Pet positioning and searching method and system

Abstract

The invention discloses a pet positioning and searching method and system, and the method comprises the steps: obtaining real-time environment data based on pet positioning data obtained based on an initial loss time sequence, and outputting a first pet motion prediction track in all the loss time sequences through a pet motion prediction model and the real-time environment data; outputting a prediction error trajectory based on the pet motion actual trajectory and the first pet motion prediction trajectory; updating a pet motion prediction model according to the prediction error trajectory to output a second pet motion prediction trajectory; outputting a user search trajectory based on the second pet motion prediction trajectory; and when the real-time pet positioning data does not conform to the predicted positioning data, updating the second pet motion prediction track according to a difference value between the real-time pet positioning data and the predicted positioning data. The method has the advantages that the user search trajectory is planned according to the predicted pet position, unnecessary search of the user is reduced, and meanwhile prediction accuracy is improved through trajectory comparison.

Description

Technical Field

[0001] The present invention relates to the field of pet positioning technology, and in particular to a pet positioning search method and system. Background Art

[0002] As the number of pets increases, the problem of pets getting lost is becoming more and more serious. Although satellite positioning technologies such as GPS and Beidou can provide high positioning accuracy (usually within 5-10 meters) in outdoor environments, the positioning accuracy will drop significantly or even be impossible in indoor environments, urban canyons, high-rise building dense areas or areas with severe signal obstruction. And usually only the real-time positioning of pets is used to guide users, but pets will not stand still, and based on the movement of most pets, users cannot keep up with the speed of pets most of the time. This will result in that when users search for pets based on the current location, the pets will deviate from the user's search direction at the next moment, causing users to need to reach a new positioning point again, affecting the user's search experience.

[0003] The Chinese patent "Pet tracking method, device, equipment and medium combining UWB positioning and vision", publication number: CN118283533A, publication date: July 2, 2024, specifically discloses: obtaining multiple frames of real-time images in the pet care scene and UWB positioning signals emitted by the transmitter worn by the pet; processing the UWB positioning signals to determine the pet's location information; determining the car's motion parameters based on the pet-finding car's location information and the pet's location information; inputting each frame of the real-time image into the pre-trained pet target detection model and outputting the pet target detection result; controlling the pet-finding car to track the pet in real time based on the pet-finding car's location information, the pet's location information, the car's motion parameters and the pet target detection result. Although this solution obtains the pet target detection result through the pet target detection model, it is actually tracking the pet target, and there is still a problem of high difficulty in searching. In addition, although the use of the pet-finding car to track the pet in real time solves the problem that the user cannot keep up with the pet's movement, it has poor applicability.

[0004] The Chinese patent "A pet positioning method and device", publication number: CN 117560627 A, publication date: February 13, 2024, specifically discloses that a wireless connection is established between a pet terminal and a mobile terminal held by a pet owner, the target pet wears the pet terminal, the pet terminal has a built-in GPS positioning module, and the mobile terminal is installed with a pet positioning application; when the target pet is determined to be lost based on the continuous interruption of the wireless connection, the real-time positioning data of the target pet's location is obtained through the GPS positioning module, and the real-time positioning data is position-screened to obtain the target positioning point of the target pet. Similarly, this solution also has the technical problem that it is not easy for users to directly find pets. Summary of the invention

[0005] In view of the problem in the prior art that lost pet searches based on real-time positioning data result in poor user search experience, the present invention provides a pet positioning search method and system, which plans user search trajectories based on predicted pet locations, reduces user unnecessary searches, and improves prediction accuracy by trajectory comparison.

[0006] To achieve the above technical objectives, the present invention provides a technical solution, which is a pet positioning search method, comprising the following steps: S1: obtaining pet historical motion data, and constructing a pet motion prediction model based on the pet historical motion data; S2: obtaining the environmental range corresponding to the pet type based on the pet positioning data obtained at the initial lost time series, obtaining real-time environmental data according to the environmental range, and outputting the first pet motion prediction trajectory in all lost time series using the pet motion prediction model and the real-time environmental data; S3: constructing the actual pet motion trajectory with the pet positioning data obtained at all lost time series, and outputting the prediction error trajectory based on the actual pet motion trajectory and the first pet motion prediction trajectory; S4: updating the pet motion prediction model with the prediction error trajectory, and outputting the second pet motion prediction trajectory in the future time series with the updated pet motion prediction model; S5: obtaining the user historical motion data and the user's current position, and outputting the user search trajectory based on the second pet motion prediction trajectory; S6: comparing the real-time pet positioning data with the predicted positioning data of the corresponding time series of the second pet motion prediction trajectory, and when the real-time pet positioning data does not match the predicted positioning data, updating the second pet motion prediction trajectory with the difference value between the real-time pet positioning data and the predicted positioning data, and re-executing S5.

