Hangar space target real-time position monitoring platform and method based on ai driving

Abstract

The application discloses a hangar space target real-time position monitoring platform and method based on AI driving, relates to the technical field of position monitoring, and is used for improving the monitoring delay problem caused by few hangar space targets in multiple monitoring pictures, and comprises the following steps: identifying and updating an image database for the targets entering and leaving the hangar, dividing a shared screen, marking all targets in the hangar and acquiring target images, merging and sorting the marked targets in regions, calculating the hangar space margin according to the target sorting order, setting the target monitoring stage number by using a logic regression algorithm in combination with the target size, constructing a multi-layer perception machine model according to the target monitoring stage number to calculate the monitoring area adjustment ratio, adjusting and updating the merged regions for the target monitoring images, determining the target monitoring number, adding an interactive function for interactive adjustment, detecting the volume of the marked target position and the speed increment of the marked target, and judging whether the target needs to be monitored and transferred.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of position monitoring, more particularly, the present application relates to an AI-driven hangar space target real-time position monitoring platform and method. BACKGROUND

[0002] Position monitoring technology is a technology for real-time tracking and positioning of objects or personnel. The application of position monitoring technology combined with AI in real-time monitoring of hangar space can effectively mobilize or adjust each target in the hangar, improving the efficiency of hangar work.

[0003] The prior art has the following disadvantages:

[0004] In the past, when monitoring the targets in the hangar space, monitoring devices were installed at multiple preset positions, and the monitoring pictures of each monitoring device were arranged in the same shared screen. If the monitoring devices are installed at few positions, there will be more monitoring blind spots. If the monitoring devices are installed at many positions, the display on the same shared screen will be complicated. When a small number of hangar space targets have a sudden condition, it is difficult to identify and handle them in the first time, and the monitoring safety risk is increased. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide an AI-driven hangar space target real-time position monitoring method. The method identifies and marks the targets entering and leaving the hangar, analyzes the behavior of each marked target, and displays the marked target in the corresponding state in the shared screen area to solve the problems raised in the background art.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0007] The AI-driven hangar space target real-time position monitoring method comprises the following steps:

[0008] Step S1: identifying the targets entering and leaving the hangar, updating the image database according to the identification result, dividing the shared screen into dynamic and static areas, setting the default area according to the number of parking positions in the hangar, marking all targets in the hangar and obtaining target images;

[0009] Step S2: merging the targets according to the target displacement change in the hangar, sorting the targets, calculating the hangar space margin according to the target sorting order, obtaining the size of each target, and setting the target monitoring level using a logistic regression algorithm by combining the hangar space margin and the target size;

[0010] Step S3: According to the target displacement change and the target monitoring number, a multi-layer perception model is constructed to calculate the monitoring area adjustment ratio, the monitoring area of each target in the dynamic area is adjusted according to the monitoring area adjustment ratio, and the merged area is updated, the target monitoring number of the dynamic area is determined to monitor and transfer;

[0011] Step S4: Using the interactive device to add interactive function to the dynamic area or static area, comparing the marked target position with the preset parking position to adjust the interaction, detecting the volume of the marked target position and the speed increment of the marked target, and judging whether the target needs to be monitored and transferred.

[0012] In a preferred embodiment, in step S1, the RFID system is used to identify the target entering or leaving the garage, when the target enters the garage, the target is identified and marked, and the identified target is transmitted into the image database, when the target leaves the garage, the target mark is deleted from the image database;

[0013] The shared screen is divided into two parts according to the preset segmentation ratio, one part is the dynamic area, and the other part is the static area;

[0014] The area of the dynamic area or the static area is proportional to the number of parking positions in the garage, which is the default area of the corresponding area, and the default area is used as the starting monitoring image area of each target in the garage. The target image is marked and obtained.

