A dump truck and a method for detecting the loading status of the dump truck's cargo box.

Abstract

This invention discloses a method for detecting the loading status of a dump truck and its cargo compartment, comprising: acquiring real-time images of the cargo compartment captured by a camera module and inertial data detected by an inertial sensor; determining whether the real-time cargo compartment image is a valid image based on its quality; if not, entering an inertial detection mode, in which the current loading status of the cargo compartment is determined based on the inertial data and an inertial detection model; if yes, entering a fusion model detection mode, in which the current loading status of the cargo compartment is determined based on the real-time cargo compartment image, and the inertial data and the current loading status of the cargo compartment are input into the inertial detection model to update the inertial detection model; the technical solution of this application solves the technical problems of camera obstruction, visual detection failure in low-light environments, poor generalization of offline unified inertial detection models, and the infeasibility of personalized offline modeling.

Description

Technical Field

[0001] This invention belongs to the field of dump truck inspection technology, and relates to dump trucks and methods for detecting the loading status of dump truck bodies. Background Technology

[0002] Dump trucks are widely used in construction sites, mining areas, logistics transportation, and other scenarios. Accurate detection of the empty and loaded status of the cargo box is a crucial foundation for vehicle scheduling, load monitoring, and safe operation. Currently, existing technologies often use cameras combined with visual algorithms to detect empty and loaded cargo boxes. However, this approach has significant technical drawbacks: cameras are easily obstructed by debris or human intervention, and image acquisition is poor in low-light environments such as at night, causing visual algorithms to fail to effectively identify the cargo box status and resulting in detection interruptions. Meanwhile, some technologies attempt to use offline-trained unified IMU models for empty and loaded detection, but these models suffer from severe generalization defects and are difficult to adapt to complex real-world working conditions.

[0003] The core reason for the poor generalization of the unified IMU empty-load detection model trained offline is that the IMU signal characteristics are affected by multiple dimensions of factors, and the differences between these factors are significant: First, the inherent attributes of the vehicles themselves differ. Different models, loads, curb weights, and vehicle conditions due to differences in service life and maintenance levels directly change the characteristics of the inertial signals collected by the IMU. Second, there are differences in cargo-related factors. Different cargoes carried by dump trucks, such as sand, earth, and steel, have different densities, stacking patterns, and center of gravity distributions, resulting in differences in the mechanical characteristics of the cargo box load, which affects the acquisition of IMU vibration and attitude change signals. Third, there are differences in external road conditions. Vibration interference from complex road conditions such as bumpy dirt roads and sloping surfaces in construction sites and mining areas will be superimposed on the load signal, resulting in significant differences in IMU characteristics of the same vehicle under different road conditions. Fourth, there are differences in IMU installation locations. The signal acquisition effect is different at different installation locations such as the cab, the bottom of the cargo box, and the crossbeam of the frame, and the unified model cannot take into account all installation scenarios.

[0004] Training a personalized IMU model offline for each dump truck individually would require significant manpower and resources to collect full-condition samples for each vehicle. Furthermore, vehicle conditions change dynamically over time, making it impossible to update the personalized offline model in real time. This approach is complex, costly, and impractical in real-world applications. Therefore, there is an urgent need for a detection device and method that can adapt to complex operating conditions, achieve full-process empty / heavy load detection, and adaptively adapt to the personalized characteristics of each vehicle. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for detecting the loading status of dump trucks and their cargo compartments, solving the technical problems of camera obstruction, visual detection failure in low-light environments, poor generalization of offline unified inertial detection models, and the infeasibility of personalized offline modeling.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for detecting the loading status of a dump truck's cargo box, the dump truck including a vehicle body, a camera module, an inertial sensor, and a control device, the vehicle body including a cargo box, the camera module disposed on the vehicle body for detecting images inside the cargo box, the inertial sensor disposed on the vehicle body for detecting inertial data of the vehicle body, and the control device electrically connected to the camera module and the inertial sensor, the method for detecting the loading status of the dump truck's cargo box including: Acquire real-time images of the carriage captured by the camera module and inertial data detected by the inertial sensor; Based on the quality of the real-time carriage image, it is determined whether the image is valid; If not, then enter the inertial detection mode. In the inertial detection mode, the current loading status of the carriage is determined based on the inertial data and the inertial detection model. The loading status includes empty and heavy load. If so, the system enters the fusion model detection mode. In this mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the model.

[0008] Optionally, the dump truck further includes a positioning module disposed on the vehicle body, and the control device is electrically connected to the positioning module; The steps for acquiring real-time images of the carriage interior captured by the camera module and inertial data detected by the inertial sensor include: Obtain the current position data of the vehicle body detected by the positioning module; Based on the current position data of the vehicle body, the sampling frequency of the camera module and the detection frequency of the inertial sensor are matched; The camera module captures real-time images of the carriage according to the sampling frequency, and the inertial sensor detects inertial data according to the detection frequency.

[0009] Optionally, the step of matching the sampling frequency of the camera module and the detection frequency of the inertial sensor based on the current position data of the vehicle body includes: Based on the current position data of the vehicle body, determine the current position of the vehicle body on the travel trajectory; Based on the current position of the travel trajectory, the sampling frequency of the camera module and the detection frequency of the inertial sensor are matched.

[0010] Optionally, the step of matching the sampling frequency of the camera module and the detection frequency of the inertial sensor based on the current position of the travel trajectory includes: When the travel trajectory is in a flat position, the first sampling frequency and the first detection frequency are matched; When the travel trajectory is in a downhill position, the second sampling frequency and the second detection frequency are matched; Wherein, the second sampling frequency is greater than the first sampling frequency, and the second detection frequency is greater than the first detection frequency.

[0011] Optionally, the step of acquiring real-time images of the carriage interior captured by the camera module and inertial data detected by the inertial sensor includes: The change in inertial data is obtained based on the inertial data detected by the inertial sensor in two recent consecutive measurements. When the change exceeds the first set value, the camera module is controlled to increase the current sampling frequency, and the inertial sensor is controlled to increase the current detection frequency. When the change is less than or equal to the second set value, the camera module is controlled to reduce the current sampling frequency, and the inertial sensor is controlled to reduce the current detection frequency. When the second set value ≤ the change amount ≤ the first set value, the camera module is controlled to maintain the current sampling frequency, and the inertial sensor is controlled to maintain the current detection frequency.

