Automobile luggage loss early warning method, system, device and medium

Abstract

The application discloses an unmanned vehicle luggage loss early warning method based on multi-modal sensing and time sequence behavior analysis, which comprises the following steps: in response to the end of an order, synchronously collecting pressure distribution, vehicle state and vehicle door state time sequence data; filtering and dynamically compensating the pressure data to generate compensated pressure features; extracting multi-dimensional time sequence features such as pressure net change, change rate and time correlation with vehicle door action, and inputting the features into a lightweight machine learning model to obtain a loss probability; and comparing the probability with a preset threshold to execute a graded early warning strategy. The application eliminates the influence of vehicle vibration by introducing dynamic interference compensation, accurately distinguishes temporary placement from real loss by using multi-dimensional time sequence features and a machine learning model, and significantly reduces the false positive rate; the graded early warning mechanism (immediate alarm for high probability, review and then decide for medium probability) avoids invalid disturbance and improves user experience. The method realizes accurate, intelligent and privacy-friendly luggage loss early warning.

Description

Technical Field

[0001] This application relates to the field of autonomous vehicle technology, and in particular to an autonomous vehicle luggage loss early warning method, system, device and medium based on multimodal sensing and temporal behavior analysis. Background Technology

[0002] With the increasing popularity of ride-sharing and self-driving cars, the problem of passengers leaving their belongings behind is becoming increasingly prominent. Existing technologies mainly offer two types of solutions:

[0003] Existing technology 1: A luggage reminder device based on pressure sensors. This device detects whether there are still items on the seat after the passenger gets out of the car by installing pressure sensors in the taxi seat or trunk. When the pressure exceeds a threshold, an on-site audible and visual alarm is triggered. This solution has the following drawbacks: 1) High false alarm rate, unable to distinguish between frequently placed items and passenger luggage; 2) Limited reminder method, unable to receive alerts after the passenger leaves the vehicle; 3) Lack of user association capabilities, unable to provide accurate notifications; 4) Simple judgment logic, unable to handle complex scenarios such as vibration and differences in the material of the items.

[0004] Existing technology 2: A visual image recognition-based solution that uses an in-vehicle camera to compare images of passengers getting in and out of the vehicle to identify lost items. This solution has the following drawbacks: 1) computational complexity and high cost; 2) poor environmental adaptability (significantly affected by lighting and occlusion); 3) privacy concerns; and 4) insensitivity to small items.

[0005] Therefore, there is an urgent need for a precise, intelligent, and privacy-protecting solution for alerting users when their luggage is lost. Summary of the Invention

[0006] The technical problem this application aims to solve is to overcome the shortcomings of existing technologies, such as high false alarm rates, ineffective alerts, poor environmental adaptability, and privacy risks. It provides a method and system for early warning of lost luggage in autonomous vehicles based on multimodal sensing and temporal behavior analysis, capable of accurately distinguishing between temporary placement and permanent loss, and achieving efficient multi-channel alerts after the passenger leaves the vehicle. To achieve the above objectives, this application adopts the following technical solution.

[0007] In a first aspect, embodiments of this application provide a method for early warning of lost vehicle luggage, including:

[0008] In response to the end of the order service, time-series data of pressure distribution, vehicle status, and door status within the monitoring area are collected synchronously.

[0009] The pressure distribution time series data is filtered and combined with the vehicle state time series data for dynamic interference compensation to generate a compensated pressure feature sequence.

[0010] Based on the compensated pressure feature sequence and the door state time series data, multi-dimensional time series features characterizing changes in the state of the item are extracted; the multi-dimensional time series features include at least the net pressure change, the pressure change rate, and the temporal correlation between pressure change and door movement.

[0011] The multi-dimensional temporal features are input into a pre-trained lightweight machine learning model to obtain the probability value of item loss.

[0012] The system compares the probability value of an item being lost with a preset decision threshold range and then executes corresponding early warning strategies in different levels.