[0015] In a preferred embodiment, in step S2, a period of time is selected as the analysis time, the displacement of each marked target in the garage is monitored within the analysis time, the marked targets with non-zero displacement change are selected and merged into the dynamic area for monitoring; the marked targets with zero displacement change are merged into the static area for monitoring;

[0016] When the target enters or leaves the garage, the number of marked templates in the garage is counted, the difference between the number of parking positions in the garage and the number of marked targets is obtained, and the ratio of the remaining parking position number to the number of parking positions in the garage is used as the garage space margin when the corresponding target enters or leaves.

[0017] In a preferred embodiment, in step S2, the target size of each marked target is detected, the target size of each marked target in the garage is merged into a size data set, and the ratio of the target size of each marked target to the maximum value of the size data set is used as the size coefficient of the corresponding marked target;

[0018] The garage space margin and the size coefficient of the marked target are used to set the target monitoring number by using the logistic regression algorithm: wherein e is a natural base, z is a logistic regression parameter and z is the difference between the size factor of the marked target and the hangar space margin, and J is the target monitoring level of the corresponding marked target.

[0019] In a preferred embodiment, in step S3, the displacement change amount of each marked target in the dynamic area within the analysis time is combined into a displacement change data set, and the target monitoring level of each marked target in the dynamic area is combined into a monitoring level data set.

[0020] The displacement change data set and the monitoring level data set are integrated to construct a multi-layer perception module to calculate the monitoring area adjustment ratio of each marked target, and the specific steps are as follows:

[0021] Input data: the data in the displacement change data set and the monitoring level data set are input as input features into the input layer, and the data in the displacement change data set is standardized and then input into the hidden layer;

[0022] Initialization parameters: the initial weights and bias parameters from the input layer to the hidden layer are initialized and set by preset values;

[0023] Setting the hidden layer processing algorithm: the displacement change data set and the monitoring level data set are processed by setting the hidden layer activation function;

[0024] Setting the output layer processing algorithm: the activation function of the output layer is set to calculate the output result using the displacement change data set and the monitoring level data set processed by the hidden layer;

[0025] Downward propagation: the multi-layer perception sequentially passes the input features through the input layer, the hidden layer, and the output layer to calculate the output result as the monitoring area adjustment ratio of each marked target;

[0026] Calculate the monitoring area adjustment threshold;

[0027] Output result: according to the monitoring area adjustment threshold, the multi-layer perception model is converged, and the result of the output layer is output and used as the monitoring area adjustment ratio of each marked target.

[0028] In a preferred embodiment, in step S3, the calculation of the monitoring area adjustment threshold in the multi-layer perception is as follows:

[0029] Step A1: compare the output result of the multi-layer perception module with the preset actual label to calculate the loss function, and use the loss function to calculate the mean square error of all output results of the output layer: the formula for calculating the mean square error is: wherein MSE is the mean square error, n is the number of data in the displacement change data set or the monitoring level data set, The normalized result of the displacement variation of the i-th marker target randomly selected from the displacement variation data set and the output result of the monitoring stage of the corresponding marker target are obtained through the output layer, The i-th time label is preset.

[0030] Step A2: Calculate the influence gradient of each layer on the loss function, and take the partial derivative of the weight and bias parameters of the hidden layer on the loss function as the gradient. The gradient calculation formula of the weight and bias parameters is as follows: wherein The weight gradient and bias parameter gradient in the hidden layer activation function are respectively

[0031] Step A3: Adjust the weight and bias parameters through the weight gradient, bias parameter gradient and preset learning rate. The product of the learning rate and the corresponding weight gradient or bias parameter gradient is taken as the updated weight and bias parameter: wherein The updated weight and , The updated bias parameter is reset to execute the above steps until the model converges. The final output result of the hidden layer is taken as the monitoring area adjustment threshold and replaces the initial monitoring area adjustment threshold.

