarm command and an elevator braking command.

[0043] The electric vehicle (EV) entry monitoring and early warning method for elevators provided by this invention utilizes the YOLOX_Tiny detection model and Darknet_Tiny classification model configured within the SS chip to achieve a dual-recognition model of "detection + classification." This model combines the detection performance of YOLOX_Tiny with the advantages of the improved Darknet_Tiny classification model, performing dual judgments on the target result. For scenarios with target occlusion (e.g., wheelchairs, trolleys, electric vehicles, or crowds), the dual-recognition mode of "detection + classification" reduces the false positive rate compared to the single model of YOLOV3_Tiny, resulting in a higher probability of correct detection. In particular, a stretcher detection function is included for elevators that can accommodate stretchers. In special emergency situations, the stretcher category is given the highest priority, maximizing the detection accuracy of electric vehicles and effectively avoiding dangerous situations caused by false stretcher detection. Compared to YOLOV3_Tiny, YOLOX_Tiny significantly reduces deployment time and improves the overall performance of the algorithm. By deploying the detection and classification models of electric vehicles on the SS smart chip, and performing intelligent analysis of the monitoring images through the SS chip, computing power consumption is saved. There is no need to configure additional local servers or cloud servers, reducing the requirements for the number and configuration of servers. At the same time, it also reduces operating costs such as electricity and maintenance. In summary, it effectively solves the problems of low accuracy of monitoring results and high monitoring costs in the monitoring of electric vehicles entering elevators.

[0044] Correspondingly, the computer-readable storage medium provided by the present invention also has the above-mentioned technical effects. Attached Figure Description

[0045] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0046] Figure 1 This is a flowchart of the monitoring and early warning method for electric vehicles entering elevators provided in Embodiment 1 of the present invention;

[0047] Figure 2 The flowchart of the monitoring and early warning method for electric vehicles entering elevators in Embodiment 2 of the present invention is as follows. Figure 1 ;

[0048] Figure 3 The flowchart of the monitoring and early warning method for electric vehicles entering elevators in Embodiment 2 of the present invention is as follows. Figure 2 ;

[0049] Figure 4This is a comparison diagram of the classification network structure based on Darknet_Tiny before and after modification in Embodiment 2 of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or devices.

[0052] Currently, most video surveillance equipment used in elevators in residential communities across the country only has the single function of saving video recordings. It lacks functions such as identifying and issuing warnings for abnormal electric vehicle behavior, requiring 24-hour manual monitoring, which is a waste of manpower. Furthermore, manual monitoring is prone to fatigue and may result in missed detections. Alternatively, devices with detection algorithms based on SVM or YOLOv3 can be used for preliminary electric vehicle classification and monitoring in elevator scenes, but the accuracy of the monitoring results cannot be guaranteed. Moreover, current devices that can perform target detection in video scenes based on detection algorithms require local servers or cloud servers to support the high computing power required by the algorithms, which further increases the operating costs of community property management.

[0053] Therefore, existing technologies for monitoring electric vehicles entering elevators suffer from low accuracy and high monitoring costs.

[0054] To address the above problems, this invention provides a method for monitoring and early warning of electric vehicles entering elevators.

[0055] Example 1:

[0056] like Figure 1 As shown, this embodiment of the invention provides a monitoring and early warning method for electric vehicles entering elevators, applied to an SS chip. The SS chip is configured with a YOLOX_Tiny detection model and a Darknet_Tiny classification model. The method includes:

[0057] S1: Acquire the detection image;

[0058] S2: Receive the preset detection area;

[0059] S3: The detection image is detected and identified using the YOLOX_Tiny detection model to obtain the detection results;

[0060] S4: Based on the detection results, determine whether there is no stretcher category, but there is an electric vehicle category, and the detection score threshold is reached;

[0061] If yes, proceed to step S5; otherwise, return to step S1.

[0062] S5: Extract the target bounding box region in the current detection image to obtain the target bounding box image. Then, classify and recognize the target bounding box image using the Darknet_Tiny classification model to obtain the classification result.

