Method and device for detecting movement pattern of container

Abstract

The invention relates to a method and an apparatus for identifying a movement pattern of a container. The method is executed by monitoring equipment on the container side and can comprise the steps that a data set comprising a plurality of continuous sample points is collected; segmenting the data set according to the sliding window to obtain a plurality of data blocks corresponding to the sliding window, wherein the data blocks are subsets of the data set; for each data block in the plurality of data blocks, extracting one or more features representing a movement mode of the container, wherein the features comprise one or more of an intensity feature, a direction correlation feature, a zero-crossing rate feature, a frequency and energy feature and an impact and morphological feature; inputting the extracted features into a model to predict a movement mode of the container corresponding to each data block; and identifying a motion pattern of the container corresponding to the data set based on the predicted motion pattern of the container corresponding to each data block. In this way, the movement mode of the container can be recognized with low power consumption and high accuracy.

Description

Technical Field

[0001] This application relates to the monitoring of containers, and more particularly to the recognition of container movement patterns. Background Technology

[0002] Container shipping is a crucial component of modern logistics. Identifying container movement patterns is often necessary during transport. While satellite positioning can be used, the lack of an external power source on the container results in high power consumption. Other solutions include positioning the container at fixed intervals, but this method still consumes significant power while maintaining real-time performance. Alternatively, comparing acceleration to a threshold can determine the movement pattern, but this approach lacks sufficient accuracy. Summary of the Invention

[0003] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form to prepare for the more detailed descriptions that follow.

[0004] To overcome the aforementioned deficiencies in the existing technology, this application provides a method and apparatus for achieving high-accuracy identification of the motion patterns of containers in a low-power manner.

[0005] According to a first aspect of this application, a method for identifying the motion pattern of a container, performed by a monitoring device on the container side, is provided. The method may include: acquiring a dataset comprising multiple consecutive sample points; segmenting the dataset according to a sliding window to obtain multiple data blocks corresponding to the sliding window, each data block being a subset of the dataset; for each of the multiple data blocks: extracting one or more features characterizing the motion pattern of the container, the features including one or more of intensity features, direction-related features, zero-crossing rate features, frequency and energy features, and impact and morphological features; inputting the extracted features into a model to predict the motion pattern of the container corresponding to each data block; and identifying the motion pattern of the container corresponding to the dataset based on the predicted motion pattern of the container corresponding to each data block.

[0006] In some examples of the first aspect, the motion modes may include: rail transport, freight transport, ocean transport, and stationary transport.

[0007] In some examples of the first aspect, the method may further include: waking up the controller module of the monitoring device in response to the container's acceleration exceeding a first threshold for a first duration, and waiting for a second duration before collecting a data set.

[0008] In some examples of the first aspect, the method may further include: waking up a controller module of the monitoring device in response to a third duration elapsed since the previous wake-up, and collecting a data set.

[0009] In some examples of the first aspect, the sample points may include axial accelerations in the x, y, and z axes acquired by the sensor modules of the monitoring device.

[0010] In some examples of the first aspect, strength characteristics can be used to characterize whether the container is stationary, and these characteristics may include the mean and standard deviation of axial acceleration, the resultant acceleration, the mean of the resultant acceleration, and the variance of the resultant acceleration.

[0011] In some examples of the first aspect, impact and morphological characteristics can be used to characterize whether a container is transported by rail or by truck, and may include the maximum value of axial acceleration, the minimum value of axial acceleration, the peak and trough values ​​of axial acceleration, the maximum value of the resultant acceleration, the minimum value of the resultant acceleration, and the peak and trough values ​​of the resultant acceleration.

[0012] In some examples of the first aspect, directional correlation features, zero-crossing rate features, and frequency and energy features can be used to characterize whether a container is in ocean transport, and the zero-crossing rate feature may include the zero-crossing rate of the composite acceleration, the frequency and energy features may include the wavelet variation energy of the axial acceleration, and the directional correlation feature may include the correlation coefficient between axial accelerations.

[0013] In some examples of the first aspect, the method may further include: performing filtering on the data block before extracting features for each data block, wherein the filtering coefficients used in the filtering are updated based on a second threshold.

[0014] In some examples of the first aspect, the filtering is a first-order inertial low-pass filter.

[0015] In some examples of the first aspect, the model is a lightweight model received from the server and deployed at the monitoring device.

[0016] In some examples of the first aspect, the model is a random forest model.

[0017] In some examples of the first aspect, the method may further include: transmitting the identified motion patterns of the containers and the corresponding dataset to a server; receiving an updated model from the server; and deploying the updated model for the identification of the motion patterns of the containers.

[0018] According to a second aspect of this application, a monitoring device for identifying the motion patterns of a container is provided. The monitoring device may include: a sensor module configured to acquire a data set comprising multiple consecutive sample points; and a controller module configured to: segment the data set according to a sliding window to obtain multiple data blocks corresponding to the sliding window, wherein each data block is a subset of the data set; for each of the multiple data blocks: extract one or more features characterizing the motion pattern of the container, including one or more of intensity features, orientation correlation features, zero-crossing rate features, frequency and energy features, and impact and morphological features; input the extracted features into a model to predict the motion pattern of the container corresponding to each data block; and identify the motion pattern of the container corresponding to the data set based on the predicted motion pattern of the container corresponding to each data block.

[0019] In some examples of the second aspect, the motion modes may include: rail transport, freight transport, ocean transport, and stationary.