[0012] Optionally, the dump truck further includes a communication module, which is electrically connected to the control device; If so, the system enters the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the inertial detection model. The system also includes controlling the communication module to upload the inertial data, the loading status of the current carriage, and the updated inertial detection model to the cloud.

[0013] Optionally, the dump truck further includes a communication module, which is electrically connected to the control device; If so, then enter the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the inertial detection model. The steps include: If so, the system enters the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image. The inertial data, the loading status of the current carriage, and the inertial data and loading status received through the communication module are input into the inertial detection model to update the inertial detection model.

[0014] The present invention also provides a dump truck, the dump truck including a vehicle body, a camera module, an inertial sensor and a control device, the vehicle body including a cargo box, the camera module being disposed on the vehicle body for detecting images inside the cargo box, the inertial sensor being disposed on the vehicle body for detecting inertial data of the vehicle body, and the control device being electrically connected to the camera module and the inertial sensor; The control device includes a memory, a processor, and a loading status detection program for the dump truck's cargo compartment stored in the memory and executable on the processor. The loading status detection program for the dump truck's cargo compartment is configured to implement the following steps of a loading status detection method for the dump truck's cargo compartment: Acquire real-time images of the carriage captured by the camera module and inertial data detected by the inertial sensor; Based on the quality of the real-time carriage image, it is determined whether the image is valid; If not, then enter the inertial detection mode. In the inertial detection mode, the current loading status of the carriage is determined based on the inertial data and the inertial detection model. The loading status includes empty and heavy load. If so, the system enters the fusion model detection mode. In this mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the model.

[0015] Optionally, the dump truck further includes a positioning module disposed on the vehicle body, and the control device is electrically connected to the positioning module; and / or, The dump truck also includes a communication module mounted on the vehicle body, and the control device is electrically connected to the communication module.

[0016] Optionally, the positioning module includes a UWB module.

[0017] In the technical solution of this invention, the method for detecting the loading status of the dump truck's cargo compartment includes: acquiring real-time images of the cargo compartment captured by a camera module and inertial data detected by an inertial sensor; determining whether the real-time cargo compartment image is a valid image based on its quality; if not, entering an inertial detection mode, determining the current loading status of the cargo compartment based on the inertial data and an inertial detection model, wherein the loading status includes empty and heavy load; if yes, entering a fusion model detection mode, determining the current loading status of the cargo compartment based on the real-time cargo compartment image, and inputting the inertial data and the current loading status of the cargo compartment into the inertial detection model to update the inertial detection model; the technical solution of this application solves the technical problems of camera obstruction, visual detection failure in low-light environments, poor generalization of offline unified inertial detection models, and the infeasibility of personalized offline modeling. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 A flowchart illustrating an embodiment of the loading status detection method for the cargo compartment of a dump truck provided by the present invention; Figure 2 is a schematic diagram of the overall hardware and software architecture of an embodiment of the dump truck detection system of the present invention; Figure 3 is a schematic diagram of the software layer functional module architecture of an embodiment of the dump truck detection system of the present invention; Figure 4 This is a flowchart illustrating a specific embodiment of the method for detecting the loading status of the dump truck's cargo compartment provided by the present invention. Figure 5 This is a flowchart illustrating another specific embodiment of the method for detecting the loading status of the dump truck's cargo compartment provided by the present invention. Figure 6 This is a flowchart illustrating another specific embodiment of the method for detecting the loading status of the dump truck's cargo compartment provided by the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] This invention provides a method for detecting the loading status of the cargo compartment of a dump truck, such as... Figure 1 As shown, the dump truck includes a vehicle body, a camera module, an inertial sensor, and a control device. The vehicle body includes a cargo box. The camera module is located on the vehicle body for detecting images inside the cargo box. The inertial sensor is located on the vehicle body for detecting inertial data of the vehicle body. The control device is electrically connected to the camera module and the inertial sensor. The method for detecting the loading status of the dump truck's cargo box includes: Step S10: Obtain real-time images of the carriage taken by the camera module and inertial data detected by the inertial sensor; Step S20: Determine whether the real-time carriage image is a valid image based on its quality; Step S30: If not, then enter the inertial detection mode. In the inertial detection mode, the current loading status of the carriage is determined according to the inertial data and the inertial detection model. The loading status includes empty and heavy load. Step S40: If yes, then enter the fusion model detection mode. In the fusion model detection mode, determine the current loading status of the carriage based on the real-time carriage image, and input the inertial data and the current loading status of the carriage into the inertial detection model to update the inertial detection model.

[0022] In the technical solution of this invention, the method for detecting the loading status of the dump truck's cargo compartment includes: acquiring real-time images of the cargo compartment captured by a camera module and inertial data detected by an inertial sensor; determining whether the real-time cargo compartment image is a valid image based on its quality; if not, entering an inertial detection mode, determining the current loading status of the cargo compartment based on the inertial data and an inertial detection model, wherein the loading status includes empty and heavy load; if yes, entering a fusion model detection mode, determining the current loading status of the cargo compartment based on the real-time cargo compartment image, and inputting the inertial data and the current loading status of the cargo compartment into the inertial detection model to update the inertial detection model; the technical solution of this application solves the technical problems of camera obstruction, visual detection failure in low-light environments, poor generalization of offline unified inertial detection models, and the infeasibility of personalized offline modeling.