[0013] Furthermore, the dynamic interference compensation specifically includes:

[0014] Acquire vehicle vibration data synchronously collected by the inertial measurement unit;

[0015] Establish the interference transfer function between vehicle vibration and pressure sensor readings, and estimate the pressure fluctuation component caused by vehicle vibration;

[0016] The pressure fluctuation component is subtracted from the original pressure distribution time series data to obtain the compensated pressure feature sequence.

[0017] Furthermore, the multi-dimensional temporal features also include:

[0018] Spatial occupancy characteristics and static / dynamic attribute characteristics of objects within the monitoring area are extracted based on millimeter-wave radar point cloud data.

[0019] Furthermore, the lightweight machine learning model is an ensemble learning model based on decision trees, which is obtained through the following steps:

[0020] Acquire historical order data; the historical order data includes synchronously collected multimodal sensor data and corresponding luggage loss tracking tags;

[0021] The multi-dimensional temporal features are extracted from the multimodal sensor data to form a training sample set;

[0022] The gradient boosting decision tree model is trained using the training sample set to minimize the loss function between the predicted missing probability and the true label.

[0023] Furthermore, the tiered execution of corresponding early warning strategies specifically includes:

[0024] If the probability of the item being lost exceeds a preset high threshold, a first-level warning will be triggered immediately. The first-level warning includes a local audible and visual alarm and a notification of the lost item being pushed to the user's terminal.

[0025] If the probability value of the lost item is between a preset low threshold and a preset high threshold, a second-level warning is triggered. The second-level warning includes activating an auxiliary sensor for verification, and if the probability value is still higher than the preset low threshold after a preset waiting time, it is upgraded to the first-level warning.

[0026] Furthermore, the step of pushing the lost notification to the user terminal specifically includes:

[0027] Through the cloud service platform, the user's identity and lost item information associated with the order will be sent to the corresponding user terminal via one or more methods, such as APP push, SMS, or telephone robot.

[0028] Furthermore, the monitoring area includes one or more of the vehicle seat area, floor area, and trunk area;

[0029] The time-series data of the pressure distribution is collected by a flexible thin-film pressure sensor array deployed in the monitoring area.

[0030] Secondly, embodiments of this application provide an early warning system capable of implementing any of the aforementioned early warning methods, comprising:

[0031] The multimodal sensing module is used to synchronously collect time-series data on pressure distribution, vehicle status, and door status within the monitoring area in response to the end of the order service.

[0032] The signal processing module is used to filter the pressure distribution time series data and perform dynamic interference compensation in combination with the vehicle state time series data to generate a compensated pressure feature sequence.

[0033] The feature extraction module is used to extract multi-dimensional temporal features representing changes in the state of an item based on the compensated pressure feature sequence and the time-series data of the door state; the multi-dimensional temporal features include at least the net pressure change, the pressure change rate, and the temporal correlation between pressure change and door movement;

[0034] The intelligent decision-making module is used to input the multi-dimensional temporal features into a pre-trained lightweight machine learning model to obtain the probability value of item loss.

[0035] The graded early warning execution module is used to compare the probability value of the lost item with a preset decision threshold range and execute corresponding early warning strategies in different grades.

[0036] Thirdly, embodiments of this application provide an electronic device, including: one or more processors;

[0037] A memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors are able to implement the steps in the warning method described in any of the preceding claims.

[0038] Fourthly, embodiments of this application provide a computer-readable medium storing a computer program, which, when executed by a processor, can implement the steps of the early warning method described in any of the preceding claims.