[0020] In some examples of the second aspect, the monitoring device may further include a timer module that can be configured to wake up the controller module in response to the acceleration of the container from the sensor module exceeding a first threshold for a first duration, and to instruct the sensor module to wait for a second duration before acquiring a data set.

[0021] In some examples of the second aspect, the monitoring device may further include a timer module that can be configured to wake up the controller module in response to a third duration elapsed since the previous wake-up, and to instruct the sensor module to begin collecting a data set.

[0022] In some examples of the second aspect, the sample points may include axial accelerations in the x, y, and z axes acquired by the sensor module.

[0023] In some examples of the second aspect, strength characteristics can be used to characterize whether the container is stationary, and these characteristics may include the mean and standard deviation of axial acceleration, the resultant acceleration, the mean of the resultant acceleration, and the variance of the resultant acceleration.

[0024] In some examples of the second aspect, impact and morphological characteristics can be used to characterize whether a container is transported by rail or by truck, and may include the maximum value of axial acceleration, the minimum value of axial acceleration, the peak and trough values ​​of axial acceleration, the maximum value of the resultant acceleration, the minimum value of the resultant acceleration, and the peak and trough values ​​of the resultant acceleration.

[0025] In some examples of the second aspect, directional correlation features, zero-crossing rate features, and frequency and energy features can be used to characterize whether a container is in ocean transport, and the zero-crossing rate feature may include the zero-crossing rate of the composite acceleration, the frequency and energy feature may include the wavelet variation energy of the axial acceleration, and the directional correlation feature may include the correlation coefficient between axial accelerations.

[0026] In some examples of the second aspect, the controller module may be further configured to perform filtering on the data blocks before extracting features for each data block, wherein the filtering coefficients used in the filtering are updated based on a second threshold.

[0027] In some examples of the second aspect, the filtering is a first-order inertial low-pass filter.

[0028] In some examples of the second aspect, the model is a lightweight model received from the server and deployed at the monitoring device.

[0029] In some examples of the second aspect, the model is a random forest model.

[0030] In some examples of the second aspect, the monitoring device may further include a communication module configured to: transmit the identified movement patterns of the containers and the corresponding data set to a server; and receive an updated model from the server; and the controller module may be further configured to deploy the updated model for the identification of the movement patterns of the containers.

[0031] According to a third aspect of this application, an apparatus for identifying the movement patterns of a container is provided. The apparatus may include: a memory; the memory and a computer program stored in the memory, the computer program being executable by the processor to implement the method as described in the first aspect of this application.

[0032] According to a fourth aspect of this application, a non-transient computer storage medium is provided having computer-executable instructions stored thereon, which, when executed by a computer, cause the computer to perform the method as described in the first aspect of this application.

[0033] Compared to existing technologies, this application triggers the container controller's wake-up based on events and / or time to perform motion pattern recognition, reducing energy consumption. This application utilizes a sliding window to obtain data blocks containing multiple continuous triaxial accelerations, and adds a steady-state operation before data acquisition, using thresholds to update filter coefficients. This effectively removes noise from the data and increases the amount of data used to identify motion patterns, significantly improving the purity of the data source and thus enhancing the accuracy of motion pattern recognition compared to existing solutions. This application extracts features that effectively characterize container motion patterns from the data blocks, effectively distinguishing between ocean and non-ocean transport, and between truck and rail transport within non-ocean transport, improving the accuracy and granularity of motion pattern recognition compared to existing solutions. This application utilizes a random forest model and adjusts parameters such as the number of subtrees, maximum depth, and minimum number of leaf samples to achieve high accuracy in motion pattern recognition while considering model size and feature quantity, and achieves continuous model iteration through cloud collaboration. Attached Figure Description

[0034] The above-described features and advantages of this application can be better understood after reading the following detailed description of the embodiments in conjunction with the accompanying drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0035] Figure 1 This is a schematic architecture diagram illustrating a system for identifying the movement patterns of a container according to various aspects of this application; Figure 2 This is a flowchart illustrating a method for identifying the movement patterns of a container according to various aspects of this application; Figure 3 This is an exemplary block diagram of a computer system suitable for implementing various aspects of this application. Detailed Implementation

[0036] The following specific embodiments illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Although the description of this application will be presented in conjunction with preferred embodiments, this does not mean that the features of this application are limited to these embodiments. On the contrary, the purpose of describing the application in conjunction with embodiments is to cover other options or modifications that may be derived based on the claims of this application. To provide a thorough understanding of this application, many specific details will be included in the following description. This application may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of this application, some specific details will be omitted in the description.

[0037] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0039] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0040] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0041] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0042] As mentioned earlier, identifying container movement patterns via satellite positioning consumes a significant amount of power. Even with real-time positioning, the power consumption remains high when locating containers at fixed intervals. Furthermore, the method of determining container movement patterns by comparing acceleration with a threshold has relatively low accuracy in movement pattern recognition.