[0007] The present invention is further configured such that S1 includes: obtaining historical movement data of lost pets, performing data preprocessing on the historical movement data of lost pets; constructing a basic deep learning model based on a convolutional neural network; training the basic deep learning model using the historical movement data of lost pets, and outputting a pet movement prediction model.

[0008] The present invention is further configured such that the pet location data acquired based on the initial loss time sequence acquires the environment range corresponding to the pet type, including: The maximum movement speed of the pet is obtained based on the pet type; the movement timing is obtained based on the preset search timing and loss timing; and the environment range corresponding to the pet type is obtained based on the movement timing and the maximum movement speed of the pet.

[0009] The present invention is further configured such that obtaining real-time environmental data according to the environmental range also includes: obtaining environmental characteristics based on the pet type; and searching the environment corresponding to the environmental characteristics in the basic environmental area to obtain corresponding real-time environmental data.

[0010] The present invention is further configured such that the pet's actual movement trajectory is constructed using the pet positioning data obtained from all lost time sequences, including: using the pet positioning data obtained from all lost time sequences as actual movement nodes; drawing a simulated movement trajectory between actual movement nodes based on the positioning sampling interval and the pet type; and outputting the pet's actual movement trajectory using the actual movement nodes and the simulated movement trajectory.

[0011] The present invention is further configured such that the output of the predicted error trajectory based on the actual pet movement trajectory and the first pet movement predicted trajectory includes: matching the actual movement nodes in the actual pet movement trajectory and the predicted movement nodes in the first pet movement predicted trajectory; outputting the predicted error trajectory with the simulated movement trajectory corresponding to the unmatched actual movement nodes and the predicted movement trajectory corresponding to the predicted movement nodes.

[0012] The present invention is further configured such that the output of the prediction error trajectory using the simulated motion trajectory corresponding to the unmatched actual motion node and the predicted motion trajectory corresponding to the predicted motion node includes: using the predicted motion trajectory corresponding to the first pet motion prediction trajectory as the actual motion trajectory that matches the actual motion node; constructing a motion loss curve using the actual motion trajectory, obtaining the pet motion loss upper limit based on the pet type, and obtaining the simulated motion trajectory corresponding to the unmatched actual motion node using the motion loss curve and the pet motion loss upper limit.

[0013] The present invention is further configured such that the pet's actual movement trajectory is constructed using the pet positioning data obtained from all lost time sequences, including: using the pet positioning data obtained from all lost time sequences as actual movement nodes; matching the actual movement nodes and the predicted movement nodes in the first pet movement prediction trajectory; and constructing a simulated movement trajectory using the actual movement nodes that do not match the predicted movement nodes and the positioning sampling intervals.

[0014] The present invention is further configured such that S1 also includes: constructing a positioning correction model based on the pet's historical positioning data and the pet's historical actual position data.

[0015] As another technical solution provided by the present invention, a pet positioning search system is connected to the pet positioning system and is used to implement the method as mentioned above, including: a data acquisition module, which is used to pre-store the pet's historical motion data and the real-time pet positioning data transmitted by the pet positioning system offline; a data modeling module, which is used to construct a pet motion prediction model based on the pet's historical motion data and output the pet's predicted motion trajectory; a trajectory construction module, which is used to construct the pet's actual motion trajectory based on the lost time series; an error analysis module, which is used to output a predicted error trajectory or a difference value between the real-time pet positioning data and the predicted positioning data based on the pet's predicted motion trajectory, the real-time pet positioning data and the pet's actual motion trajectory; a trajectory planning module, which is used to output the user search trajectory according to the pet's predicted motion trajectory.