[0039] This application discloses a method for early warning of lost luggage in autonomous vehicles based on multimodal sensing and temporal behavior analysis. The method includes: upon order completion, simultaneously collecting temporal data on pressure distribution, vehicle status, and door status; filtering and dynamically compensating for pressure data to generate compensated pressure features; extracting multi-dimensional temporal features such as net pressure change, rate of change, and temporal correlation with door movement, and inputting these features into a lightweight machine learning model to obtain the probability of loss; and implementing a tiered early warning strategy based on a comparison of this probability with a preset threshold. This application eliminates the impact of vehicle vibration by introducing dynamic interference compensation, accurately distinguishes between temporary placement and genuine loss using multi-dimensional temporal features and a machine learning model, significantly reducing the false alarm rate; the tiered early warning mechanism (immediate alarm for high probability, and further review for medium probability) avoids unnecessary disturbances and improves user experience. This method achieves accurate, intelligent, and privacy-friendly early warning of lost luggage. Attached Figure Description

[0040] Figure 1 A core flowchart of a method for early warning of lost luggage in unmanned vehicles based on multimodal sensing and temporal behavior analysis provided in this application embodiment;

[0041] Figure 2 A flowchart illustrating an unmanned vehicle luggage loss warning method based on multimodal sensing and temporal behavior analysis, provided for an embodiment of this application;

[0042] Figure 3 A schematic diagram of the module structure of an unmanned vehicle luggage loss early warning system based on multimodal sensing and temporal behavior analysis provided in an embodiment of this application;

[0043] Figure 4 This is a structural block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0044] To enable those skilled in the art to better understand the technical solutions of this application, exemplary embodiments of this application are described below with reference to the accompanying drawings, including various details of the embodiments of this application to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. Unless otherwise specified, the various embodiments of this application and the features within those embodiments can be combined with each other.

[0045] As used herein, the term "and / or" includes any and all combinations of one or more of the associated enumerated entries. The terminology used herein is for describing particular embodiments only and is not intended to limit the application. As used herein, the singular forms "a" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that when the terms "comprising" and / or "made of" are used herein, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as "connected" or "linked" are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0046] Unless otherwise specified, all terms used in this application (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It should also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this application, and will not be interpreted as having an idealized or overly formal meaning, unless expressly so defined in this application.

[0047] The vehicle luggage loss warning system described in this application comprises an onboard unit, a cloud service platform, and a user-end application. The onboard unit includes multimodal sensors (pressure sensor array, IMU, door sensors, millimeter-wave radar), a decision-making module (operational signal processing, feature extraction, and intelligent modeling), and an execution module (local audible and visual alarm, T-Box). The cloud service platform is responsible for model training, order association, and multi-channel push notifications. The user-end application is a passenger mobile app or SMS.

[0048] refer to Figures 1-3 One embodiment of this application proposes a method for early warning of lost luggage in autonomous vehicles based on multimodal sensing and temporal behavior analysis. The early warning method may specifically include the following steps.

[0049] S1. In response to the end of the order service, synchronously collect time-series data of pressure distribution, vehicle status, and door status within the monitoring area.

[0050] Specifically, in response to the completion of the order service (e.g., after the vehicle is put into Park and all doors are closed), the system automatically triggers the luggage loss detection process. The multimodal perception module simultaneously collects the following multi-source data:

[0051] Pressure distribution time-series data: A flexible thin-film pressure sensor array deployed on the seat cushions, rear floor, and trunk floor continuously collects pressure data at a frequency of 20 Hz. This array contains 16×16 sensing units, with a thickness of only 0.3 mm, enabling precise sensing of the placement of objects and weight distribution.

[0052] Vehicle status timing data: Vehicle gear status, vehicle speed signal, and inertial measurement unit (IMU) data (three-axis acceleration and angular velocity) are acquired via the vehicle's CAN bus at a frequency of 50 Hz. This data is used to determine the vehicle's motion status and identify vibration interference.

[0053] Door status timing data: Collect timing signals of opening / closing of the four doors to accurately record the timing of door actions.

[0054] Auxiliary perception data: The 4D imaging millimeter-wave radar integrated into the roof lining synchronously collects point cloud data within the monitoring area for subsequent analysis of object space occupancy and static / dynamic attributes.

[0055] S2. Filter the pressure distribution time series data and perform dynamic interference compensation in combination with the vehicle state time series data to generate a compensated pressure feature sequence.