[0063] S6: Based on the classification results, determine whether it belongs to the electric vehicle category and whether it reaches the classification score threshold;

[0064] If yes, proceed to step S7; otherwise, return to step S1.

[0065] S7: Determine whether the target bounding box for the electric vehicle category in the current detection image is located within the detection area;

[0066] If it is within the detection area, proceed to step S8;

[0067] S8: The cumulative count of alarm categories starts based on the current detected image;

[0068] S9: When the cumulative count of the alarm category reaches the preset number of times within the preset frame, send an alarm command and an elevator braking command;

[0069] S10: Obtain the current image and store it.

[0070] The monitoring and early warning method for electric vehicles entering elevators provided in this invention utilizes the YOLOX_Tiny detection model and Darknet_Tiny classification model configured within the SS chip to achieve a dual-recognition model of "detection + classification." This model combines the detection performance of YOLOX_Tiny with the advantages of the improved Darknet_Tiny classification model, performing dual judgments on the target result. For scenarios with target occlusion (e.g., wheelchairs, trolleys, electric vehicles, or crowds), the dual-recognition mode of "detection + classification" reduces the false positive rate compared to the single model of YOLOV3_Tiny, resulting in a higher probability of correct detection. In particular, a stretcher detection function is included for elevators that can accommodate stretchers. In special emergency situations, the stretcher category is prioritized, maximizing the detection accuracy of electric vehicles and effectively avoiding dangerous situations caused by false stretcher detection. Compared to YOLOV3_Tiny, YOLOX_Tiny significantly reduces deployment time and improves the overall performance of the algorithm. By deploying the detection and classification models of electric vehicles on the SS smart chip, and performing intelligent analysis of the monitoring images through the SS chip, computing power consumption is saved. There is no need to configure additional local servers or cloud servers, reducing the requirements for the number and configuration of servers. At the same time, it also reduces operating costs such as electricity and maintenance. In summary, it effectively solves the problems of low accuracy of monitoring results and high monitoring costs in the monitoring of electric vehicles entering elevators.

[0071] In one possible implementation, the step of acquiring the detection image includes:

[0072] The monitoring video stream is captured at a rate of 8 frames per second. This 8-frame-per-second image extraction rate ensures effective monitoring of the elevator while keeping the data volume within a reasonable range.

[0073] In one possible implementation, the preset frame is 24 frames and the preset number of alarms is 16. If the number of alarms reaches 16 within 24 frames, the alarm command is executed.

[0074] In one possible implementation, after step S9, the method further includes:

[0075] Determine whether the electric vehicle has left the detection area;

[0076] If so, then stop sending alarm commands and elevator braking commands;

[0077] If not, return to the step of determining whether the electric vehicle has left the detection area.

[0078] In one possible implementation, the training method for the YOLOX_Tiny detection model is as follows:

[0079] A detection dataset containing both positive and negative samples was collected and data augmentation was performed. The categories of the detection dataset included electric vehicles, bicycles, people, other vehicles, and stretchers.

[0080] Model training is performed based on training images from the detection dataset;

[0081] The training image size is 576x320, the mini_batch is 32 per training session, the network depth parameter is set to self.depth = 0.33, and the network width parameter is set to self.width = 0.25.

[0082] In one possible implementation, the Darknet_Tiny classification model is generated as follows:

[0083] The Darknet_Tiny classification model is generated using methods for tuning and training.

[0084] The adjustment method for the Darknet_Tiny classification model is as follows:

[0085] Based on the original Darknet_Tiny classification model, when calculating the target value of logits of the input softmax layer, boundary processing is performed on the angle θ after adding the combined margin, and boundary protection is performed on the gradient during backpropagation (BP) of the error, resulting in a Darknet_Tiny classification model with a structure of 15 convolutional layers, 6 pooling layers, 1 InnerProduct fully connected layer, 1 softmax loss function layer, 1 Normalize layer, 1 Combined Margin layer and scale layer;

[0086] The training method for the Darknet_Tiny classification model is as follows:

[0087] The images in the detection dataset are cut out to obtain a classification dataset, which includes categories for electric vehicles and non-electric vehicles.