[0016] The beneficial effects of the present invention are as follows: the pet movement prediction model is updated through the pet's behavioral habits during the period of loss, so that the pet movement prediction model is closer to the pet's behavior when lost, thereby improving the accuracy of the predicted movement trajectory. At the same time, after obtaining the real-time positioning data, the second pet movement trajectory is updated only according to the difference value between the real-time positioning data and the predicted positioning data, avoiding model deviation caused by the pet's occasional choice of updating the model, thereby further improving the prediction accuracy.

[0017] The pet movement prediction model is based on the private data of the user's own pet and the public data of all users' pets in non-private areas. It uses the habits of most pets and their performance in different areas to supplement the performance of corresponding pet types in different areas, thereby improving the accuracy of model construction while protecting user privacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the process of the pet location search method of the present invention. DETAILED DESCRIPTION

[0019] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, as the first embodiment of the present invention, the pet location search method includes the following steps: S1: Obtain pet historical movement data, and build a pet movement prediction model based on the pet historical movement data; S2: obtaining the environment range corresponding to the pet type based on the pet positioning data obtained from the initial lost time sequence, obtaining real-time environment data according to the environment range, and outputting the first pet movement prediction trajectory within all lost time sequences using the pet movement prediction model and the real-time environment data; S3: constructing the actual pet movement trajectory using the pet positioning data acquired from all lost time series, and outputting a prediction error trajectory based on the actual pet movement trajectory and the first pet movement prediction trajectory; S4: updating the pet motion prediction model with the prediction error trajectory, and outputting a second pet motion prediction trajectory in the future time series with the updated pet motion prediction model; S5: Obtain the user's historical movement data and the user's current location, and output the user's search trajectory based on the second pet's predicted movement trajectory; S6: Compare the real-time pet positioning data with the predicted positioning data of the corresponding time series of the second pet movement prediction trajectory. When the real-time pet positioning data does not match the predicted positioning data, update the second pet movement prediction trajectory with the difference value between the real-time pet positioning data and the predicted positioning data, and re-execute S5.

[0021] In this embodiment, the pet's historical operation data is obtained to construct a pet movement prediction model. The maximum range of the pet's movement within the preset time sequence, i.e., the environmental range, can be obtained based on the pet positioning data obtained at the initial loss time sequence and the pet type. The first pet movement prediction trajectory based on the pet's historical behavior habits is obtained using the real-time environmental data of the environmental range and the pet movement prediction model. All positioning information uploaded by the locator since the pet was lost is obtained to construct the pet's actual movement trajectory. A prediction error trajectory is output based on the actual pet movement trajectory and the first pet movement prediction trajectory. The environmental data error is output based on the actual difference between the environment and the predicted environment corresponding to the prediction error trajectory. The pet movement prediction model is updated and a second pet movement prediction trajectory is output. The real-time uploaded positioning data is received and compared with the predicted positioning data of the corresponding time sequence of the second pet movement prediction trajectory. The second pet movement prediction trajectory is updated with the difference value, and the user search trajectory is output through the second pet movement prediction trajectory. The pet movement prediction model is updated through the pet's behavioral habits during the period of loss, so that the pet movement prediction model is closer to the pet's behavior when lost, thereby improving the accuracy of the predicted movement trajectory. At the same time, after obtaining the real-time positioning data, the second pet movement trajectory is updated only based on the difference between the real-time positioning data and the predicted positioning data, avoiding model deviation caused by the pet's occasional choice of updating the model, thereby further improving the prediction accuracy.

[0022] The pet's historical movement data at least includes lost pet's historical movement data. Obtaining the pet's historical movement data and building a pet's movement prediction model based on the pet's historical movement data include: Acquire historical movement data of the lost pet, and perform data preprocessing on the historical movement data of the lost pet; Build a basic deep learning model based on convolutional neural network; The basic deep learning model is trained using the historical movement data of lost pets to output a pet movement prediction model.

[0023] The lost pet historical movement data at least includes the lost pet historical movement speed, the lost pet historical position and the lost pet historical movement environment. The basic deep learning model is trained based on the lost pet historical movement speed, the lost pet historical position and the lost pet historical movement environment, and the lost pet movement speed environment weight is obtained according to the correlation between the lost pet historical movement speed and the lost pet historical movement environment, and the lost pet movement location environment weight is obtained according to the correlation between the lost pet historical position and the lost pet historical movement environment, and the lost pet movement speed environment weight and the lost pet movement location environment weight are used to construct a pet movement prediction model based on the basic deep learning model. It can be understood that in the remaining embodiments, the basic deep learning model can also be constructed based on a recurrent neural network (RNN) or LSTM.