[0056] Specifically, the signal processing module first preprocesses the raw pressure data. A Kalman filter algorithm is used to smooth and denoise the pressure time-series data P(t), resulting in a pre-filtered P'(t). However, vibrations during vehicle movement or idling (measured by the IMU) introduce interference components into the pressure sensor, leading to misjudgments. Therefore, this application introduces a dynamic interference compensation mechanism, which includes the following sub-steps.

[0057] S21. Acquire vehicle vibration data (such as vertical acceleration) synchronously collected by the IMU.

[0058] S22. Establish the interference transfer function between vehicle vibration and pressure sensor readings. Specifically, under vehicle no-load conditions, collect pressure sensor noise data corresponding to different vibration intensities experimentally. Use system identification methods (such as the least squares method) to fit a second-order transfer function model, which is used to estimate the pressure fluctuation component ΔP caused by vibration in real time. vib (t).

[0059] S23. Subtract this fluctuation component from the original pressure data to obtain the compensated pressure characteristic sequence P. comp (t)=P'(t)–ΔP vib (t). For example, when the IMU measures a vertical acceleration of 0.2g, the model estimates the fluctuation component as ±50g. If the original pressure reading is 2.5 kg, the compensated net weight is 2.5 kg ± 0.05 kg, thus accurately identifying the true weight of the item.

[0060] S3. Based on the compensated pressure feature sequence and the door state time series data, extract multi-dimensional time series features that characterize the changes in the state of the item; the multi-dimensional time series features include at least the net pressure change, the pressure change rate, and the time correlation between pressure change and door movement.

[0061] Specifically, the feature extraction module extracts multi-dimensional temporal features based on the compensated pressure feature sequence, door state time-series data, and millimeter-wave radar point cloud data. These multi-dimensional temporal features mainly include:

[0062] Pressure characteristics: Calculate the net pressure change ΔP, pressure change rate ΔP / Δt, and variance after pressure stabilization for each monitoring area (e.g., rear seats) before and after passengers get off the vehicle.

[0063] Temporal correlation characteristics: Precisely correlate the time difference between pressure changes and door movements. For example, if the left rear door opens within 3 seconds after the pressure disappears (the item is picked up), it indicates that the item was likely taken by a passenger; conversely, if the pressure persists and the door is closed, the item may have been left behind.

[0064] Radar characteristics: Spatial occupancy features (such as the volume and location of the point cloud cluster) and static / dynamic attribute features are extracted from millimeter-wave radar point clouds (micro-Doppler analysis is used to determine whether an object has signs of life). For example, if there is a static point cloud cluster measuring 20 cm long × 10 cm wide × 5 cm high in the rear seat area, and it does not move, it is identified as suspected luggage; if the point cloud cluster shows periodic micro-movements (corresponding to breathing), it is identified as left by a child.

[0065] Contextual features include vehicle status (in P gear, speed at 0) and whether the user has moved away based on IMU data.

[0066] The above features are combined into a multidimensional feature vector, which is then input into the subsequent decision model.

[0067] S4. Input the multi-dimensional temporal features into a pre-trained lightweight machine learning model to obtain the item loss probability value.

[0068] Specifically, the intelligent decision-making module is equipped with a pre-trained lightweight machine learning model. This embodiment uses an ensemble learning model based on decision trees—XGBoost (Extreme Gradient Boosting)—and its training process is as follows:

[0069] S41. Collect historical order data on the cloud big data platform, including multimodal sensor data and real labels indicating whether items were lost or not, based on customer service follow-ups or user feedback.

[0070] S42. Extract features from each historical data point and construct a training sample set (feature matrix + label).

[0071] S43. The XGBoost algorithm is used to train the model to minimize the log loss function between the predicted probability and the true label. During training, the algorithm automatically learns the weights of different features. For example, it is found that the combination of "radar point cloud volume + net pressure change" contributes the most to the missing data judgment.