[0023] In practice, determining whether a real-time carriage image is valid typically involves a comprehensive assessment of three dimensions: image clarity, carriage area proportion, and target visibility. For example, the Laplacian variance algorithm is first used to calculate the image's blurriness. If the variance value is below a set clarity threshold, the image is directly deemed invalid. If the clarity meets the requirements, a pre-trained semantic segmentation model is used to identify the carriage area in the image and calculate its proportion of the entire image. If the proportion is below a set proportion threshold, it indicates that the camera angle is off or the image is partially obscured by foreign objects, and it is also deemed invalid. Finally, the visibility proportion inside the carriage is checked. If dust or cargo accumulation obscures the view beyond a set visibility proportion threshold, the image is also deemed invalid. Only when all three dimensions meet the requirements is the image deemed valid and enters the fusion model detection mode.

[0024] In inertial detection mode, the inertial detection model first preprocesses the input inertial data, removing instantaneous abnormal interference values ​​caused by sudden braking and speed bumps. Then, it extracts the signal mean, variance, and peak value in the time domain, as well as the energy proportion of different frequency bands in the frequency domain. These features are then input into the trained classification model, which directly outputs the detection result of whether the current truck bed is empty or loaded. The entire detection process can be completed on the vehicle, with a response time of no more than 100 milliseconds, fully meeting the real-time detection requirements of dump truck operation scenarios.

[0025] In the fusion model detection mode, the visual loading status determination itself has high accuracy. When the image is valid, the visual detection result can be used as the true label of the inertial detection model. Each valid detection will automatically add the corresponding inertial data and label to the model's update dataset. When the sample size of the update dataset reaches the set batch threshold, the model will automatically use a small learning rate for incremental learning, quickly fine-tuning the original model parameters to complete the update of its own parameters. There is no need for manual sample labeling or retraining from scratch. The entire update process is automatically completed in the background of the vehicle terminal, without consuming too many computing resources or affecting the operation of normal detection functions.

[0026] This online update mechanism allows the inertial detection model to continuously adapt to the individual characteristics of the current vehicle. For example, as the vehicle ages, the aging of components causes the overall vibration characteristics to drift. The model can automatically learn this drift change through continuous incremental updates. When a new vehicle replaces inertial sensors with those installed in different locations, there is no need for re-manual annotation and training. After only a few effective visual inspections, the model can automatically adapt to the signal characteristics of the new installation location. Even if the type of transported goods changes over a long period, the model can gradually learn to adapt to the load signal characteristics of the new goods. This solves the problem of poor generalization of offline unified models and eliminates the need for large-scale offline annotation training for each vehicle, significantly reducing deployment and maintenance costs.

[0027] In this embodiment, after the vehicle obtains its current location data through the positioning module, it will be further optimized by combining the common operating route characteristics of dump trucks. The typical operating route of a dump truck is usually "loading point - transport road - unloading point - empty return". The operating scenarios and working conditions of different road sections are significantly different. When the positioning data determines that the vehicle is currently near the loading point or unloading point, the system will automatically increase the sampling frequency of the camera module and inertial sensor. This is because loading and unloading are critical nodes where the loading state changes. Increasing the sampling frequency can capture the state changes more quickly and update the model in a timely manner. When the vehicle is traveling at a constant speed in the middle section of the fixed transport route, the sampling frequency can be appropriately reduced to reduce unnecessary computing power consumption and data storage occupation, and extend the service life of the equipment.

[0028] For fleet operation scenarios, after the labeled samples of a single vehicle and the updated model parameters are uploaded to the cloud through the communication module, the cloud can also aggregate and organize all samples of the entire fleet, and periodically output optimized general model parameters and then distribute them to each vehicle, realizing fleet-level model collaborative evolution: for example, when a new vehicle of the same model joins the fleet, it does not need to start training from scratch. It can directly download the optimized model adapted to the model, and only a few adaptive updates are needed to achieve high detection accuracy, further improving the efficiency of large-scale deployment and reducing the overall application cost.

[0029] Compared with existing technologies, this invention combines the advantages of visual inspection and inertial inspection. When visual inspection is effective, it provides accurate results using visual inspection while automatically updating the inertial model online. When visual inspection fails, it uses an inertial model that has been adapted to the current vehicle to complete the inspection. This solves the problem of visual inspection failure in scenarios such as low light, occlusion, and dust, and also addresses the pain points of poor generalization of offline inertial models and high cost of personalized modeling. It is adaptable to various complex working conditions, simple to install and deploy, and has low maintenance costs, making it very suitable for large-scale operation of dump trucks in complex scenarios such as construction sites and mining areas.

[0030] In one embodiment, the dump truck further includes a positioning module disposed on the vehicle body, and the control device is electrically connected to the positioning module; The steps for acquiring real-time images of the carriage interior captured by the camera module and inertial data detected by the inertial sensor include: Obtain the current position data of the vehicle body detected by the positioning module; Based on the current position data of the vehicle body, the sampling frequency of the camera module and the detection frequency of the inertial sensor are matched; The camera module captures real-time images of the carriage according to the sampling frequency, and the inertial sensor detects inertial data according to the detection frequency.

[0031] Different sampling frequencies and detection frequencies are used at different locations to meet different needs.

[0032] In one embodiment, specifically, the step of matching the sampling frequency of the camera module and the detection frequency of the inertial sensor based on the current position data of the vehicle body includes: Based on the current position data of the vehicle body, determine the current position of the vehicle body on the travel trajectory; Based on the current position of the travel trajectory, the sampling frequency of the camera module and the detection frequency of the inertial sensor are matched.

[0033] The trajectory of the dump truck is fixed. By matching the sampling frequency of the camera module and the detection frequency of the inertial sensor at the current position of the trajectory, the accuracy is improved.

[0034] Furthermore, the step of matching the sampling frequency of the camera module and the detection frequency of the inertial sensor based on the current position of the travel trajectory includes: When the travel trajectory is in a flat position, the first sampling frequency and the first detection frequency are matched; When the travel trajectory is in a downhill position, the second sampling frequency and the second detection frequency are matched; Wherein, the second sampling frequency is greater than the first sampling frequency, and the second detection frequency is greater than the first detection frequency.

[0035] In flat locations, inertial data is relatively small, so the differences between different loading states are relatively small. Compared to flat locations, inertial data corresponding to downhill locations is more valuable. Therefore, in this embodiment, the second sampling frequency is greater than the first sampling frequency, and the second detection frequency is greater than the first detection frequency, meaning that more data is collected in downhill locations.