[0032] In a preferred embodiment, in step S3, the product of the default area of each marker target in the dynamic area and the monitoring area adjustment ratio of the corresponding marker target is taken as the monitoring area of the corresponding marker target. The monitoring area sizes of each marker target in the dynamic area are calculated and summed, and then compared with the maximum monitoring area of the dynamic area to determine whether the merging area needs to be updated. The rules are as follows:

[0033] Rule 1: When the sum of the monitoring area of each marker target in the dynamic area is lower than the maximum monitoring area of the dynamic area, the merging area is not updated.

[0034] Rule 2: When the sum of the monitoring area of each marker target in the dynamic area exceeds the maximum monitoring area of the dynamic area, the monitoring areas of each marker target in the dynamic area are arranged from large to small, and each marker target monitoring area is added from large to small until the addition of the marker target monitoring area causes the cumulative result to exceed the maximum monitoring area. The marker targets participating in the accumulation are retained and the number of retained marker targets is recorded. The remaining marker targets are all transferred to the static area.

[0035] ​​​In a preferred embodiment, in step S4, when the interactive device is used to monitor the external interaction, if the marked target is in the dynamic area, the interactive device simultaneously starts the incoming and outgoing functions; if the marked target is in the static area, the interactive device only starts the outgoing function;

[0036] When the marked target is in the static area, the position of the marked target is compared with the preset parking position, if the positions do not coincide, the interactive device simultaneously starts the incoming and outgoing functions for the marked target;

[0037] If the positions coincide, the volume of the position of the marked target and the speed increment of the target are detected, and the momentum index of the marked target is calculated by using the geometric mean method: Wherein D is the momentum index of the position of the marked target, Y is the volume of the position of the marked target, and V is the speed increment of the target;

[0038] When the momentum index of the marked target exceeds the preset momentum threshold value, the marked target is regarded as a candidate target, a plurality of candidate targets are screened out by calculating the momentum indexes of the marked targets in the static area, the monitoring areas of the candidate targets are obtained and arranged from small to large, and then the candidate targets are sequentially transferred to the dynamic area, and when the dynamic area continues to add candidate targets, the total sum of the monitoring areas of the marked targets in the dynamic area exceeds the maximum monitoring area, the transfer is stopped.

[0039] The hangar space target real-time position monitoring platform based on AI driving is used to realize the hangar space target real-time position monitoring method based on AI driving, and comprises the following devices and systems:

[0040] An RFID identification system is used to identify the target types entering and leaving the hangar and mark them differently;

[0041] An image database is used to record or delete the target types in the hangar and adjust the monitoring areas of the target types in the hangar;

[0042] A shared screen is used to monitor the states of the marked targets in the hangar in real time;

[0043] An interactive device is used to call in and out the position of the marked target, interact information, and assist in reminding or processing the state of the marked target.

[0044] The hangar space target real-time position monitoring platform and method based on AI driving have the following technical effects and advantages:

[0045] The application identifies the target entering and exiting the hangar, updates the image database, divides the shared screen, marks all targets in the hangar and obtains target images, merges and sorts the marked targets by area, sorts the targets to facilitate subsequent targeted adjustment of different marked targets, reduces adjustment cost, calculates the hangar space margin according to the target sorting order, obtains the target size, sets the target monitoring stage, adjusts and updates the merged area of the target monitoring image according to the target monitoring stage, merges the area according to the monitoring image area, closes the monitoring function, improves the monitoring inclusiveness, determines the target monitoring quantity, adds an interactive function using an interactive device, compares the marked target position with the preset parking position for interactive adjustment, detects the volume of the marked target position and the speed increment of the marked target, judges whether the target needs to be monitored and transferred, reduces the monitoring cost while improving the monitoring efficiency of the hangar target. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The application is based on an AI-driven hangar space target real-time position monitoring method.

[0047] Figure 2 The application is based on an AI-driven hangar space target real-time position monitoring platform flowchart. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0049] The application identifies the target entering and exiting the hangar, updates the image database, divides the shared screen, marks all targets in the hangar and obtains target images, merges and sorts the marked targets by area, calculates the hangar space margin according to the target sorting order, obtains the target size, sets the target monitoring stage, adjusts and updates the merged area of the target monitoring image according to the target monitoring stage, determines the target monitoring quantity, adds an interactive function using an interactive device, compares the marked target position with the preset parking position for interactive adjustment, detects the volume of the marked target position and the speed increment of the marked target, judges whether the target needs to be monitored and transferred, reduces the monitoring cost while improving the monitoring efficiency of the hangar target.