[0072] After training, the model is converted to a lightweight format (such as ONNX) and deployed to the vehicle-mounted domain controller for real-time inference. The model input is the currently extracted feature vector, and the output is the omission probability value P between 0 and 1. 遗落 .

[0073] S5. Based on the comparison between the probability value of the lost item and the preset decision threshold range, the corresponding early warning strategy is executed in stages.

[0074] Specifically, the tiered early warning execution module executes the missing probability value P output by the model. 遗落 It is compared with a preset threshold. In this embodiment, a high threshold Th is set. high =0.85, low threshold Th low =0.4.

[0075] If P 遗落 A reading >0.85 indicates a high probability of loss, immediately triggering a Level 1 warning: a local audible and visual alarm (buzzer sounds, roof light flashes yellow), and simultaneously uploading event information (order ID, vehicle location, and estimated item type) to the cloud service platform via the T-Box. The cloud service platform then links the order ID to the passenger's mobile app, pushes a strong reminder notification, and sends a backup SMS message. For emergency scenarios (such as leaving a child behind), an AI-powered telephone robot can be activated to automatically contact the passenger.

[0076] If 0.4 <P 遗落If the probability is ≤0.85, it is considered a medium probability and triggers a Level 2 warning: The system activates millimeter-wave radar for a re-scan, performs a detailed analysis of the point cloud features of the suspicious area, and starts a 30-second waiting timer. If the probability value drops below 0.4 within 30 seconds, it is determined to be a temporary placement and no alarm is triggered; if the probability is still above 0.4 after 30 seconds, it is upgraded to a Level 1 warning, and a notification is sent to the user.

[0077] If P 遗落 If the value is ≤0.4, it is considered a safe state, and no operation is performed.

[0078] refer to Figure 3 An embodiment of this application also proposes an unmanned vehicle luggage loss early warning system based on multimodal sensing and temporal behavior analysis. The early warning system is integrated into the unmanned vehicle and may specifically include the following modules.

[0079] (1) Multimodal sensing module: It consists of a flexible pressure sensor array, IMU, door sensor, millimeter-wave radar and CAN bus interface, and is used to collect the multi-source data mentioned in step S1 above.

[0080] (2) Signal processing module: running on the MCU core of the intelligent cockpit domain controller, realizing Kalman filtering and dynamic interference compensation.

[0081] (3) Feature extraction module: running on the high-performance CPU core of the domain controller, it calculates multi-dimensional time-series features in real time.

[0082] (4) Intelligent decision-making module: Deploy a lightweight XGBoost model and output the probability value of item loss.

[0083] (5) Graded early warning execution module: Controls local sound and light alarm and T-Box communication unit according to the probability value of item loss, and works with the cloud to complete multi-channel push.

[0084] All of the above modules work together to achieve a complete closed loop from perception to early warning.

[0085] The monitoring area of ​​this application can be extended to the seats, floor, and trunk, with full coverage through a flexible thin-film pressure sensor array to avoid missed detections. Vehicle status signals, including gear position, speed, and IMU data, provide accurate context for decision-making. Furthermore, alternative solutions such as pressure-vision fusion can be considered as an enhancement to this solution. Specifically, when pressure judgment is uncertain (in the medium probability range), a low-power camera can be activated for instantaneous auxiliary verification, but the core logic remains primarily based on non-visual sensing and intelligent algorithms.

[0086] Overall, the advantages of this application compared to the prior art include:

[0087] 1. Significantly reduces false alarm rate and accurately distinguishes between temporary placement and actual loss.

[0088] Some existing technologies rely solely on a single pressure threshold for judgment, failing to distinguish between frequently placed items inside the vehicle (such as bottled water or first-aid kits) and passenger luggage, making them prone to false alarms triggered by vehicle vibrations or these items. This application overcomes this deficiency through the following technical means:

[0089] Dynamic interference compensation: The vehicle vibration is monitored in real time using an IMU, a vibration-pressure interference transfer function is established, the vibration component is deducted from the pressure signal, and the interference of vehicle motion on the pressure reading is eliminated.