[0036] In another embodiment, a positioning module is not required; the current position is determined directly by the change in inertial data. The steps of acquiring real-time images of the carriage from the camera module and inertial data detected by the inertial sensor include: The change in inertial data is obtained based on the inertial data detected by the inertial sensor in two recent consecutive measurements. When the change exceeds the first set value, the camera module is controlled to increase the current sampling frequency, and the inertial sensor is controlled to increase the current detection frequency. When the change is less than or equal to the second set value, the camera module is controlled to reduce the current sampling frequency, and the inertial sensor is controlled to reduce the current detection frequency. When the second set value ≤ the change amount ≤ the first set value, the camera module is controlled to maintain the current sampling frequency, and the inertial sensor is controlled to maintain the current detection frequency.

[0037] In one embodiment, the dump truck further includes a communication module, which is electrically connected to the control device; If so, the system enters the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the inertial detection model. The system also includes controlling the communication module to upload the inertial data, the loading status of the current carriage, and the updated inertial detection model to the cloud.

[0038] In this way, the cloud can aggregate the detection data of multiple dump trucks, complete larger-scale model iteration and optimization, and the updated general model in the cloud can also be distributed to the control device of each dump truck, so that the detection accuracy of all devices can be continuously improved, which is suitable for application scenarios of fleet deployment.

[0039] In another embodiment, if the condition is met, the system enters a fusion model detection mode. In this mode, the loading status of the current truck bed is determined based on the real-time truck bed image, and the inertial data and the loading status of the current truck bed are input into the inertial detection model to update the model. Specifically, this includes: after the local inertial detection model completes the parameter update, the updated parameters are uploaded to the cloud synchronously via a communication module. At the same time, the system receives global update parameters obtained by aggregating data uploaded by other dump trucks from the cloud. The local inertial detection model is further adjusted using the global update parameters. This approach preserves the model adaptability brought by the operating characteristics of a single dump truck while leveraging the operating data of all vehicles to improve the overall generalization ability of the model, thus avoiding the overfitting problem caused by insufficient data from a single vehicle.

[0040] In one embodiment, the dump truck further includes a communication module, which is electrically connected to the control device; If so, then enter the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the inertial detection model. The steps include: If so, the system enters the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image. The inertial data, the loading status of the current carriage, and the inertial data and loading status received through the communication module are input into the inertial detection model to update the inertial detection model.

[0041] In this embodiment, the cloud can preprocess the initial inertial data and corresponding loading status labels collected from multiple dump trucks of the same model and operating scenario to train an initial general inertial detection model. This model is then distributed to the local control device of each dump truck as the initial model, avoiding the problem of a large amount of initial data collection time required to train a model from scratch for a single device. This allows the device to have basic inertial detection capabilities immediately after deployment, significantly shortening the model initialization cycle of new devices. Simultaneously, during subsequent operation, each device can continuously upload locally collected and labeled new data to the cloud via a communication module. The cloud periodically aggregates all new data to retrain and optimize the model, and then distributes the optimized new model back to the local device, achieving synchronous iterative upgrades of the detection capabilities of all devices in the entire fleet.

[0042] The present invention also provides a dump truck, the dump truck including a vehicle body, a camera module, an inertial sensor and a control device, the vehicle body including a cargo box, the camera module being disposed on the vehicle body for detecting images inside the cargo box, the inertial sensor being disposed on the vehicle body for detecting inertial data of the vehicle body, and the control device being electrically connected to the camera module and the inertial sensor; The control device includes a memory, a processor, and a loading status detection program for the dump truck's cargo compartment stored in the memory and executable on the processor. The loading status detection program for the dump truck's cargo compartment is configured to implement the following steps of a loading status detection method for the dump truck's cargo compartment: Acquire real-time images of the carriage captured by the camera module and inertial data detected by the inertial sensor; Based on the quality of the real-time carriage image, it is determined whether the image is valid; If not, then enter the inertial detection mode. In the inertial detection mode, the current loading status of the carriage is determined based on the inertial data and the inertial detection model. The loading status includes empty and heavy load. If so, the system enters the fusion model detection mode. In this mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the model.

[0043] In one embodiment, the dump truck further includes a positioning module disposed on the vehicle body, and the control device is electrically connected to the positioning module; and / or, The dump truck also includes a communication module mounted on the vehicle body, and the control device is electrically connected to the communication module.

[0044] Specifically, the positioning module includes a UWB module. UWB, as a positioning module, has the following core advantages: 1. centimeter-level positioning accuracy, 2. strong anti-interference capability, 3. low power consumption and real-time performance, and 4. high deployment flexibility. Specific Implementation Example 1: A detection system for dump trucks is disclosed. This system features an integrated hardware and software architecture, with all hardware components using automotive-grade parts. The software layer encapsulates functions based on the hardware layer, comprehensively meeting the stability, real-time performance, and environmental adaptability requirements of complex on-board operating conditions. It can be directly integrated into various types of dump trucks and comprises both hardware and software layers. Hardware layer: The hardware layer serves as the physical execution carrier of the device, realizing data acquisition, processing, storage, communication, output, and power supply functions. It includes a core processing unit, storage unit, sensing peripherals, communication peripherals, output peripherals, and a power supply unit. These units work together to provide physical support for the execution of the detection method. Core processing unit: It adopts automotive-grade microprocessors or embedded chips (such as ARM Cortex-A series, FPGA, etc.), supports floating-point operations and neural network inference, and has high computing performance, low power consumption and anti-interference characteristics. It is the computing core of the device, used to run embedded operating system and no-load / heavy-load detection application, and execute all core algorithm logic such as data processing, model training, state determination, and mode switching.

[0046] Storage Unit: A hierarchical storage design is adopted, including volatile and non-volatile memory. The volatile memory is DDR series memory, used for high-speed caching of real-time acquired sensing data and intermediate data of algorithm calculations to ensure the real-time performance of data processing. The non-volatile memory includes eMMC, Flash or SD card, used for persistent storage of firmware for the no-load detection application, trained IMU model files, and full-process running data (IMU raw data, visual labels, detection results, etc.). The storage capacity can be flexibly configured according to actual needs.