[0043] To address this, this application provides a method and system for achieving high-accuracy identification of container motion patterns in a low-power manner. This application triggers the wake-up of the monitoring equipment controller module based on events and / or time to perform motion pattern recognition, ensuring that the monitoring equipment is in a deep sleep mode with power consumption at the microamp level most of the time, significantly reducing energy consumption compared to existing solutions. This application utilizes a sliding window to obtain data blocks containing multiple continuous triaxial accelerations, and adds a steady-state operation before data acquisition, using thresholds to update filter coefficients, effectively removing noise from the data and increasing the amount of data used to identify motion patterns. Compared to existing solutions, this significantly improves the purity of the data source, thereby enhancing the accuracy of motion pattern recognition. This application extracts features that effectively characterize container motion patterns from the data blocks, effectively distinguishing between ocean and non-ocean transport, and between truck and rail transport within non-ocean transport, improving the accuracy and granularity of motion pattern recognition compared to existing solutions. This application utilizes a random forest model and adjusts parameters such as the number of subtrees, maximum depth, and minimum number of leaf samples to achieve high accuracy in motion pattern recognition while considering model size and feature quantity, and achieves continuous model iteration through cloud collaboration.

[0044] The various aspects of this application will now be described with reference to the accompanying drawings.

[0045] First refer to Figure 1 , Figure 1 This is a schematic architecture diagram illustrating a system 100 for identifying the movement patterns of a container according to various aspects of this application. For example... Figure 1 As shown, system 100 may include container 105, monitoring equipment 110 and / or optional server 115.

[0046] In some examples, container 105 and monitoring equipment 110 are physically located together, or monitoring equipment 110 may be referred to as being on the side of container 105. For example, monitoring equipment 110 may be implemented as a separate component from container 105, such as being electrically connected to container 105 via a wired connection. Alternatively, monitoring equipment 110 may be implemented as an integrated component built into container 105. In some examples, monitoring equipment 110 may be communicatively coupled to container 105 via wired or wireless means to achieve bidirectional communication with container 105. Additionally, monitoring equipment 110 may be communicatively coupled to a remotely located server 115 via wired or wireless means to achieve bidirectional communication with server 115. It should be noted that the number of containers 105, monitoring equipment 110, and / or servers 115 may vary, and Figure 1 The examples shown are merely illustrative and not limiting. Those skilled in the art will understand that system 100 may include... Figure 1The different numbers and combinations of containers, monitoring equipment, and servers shown are arranged without departing from the scope of this application. For example, system 100 may include three containers 105 and one monitoring device 110 located therewith, but not server 115.

[0047] In some examples, the monitoring device 110 may include a sensor module 1101, a controller module 1102, a storage module 1103, a communication module 1104, and a timer module 1105, which are communicatively coupled to each other. In some examples, the sensor module 1101 may be configured to acquire status data of the monitoring device 110 and, consequently, status data of the container 105. The number of sensor modules 1101 may be one or more, and examples may include, but are not limited to, a triaxial accelerometer, a triaxial angular velocity sensor, a temperature sensor, a humidity sensor, a positioning sensor, etc. In some examples, the sensor module 1101 may be configured to acquire a dataset comprising multiple consecutive sample points. In this example, the sample points may include axial accelerations in the x, y, and z axes acquired by the triaxial accelerometer. In some examples, the sensor module 1101 may be configured to periodically or with low power sample the motion state of the container under the control of the timer module 1105 and output the raw sampled data. In some examples, sensor module 1101 may be further configured to wait for a second duration before acquiring a dataset comprising multiple consecutive sample points in response to an instruction from timer module 1105.

[0048] In some examples, timer module 1105 can be configured to remain operational to perform timing management on various modules within the monitoring device. For instance, timer module 1105 can be configured to control aspects such as the sampling period, statistical window, sampling sequence, and duration threshold for data sampling by sensor module 1101. As another example, timer module 1105 can be configured to control the wake-up of controller module 1102. In some examples, timer module 1105 can be configured to trigger the switching of controller module 1102 from a sleep state to an active state when sensor data samples collected by sensor module 1101 meet a threshold condition and continue to reach a duration threshold under the timing control of timer module 1105. In some examples, timer module 1105 can be configured to wake up controller module 1102 in response to the acceleration of a container from sensor module 1101 exceeding a first threshold for a first duration, and to instruct sensor module 1101 to wait for a second duration before collecting a data set of multiple consecutive sample points. In some examples, timer module 1105 may be configured to wake up controller module 1102 in response to a third duration elapsed since the previous wake-up, and to instruct sensor module 1101 to begin acquiring a data set.

[0049] In some examples, controller module 1102 is in a sleep state by default. In some examples, controller module 1102 may be configured to control the operation of monitoring device 110 and / or container 105 after being woken up to identify the motion patterns of container 105, such as performing feature calculations, motion pattern recognition, and corresponding status processing or data output operations. For example, controller module 1102 may be configured to interact with one or more of sensor module 1101, storage module 1103, communication module 1104, and timer module 1105 and / or container 105 and server 115 to control the operation of one or more of sensor module 1101, storage module 1103, communication module 1104, and timer module 1105 to identify the motion patterns of container 105.

[0050] In some examples, storage module 1103 may be configured to store data and / or instructions. Examples of data may include state data of container 105 obtained from sensor module 1101, such as axial acceleration, current geographic location, temperature, humidity, etc. Examples of instructions may include lightweight models deployed at monitoring device 110 in the form of processor-executable instructions and / or code stored in storage module 1103, such as a lightweight random forest model for implementing motion pattern recognition of the container.

[0051] In some examples, communication module 1104 may be configured to enable bidirectional communication between monitoring device 110 and container 105 and server 115 via wired and / or wireless means. For example, communication module 1104 may be configured to receive a lightweight model to be deployed from server 115, and / or transmit low-confidence prediction samples to server 115 for further training of the model, and receive an updated model from the server. A detailed description of monitoring device 110 is provided below in conjunction with... Figure 2 The content described.