[0090] Multi-dimensional temporal feature fusion: Extract refined features such as net pressure change, rate of change, and temporal correlation between pressure and door movement, enabling the model to understand the interaction between "people-objects-vehicles".

[0091] Millimeter-wave radar assistance: By analyzing the spatial occupancy and static / dynamic attributes of items through point cloud analysis, luggage can be distinguished from living beings, further reducing misjudgments.

[0092] The system can accurately identify commonly placed items (such as first aid kits fixed in the vehicle) and ignore their interference, triggering an alarm only when a passenger's personal belongings are left behind, reducing the false alarm rate by more than 80%.

[0093] 2. Achieve precise reach after the driver leaves the vehicle, resolving the pain point of "ineffective reminders".

[0094] Existing technologies rely solely on on-site audio-visual alarms, which fail to provide alerts if passengers have already left the vehicle. This application proposes a vehicle-cloud collaborative system:

[0095] Orders are linked to user identities: The system automatically associates order IDs with user accounts, and after a lost item is uploaded to the cloud, it can be accurately located to the corresponding passenger.

[0096] Multi-channel redundant push: Reach users through multiple methods such as APP push, SMS, and AI call robots to ensure that notifications are delivered.

[0097] Even if passengers have been away from the vehicle for several kilometers, they can still receive an immediate alert that "you may have left your backpack behind," which can increase the item retrieval rate by more than 90%.

[0098] 3. Excellent environmental adaptability and comprehensive privacy protection.

[0099] Existing visual solutions are greatly affected by light and occlusion, and continuous shooting raises privacy concerns. This application primarily uses non-visual sensors:

[0100] Pressure array + millimeter-wave radar: Unaffected by light, smoke, or obstruction, it can still work stably in dim environments such as nighttime and underground parking lots.

[0101] Privacy protection design: No image data is uploaded throughout the process; only anonymized feature vectors and point clouds are transmitted, fundamentally eliminating the risk of privacy leakage.

[0102] The system can accurately identify small items such as cell phones and keys left in the back seat in a completely dark environment, while passengers do not need to worry about their privacy being recorded inside the vehicle.

[0103] 4. Intelligent hierarchical early warning system to avoid unnecessary disturbances.

[0104] Existing "one-size-fits-all" alarm technologies often trigger false alarms due to passengers temporarily placing items on their seats (such as placing their phones on the seat while retrieving a card), resulting in a poor user experience. This application proposes a tiered early warning mechanism:

[0105] High probability of immediate alarm: When the model output is >0.85, the confidence level is extremely high, and an immediate audible and visual alarm + remote notification will be triggered.

[0106] Medium probability re-verification and decision: When the probability is in the range of 0.4 to 0.85, the radar re-verification is initiated and waits for 30 seconds. If it is a temporary placement (such as briefly putting down the item and then picking it up again), the alarm will be automatically canceled.

[0107] Passengers will not be disturbed when temporarily leaving items, and will only receive a reminder when they actually leave something behind. This significantly improves the intelligent experience and increases user satisfaction.

[0108] 5. The system has the ability to evolve, becoming smarter the more it is used.

[0109] Existing technical solutions are rigid and inflexible, unable to adapt to new scenarios. This application adopts a machine learning architecture of cloud-based training and vehicle-side deployment:

[0110] Continuous learning: The cloud aggregates massive amounts of real order data (including missing positive samples and normal negative samples) and iterates and optimizes the model regularly.

[0111] Lightweight deployment: The model is compressed and deployed on the vehicle, enabling real-time inference without relying on the network, and can also be upgraded via OTA.

[0112] The system can continuously learn new types of items (such as new types of suitcases and shopping bags) and abandoned scenarios (such as child safety seats left behind). The recognition accuracy continues to improve with the accumulation of data, forming a technological moat.