[0047] Peripheral sensing devices: These are the "sensing organs" of the device, serving as the source of cargo box status data. They include a camera module and an IMU sensor (i.e., an inertial measurement unit). The camera module is a vehicle-mounted high-definition industrial camera, which can be equipped with an infrared supplementary lighting component to collect real-time image data of the dump truck's cargo box, providing a basis for visual status judgment and empty / heavy load label output. The IMU sensor is a six-axis or nine-axis inertial measurement unit, capable of simultaneously acquiring multi-dimensional data such as acceleration, angular velocity, and attitude angle. The sampling frequency is adjustable, and it is used to continuously collect inertial data of the cargo box, providing data support for IMU model detection.

[0048] Communication peripherals (i.e., communication modules): including one or more of the following: 4G / 5G wireless communication module, Bluetooth module, WiFi module, and CAN bus module, to realize multi-scenario data interaction of the device; 4G / 5G and WiFi modules are used for bidirectional data interaction between the end side and the cloud (uploading of high-value samples, downloading of cloud models) and OTA remote model upgrades; Bluetooth module is used for local debugging, parameter configuration, and data export of the device; CAN bus module is used for data interaction with the dump truck's on-board terminal to realize the linkage between detection results and the on-board dispatch system.

[0049] Output peripherals include a display screen, an audible and visual alarm, and indicator lights, enabling local visualization of detection results and alerts for abnormal states. The display screen is a small, in-vehicle high-definition touchscreen used to display information such as the cargo box empty / heavy load determination result, result confidence level, and device operating mode. The audible and visual alarm is a combination of a buzzer and a warning light, which triggers audible and visual alerts in abnormal states such as low confidence level of detection results or visual failure. Indicator lights use different colors / flash frequencies to indicate the device's operating status (such as fusion mode, IMU single module mode, standby mode, and abnormal mode).

[0050] Power supply unit: This is a vehicle power adapter module that supports a wide voltage input of 12V / 24V vehicle power supply. It has overvoltage, overcurrent, short circuit and reverse connection protection functions, and provides a stable and safe DC power supply for all hardware components of the device, ensuring that the device can work normally when the vehicle power supply voltage fluctuates.

[0051] Software layer: The software layer runs on top of the hardware layer and serves as the logical implementation carrier for the detection method. It includes an embedded operating system and a dedicated application for no-load / heavy-load detection, encapsulating the entire process of hardware driving, data processing, and algorithm execution. The embedded operating system adopts an automotive-grade real-time embedded system (such as RT-Thread, FreeRTOS, QNX, Linux, etc.), which supports multi-task real-time scheduling, multi-peripheral driver management, and high-precision clock synchronization. It has high stability, low latency, and strong anti-interference characteristics, meeting the real-time requirements of synchronous acquisition of sensing data and high-speed algorithm computation, and providing a stable operating environment for the dedicated application for no-load / heavy-load detection.

[0052] The dedicated application for empty / heavy load detection is developed based on an embedded operating system. It is the core functional software of the device, encapsulating six major functional modules: image analysis, IMU data processing, sample processing and model training, mode switching, detection result output, and data storage. These modules work together to realize the entire process logic of empty / heavy load detection, and all modules support flexible parameter configuration to adapt to different operating conditions of dump trucks. The image analysis module performs real-time analysis on the cargo box image data collected by the camera module. It determines the visual working status through image clarity detection and occlusion recognition algorithms. The judgment criteria are no occlusion, sufficient light, and clear image. If the criteria are met, the vision is considered valid; otherwise, the vision is considered invalid. When the vision is valid, the visual recognition algorithm outputs a visual label indicating whether the cargo box is "empty" or "heavy" and synchronizes the label to the IMU data processing module.

[0053] The IMU data processing module synchronizes image data with IMU data in time, binding visual labels to the raw IMU data of the corresponding time window. At the same time, it preprocesses the inertial data acquired by the IMU sensor, including denoising, filtering, and time / frequency domain feature extraction, to obtain standardized IMU feature data, providing a data foundation for sample selection and model detection.

[0054] The sample processing and model training module processes the original samples of "IMU feature data + visual labels". First, it uses a clustering algorithm to remove redundant and invalid samples with high similarity. Then, it uses an active learning algorithm to select high-value samples that contribute significantly to improving model accuracy, such as boundary samples and difficult samples. It also supports two IMU model training / update modes: on-device online incremental training and cloud-based incremental training. After training, the model accuracy is verified by samples. If the accuracy does not reach the preset threshold, the samples are re-selected until the accuracy meets the standard.

[0055] The mode switching module receives the visual state determination result from the image analysis module as the trigger for mode switching. When vision is effective, the control device is in vision + IMU fusion mode and executes sample acquisition and model training / update logic. When vision fails, the control device automatically switches to IMU single module detection mode and executes IMU model independent detection logic. When the visual state returns to normal, the control device switches back to fusion mode, realizing uninterrupted automatic mode switching.

[0056] The detection result output module operates in different modes, receiving the no-load / overload judgment result and confidence level, and comparing the result with the preset confidence level threshold. If the confidence level is higher than the threshold, the judgment result is synchronized to the output peripheral for local output, and remotely uploaded to the vehicle terminal / cloud backend through the communication peripheral. If the confidence level is lower than the threshold, uncertain status information is output.

[0057] The data storage module calls the storage unit to complete the persistent storage and management of data throughout the entire process. The stored data includes IMU raw data, visual labels, high-value samples, model training logs, detection results, state switching records, etc. It also supports local data export and cloud upload, realizing data closure and providing data support for subsequent offline model optimization, equipment fault tracing, and sample library expansion.