[0052] In some examples, server 115 may be located remotely from container 105 and monitoring equipment 110, and may be interchangeably referred to as a remote server, cloud server, etc. In some examples, server 115 may include a communication module 1151 and a model training module 1152, which are communicatively coupled to each other. In some examples, communication module 1151 may be configured to enable bidirectional communication between server 115 and monitoring equipment 110 and / or container 105 via wired and / or wireless means. For example, communication module 1151 may be configured to transmit a lightweight model to be deployed to monitoring equipment 110, and / or receive low-confidence prediction samples from monitoring equipment 110 and transmit updated models to monitoring equipment 110. For example, communication module 1151 may be configured to receive collected datasets and corresponding identified motion patterns from one or more monitoring devices 110. In some examples, model training module 1152 may be configured to initially train the model to provide a lightweight model for deployment to monitoring equipment 110. In some examples, the model training module 1152 may be configured to retrain the model based on low-confidence prediction samples received via the communication module 1151 from one or more monitoring devices 110. For example, the model training module 1152 may be configured to retrain the model to generate an updated model based on the recognition results of motion patterns from one or more monitoring devices 110 and the corresponding datasets.

[0053] Next, refer to Figure 2 , Figure 2 A flowchart illustrating a method 200 for identifying the movement pattern of a container according to various aspects of this application is provided. Method 200 can be executed by monitoring equipment on the container side, specifically by, for example... Figure 1 The monitoring equipment 110 shown in the illustration performs the function, and / or in conjunction with container 105 and server 115.

[0054] In some examples, method 200 may include: acquiring a dataset 205 comprising multiple consecutive sample points. In some examples, the motion patterns include: rail transport, freight transport, ocean transport, and stationary motion. In some examples, the sample points may include axial accelerations in the x, y, and z axes, acquired, for example, by a sensor module (such as an accelerometer) of a monitoring device.

[0055] In some examples, method 200 may optionally include: waking up the container's controller module in response to the container's acceleration exceeding a first threshold for a first duration, and waiting for a second duration before acquiring a data set. This application utilizes the low-power characteristics of sensor modules (such as accelerometers) in monitoring equipment as the trigger in an event triggering mechanism. In the example of an accelerometer, ambient noise is filtered by configuring the accelerometer's internal registers to set an acceleration threshold (i.e., the first threshold) and a duration (i.e., the first duration). To reduce overall system power consumption, the controller module remains in a dormant state unless necessary, only being woken up when a valid motion signal is detected. The acceleration threshold, or first threshold, can be used to distinguish between static noise and genuine motion trend changes. In some examples, the acceleration threshold, or first threshold, can be between 0.05g and 0.2g. When the acceleration change is below this threshold, it is considered to correspond to ambient noise or minor disturbances, while when the acceleration change exceeds this threshold, it is considered to correspond to genuine motion trend changes. The first duration can be used to prevent transient interference from triggering the controller module. In some examples, the first duration can be in the range of 1-10 seconds. The controller module is only triggered to wake up when the acceleration change persists for more than this first duration threshold. In some examples, the container is determined to be non-stationary in response to its acceleration exceeding the first threshold for the first duration, i.e., a valid motion signal is detected. When a valid motion signal is detected, the controller module (such as a microcontroller (MCU)) of the monitoring equipment can be woken up, for example, by an interrupt, to perform motion pattern recognition of the container. Under the event-triggered mechanism, a second duration or a steady-state operation phase is entered before acquiring a data set for motion pattern recognition. In some examples, the second duration can be 2 ± 0.5 seconds. During the second duration, no data sampling is performed to avoid transient signal interference from potential initial wake-up events. After the second duration (e.g., 2 seconds), a data acquisition phase begins, during which a data set including multiple consecutive sample points is acquired. In some examples, after a 2-second steady-state operation phase, an 8-second data acquisition phase begins. In some examples, during the data acquisition phase, the data obtained by the accelerometer is continuously sampled at a sampling frequency of 30 Hz to capture a dataset containing 240 consecutive axial accelerations for subsequent motion pattern recognition.

[0056] In some examples, method 200 may optionally include waking the controller module of the monitoring device in response to a third duration elapsed since the previous wake-up, and acquiring a data set. In some examples, even when no valid motion signal is detected, the sensor module of the monitoring device may wake the controller module in response to a third duration elapsed since the previous wake-up of the controller module, and begin acquiring a data set including multiple consecutive sample points. This method of waking the controller module can be referred to as a time-based mechanism. In this way, even when the container is stationary, the controller module can be periodically woken up to identify the container's motion patterns. Thus, method 200's identification of container motion patterns covers both stationary and non-stationary states. Furthermore, the combination of the event-triggered mechanism and the time-triggered mechanism ensures that the monitoring device consumes power at the microamp level most of the time, i.e., in a deep sleep mode (where only the sensor module operates, and the controller module does not), significantly reducing energy consumption.