[0113] 6. Ensure comprehensive coverage of the monitoring area and eliminate blind spots.

[0114] Existing pressure monitoring solutions typically only monitor the seats, often neglecting areas such as the trunk and floor. This application expands the monitoring scope:

[0115] Flexible film pressure array: laid on the seats, rear floor, and bottom of the trunk to achieve full coverage of the vehicle without blind spots.

[0116] Millimeter-wave radar supplement: It can penetrate seat cover and detect concealed items (such as a mobile phone that has slipped into the seat gap).

[0117] The system can accurately identify whether passengers place their luggage in the trunk, rear floor, or seat gaps, with a false negative rate approaching zero.

[0118] 7. Low cost and low power consumption, easy to mass-produce and deploy.

[0119] Compared to vision solutions that require high-performance GPUs, this application has a significant cost advantage:

[0120] Sensor selection: Pressure arrays, IMUs, and millimeter-wave radars are all mature, low-cost devices, with a total cost lower than that of a high-definition camera plus a computing unit.

[0121] Lightweight models: Traditional machine learning models such as XGBoost do not require GPU acceleration and can run in real time on MCUs or low-power CPUs.

[0122] The overall hardware cost of the system can be reduced by more than 60%, and the power consumption can be lower than 5W, making it suitable for large-scale pre-installation mass production and providing a cost-effective solution for operators of driverless taxis.

[0123] This application systematically addresses the core pain points of existing technologies, such as high false alarms, ineffective alerts, privacy risks, and poor environmental adaptability, through a combination of technologies including "multimodal perception, dynamic interference compensation, temporal feature fusion, machine learning decision-making, and hierarchical early warning." It achieves accurate, intelligent, privacy-friendly, and evolvable baggage loss early warning, providing key technical support for shared mobility and autonomous driving operations.

[0124] The embodiments of the aforementioned early warning method and the embodiments of the aforementioned early warning system are identical or related in technical concept, and they can be referenced and learned from each other in terms of technical details and technical effects, which will not be repeated here.

[0125] Based on the same inventive concept, embodiments of this application also provide an electronic device. Figure 4 This is a structural block diagram of an electronic device provided in an embodiment of this application. Figure 4 As shown in the embodiments of this application, an electronic device includes: one or more processors 101, a memory 102, and one or more I / O interfaces 103. The memory 102 stores one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement any of the early warning methods described in the above embodiments; the one or more I / O interfaces 103 are connected between the processor and the memory, configured to enable information interaction between the processor and the memory.

[0126] The processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read / write interface) 103 is connected between the processor 101 and the memory 102, and can realize information interaction between the processor 101 and the memory 102, including but not limited to a data bus (Bus).

[0127] In some embodiments, the processor 101, memory 102, and I / O interface 103 are interconnected via bus 104, and thus connected to other components of the computing device.

[0128] In some embodiments, the one or more processors 101 include a field-programmable gate array.

[0129] This application also provides a computer-readable medium. The computer-readable medium stores a computer program, which, when executed by a processor, implements the steps of any of the warning methods described in the above embodiments. The computer-readable storage medium may be volatile or non-volatile.

[0130] This application also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-mentioned warning method.

[0131] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0132] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0133] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0134] The computer program instructions used to perform the operations of this application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuits, such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), are personalized by utilizing the status information of the computer-readable program instructions. These electronic circuits can execute the computer-readable program instructions to implement various aspects of this application.

[0135] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0136] Various aspects of this application are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0137] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0138] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0139] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. 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 the block diagrams and / or flowcharts, and combinations of blocks in the 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.

[0140] Exemplary embodiments have been disclosed in this application, and while specific terminology has been used, it is used only and should be interpreted in a general illustrative sense and is not intended to be limiting. In some embodiments, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of this application as set forth by the appended claims.