[0050] Embodiment 1, an AI-driven hangar space target real-time position monitoring method, as shown in Figure 1 includes the following steps:

[0051] Step S1: identify the target entering or leaving the garage, update the image database according to the identification result, divide the shared screen into dynamic area and static area, set the default area according to the number of parking positions in the garage, mark all targets in the garage and obtain target images;

[0052] Step S2: merge the targets according to the target displacement change in the garage, sort the targets, calculate the garage space margin according to the target sorting order, obtain the target size, and set the target monitoring level using the logistic regression algorithm by combining the garage space margin and the target size;

[0053] Step S3: construct a multi-layer perception machine model according to the target displacement change and the target monitoring level to calculate the monitoring area adjustment ratio, adjust the monitoring area of each target in the dynamic area according to the monitoring area adjustment ratio and update the merged area, and determine the target monitoring number of the dynamic area for monitoring transfer;

[0054] Step S4: add interactive function to the dynamic area or static area using interactive device, compare the marked target position with the preset parking position for interactive adjustment, detect the volume of the marked target position and the speed increment of the marked target, and judge whether the target needs monitoring transfer.

[0055] The specific implementation is as follows:

[0056] In step S1, the target entering or leaving the garage is identified by the RFID system, and when the target enters the garage, the target is identified and marked and then transmitted into the image database, and when the target leaves the garage, the target mark is deleted from the image database;

[0057] The shared screen is a mobile device screen for monitoring the target of the garage, which is used to display real-time monitoring content; the shared screen is divided into two parts according to the preset segmentation ratio, one part is the dynamic area, and the other part is the static area, and the default area is set according to the number of parking positions in the garage;

[0058] The area of the dynamic area or the static area and the ratio of the number of parking positions in the garage are used as the default area of the corresponding area, and the default area is used as the starting monitoring image area of each target in the garage, all targets in the garage are marked and target images are obtained.

[0059] It should be noted that the RFID system is a technology for automatically identifying and tracking objects, which is used to identify the target entering or leaving the garage, the dynamic area is the target monitoring area with large motion in the garage, the static area is the target monitoring area with static or small motion in the garage, and the segmentation ratio is not unique, for example, the segmentation ratio is set to 1 / 2, then the shared screen is divided into dynamic area and static area, etc., which is not described here.

[0060] In step S2, a period of time is selected as an analysis time, the displacement of each marker target in the hangar is monitored within the analysis time, the marker targets with a displacement change amount of 0 are screened out and merged into a static area for monitoring, and the marker targets with a displacement change amount of 0 are merged into a static area for monitoring;

[0061] According to the time sequence of the target entering the hangar, the number of targets in the hangar is increased by 1 each time the target enters the hangar, and the number of targets in the hangar is decreased by 1 each time the target leaves the hangar;

[0062] When the target enters or leaves the hangar, the number of marker templates in the garage is counted, the number of remaining parking positions is obtained by subtracting the number of marker targets from the number of parking positions in the hangar, and the ratio of the number of remaining parking positions to the number of parking positions in the hangar is taken as the space margin of the hangar when the corresponding target enters or leaves;

[0063] The target size of each marker target is detected by a laser range finder, the target size of each marker target in the hangar is merged into a size data set, and the ratio of the target size of each marker target to the maximum value of the size data set is taken as the size coefficient of the corresponding marker target;

[0064] The hangar space margin and the size coefficient of the marker target are used to set the target monitoring level by using a logistic regression algorithm: where e is the natural base, z is the logistic regression parameter and z is the difference between the size coefficient and the hangar space margin of the marker target, and J is the target monitoring level of the corresponding marker target.