[0024] The method of using the lost pet's historical movement data to train the basic deep learning model and outputting the pet movement prediction model also includes: The supervised learning method is used to optimize the weight parameters of the basic deep learning model based on the back propagation algorithm according to the historical movement data of the lost pet, and the pet movement prediction model is output.

[0025] The basic deep learning model includes at least an input layer, a hidden layer and an output layer. The input layer receives the historical movement data of the lost pet, the hidden layer extracts environmental features and pet movement features, and the output layer outputs predicted positioning data.

[0026] The environment range corresponding to the pet type obtained based on the pet positioning data obtained from the initial loss sequence includes: Get the pet's maximum movement speed based on the pet type; Get the duration of the movement with the preset search sequence and loss sequence; The environmental range of the corresponding pet type is obtained based on the exercise duration and the pet's maximum movement speed.

[0027] The preset search sequence can be set according to the sequence corresponding to the pet's historical movement cycle, or it can be set by the user. Usually, the pet movement cycle is different based on different pet types, that is, the pet must have active time and rest time, and the position will not change during the rest time. Based on the pet movement cycle, the duration of the pet activity-rest cycle is obtained, and this duration is used as the preset search duration, so as to reduce the unnecessary search range while ensuring sufficient search time. Based on the movement duration and the pet's maximum movement speed, the movement range corresponding to the pet's continuous movement at the maximum movement speed within the movement duration, that is, the environmental range, is simulated. Real-time environmental data within the environmental range is obtained by using the environmental range, which can reduce the amount of calculation while reducing the environmental range, and more detailed environmental data within the corresponding environmental range can be obtained. For example, the speed of the corresponding pet dog and pet turtle limits the different movement ranges. At this time, if the movement trajectory of the pet turtle is calculated based on the movement range of the pet dog, it will not only increase the amount of calculation but also ignore some environmental features, resulting in prediction trajectory deviation.

[0028] In another case, the different movement environments of the pets must also be taken into account. In this case, obtaining real-time environmental data based on the environmental range also includes: Get environmental features based on pet type; Search the environment with corresponding environmental features in the basic environment area to obtain corresponding real-time environmental data.

[0029] In this case, different environmental features are obtained for different pet types. For example, large dogs cannot enter narrow passages under normal circumstances. In this case, when searching the basic environmental area, the locations that the pet cannot reach can be abandoned, thereby further narrowing the search range, reducing the amount of calculation, and improving calculation efficiency. For some small pets, such as hamsters, additional consideration of small areas is required. In this case, when searching the basic environmental area, it is necessary to mark the corresponding small locations, such as sewer outlets, etc. Obtaining real-time environmental data based on the adaptability of the pet type can not only improve the accuracy of trajectory prediction, but also improve the efficiency of trajectory prediction output, making it easier to quickly find lost pets.

[0030] The actual trajectory of the pet's movement is constructed using the pet's positioning data obtained from all lost time series, including: The pet positioning data obtained from all lost time series are used as the actual motion nodes; Draw the simulated motion trajectory between actual motion nodes based on the positioning sampling interval and the pet type; The actual movement trajectory of the pet is output with actual movement nodes and simulated movement trajectory.

[0031] At this time, the actual trajectory of the pet's movement at least includes multiple movement trajectories between movement nodes that are in line with the pet's achievable movement trajectory in the positioning sampling interval. That is, the maximum movement speed is obtained according to the pet type, and the trajectory that can reach the next movement node at the speed within the maximum movement speed and the positioning sampling interval time is used as the simulated movement trajectory.

[0032] Outputting a prediction error trajectory based on the actual pet movement trajectory and the first pet movement prediction trajectory includes: Matching actual motion nodes in the actual pet motion trajectory and predicted motion nodes in the first pet motion predicted trajectory; The prediction error trajectory is outputted based on the simulated motion trajectory corresponding to the unmatched actual motion node and the predicted motion trajectory corresponding to the predicted motion node.