[0058] A detection method based on the above-mentioned detection system, implemented using the above-mentioned detection device, integrates the advantages of visual detection and IMU inertial detection technologies, and adopts the core working logic of "visual effective training model and visual failure application model" to achieve full-condition adaptive detection of dump truck cargo boxes, whether empty or loaded. The method specifically includes the following four steps: S1. After the system initialization device is powered on, the embedded operating system automatically starts and completes the driver loading and parameter initialization configuration of the core processing unit, storage unit, sensing peripherals, communication peripherals, and output peripherals. At the same time, the camera module and IMU sensor are started, and the no-load heavy-load detection dedicated application is started to load the image analysis module, IMU data processing module, and sample processing and model training module. The high-precision real-time clock of the operating system completes the timestamp alignment of each sensing peripheral, unifies the data acquisition frequency, and ensures that the image data of the camera and the inertial data of the IMU are acquired synchronously, so as to provide a guarantee for the validity of subsequent samples and the accuracy of detection.

[0059] S2. Sample acquisition, processing, and model training / updating under effective visual operating conditions. (1) Visual state determination: The image analysis module analyzes the real-time images of the cargo box collected by the camera module and determines that the visual working state is valid; (2) Data synchronous acquisition: The core processing unit controls the perception peripheral to synchronously acquire the IMU raw data and the visual labels output by the image analysis module within the preset time window. The IMU data processing module completes data binding and preprocessing to obtain the raw samples of "IMU feature data + visual labels", which are temporarily cached in volatile memory; (3) High-value sample screening: The sample processing and model training module calls the computing resources of the core processing unit to perform clustering processing on the raw samples to remove redundant samples, and then selects high-value samples through active learning processing. The selected high-value samples are transferred from volatile memory to non-volatile memory for persistent storage; (4) Model training / update: According to the operation scenario of the dump truck (single vehicle operation / large-scale fleet operation), the edge online training or cloud training + OTA upgrade mode is selected to complete the incremental training / update of the IMU model: Online training on the edge: The core processing unit directly calls high-value samples in the local non-volatile memory to complete the incremental training of the IMU model, update the model parameters in real time, and store the updated model back to the non-volatile memory without uploading data, quickly adapting to the personalized working conditions of a single vehicle. Cloud training + OTA upgrade: The core processing unit uploads high-value samples to the cloud server through the communication peripheral. After the incremental iteration training of the model is completed in the cloud, the new version of the IMU model is downloaded to the non-volatile memory on the edge through OTA remote upgrade to complete the model update, taking into account the working conditions of multiple vehicles and the unified optimization of the model; (5) Model accuracy verification: After each training / update, a small number of newly collected high-value samples are selected to verify the accuracy of the IMU model. If the accuracy is lower than the preset threshold, return to step (3) to re-optimize the screening parameters and screen samples until the model accuracy meets the standard. After meeting the standard, the model is put into use directly.

[0060] S3. Independent detection of IMU model under visual failure conditions (1) Visual state determination: The image analysis module analyzes the image data collected by the camera module and determines that the visual working state is failure (occlusion / dark light). The visual algorithm stops outputting labels. (2) Mode switching: The mode switching module receives the visual failure signal and triggers the mode switching command. The device automatically switches from "visual + IMU fusion mode" to "IMU single module detection mode". (3) Data processing and model detection: The inertial data of the cargo box is continuously collected by the IMU sensor. After the IMU data processing module removes noise and extracts features, it is input into the latest IMU model that has been trained in the non-volatile memory. The model outputs the judgment result of the cargo box being "empty" or "heavy", and outputs the confidence level of the result simultaneously. (4) Result judgment and output: The preset threshold is used to judge the empty and heavy load judgment results and confidence level by comparing them with the preset threshold. If the confidence level is higher than the preset threshold, the empty reload status is output locally through the output peripheral and the result is uploaded remotely through the communication peripheral. If the confidence level is lower than the threshold, the empty reload status information is not output until the visual state returns to normal or the model is optimized and then re-evaluated.

[0061] S4, State Switching and Data Loop (1) State Switching Back: When the image analysis module detects that the visual state has returned to normal (occlusion removed / sufficient light), the mode switching module immediately triggers an instruction, and the device switches from "IMU single module detection mode" back to "vision + IMU fusion mode", returns to step S2, and continues to execute sample collection, processing and model training / update logic to continuously optimize the adaptability of the IMU model; (2) Data Loop: The data storage module transfers the entire process running data (IMU raw data, visual labels, high-value samples, model training logs, detection results, state switching records, etc.) from volatile memory to non-volatile memory for persistent storage; according to the preset cycle or manual instruction, the local stored data can be uploaded to the cloud server through the communication peripheral for offline optimization of the model, equipment fault tracing and sample library expansion, realizing the closed-loop utilization of data, so that the IMU model has continuous optimization capabilities.

[0062] Compared with the prior art, the present invention has the following significant advantages: Addressing the pain points of full-condition inspection and achieving uninterrupted inspection: Integrating visual inspection and IMU inertial detection technologies, and adopting the logic of "training the model effectively and using the model when visual inspection fails", it effectively solves the problems of visual inspection failure in low-light environments and camera obstruction, and achieves full-condition, uninterrupted adaptive inspection of dump truck cargo boxes, with 100% inspection coverage.

[0063] To address the generalization problem of IMU models and adapt to personalized operating conditions: high-value samples are selected through clustering and active learning to improve model training efficiency and accuracy; two modes are designed: edge-side online training and cloud training + OTA upgrade. Edge-side training can adapt to the personalized operating conditions of a single vehicle in real time, while cloud training can achieve large-scale model optimization for a fleet; incremental training mode is adopted to update the model, retaining historical effective feature parameters and adapting to dynamic changes in vehicle operating conditions (such as suspension wear and cargo type adjustment), completely solving the problems of poor generalization of offline unified IMU models and the impracticality of personalized offline modeling.

[0064] Integrated hardware and software architecture, adapted to complex vehicle operating conditions: The device adopts automotive-grade components and a standardized integrated hardware and software architecture design. The hardware layer features hierarchical storage and multi-peripheral collaboration, while the software layer features a real-time operating system and dedicated application packaging. It has high stability, high real-time performance, low power consumption, and strong anti-interference characteristics. It can adapt to complex operating conditions such as high temperature, bumps, and electromagnetic interference in vehicles, and is easy to install, and can be directly mounted on various dump trucks.