[0065] It should be noted that the fewer the number of remaining parking positions, the longer the time spent by the target in finding a parking position, and the higher the safety risk. The larger the target size, the more likely it is to cause bumps when parking, and the higher the safety risk.

[0066] In step S3, the displacement change amount of each marker target in the dynamic area is counted within the analysis time and merged into a displacement change data set, and the target monitoring level of each marker target in the dynamic area is merged into a monitoring level data set;

[0067] The displacement change data set and the monitoring level data set are integrated to construct a multi-layer perception machine module to calculate the monitoring area adjustment ratio of each marker target. The multi-layer perception machine model is composed of an input layer, a hidden layer and an output layer. The input layer is used to receive input features and perform preprocessing, the hidden layer is used to calculate and process the preprocessed data, and the output layer is used to output the calculation result. The specific steps are as follows:

[0068] Input data: the data in the displacement change dataset and the monitoring stage dataset are input into the input layer as input features, the data in the displacement change dataset are input into the hidden layer after Max-Min standardization, and the Max-Min standardization formula is: , is the data in the displacement change dataset, is the maximum value in the displacement change dataset, is the minimum value in the displacement change dataset, is the standardized result of the data in the displacement change dataset;

[0069] Initialization parameters: the initial weights and bias parameters from the input layer to the hidden layer are initialized by preset values;

[0070] Setting the hidden layer processing algorithm: the displacement change dataset and the monitoring stage dataset are processed by setting the hidden layer activation function: , wherein A is the output of the hidden layer, a is the normalized result of the displacement change amount of the marked target in the displacement change dataset, b is the monitoring stage of the corresponding marked target in the monitoring stage dataset, and are two initial weights set in the input layer to the hidden layer, and c is the bias parameter set in the input layer to the hidden layer;

[0071] Setting the output layer processing algorithm: the activation function of the output layer is set as: , wherein B is the output result of the output layer, is the initial monitoring area adjustment threshold, and is the output result of the hidden layer calculated by the normalized result of the displacement change amount of the marked target and the average monitoring stage of the corresponding marked target;

[0072] Downward propagation: the input features are sequentially passed through the input layer, the hidden layer and the output layer by the multi-layer perception, and the output result is calculated as the monitoring area adjustment ratio of each marked target;

[0073] Calculating the monitoring area adjustment threshold:

[0074] Step A1: comparing the output result of the multi-layer perception module with the preset actual label to calculate the loss function, and calculating the mean square error of all output results of the output layer by using the loss function: the formula for calculating the mean square error is: , wherein MSE is the mean square error, n is the number of data in the displacement change dataset or the monitoring stage dataset, is the normalized result of the displacement change amount of the i-th marked target randomly selected in the displacement change dataset and the output result of the corresponding marked target through the output layer, preset i-th actual label;

[0075] Step A2: Calculate the gradient of the influence of each layer on the loss function, and the partial derivative of the weight and bias parameters of the hidden layer on the loss function as the gradient, then the gradient calculation formula of the weight and bias parameters is respectively: , wherein , The weight gradient and bias parameter gradient in the hidden layer activation function are respectively:

[0076] Step A3: Set the learning rate, adjust the weight and bias parameters by the weight gradient, bias parameter gradient and learning rate; The product of the learning rate and the corresponding weight gradient or bias parameter gradient is taken as the updated weight and bias parameter: , wherein, The updated weight and , The updated bias parameter; reset to execute the above steps until the model converges, and the final output result of the hidden layer is taken as the monitoring area adjustment threshold and replaces the initial monitoring area adjustment threshold;

[0077] Output result: After the multilayer perceptron model converges, the output of the output layer is output as the monitoring area adjustment proportion of each label target.

[0078] It should be noted that the initialization parameters, actual labels and learning rate in the multilayer perceptron are given by professionals in the field according to the actual situation, which is not analyzed here.