[0088] Model training is performed based on training images from a classification dataset.

[0089] The training image size is 128x128, and each training batch has a mini_batch size of 32.

[0090] Example 2:

[0091] like Figure 2 and Figure 3 The embodiment of the present invention shown provides a method for monitoring and early warning of electric vehicles entering elevators.

[0092] Step 1: Collect the dataset for model training. Detection model dataset collection: A scenario of an electric vehicle entering an elevator is constructed. On-site video data is acquired from monitoring equipment (including positive samples when an electric vehicle is in the elevator and negative samples when no electric vehicle is in the elevator). The acquired dataset is then subjected to data augmentation processing (mirroring, Gaussian blur, video rotation, etc.). Manually labeled detection categories are divided into five classes: Rider, Bicycle, Person, otherbike, and Stretcher. The ratio of training to test set samples in the dataset is 10:1. Classification model dataset collection: The detection dataset is cut out and saved. Classification categories are divided into two types: Rider and noRider. The ratio of training to test set samples in the dataset is 10:1.

[0093] Step 2: Install and compile the YOLOX and Caffe training projects. The YOLOX training project directly uses the open-source training process on GitHub. Because the classification network needs to transform the softmax loss to the angle space cosθ for calculation during training, and the weights need to be normalized, the Caffe source code of the fully connected layer InnerProduct preceding the loss layer needs to be modified. The modified part involves improving the combined margin loss function on top of the original softmax layer, with the aim of correcting the classification task to the angle space.

[0094] Step 3: Set up the training environment for the YOLOX_Tiny-based detection network and the improved Darknet_Tiny-based classification network, and train the models. The YOLOX_Tiny detection network is trained using the open-source YOLOX project, with only minor adjustments to some parameters. The classification model training involves modifying the network architecture based on the Darknet_Tiny-based classification network, such as... Figure 4 As shown, the original loss calculation structure (1) of the network is changed to structure (2). The main difference is that when calculating the target value of logits of the input softmax layer, the angle θ after adding the combined margin is processed, and the gradient during backpropagation (BP) is protected to prevent the gradient from being too large or too small. Because the classification network needs to calculate the similarity of samples of unknown categories during training, it is necessary to reduce the difference between samples within the class and increase the difference between samples between classes while ensuring that the known categories are classified correctly. This is to improve the generalization performance of the model, and these are not the direct optimization targets of softmax loss.

[0095] The YOLOX_Tiny-based detection model is trained using random cropping with a size of 576x320. Each training batch has a mini_batch size of 32. The network depth parameter is set to self.depth = 0.33, and the network width parameter is set to self.width = 0.25. The Darknet_Tiny-based classification model is trained by resizing the input and output images to a size of 128x128. The network has 15 convolutional layers, 6 pooling layers, 1 InnerProduct fully connected layer, 1 softmax loss function layer, 1 Normalize layer, and 1 Combined Margin layer + scale layer structure. Each training batch has a mini_batch size of 32.

[0096] Step 4: Migrate the electric vehicle detection and classification models to the intelligent analysis module of the SS chip. Since the SS chip supports Caffe models but not Darknet weights format models, the trained models need to be converted to Caffe models first. Then, the converted YOLOX_Tiny detection model needs to be quantized into yolox_tiny_fixed.sim_sgsimg.img format, and the Darknet_Tiny classification model needs to be quantized into ebike_fixed.sim_sgsimg.img format. Finally, package the quantized models into a bin format upgrade package and upload it to the SS chip for normal model access.

[0097] Step 5: Acquire the image information to be detected. Directly acquire the monitoring video stream from the monitoring equipment of the elevator scene to be monitored, at a speed of 8 frames per second.