Claims

1. A method for early warning of lost luggage in a car, characterized in that, include: In response to the end of the order service, time-series data of pressure distribution, vehicle status, and door status within the monitoring area are collected synchronously. The pressure distribution time series data is filtered and combined with the vehicle state time series data for dynamic interference compensation to generate a compensated pressure feature sequence. Based on the compensated pressure feature sequence and the time series data of the vehicle door state, multi-dimensional time series features representing changes in the state of the item are extracted. The multi-dimensional temporal features include at least the net pressure change, the pressure change rate, and the temporal correlation between pressure change and door movement. The multi-dimensional temporal features are input into a pre-trained lightweight machine learning model to obtain the probability value of item loss. The system compares the probability value of an item being lost with a preset decision threshold range and then executes corresponding early warning strategies in different levels.

2. The early warning method according to claim 1, characterized in that, The dynamic interference compensation specifically includes: Acquire vehicle vibration data synchronously collected by the inertial measurement unit; Establish the interference transfer function between vehicle vibration and pressure sensor readings, and estimate the pressure fluctuation component caused by vehicle vibration; The pressure fluctuation component is subtracted from the original pressure distribution time series data to obtain the compensated pressure feature sequence.

3. The early warning method according to claim 1, characterized in that, The multi-dimensional temporal features also include: Spatial occupancy characteristics and static / dynamic attribute characteristics of objects within the monitoring area are extracted based on millimeter-wave radar point cloud data.

4. The early warning method according to claim 1, characterized in that, The lightweight machine learning model is an ensemble learning model based on decision trees, which is obtained through the following steps: Acquire historical order data; the historical order data includes synchronously collected multimodal sensor data and corresponding luggage loss tracking tags; The multi-dimensional temporal features are extracted from the multimodal sensor data to form a training sample set; The gradient boosting decision tree model is trained using the training sample set to minimize the loss function between the predicted missing probability and the true label.

5. The early warning method according to claim 1, characterized in that, The tiered execution of corresponding early warning strategies specifically includes: If the probability of the item being lost exceeds a preset high threshold, a first-level warning will be triggered immediately. The first-level warning includes a local audible and visual alarm and a notification of the lost item being pushed to the user's terminal. If the probability value of the lost item is between a preset low threshold and a preset high threshold, a second-level warning is triggered. The second-level warning includes activating an auxiliary sensor for verification, and if the probability value is still higher than the preset low threshold after a preset waiting time, it is upgraded to the first-level warning.

6. The early warning method according to claim 5, characterized in that, The process of pushing lost notifications to user terminals specifically includes: Through the cloud service platform, the user's identity and lost item information associated with the order will be sent to the corresponding user terminal via one or more methods, such as APP push, SMS, or telephone robot.

7. The early warning method according to claim 1, characterized in that, The monitoring area includes one or more of the vehicle seat area, floor area, and trunk area; The time-series data of the pressure distribution is collected by a flexible thin-film pressure sensor array deployed in the monitoring area.

8. An early warning system capable of implementing the early warning method according to any one of claims 1-7, characterized in that, include: The multimodal sensing module is used to synchronously collect time-series data on pressure distribution, vehicle status, and door status within the monitoring area in response to the end of the order service. The signal processing module is used to filter the pressure distribution time series data and perform dynamic interference compensation in combination with the vehicle state time series data to generate a compensated pressure feature sequence. The feature extraction module is used to extract multi-dimensional temporal features representing changes in the state of an item based on the compensated pressure feature sequence and the time-series data of the door state; the multi-dimensional temporal features include at least the net pressure change, the pressure change rate, and the temporal correlation between pressure change and door movement; The intelligent decision-making module is used to input the multi-dimensional temporal features into a pre-trained lightweight machine learning model to obtain the probability value of item loss. The graded early warning execution module is used to compare the probability value of the lost item with a preset decision threshold range and execute corresponding early warning strategies in different grades.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the steps in the warning method as described in any one of claims 1 to 7.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it can implement the steps of the early warning method as described in any one of claims 1 to 7.

Images