[0065] Sensor installation has no special requirements, reducing deployment costs: The detection method has no special requirements for the installation position and angle of the IMU sensor. Relying on the online training mechanism, the IMU model can adaptively learn the IMU signal characteristics under different installation positions and angles, automatically offsetting the impact of installation differences on signal acquisition. No additional installation and debugging operations are required, which greatly reduces the on-site deployment cost.

[0066] Data closed-loop design and continuous model optimization: The data storage module enables persistent storage and cloud upload of data throughout the entire process, building a complete data closed loop. The stored data can be used for offline model optimization, fault tracing, and sample library expansion, enabling the IMU model to have continuous optimization capabilities and gradually improve detection accuracy over time.

[0067] Multi-scenario data output to meet diverse monitoring needs: Detection results support local visualization output (display screen / indicator light) and remote upload (vehicle terminal / cloud backend). Abnormal conditions trigger audible and visual alerts, meeting the dual monitoring needs of on-site operators and remote dispatch managers, and improving the level of intelligence in dump truck operation.

[0068] Specific implementation 2, such as Figures 2 to 6 Hardware and software configuration of the detection system The detection system in this embodiment is an integrated hardware and software design, adapted to the complex operating conditions of dump trucks in mining areas. The specific configuration is as follows: 1. Hardware layer configuration Core processing unit 11: adopts automotive-grade ARM Cortex-A53 embedded chip with a main frequency of 1.5GHz, supports floating-point operations and neural network inference, and meets the computing needs of model training and data processing; Storage unit 12: Volatile memory 121 is 2GB DDR4 memory, used for caching real-time data; non-volatile memory 122 is 32GB eMMC, used for storing application programs, model files and historical running data; Perception peripherals 13: Camera module 131 is a 2-megapixel automotive high-definition industrial camera equipped with an infrared fill light component and a frame rate of 25fps; IMU sensor 132 is a six-axis inertial measurement unit with a sampling frequency of 100Hz, which can collect multi-dimensional data such as acceleration, angular velocity, and attitude angle. Communication peripherals 14: integrates a 4G wireless communication module, a CAN bus module, and a Bluetooth 5.0 module. 4G is used for end-to-cloud interaction and OTA upgrades, the CAN bus is used for linkage with the vehicle terminal, and Bluetooth is used for local debugging; Output peripherals 15: 1.8-inch in-vehicle high-definition touch screen display, buzzer and light alarm, red / green / blue three-color status indicator (green - fusion mode, red - IMU single module mode, blue - standby, red flashing - abnormal); Power supply unit 16: 24V vehicle power adapter module, with overvoltage, overcurrent and short circuit protection, outputting a stable 5V / 3A DC voltage.

[0069] 2. Software layer configuration Embedded Operating System 21: Employs RT-Thread real-time embedded operating system, supporting real-time scheduling of multiple tasks, with clock synchronization accuracy of ±0.01s; Application 22 for no-load / heavy-load detection: Based on RT-Thread, with a preset model accuracy threshold of 95% and a detection result confidence threshold of 90%. The parameters of the six functional modules can be configured locally via Bluetooth.

[0070] Specific steps of the detection method Based on the detection system of Example 2, the specific execution steps of the detection method in this example are as follows: S1. After the device is powered on, the RT-Thread operating system automatically starts and completes the driver loading and parameter configuration of all hardware components. It starts the camera module 131, IMU sensor 132 and the six major functional modules of the software layer. The timestamps of the camera and IMU are aligned through the real-time clock of the operating system. The camera frame rate is 25fps and the IMU sampling frequency is 100Hz to ensure synchronous data acquisition. After initialization, the blue status indicator light stays on and the device enters the standby state.

[0071] S2. Sample collection, processing and model training / updating under visually effective working conditions: The daytime light in the mining area is sufficient and the camera is unobstructed. The image analysis module 221 determines that the vision is effective, the green status indicator light is always on, and the device enters the "vision + IMU fusion mode": (1) S21: The image analysis module 221 recognizes the cargo box image and outputs the "empty / heavy" visual label; (2) S22: The core processing unit 11 controls the sensing peripheral to synchronously collect the IMU raw data and visual label within a 10s time window. After the IMU data processing module 222 performs noise reduction and feature extraction, it is bound as the original sample and cached in DDR4 memory; (3) S23: The sample processing and model training module 223 removes redundant samples through the K-means clustering algorithm and then selects high-value samples through the active learning algorithm and transfers them to 32GB eMMC; (4) S24: In this embodiment, the single vehicle operation is selected. The end-side online training mode is selected, and the core processing unit 11 calls the eMMC. (5) S25: Select the latest 5% of high-value samples to verify the model accuracy. If the accuracy is <95%, return to S23 to re-select samples until the accuracy meets the standard.

[0072] S3. Under visual failure conditions, the IMU model independently detects the mining area at night without lighting. The image analysis module 221 determines visual failure, the red status indicator light stays on, and the device automatically switches to "IMU single module detection mode": (1) S31: The image analysis module 221 stops outputting visual labels and sends a visual failure signal to the mode switching module 224; (2) S32: The mode switching module 224 triggers the command, and the device completes the mode switching; (3) S33: The IMU sensor 132 continuously collects the inertial data of the cargo box. After the IMU data processing module 222 denoises and extracts features, it is input to the IMU model trained in the eMMC. The model outputs the judgment result and confidence level; (4) S34: If the confidence level is ≥90%, the display shows "no load / heavy load + confidence level value", which is uploaded to the vehicle terminal via the CAN bus and uploaded to the cloud backend via the 4G module; if the confidence level is <90%, the display shows "uncertain".