[0033] In this embodiment, the simulated motion trajectory that reaches the actual motion node that does not match the predicted motion node and the simulated motion trajectory that starts from the actual motion node that does not match the predicted motion node are used as the simulated motion trajectory corresponding to the actual motion node. Similarly, the predicted motion trajectory that reaches the predicted motion node that does not match the actual motion node and the predicted motion trajectory that starts from the predicted motion node that does not match the actual motion node are used as the predicted motion trajectory corresponding to the predicted motion node. The predicted error trajectory at least includes the predicted environment error, the predicted position error and the predicted speed error. The predicted environment error is obtained according to the environmental difference between the simulated motion trajectory and the predicted motion trajectory, the predicted speed error is obtained according to the pet speed difference between the simulated motion trajectory and the predicted motion trajectory, and the position difference is obtained according to the position difference between the actual motion node and the predicted motion node. Since there are multiple simulated motion trajectories, the predicted error trajectory output at this time also contains the errors of multiple trajectories, thereby forming an error correction training set, performing error correction with multiple trajectories that can reach the final target position, increasing the data used to correct the error, and avoiding the correction deviation caused by a single error data. It is understandable that when there are multiple unmatched actual motion nodes, the simulated motion trajectories corresponding to the multiple actual motion nodes must have the same characteristics. At this time, the simulated motion trajectories corresponding to the multiple actual motion nodes can further obtain the behavioral bias weight after the pet is lost, and realize the correction of local deviation and global deviation from the actual local motion trajectory and the global deviation motion trajectory at the same time, thereby improving the correction accuracy.

[0034] Outputting a prediction error trajectory using a simulated motion trajectory corresponding to an unmatched actual motion node and a predicted motion trajectory corresponding to a predicted motion node includes: The corresponding predicted motion trajectory in the first pet motion prediction trajectory is used as the actual motion trajectory that matches the actual motion node; a motion loss curve is constructed using the actual motion trajectory, and the pet motion loss upper limit is obtained based on the pet type; the motion loss curve and the pet motion loss upper limit are used to obtain the simulated motion trajectory corresponding to the unmatched actual motion node.

[0035] The pet's motion loss is reflected through the actual motion trajectory, so as to screen out the simulated motion trajectory that does not match the pet's actual motion situation, thereby improving the efficiency of subsequent corrections. In other cases, the pet motion prediction model can be corrected based on all the simulated motion trajectories of the unmatched actual motion nodes, and the predicted motion trajectory can be output again with the corrected pet motion prediction model, and then the actual motion trajectory of the remaining matched actual motion nodes can be output with the predicted motion trajectory output again, and then the simulated motion trajectory of the unmatched actual motion nodes can be screened out, and the iteration can be continued until the number of screened trajectories is 0 or the maximum number of iterations is reached.

[0036] In other embodiments, constructing the actual trajectory of the pet's movement using all the pet positioning data acquired in lost time sequence includes: The pet positioning data obtained from all lost time series are used as the actual motion nodes; Matching the actual motion nodes and the predicted motion nodes in the predicted motion trajectory of the first pet; The simulated motion trajectory is constructed with actual motion nodes and positioning sampling intervals that do not match the predicted motion nodes.

[0037] In this embodiment, the actual motion nodes are matched with the predicted motion nodes first, and only the simulated motion trajectories corresponding to the unmatched actual motion nodes are output, thereby reducing the output of unnecessary simulated motion trajectories, reducing the amount of calculation, and improving efficiency.

[0038] After the second pet motion prediction trajectory in the future time sequence is output based on the updated pet motion prediction model, the user's motion speed in the corresponding time sequence is obtained according to the user's historical motion data, and this is combined with the user's current position. The position that the user and the pet can reach at the same time sequence is used as the user search trajectory, and the user search trajectory is constructed with the trajectory and speed that can be reached at the same time sequence.