[0079] The product of the default area of each label target in the dynamic area and the monitoring area adjustment proportion of the corresponding label target is taken as the monitoring area of the corresponding label target, the monitoring area of each label target in the dynamic area is calculated and summed, and then compared with the maximum monitoring area of the dynamic area to determine whether the merged area needs to be updated, the rules are as follows:

[0080] Rule 1: When the sum of the monitoring area of each label target in the dynamic area is lower than the maximum monitoring area of the dynamic area, the merged area is not updated;

[0081] Rule 2: When the sum of the monitoring area of each label target in the dynamic area exceeds the maximum monitoring area of the dynamic area, arrange the monitoring area of each label target in the dynamic area from large to small, and add each label target monitoring area from large to small, until the addition of the label target monitoring area leads to the accumulation result exceeding the maximum monitoring area, then the label target participating in the accumulation is reserved and the number of reserved label targets is recorded, and the rest of the label targets are all transferred to the static area.

[0082] It should be noted that the greater the displacement change of the marked target or the target monitoring level, the more attention the marked target needs to be paid to the marked target, and the monitoring area of the marked target increases.

[0083] In step S4, when monitoring external interaction using the interactive device, if the marked target is in the dynamic area, the interactive device simultaneously starts the incoming and outgoing functions; if the marked target is in the static area, the interactive device only starts the outgoing function;

[0084] The incoming function of the interactive device is a device that can receive the sound at the location of the marked target, which is installed in the hangar in advance for receiving the sound on site; the incoming function of the interactive device is a device that can transmit the sound in the monitoring room to the location of the marked target for information transmission;

[0085] When the marked target is in the static area, the position of the marked target is compared with the preset parking position, if the positions do not coincide, the interactive device simultaneously starts the incoming and outgoing functions for the marked target; if the positions coincide, the volume of the marked target position and the speed increment of the target are detected, and the momentum index of the marked target is calculated by using the geometric mean method: , wherein D is the momentum index of the marked target position, Y is the volume of the marked target position, and V is the speed increment of the marked target.

[0086] When the momentum index of the marked target exceeds the preset momentum threshold, the marked target is selected as a candidate target, a plurality of candidate targets are selected by calculating the momentum index of each marked target in the static area, the monitoring area of the candidate target is obtained and arranged from small to large, and then transferred to the dynamic area in sequence, when the dynamic area continues to add candidate targets, resulting in that the total monitoring area of the marked targets in the dynamic area exceeds the maximum monitoring area, the transfer is stopped.

[0087] It should be noted that the volume of the marked target position and the speed increment of the marked target can be obtained by a sound sensor and a speed sensor respectively, the greater the volume of the marked target position or the greater the speed increment of the marked target, the more the marked target needs to be monitored in the state of changing the static state or in the state of increasing the amplitude.

[0088] Embodiment 2, a hangar space target real-time position monitoring platform based on AI driving, as shown in Figure 2 , comprising the following devices or systems:

[0089] RFID identification system: for identifying the target model entering and leaving the hangar and marking differently;

[0090] Image database: for recording or deleting the target model in the hangar and adjusting the monitoring area of the target model in the hangar;

[0091] Shared screen: for real-time monitoring of the state of each marked target in the hangar;

[0092] Interaction device: for incoming and outgoing call to mark target position, information interaction, auxiliary reminder or processing to the state of the marked target.

[0093] The above examples can be realized by software, hardware, firmware or any combination thereof, in whole or in part. When realized by software, the above examples can be realized in the form of a computer program product, in whole or in part.