[0057] In some examples, method 200 may include segmenting the dataset according to a sliding window to obtain multiple data blocks corresponding to the sliding window, the data blocks being subsets 210 of the dataset. In some examples, the sliding window may be configured to have an overlap rate and a window width. In an example where a dataset containing 240 continuous axial accelerations is captured at a sampling frequency of 30Hz during an 8-second data acquisition phase after a 2-second stabilization operation phase, the overlap rate of the sliding window may be configured to 50%, and the window width may be configured to 2 seconds. In this example, each data block may include 60 continuous axial accelerations, and there are 30 overlapping samples between consecutive data blocks. This application does not identify motion patterns based on a single acceleration value, but rather based on data blocks containing multiple continuous sample points, and determines the motion pattern corresponding to the entire dataset based on the motion pattern identified for each data block. This significantly increases the amount of data used to identify motion patterns, effectively avoids repeated jumps in the identified motion patterns, and ensures the stability and accuracy of motion pattern recognition.

[0058] In some examples, method 200 may optionally include: performing data preprocessing after obtaining the data block (in... Figure 2 (Not shown in the image). In some examples, the data block is filtered before it is acquired and before subsequent operations are performed on it, with the filter coefficients updated based on a second threshold. In some examples, the filtering may be a first-order inertial low-pass filter, which can be expressed as follows: (1) in This is the current filter output value. For sampled values, The previous filter output value is used, and the current filter output value is a weighted sum of the current sampled value and the previous filter output value. The filter coefficients are updated based on the following formula: (2) Where Δ is the difference between the current filter output value and the previous filter output value, i.e. . The second threshold, referred to above, is an empirical threshold used to distinguish between "signal noise / minor jitter" and "real motion pattern changes," and its value can be between 0.01 and 0.05. When the difference between the current filtered output value and the previous filtered output value is lower than this threshold, it indicates that the filtered output mainly contains static noise. When the difference between the current filtered output value and the previous filtered output value exceeds this threshold, it indicates that the container exhibits a real motion trend change. This is a hyperparameter, based on the characteristics of a first-order inertial filter, and its value can be 0.5 to control the filter coefficients. . It is updated only when the above expression is positive, i.e. Only when Δ is greater than The values ​​are updated only occasionally (generally greater than 0 and less than 1), and are updated as the sliding window moves, thus varying from data block to data block. This application utilizes a sliding window to obtain data blocks containing multiple continuous triaxial accelerations, and adds a steady-state operation before data acquisition and uses a threshold to update the filter coefficients, effectively removing noise from the data, significantly improving the purity of the data source, and thus improving the accuracy of motion pattern recognition.

[0059] In some examples, method 200 may include: for each of a plurality of data blocks: extracting one or more features 2151 characterizing the motion pattern of the container. In some examples, these features may include one or more of intensity features, orientation-related features, zero-crossing rate features, frequency and energy features, and impact and morphological features.

[0060] In some examples, intensity features can be used to characterize whether a container is stationary, and these can include the mean and standard deviation of axial acceleration, the resultant acceleration, the mean of the resultant acceleration, and the variance of the resultant acceleration. In other examples, intensity features can be used to quantify the overall energy and intensity of motion, effectively distinguishing between a motion state with significant vibration and a stationary state.

[0061] In some examples, impact and morphological features can be used to characterize whether a container is being transported by rail or truck. These features may include the maximum and minimum values ​​of axial acceleration, the peak and trough values ​​of axial acceleration, and the maximum, minimum, and peak and trough values ​​of the resultant acceleration. In other examples, impact and morphological features can be used to capture instantaneous, severe acceleration impacts within a sliding window. Such features can be used to identify bumps caused by uneven road surfaces during land transport, or characteristic impacts when a train passes over a track joint.

[0062] In some examples, directional correlation features, zero-crossing rate features, and frequency and energy features can be used to characterize whether a container is in ocean transport. Zero-crossing rate features can include the zero-crossing rate of the composite acceleration. Zero-crossing rate features can indicate the number of times a signal crosses its mean line. Zero-crossing rate features can be used to calculate the vibration frequency by the number of times the signal crosses its mean line. Since ocean transport involves large-amplitude, low-frequency swaying, the zero-crossing rate is relatively low, while in truck transport, high-frequency jitter from sources such as engines and road surfaces produces a higher zero-crossing rate. Frequency and energy features can include wavelet variations in the energy of axial acceleration. Frequency and energy features can be used to analyze the energy distribution of the motion signal across different frequency bands. This is key to distinguishing ocean transport from non-ocean transport, as the energy in ocean transport is mainly concentrated in the extremely low-frequency region, while the vibration spectrum of land transport is more widely distributed, containing more mid- and high-frequency components. Directional correlation features can include correlation coefficients between axial accelerations. Directional correlation features (such as the correlation coefficients between pairs of axial accelerations) can be used to describe the coupling relationship of motion in different directions. In one example, the swaying of the hull in ocean transport leads to a strong correlation in specific axial accelerations, which is significantly different from the vibration patterns in land transport.

[0063] In some examples, the mean values ​​of the axial accelerations in the x, y, and z directions can be determined using the following formula: (3) (4) (5) in , , Let be the axial acceleration in the x, y, and z directions obtained at the k-th sampling point in a sliding window; n is the number of sampling points in each sliding window.

[0064] In some examples, the composite acceleration can be defined as the vector consisting of the axial accelerations in the x, y, and z directions collected by the accelerometer at the same sampling moment. The magnitude of the composite acceleration can be defined as the magnitude of the composite acceleration vector, reflecting the overall motion intensity of the container at that sampling moment. The magnitude and mean of the composite acceleration can be determined using the following formula: (6) (7) in Let the magnitude of the resultant acceleration of the axial accelerations in the x, y, and z directions obtained at the k-th sampling point within a sliding window be... Substituting into equation (7) yields the mean value of the resultant acceleration. .