[0039] During the user search process, the pet's positioning data is obtained in real time, and the real-time pet positioning data is compared with the predicted positioning data of the corresponding time series of the second pet's motion prediction trajectory. When the real-time pet positioning data does not match the predicted positioning data, the second pet's motion prediction trajectory is updated with the difference between the real-time pet positioning data and the predicted positioning data. The real-time pet positioning data at least includes the real-time pet position, and the predicted positioning data at least includes the predicted pet position. Based on the deviation coefficient between the real-time pet position and the predicted pet position, such as the latitude and longitude deviation ratio, the deviation value of the subsequent predicted positioning data is calculated, and the second pet's motion prediction trajectory is updated accordingly. It can be understood that at this time, the real-time pet position deviation is used as a local deviation event, and the second pet's motion prediction trajectory is continuously updated accordingly, which can avoid the overall model being affected by local data, and at the same time, the second pet's motion prediction trajectory can be fine-tuned in real time to improve the search accuracy.

[0040] As a second embodiment of the present invention, due to the influence of the environment, the pet positioning data may also have an impact. Therefore, the pet historical movement data at least includes the pet historical positioning data and the pet historical actual position data. At this time, step S1 also includes: A positioning correction model is constructed based on the pet's historical positioning data and the pet's historical actual location data.

[0041] Step S2 also includes: Obtain all pet positioning data within the lost time sequence, and correct all pet positioning data within the lost time sequence based on the positioning correction model.

[0042] Step S6 also includes: The real-time pet positioning data is corrected based on the positioning correction model.

[0043] The final pet motion prediction trajectory is output based on the corrected pet positioning data within the lost time series and the real-time pet positioning data, further improving the trajectory prediction accuracy and avoiding prediction deviations caused by positioning errors in special environments.

[0044] As a third embodiment of the present invention, a pet positioning search system is connected to a pet positioning system, comprising: The data acquisition module is used to pre-store the pet's historical movement data offline and the real-time pet positioning data transmitted by the pet positioning system; the data modeling module is used to build a pet movement prediction model based on the pet's historical movement data to output the pet's movement prediction trajectory; the trajectory construction module is used to build the pet's actual movement trajectory based on the lost time series; An error analysis module is used to output a predicted error trajectory or a difference value between the real-time pet positioning data and the predicted positioning data based on the predicted pet movement trajectory, the real-time pet positioning data and the actual pet movement trajectory; The trajectory planning module is used to output the user's search trajectory based on the pet's movement prediction trajectory.

[0045] Among them, the data acquisition module is connected to the pet positioning system, the data modeling module, the trajectory construction module and the error analysis module are connected to the data acquisition module, the error analysis module is connected to the data modeling module and the trajectory construction module, and the error analysis module is connected to the data modeling module.

[0046] Generally speaking, the pet locator will upload the positioning data to the pet positioning cloud terminal. However, in order to ensure user privacy, the data collection module is set as an offline pre-storage module in this embodiment. When the positioning data is uploaded to the pet positioning cloud terminal, it is collected to the offline pre-storage module required by the user, and the pet positioning cloud terminal information is deleted at the same time to protect user privacy. It is understandable that the data collection module can be set to delete data regularly and delete long-term positioning information to avoid memory occupation by useless data.

[0047] In other cases, the data collection module at least includes collecting the lost pet's historical movement data in the offline pre-storage module and all pets' historical movement data in the pet positioning cloud terminal. In this embodiment, the data collection module is connected to the offline pre-storage module and the pet positioning cloud terminal respectively. At this time, the pet positioning cloud terminal is included in the pet positioning search system, which is configured to selectively retain the pet's historical movement data in the non-privacy area according to the user's privacy area division. At this time, the pet movement prediction model is based on the private domain data of the user's own pet and the public domain data of all user's pets in the non-privacy area. The habits of most pets and their performance in different areas are used to supplement the performance of the corresponding pet type in different areas, thereby improving the accuracy of model construction while protecting the user's privacy.

[0048] The specific implementation described above is a preferred implementation of the pet positioning search method and system of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A pet location search method, characterized by: The steps include: S1: Obtain pet historical movement data, and build a pet movement prediction model based on the pet historical movement data; S2: obtaining the environment range corresponding to the pet type based on the pet positioning data obtained from the initial lost time sequence, obtaining real-time environment data according to the environment range, and outputting the first pet movement prediction trajectory within all lost time sequences using the pet movement prediction model and the real-time environment data; S3: constructing the actual pet movement trajectory using the pet positioning data acquired from all lost time series, and outputting a prediction error trajectory based on the actual pet movement trajectory and the first pet movement prediction trajectory; S4: updating the pet motion prediction model with the prediction error trajectory, and outputting a second pet motion prediction trajectory in the future time series with the updated pet motion prediction model; S5: Obtain the user's historical movement data and the user's current location, and output the user's search trajectory based on the second pet's predicted movement trajectory; S6: Compare the real-time pet positioning data with the predicted positioning data of the corresponding time series of the second pet movement prediction trajectory. When the real-time pet positioning data does not match the predicted positioning data, update the second pet movement prediction trajectory with the difference value between the real-time pet positioning data and the predicted positioning data, and re-execute S5.