[0073] S4, State Switching and Data Loop (1) State Switching Back: The next day, when the light in the mining area is restored, the image analysis module 221 determines that the visual state is normal, and the device automatically switches back to the "Vision + IMU Fusion Mode". The green indicator light stays on, and it returns to S2 to continue to perform sample collection and model training; (2) Data Loop: The data storage module 226 transfers the entire process data from DDR4 memory to eMMC. The preset cycle is configured to be early morning every day. Every early morning, the data of the previous day is uploaded to the cloud through the 4G module for offline model optimization and sample library expansion.

[0074] Model training mode for large-scale fleet operation For large-scale operation scenarios of dump truck fleets in mining areas, the model training mode of the detection device is switched to cloud training + OTA upgrade: The core processing unit 11 of each vehicle uploads the selected high-value samples to the cloud server in the mining area via 4G module. After the unified incremental iterative training of the IMU model of all vehicles is completed in the cloud, the new version of the unified IMU model is downloaded in batches to the eMMC of all vehicles in the fleet through OTA remote upgrade, completing the unified model update, improving the overall detection accuracy of the fleet, and reducing the fleet management cost.

[0075] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0076] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of detecting a loading state of a vehicle body of a self-dumping vehicle, characterized by, The dump truck includes a vehicle body, a camera module, an inertial sensor, and a control device. The vehicle body includes a cargo box. The camera module is located on the vehicle body for detecting images inside the cargo box. The inertial sensor is located on the vehicle body for detecting inertial data of the vehicle body. The control device is electrically connected to the camera module and the inertial sensor. The method for detecting the loading status of the dump truck's cargo box includes: Acquire real-time images of the carriage captured by the camera module and inertial data detected by the inertial sensor; Based on the quality of the real-time carriage image, it is determined whether the image is valid; If not, then enter the inertial detection mode. In the inertial detection mode, the current loading status of the carriage is determined based on the inertial data and the inertial detection model. The loading status includes empty and heavy load. If so, the system enters the fusion model detection mode. In this mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the model.

2. The loading state detection method of a carriage of a self-dumping vehicle according to claim 1, characterized in that, The dump truck also includes a positioning module mounted on the vehicle body, and the control device is electrically connected to the positioning module; The steps for acquiring real-time images of the carriage interior captured by the camera module and inertial data detected by the inertial sensor include: Obtain the current position data of the vehicle body detected by the positioning module; Based on the current position data of the vehicle body, the sampling frequency of the camera module and the detection frequency of the inertial sensor are matched; The camera module captures real-time images of the carriage according to the sampling frequency, and the inertial sensor detects inertial data according to the detection frequency.

3. The method of claim 2, wherein The step of matching the sampling frequency of the camera module and the detection frequency of the inertial sensor based on the current position data of the vehicle body includes: Based on the current position data of the vehicle body, determine the current position of the vehicle body on the travel trajectory; Based on the current position of the travel trajectory, the sampling frequency of the camera module and the detection frequency of the inertial sensor are matched.

4. The method of claim 3, wherein The step of matching the sampling frequency of the camera module with the detection frequency of the inertial sensor based on the current position of the travel trajectory includes: When the travel trajectory is in a flat position, the first sampling frequency and the first detection frequency are matched; When the travel trajectory is in a downhill position, the second sampling frequency and the second detection frequency are matched; Wherein, the second sampling frequency is greater than the first sampling frequency, and the second detection frequency is greater than the first detection frequency.

5. The method of claim 1, wherein The steps for acquiring real-time images of the carriage interior captured by the camera module and inertial data detected by the inertial sensor include: The change in inertial data is obtained based on the inertial data detected by the inertial sensor in two recent consecutive measurements. When the change exceeds the first set value, the camera module is controlled to increase the current sampling frequency, and the inertial sensor is controlled to increase the current detection frequency. When the change is less than or equal to the second set value, the camera module is controlled to reduce the current sampling frequency, and the inertial sensor is controlled to reduce the current detection frequency. When the second set value ≤ the change amount ≤ the first set value, the camera module is controlled to maintain the current sampling frequency, and the inertial sensor is controlled to maintain the current detection frequency.

6. The method of claim 1, wherein The dump truck also includes a communication module, which is electrically connected to the control device; If so, the system enters the fusion model detection mode. In this mode, the steps of determining the current loading status of the carriage based on the real-time carriage image and inputting the inertial data and the current loading status of the carriage into the inertial detection model to update the inertial detection model further include: The communication module is controlled to upload the inertial data, the current loading status of the carriage, and the updated inertial detection model to the cloud.

7. The method for detecting the loading status of the cargo compartment of a dump truck according to claim 1, characterized in that, The dump truck also includes a communication module, which is electrically connected to the control device; If so, then enter the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image, and the inertial data and the loading status of the current carriage are input into the inertial detection model to update the inertial detection model. The steps include: If so, the system enters the fusion model detection mode. In the fusion model detection mode, the loading status of the current carriage is determined based on the real-time carriage image. The inertial data, the loading status of the current carriage, and the inertial data and loading status received through the communication module are input into the inertial detection model to update the inertial detection model.

8. A dump truck, characterized in that, The dump truck includes a vehicle body, a camera module, an inertial sensor, and a control device. The vehicle body includes a cargo box. The camera module is located on the vehicle body and is used to detect images inside the cargo box. The inertial sensor is located on the vehicle body and is used to detect inertial data of the vehicle body. The control device is electrically connected to the camera module and the inertial sensor. The control device includes a memory, a processor, and a loading status detection program for the dump truck's cargo compartment stored in the memory and executable on the processor. The loading status detection program for the dump truck's cargo compartment is configured to implement the steps of the loading status detection method for the dump truck's cargo compartment as described in any one of claims 1 to 7.

9. A method for detecting the loading status of a dump truck's cargo compartment according to claim 8, characterized in that, The dump truck also includes a positioning module mounted on the vehicle body, and the control device is electrically connected to the positioning module; and / or, The dump truck also includes a communication module mounted on the vehicle body, and the control device is electrically connected to the communication module.

10. A method for detecting the loading status of a dump truck's cargo compartment according to claim 9, characterized in that, The positioning module includes a UWB module.

Images