[0098] Step Six: Input the acquired image data into the SS intelligent module to determine if there is an electric vehicle. 1) After enabling the algorithm, draw the detection area within the monitoring field of view. 2) Input the acquired image data into the YOLOX_Tiny detection model. a. If the current image output target category contains the "stretcher" category, proceed to the next frame image detection. b. If the current image output target category does not contain the Stretcher category and does not contain the Rider category, or does not contain the Stretcher category but contains the Rider category, but the Rider category score does not reach the set detection score threshold, proceed to the next frame image detection. c. If the current image output target category does not contain the Stretcher category but contains the Rider category, and the set detection score threshold is reached, the target box region of the Rider category is extracted and sent to the Darknet_Tiny classification model for judgment. If the classification result outputs a "noRider" category or a "Rider" category but the category score does not reach the set classification score threshold, then the next frame image detection is performed. If the classification result outputs a "Rider" category and reaches the set classification score threshold, then it is further determined whether the target box of the current "Rider" category is within the detection area. If it is not within the detection area, then the next frame image detection is performed. If it is within the detection area, then the alarm category cumulative count begins. Considering that the space inside the elevator in the community is small and prone to obstruction, the present invention sets the alarm condition to be that the cumulative count of meeting the alarm condition reaches more than 16 times within 24 frames, then an alarm is triggered.

[0099] Step Seven: Post-processing of alarm results. When an alarm is detected during Step Six, the electric vehicle entering the elevator is captured on camera, and an audible alarm is activated. Simultaneously, the elevator control system is activated to brake the elevator. This invention sets the alarm duration to a fixed 2 seconds. If the electric vehicle leaves the designated elevator detection area within the fixed alarm time, the alarm is deactivated and the elevator resumes normal operation. If the electric vehicle remains within the designated elevator detection area within the fixed alarm time, the alarm continues and the elevator brakes until the electric vehicle leaves the elevator detection area, at which point the alarm is deactivated and the elevator resumes normal operation.

[0100] The calculation method for the target logit value in the CombinedMargin layer of the Darknet_Tiny classification network in step three is shown in the following formula:

[0101] (1)

[0102] (2)

[0103] The CombinedMargin layer obtains the target logit value and the original θ through equation (1). After the CombinedMargin layer, a scale layer needs to be added, which becomes equation (2) to solve the problem of non-convergence. The parameters are set m1=1.0, m2=0.2, and m3=0.3 for training.

[0104] This invention employs a dual-recognition model of "detection + classification" deployed on the SigmaStar smart chip. The SS chip performs intelligent analysis of the monitoring footage, ensuring accurate electric vehicle identification while eliminating the need for additional local or cloud servers, thus reducing the load on internet-connected cameras. Specifically, since my country's "Residential Design Code" clearly stipulates that "residential buildings with twelve or more floors should have no fewer than two elevators per building, one of which should be capable of accommodating a stretcher," and given the prevalence of buildings with more than twelve floors in my country, we have designed a stretcher detection function for buildings with elevators capable of accommodating stretchers. When a stretcher is present in the elevator, the electric vehicle entry behavior judgment and warning function is not activated, ensuring that personal safety is paramount in emergencies.

[0105] This invention also provides a computer-readable storage medium storing computer-executable instructions. When called and executed by a processor, the computer-executable instructions cause the processor to perform the method provided in the above embodiments.