2. The pet location search method according to claim 1, characterized in that: The S1 includes: Acquire historical movement data of the lost pet, and perform data preprocessing on the historical movement data of the lost pet; Build a basic deep learning model based on convolutional neural network; The basic deep learning model is trained using the historical movement data of lost pets to output a pet movement prediction model.

3. The pet location search method according to claim 1, characterized in that: The pet location data acquired based on the initial loss time sequence acquires the environment range corresponding to the pet type, including: Get the pet's maximum movement speed based on the pet type; Get motion timing with preset search timing and loss timing; The environment range corresponding to the pet type is obtained by using the movement timing and the pet's maximum movement speed.

4. The pet location search method according to claim 1, characterized in that: The acquiring of real-time environmental data according to the environmental range also includes: Get environmental features based on pet type; Search the environment with corresponding environmental features in the basic environment area to obtain corresponding real-time environmental data.

5. The pet location search method according to claim 1, characterized in that: The pet positioning data acquired from all lost time sequences are used to construct the actual trajectory of the pet's movement, including: The pet positioning data obtained from all lost time series are used as the actual motion nodes; Draw the simulated motion trajectory between actual motion nodes based on the positioning sampling interval and the pet type; The actual motion trajectory of the pet is output with actual motion nodes and simulated motion trajectory.

6. The pet location search method according to claim 5, characterized in that: The outputting of a prediction error trajectory based on the actual pet movement trajectory and the first pet movement prediction trajectory comprises: Matching actual motion nodes in the actual pet motion trajectory and predicted motion nodes in the first pet motion predicted trajectory; The prediction error trajectory is outputted based on the simulated motion trajectory corresponding to the unmatched actual motion node and the predicted motion trajectory corresponding to the predicted motion node.

7. The pet location search method according to claim 6, characterized in that: Outputting a prediction error trajectory using a simulated motion trajectory corresponding to an unmatched actual motion node and a predicted motion trajectory corresponding to a predicted motion node comprises: Using the predicted motion trajectory corresponding to the first pet's predicted motion trajectory as the actual motion trajectory matching the actual motion node; A motion loss curve is constructed based on the actual motion trajectory, and the upper limit of the pet's motion loss is obtained based on the pet type. The motion loss curve and the upper limit of the pet's motion loss are used to obtain the simulated motion trajectory corresponding to the unmatched actual motion node.

8. The pet location search method according to claim 1, characterized in that: The pet positioning data acquired from all lost time sequences are used to construct the actual trajectory of the pet's movement, including: The pet positioning data obtained from all lost time series are used as the actual motion nodes; Matching the actual motion nodes and the predicted motion nodes in the predicted motion trajectory of the first pet; The simulated motion trajectory is constructed with actual motion nodes and positioning sampling intervals that do not match the predicted motion nodes.

9. The pet location search method according to claim 1, characterized in that: The S1 further comprises: A positioning correction model is constructed based on the pet's historical positioning data and the pet's historical actual location data.

10. A pet positioning search system, connected to a pet positioning system, for implementing the pet positioning search method according to any one of claims 1 to 9, characterized in that: The pet location search system includes: The data acquisition module is used to pre-store the pet's historical movement data offline and the real-time pet positioning data transmitted by the pet positioning system; the data modeling module is used to build a pet movement prediction model based on the pet's historical movement data and output the pet's movement prediction trajectory; A trajectory construction module, used to construct the actual trajectory of the pet's movement based on the lost timing; An error analysis module is used to output a predicted error trajectory or a difference value between the real-time pet positioning data and the predicted positioning data based on the predicted pet movement trajectory, the real-time pet positioning data and the actual pet movement trajectory; The trajectory planning module is used to output the user's search trajectory based on the pet's movement prediction trajectory.

Images