[0065] In some examples, the standard deviation of axial acceleration can be determined using the following formula: (8) (9) (10) In some examples, the variance of the resultant acceleration can be determined using the following formula: (11) In some examples, the peak and trough values ​​of acceleration can be determined using the following formula: (12) In some examples, the correlation coefficient between axial accelerations can be determined using the following formula: (13) (14) (15) In some examples, the signal is in any time or space domain. The binary wavelet decomposition can be expressed by the following formula: (16) Its total energy is (17) The energy of the approximate signal of layer j and the detail signals of each layer are selected as features, and the feature vector is constructed as follows: (18) Continue to refer to Figure 2In some examples, method 200 may include: inputting the extracted features into a model to predict the motion pattern of the containers corresponding to each data block 2152. In some examples, the model may be a random forest model. In some examples, the model may be a trained lightweight random forest model and deployed to a monitoring device. In some examples, the trained lightweight random forest model outputs a classification result corresponding to each data block, i.e., the motion pattern of the containers, for the input features corresponding to each data block. As is known to those skilled in the art, random forest models are built on a decision tree basis. Each decision tree acts as a classifier and outputs a classification result, and then, based on the multiple classification results output by multiple classifiers, the random forest model uses a voting mechanism to determine the final classification result of the model for the current data block. In some examples, the random forest model contains three decision trees, namely three classifiers A, B, and C, and classifier A outputs a classification result of freight transport, classifier B outputs a classification result of rail transport, and classifier C outputs a classification result of freight transport. In this example, the random forest model determines the classification result of freight transport for the current data block based on a voting mechanism.

[0066] In some examples, method 200 may include: identifying motion patterns 220 of containers corresponding to the dataset based on the predicted motion patterns of containers corresponding to each data block. Similar to the use of a voting mechanism to determine the classification result of the final output in the random forest model described above, step 220 may include using a voting mechanism to determine the motion patterns of containers corresponding to a data combination comprising multiple data blocks, based on the determined motion patterns of containers corresponding to each data block. In an example where a dataset containing 240 consecutive axial accelerations is captured at a sampling frequency of 30 Hz during an 8-second data acquisition phase, with a sliding window overlap rate configured to 50% and a window width configured to 2 seconds, the dataset comprises 7 data blocks, each containing 60 consecutive axial accelerations, and with 30 overlapping samples between consecutive data blocks. In this example, assuming the motion pattern corresponding to the first data block is determined to be freight transport, the motion pattern corresponding to the second data block is determined to be rail transport, the motion pattern corresponding to the third data block is determined to be rail transport, the motion pattern corresponding to the fourth data block is determined to be rail transport, the motion pattern corresponding to the fifth data block is determined to be rail transport, the motion pattern corresponding to the sixth data block is determined to be rail transport, and the motion pattern corresponding to the seventh data block is determined to be freight transport, then step 220 may include using a voting mechanism to determine that the motion pattern of the container corresponding to the data set including these 7 data blocks is rail transport. This application extracts features that effectively characterize the motion pattern of the container from the data blocks, effectively distinguishing between ocean transport and non-ocean transport, and between freight transport and rail transport in non-ocean transport, thus improving the accuracy and granularity of motion pattern recognition.

[0067] In some examples, method 200 may optionally include: transmitting the identified movement patterns of the containers and the corresponding collected data set to server 230. Figure 2 (Not shown in the image). In some aspects, the confidence level of the monitoring equipment's identification of motion patterns based on the deployed model may be low. In this case, the monitoring equipment may transmit the identified motion patterns and the corresponding collected data set to a server. In some examples, method 200 may optionally include: receiving an updated model from the server, and deploying the updated model for the identification of container motion patterns 240 (…). Figure 2 (Not shown in the image). As described above, the monitoring equipment can receive an updated model retrained from the server based on the recognition results and dataset of motion patterns previously provided by the monitoring equipment. This updated model can be deployed at the monitoring equipment to replace the original model for subsequent recognition of container motion patterns.

[0068] Those skilled in the art should understand that, in conjunction with the above text Figure 2The descriptions of the range of values ​​for parameters such as sampling frequency, sampling duration, acceleration threshold, first duration, filter coefficient, empirical threshold, hyperparameter, and other aspects in the described embodiments are merely exemplary and are not intended to limit the scope of the application.

[0069] As described above, the model in step 220 is a trained lightweight model and is deployed to the monitoring device. The training and deployment of the model are then described.

[0070] The training and deployment of the model first includes the data collection and labeling phase. To ensure the model's generalization ability, the collected data must include at least 5000 valid continuous triaxial acceleration data points for each category. Categories can include, but are not limited to, real-world scenarios such as equipment type, route, weather, and sea state. The raw data is then cleaned, including removing missing, erroneous, and non-standard data. Next, based on auxiliary information such as GPS tracks and transportation logs, the collected data is labeled with time periods, and the dataset is divided into a training set (80%) and a validation set (20%).