[0094] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and the constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0095] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0096] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0097] Finally, the above is only the preferred embodiments of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. An AI-driven hangar space target real-time position monitoring method, characterized in that, Comprising the following steps, Step S1: Identify the target entering and exiting the hangar, update the image database according to the identification result, divide the shared screen into dynamic area and static area, set the default area of dynamic area and static area according to the number of parking positions in the hangar, mark all targets in the hangar and obtain target images; Step S2: According to the target displacement change in the hangar, the target is merged, the target is sorted, the space margin of the hangar is calculated according to the target sorting order, the size of each target is obtained, and the space margin of the hangar and the size of the target are used to set the target monitoring level by using the logical regression algorithm; Step S3: According to the target displacement change and the target monitoring level, a multi-layer perception machine model is constructed to calculate the monitoring area adjustment ratio, the target monitoring area in the dynamic area is adjusted according to the monitoring area adjustment ratio, and the merged area is updated, the target monitoring number of the dynamic area is determined, and the monitoring is transferred; Step S4: Use the interactive device to add interactive function to the dynamic area or static area, compare the marked target position with the preset parking position, adjust the interaction, detect the volume of the marked target position and the speed increment of the marked target, and judge whether the target needs to be transferred; In step S2, a period of time is selected as the analysis time, the displacement of each marked target in the hangar is monitored in the analysis time, the marked targets with displacement change of 0 are selected and merged into dynamic area for monitoring; The marked target with displacement change of 0 is merged into static area for monitoring; According to the time sequence of the target entering the hangar, the target number in the hangar is increased by 1 every time the target enters the hangar, and vice versa, the target number in the hangar is reduced by 1 every time the target leaves the hangar; When the target enters or leaves the hangar, the number of marked targets in the hangar is counted, the difference between the number of parking positions in the hangar and the number of marked targets is obtained to obtain the number of remaining parking positions, and the ratio of the number of remaining parking positions to the number of parking positions in the hangar is taken as the space margin of the hangar when the corresponding target enters or leaves; In step S3, the product of the default area of each marked target in the dynamic area and the monitoring area adjustment ratio of the corresponding marked target is taken as the monitoring area of the corresponding marked target, the monitoring area of each marked target in the dynamic area is calculated and summed, and then compared with the maximum monitoring area of the dynamic area to judge whether the merged area needs to be updated, the rules are as follows: Rule 1: When the sum of the monitoring area of each marked target in the dynamic area is less than the maximum monitoring area of the dynamic area, the merged area is not updated; Rule 2: When the sum of the monitoring area of each marked target in the dynamic area exceeds the maximum monitoring area of the dynamic area, arrange the monitoring area of each marked target in the dynamic area from large to small, add each marked target monitoring area from large to small, and stop adding when the cumulative result exceeds the maximum monitoring area, keep the marked targets participating in the addition and record the number of reserved marked targets, and the remaining marked targets are all transferred to the static area.

2. The AI-driven hangar space target real-time position monitoring method according to claim 1, characterized in that: In step S1, the target entering or leaving the garage is identified by the RFID system, and when the target enters the garage, the target is identified and marked, and the identified target is transmitted into the image database; when the target leaves the garage, the target mark is deleted from the image database; The shared screen is divided into two parts according to the preset segmentation ratio, one part is a dynamic area, and the other part is a static area; The area of the dynamic area or the static area is proportional to the number of parking positions in the garage, and the default area of the corresponding area is used as the starting monitoring image area of each target in the garage, and the target image is obtained by marking all targets in the garage; The shared screen is a mobile device screen for monitoring the target in the garage, which is used to display real-time monitoring content; the RFID system is an automatic identification and tracking technology for identifying targets entering or leaving the garage, and the image database is a database for storing and processing images, which is used to store or delete the marked target.

3. The AI-driven garage space target real-time position monitoring method according to claim 1, characterized in that: In step S2, the target size of each marked target is detected, the target sizes of all marked targets in the garage are combined into a size data set, and the ratio of the target size of each marked target to the maximum value of the size data set is used as the size coefficient of the corresponding marked target. The warehouse space margin of the marked target and the size coefficient are utilized to set the target monitoring stage by using a logistic regression algorithm: wherein e is a natural base, z is a logistic regression parameter and z is the difference between the size coefficient of the marked target and the warehouse space margin, and J is the target monitoring stage corresponding to the marked target.