[0106] The apparatus provided in this embodiment of the invention can be specific hardware on a device or software or firmware installed on the device. The implementation principle and technical effects of the apparatus provided in this embodiment of the invention are the same as those in the foregoing method embodiments. For the sake of brevity, any parts not mentioned in the apparatus embodiments can be referred to the corresponding content in the foregoing method embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, apparatuses, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0107] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0108] For example, the division of units is merely a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be indirect couplings or communication connections through some communication interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0109] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0110] In addition, the functional units in the embodiments provided by the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0111] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0112] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0113] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention. All should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring and early warning of electric vehicles entering elevators, characterized in that, Applied to SS chips, wherein the SS chip is configured The method includes the YOLOX_Tiny detection model and the Darknet_Tiny classification model, and comprises: Acquire the detection image; Receive from the preset detection area; The detection image is detected and identified using the YOLOX_Tiny detection model, and the detection results are obtained. Based on the test results, determine whether there is no stretcher category, but there is an electric vehicle category, and the test score threshold is reached; If so, the target bounding box region in the current detection image is cut out to obtain the target bounding box image. The target bounding box image is then classified and recognized using the Darknet_Tiny classification model to obtain the classification result. Based on the classification results, determine whether it belongs to the electric vehicle category and whether it reaches the classification score threshold; If so, determine whether the target bounding box of the electric vehicle category in the current detection image is located within the detection area; If it is within the detection area, the cumulative count of the alarm category will begin based on the current detected image; When the cumulative count of alarm categories reaches the preset number of times within a preset frame, an alarm command and an elevator braking command are sent. The training method for the YOLOX_Tiny detection model is as follows: A detection dataset containing both positive and negative samples is collected and data augmentation is performed. The categories of the detection dataset include electric vehicles, bicycles, people, other vehicles, and stretchers. Model training is performed based on training images from the detection dataset; The training image size is 576x320, the mini_batch is 32 per training session, the network depth parameter is set to self.depth=0.33, and the network width parameter is set to self.width=0.

25. The method for generating the Darknet_Tiny classification model is as follows: The method for generating the Darknet_Tiny classification model includes an adjustment method and a training method; The adjustment method for the Darknet_Tiny classification model is as follows: Based on the original Darknet_Tiny classification model, when calculating the target value of logits of the input softmax layer, boundary processing is performed on the angle θ after adding combined margin, and boundary protection is performed on the gradient during backpropagation (BP) of the error, resulting in a Darknet_Tiny classification model with a structure of 15 convolutional layers, 6 pooling layers, 1 InnerProduct fully connected layer, 1 softmax loss function layer, 1 Normalize layer, 1 CombinedMargin layer and scale layer; The training method for the Darknet_Tiny classification model is as follows: Images in the detection dataset are cut out to obtain a classification dataset, which includes categories for electric vehicles and non-electric vehicles. Model training is performed based on training images from a classification dataset. The training image size is 128x128, and each training batch has a mini_batch size of 32. The calculation method for the target logit value in the CombinedMargin layer of the Darknet_Tiny classification network is shown in the following formula: （1） （2） The CombinedMargin layer obtains the target logit value and the original θ through equation (1). After the CombinedMargin layer, a scale layer needs to be added, which becomes equation (2) to solve the problem of non-convergence. The parameters are set m1=1.0, m2=0.2, and m3=0.3 for training.

2. The monitoring and early warning method for electric vehicles entering elevators according to claim 1, characterized in that, The step of acquiring the detection image includes: The system acquires detection images from the monitoring video stream at a rate of 8 frames per second.

3. The monitoring and early warning method for electric vehicles entering elevators according to claim 1, characterized in that, After determining, based on the detection results, whether there is no stretcher category but there is an electric vehicle category, and whether the detection score threshold has been reached, the method further includes: If not, return to the step of obtaining the detected image.

4. The monitoring and early warning method for electric vehicles entering elevators according to claim 1, characterized in that, After the step of determining whether it belongs to the electric vehicle category based on the classification results and reaching the classification score threshold, the method further includes: If not, return to the step of obtaining the detected image.

5. The monitoring and early warning method for electric vehicles entering elevators according to claim 1, characterized in that, The preset frame is 24 frames, and the preset number of times is 16.

6. The monitoring and early warning method for electric vehicles entering elevators according to claim 1, characterized in that, After the step of sending an alarm command and an electric braking command when the cumulative count of the alarm category reaches a preset number of times within a preset frame, the method further includes: Determine whether the electric vehicle has left the detection area; If so, then stop sending alarm commands and elevator braking commands; If not, return to the step of determining whether the electric vehicle has left the detection area.

7. The monitoring and early warning method for electric vehicles entering elevators according to claim 1, characterized in that, The step of sending an alarm command and an elevator braking command when the cumulative count of the alarm category reaches a preset number of times within a preset frame further includes: Get the current image and save it.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 7.

Images