[0071] Model training and deployment then includes an initial training phase. An initial random forest model is trained using all candidate features generated from the data processed above. The goal of this phase is to maximize the model's discriminative power, providing a comprehensive and unbiased evaluation of the contribution of all features, rather than pursuing immediate deployment performance. In some examples, the Gini importance method can be used to extract the importance scores of each feature from the model. As is known to those skilled in the art, the basic idea of ​​the Gini importance method is that in a decision tree, the importance of a feature depends on how much "purity boost" (i.e., how much impurity decreases) it brings to that node when it splits. If a feature makes the data categories more "pure" after splitting, then its contribution is large. The total importance of a feature in a tree is the sum of the "purity boost" it contributes to all split nodes in that tree. The importance scores of all features are normalized so that their sum is 1. In some example sets, the `feature_importance_` library function in scikit-learn can be used to calculate the importance values ​​of all features. Subsequently, all candidate features are sorted from highest to lowest importance score. An iterative validation approach was adopted, starting with the features ranked highest in importance. While gradually increasing the number of features (e.g., first selecting the top 10, then the top 12, 15, 20, etc.), the curves representing "accuracy-feature count" and "model volume-feature count" were analyzed. The final selected number of features, N, was determined as the value corresponding to the "inflection point" of the curve.

[0072] Model training and deployment then include the final model training phase. In some examples, the optimal feature subset determined in the previous step is used to train the final random forest model for deployment. First, initial hyperparameter tuning is performed by controlling the number of subtrees (n_estimators) and the maximum depth (max_depth). Specifically, the optimal number of subtrees and maximum depth can be selected based on heatmaps of subtree number, maximum depth, and accuracy, and heatmaps of subtree number, maximum depth, and model size. Following the initial hyperparameter tuning is a fine-tuning step. In this step, the optimal number of subtrees and maximum depth selected in the previous step are fixed, and the minimum number of leaf samples, the maximum number of features considered when splitting nodes, and the minimum number of samples for splitting internal nodes are adjusted. Based on the results, facet plots of accuracy and model size are drawn, and the optimal hyperparameters are selected using the facet plots. The model is then saved.

[0073] Finally, model training and deployment include the model deployment and prediction phases. In some examples, a lightweight model that has been trained and validated is converted into C code and deployed to the monitoring equipment of a container.

[0074] Additionally or alternatively, model training and deployment may further include model maintenance and update phases. To maintain model effectiveness and further improve its predictive accuracy, the server may include a model training module and a communication module. The monitoring device may upload low-confidence prediction samples to the server, which periodically retrains the model. Validated, retrained models can be pushed to the monitoring device via methods such as over-the-air (OTA) firmware updates to enable continuous model iteration. In some examples, the server may include a communication module and a model training module, which are communicatively coupled to each other. In some examples, the communication module may be configured to enable bidirectional communication between the server and the monitoring device and / or container via wired and / or wireless means. For example, the communication module may be configured to transmit a lightweight model to be deployed to the monitoring device, and / or receive low-confidence prediction samples from the monitoring device and transmit updated models to the monitoring device. In some examples, the model training module may be configured to initially train the model to provide a lightweight model for deployment to the monitoring device. In some examples, the model training module may be configured to retrain the model based on low-confidence prediction samples received from the monitoring device via the communication module. For example, the model training module can be configured to retrain the model to generate an updated model based on the recognition results of motion patterns from one or more monitoring devices and the corresponding dataset.

[0075] This application utilizes a random forest model and adjusts parameters such as the number of subtrees, maximum depth, and minimum number of leaf samples to achieve high accuracy in motion pattern recognition while taking into account model size and feature quantity, and achieves continuous model iteration through cloud collaboration.

[0076] Those skilled in the art should understand that the descriptions of parameters such as the number of samples, the ratio of training set to validation set, and the number of features, as well as other aspects, in the above text are merely exemplary and this application does not impose any limitations on them.

[0077] Next, refer to Figure 3 , Figure 3 This is an exemplary block diagram of a computer system 012 suitable for implementing various aspects of this application. Figure 3 The computer system 012 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0078] like Figure 3 As shown, the computer system 012 is represented in the form of a general-purpose computing device. The components of the computer system 012 may include, but are not limited to: one or more processors or processing units 016, system memory 028, and a bus 018 connecting different system components (including system memory 028 and processing unit 016).

[0079] Bus 018 represents one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0080] Computer system 012 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer system 012, including volatile and non-volatile media, removable and non-removable media.

[0081] System memory 028 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 030 and / or cache memory 032. Computer system 012 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 034 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 3 Not shown; usually referred to as a "hard drive"). Although Figure 3As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 018 via one or more data media interfaces. Memory 028 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0082] A program / utility 040 having a set (at least one) of program modules 042 may be stored, for example, in memory 028. Such program modules 042 include—but are not limited to—an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 042 typically perform the functions and / or methods described in the embodiments of this application.

[0083] Computer system 012 can also communicate with one or more external devices 014 (e.g., keyboard, pointing device, display 024, etc.). In this application, computer system 012 communicates with external radar equipment, and can also communicate with one or more devices that enable users to interact with the computer system 012, and / or with any device that enables the computer system 012 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 022. Furthermore, computer system 012 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) through network adapter 020. As shown, network adapter 020 communicates with other modules of computer system 012 through bus 018. It should be understood that, although... Figure 3 As not shown in the diagram, it can be used in conjunction with computer system 012 with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0084] The processing unit 016 executes various functional applications and data processing by running programs stored in the system memory 028, such as implementing the method flow provided in the embodiments of this application.

[0085] The aforementioned computer program can be stored in a computer storage medium, meaning the computer storage medium is encoded with a computer program. When executed by one or more computers, this program causes the one or more computers to perform the method flows and / or apparatus operations shown in the above embodiments of this application. For example, the method flows provided in the embodiments of this application can be executed by one or more processors.