4. The AI-driven garage space target real-time position monitoring method according to claim 3, characterized in that: In step S3, the displacement change amount of each marked target in the dynamic area is calculated within the analysis time, and the displacement change data set is combined; the target monitoring level of each marked target in the dynamic area is combined into a monitoring level data set; The displacement change data set and the monitoring level data set are combined to construct a multi-layer perception module to calculate the monitoring area adjustment ratio of each marked target, and the specific steps are as follows: Input data: the data in the displacement change data set and the monitoring level data set are used as input features to input into the input layer, and the data in the displacement change data set is standardized and transmitted to the hidden layer after the input layer; Initialize parameters: the initial weights and bias parameters from the input layer to the hidden layer are initialized and set by preset values; Set the hidden layer processing algorithm: process the displacement change data set and the monitoring level data set by setting the hidden layer activation function; Set the output layer processing algorithm: set the activation function of the output layer to calculate the output result using the displacement change data set and the monitoring level data set processed by the hidden layer; Downward propagation: the input features are sequentially transmitted through the input layer, the hidden layer and the output layer in the multi-layer perception, and the output result is calculated as the monitoring area adjustment ratio of each marked target; Calculate the monitoring area adjustment threshold; Output result: according to the monitoring area adjustment threshold, the multi-layer perception model is converged, and the result of the output layer is output and used as the monitoring area adjustment ratio of each marked target.

5. The AI-driven garage space target real-time position monitoring method according to claim 4, characterized in that: In step S3, the calculation of the monitoring area adjustment threshold in the multi-layer perception is as follows: Step A1: calculate the loss function by comparing the output result of the multi-layer perception module with the preset actual label, and calculate the mean square error of the output result of all output layers by using the loss function: the formula for calculating the mean square error is: wherein MSE is the mean square error, n is the number of data in the displacement change data set or the monitoring stage data set, is the normalized result of the i-th labeled target displacement change amount randomly selected from the displacement change data set and the output result obtained by the output layer for the corresponding labeled target monitoring stage, is the i-th preset actual label; Step A2: Calculate the gradient of the effect of each layer on the loss function, , where MSE is the mean square error, is the partial derivative of the loss function MSE with respect to the hidden layer weight w, is the partial derivative of the loss function MSE with respect to the hidden layer bias parameter c; Step A3: adjusting the weight and bias parameters by the weight gradient, bias parameter gradient and preset learning rate; the product of the learning rate and the corresponding weight gradient or bias parameter gradient is taken as the updated weight and bias parameter: , , is the updated hidden layer weight; is the updated bias parameter; reset the execution of the above steps until the model converges, and the final output result of the hidden layer is taken as the monitoring area adjustment threshold and replaces the initial monitoring area adjustment threshold. 6.The AI-driven hangar space target real-time position monitoring method according to any one of claims 1 or 5, characterized in that: In step S4, when the interactive device is used to monitor external interaction, if the marked target is in the dynamic area, the interactive device simultaneously starts the incoming and outgoing call functions; if the marked target is in the static area, the interactive device only starts the outgoing call function; When the marked target is in the static area, the position of the marked target is compared with the preset parking position, and if the positions do not coincide, the interactive device simultaneously starts the incoming and outgoing call functions for the marked target; If the positions coincide, the volume of the marked target position and the velocity increment of the target are detected, and the momentum index of the marked target is calculated by using the geometric mean method: where D is the momentum index of the marked target position, Y is the volume of the marked target position, and V is the velocity increment of the target. When the momentum index of the marked target exceeds the preset momentum threshold, the marked target is selected as a candidate target, a plurality of candidate targets are selected by calculating the momentum indexes of the marked targets in the static area, the monitoring areas of the candidate targets are obtained and arranged from small to large, and then the candidate targets are sequentially transferred to the dynamic area; when the dynamic area continues to add candidate targets, causing the total monitoring area of the marked targets in the dynamic area to exceed the maximum monitoring area, the transfer is stopped.

Images