[0086] With the advancement of time and technology, the meaning of "medium" has become increasingly broad. The dissemination of computer programs is no longer limited to tangible media; they can also be downloaded directly from the internet. Any combination of one or more computer-readable media can be used.

[0087] A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of a computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0088] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0089] The program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0090] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can 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 remote computers, the remote computer can 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 it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0091] Those skilled in the art will understand that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this hardware-software interchangeability, the various illustrative components, blocks, modules, circuits, and steps are described above in a generalized manner in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in different ways for each specific application, but such implementation decisions should not be construed as departing from the scope of this application.

[0092] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or performed using a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor may be a microprocessor, but in alternatives, it may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors cooperating with a DSP core, or any other such configuration.

[0093] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor such that the processor can read and write information to / from the storage medium. In an alternative, the storage medium may be integrated into the processor. The processor and storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and storage medium may reside as discrete components in the user terminal.

[0094] The prior description of this application is provided to enable any person skilled in the art to make or use this application. Various modifications to this application will be apparent to those skilled in the art, and the general principles defined herein can be applied to other variations without departing from the spirit or scope of this application. Therefore, this application is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying the movement patterns of a container, performed by monitoring equipment on the container side, comprising: Collect a dataset that includes multiple consecutive sample points; The data set is segmented according to a sliding window to obtain multiple data blocks corresponding to the sliding window, wherein the data blocks are subsets of the data set; For each of the plurality of data blocks: Extract one or more features characterizing the motion pattern of the container, including one or more of the following: intensity features, direction correlation features, zero-crossing rate features, frequency and energy features, and impact and morphological features. The extracted features are input into the model to predict the movement pattern of the container corresponding to each data block; and The movement patterns of containers corresponding to the data set are identified based on the predicted movement patterns of containers corresponding to each data block.

2. The method as described in claim 1, characterized in that, The movement modes include: rail transport, freight transport, ocean transport, and stationary transport.

3. The method as described in claim 1, characterized in that, The method further includes: The controller module of the monitoring device is activated in response to the container's acceleration exceeding a first threshold for a first duration. Wait for a second duration before collecting the data set.

4. The method as described in claim 1, characterized in that, The method further includes: The controller module of the monitoring device is woken up in response to the third duration since the previous wake-up, and Collect the aforementioned data set.

5. The method as described in claim 2, characterized in that, The sample points include axial accelerations in the x, y, and z axes obtained by the sensor modules of the monitoring equipment.

6. The method as described in claim 5, characterized in that, The strength characteristics are used to characterize whether the container is stationary, including the mean and standard deviation of axial acceleration, the resultant acceleration, the mean of the resultant acceleration, and the variance of the resultant acceleration.

7. The method as described in claim 5, characterized in that, The impact and morphological characteristics are used to characterize whether the container is transported by rail or truck, and include the maximum value of axial acceleration, the minimum value of axial acceleration, the peak and valley values ​​of axial acceleration, the maximum value of the combined acceleration, the minimum value of the combined acceleration, and the peak and valley values ​​of the combined acceleration.

8. The method as described in claim 5, characterized in that, The directional correlation feature, the zero-crossing rate feature, and the frequency and energy feature are used to characterize whether the container is in ocean transport. The zero-crossing rate feature includes the zero-crossing rate of the composite acceleration, the frequency and energy feature includes the wavelet variation energy of the axial acceleration, and the directional correlation feature includes the correlation coefficient between axial accelerations.

9. The method as described in claim 1, characterized in that, The method further includes: Before extracting features for each data block, filtering is performed on the data block, and the filtering coefficients used in the filtering are updated based on a second threshold.

10. The method as described in claim 9, characterized in that, The filtering is a first-order inertial low-pass filter, and is expressed by the following formula: in This is the current filter output value. For sampled values, This is the output value of the previous filter. The filter coefficients are updated based on the following formula: Where Δ is the difference between the current filter output value and the previous filter output value, i.e. , The second threshold value ranges from 0.01 to 0.

05. This is a hyperparameter with a value of 0.5, where It is only updated when the above expression is positive.

11. The method as described in claim 1, characterized in that, The model is a lightweight model received from the server and deployed on the monitoring device.

12. The method as described in claim 11, characterized in that, The method further includes: The identified movement patterns of the containers and the corresponding data sets are transmitted to the server; Receive the updated model from the server; and The updated model is deployed for the identification of container movement patterns.

13. A monitoring device for identifying the movement patterns of containers, comprising: The sensor module is configured to: Collect a dataset that includes multiple consecutive sample points; The controller module is configured as follows: The data set is segmented according to a sliding window to obtain multiple data blocks corresponding to the sliding window, wherein the data blocks are subsets of the data set; For each of the plurality of data blocks: Extract one or more features characterizing the motion pattern of the container, including one or more of the following: intensity features, direction correlation features, zero-crossing rate features, frequency and energy features, and impact and morphological features. The extracted features are input into the model to predict the movement pattern of the container corresponding to each data block; and The movement patterns of containers corresponding to the data set are identified based on the predicted movement patterns of containers corresponding to each data block.

14. An apparatus for identifying the movement patterns of a container, comprising a processor, a memory, and a computer program stored in the memory, characterized in that, The computer program can be executed by the processor to implement the method according to any one of claims 1-12.

15. A non-transient computer storage medium having stored thereon computer-executable instructions, which, when executed by a computer, cause the computer to perform the operation of the method as described in any one of claims 1-12.

Images