Truck loading rate identification method

Abstract

The invention relates to the technical field of transportation management, in particular to a truck loading rate identification method, which comprises the following steps: acquiring truck three-dimensional data in a truck, and establishing a corresponding truck digital model based on the truck three-dimensional data, based on the truck digital model, selecting a regular gridding or self-adaptive gridding method to divide the interior of the truck into a plurality of cells; acquiring real-time image set data in the truck, adjusting cell parameters in the truck based on the real-time image set data, and performing occupation state judgment on each adjusted cell based on the image set data according to a pixel-level threshold value and / or a depth information threshold value; and determining cell occupancy data based on the judgment result of each cell, calculating proportion data of cells in an occupancy state based on the cell occupancy data, determining a truck loading rate based on the proportion data, and generating a corresponding visual report based on the cell occupancy data. According to the invention, the accuracy of truck loading rate identification can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of transportation management, and in particular to a method for identifying a truck loading rate. Background Art

[0002] In the modern logistics and transportation industry, optimizing truck loading efficiency is crucial to reducing logistics costs, improving transportation capacity and resource utilization. Traditional manual planning methods are not only time-consuming and labor-intensive, but also difficult to maximize space utilization, especially when dealing with trucks with complex shapes or non-standard specifications. This limitation is particularly obvious.

[0003] Although intelligent solutions based on computer-aided design (CAD), 3D modeling and image processing technologies have gradually emerged in recent years, these methods are usually limited to containers of regular shapes and are not effective enough when faced with complex boundary conditions.

[0004] From the above, we can see that how to improve the accuracy of truck loading rate recognition still needs to be solved. Summary of the invention

[0005] In order to improve the accuracy of truck loading rate recognition, the present application provides a truck loading rate recognition method.

[0006] In a first aspect, the present application provides a method for identifying a truck loading rate, which adopts the following technical solution:

[0007] A method for identifying a truck loading rate comprises: acquiring three-dimensional truck data inside the truck, establishing a corresponding truck digital model based on the three-dimensional truck data, and selecting a regular gridding or adaptive gridding method based on the truck digital model to divide the inside of the truck into a plurality of cells; acquiring real-time image set data inside the truck, adjusting cell parameters inside the truck based on the real-time image set data, and performing occupancy status judgment on each adjusted cell based on the image set data using a pixel-level threshold and / or a depth information threshold; determining cell occupancy data based on the judgment result of each cell, calculating proportion data of occupied cells based on the cell occupancy data, determining the truck loading rate based on the proportion data, and generating a corresponding visualization report based on the cell occupancy data.

[0008] By adopting the above technical solutions, the accuracy and reliability of occupancy status judgment are significantly improved by integrating high-precision 3D modeling, intelligent grid division and multi-source data fusion technology. The system can not only dynamically adapt to changes in the internal structure of the truck, but also eliminate visual blind spots through real-time image analysis and feature fusion, ensuring that the status judgment of each cell is comprehensive and accurate. Finally, the loading rate evaluation calculated based on the weighted average method is more in line with the actual loading situation, providing a detailed visual report to help users optimize cargo stacking, achieve efficient space utilization and transportation planning, and ensure that excellent recognition accuracy can be maintained in complex and changing loading environments.

[0009] Optionally, the method also includes: extracting corresponding image features from a real-time image set, classifying the current scene corresponding to the image features through a pre-trained convolutional neural network, and obtaining corresponding scene classification results, wherein the image features include color distribution, texture information, edge contour and depth information; predicting the optimal pixel level threshold and / or depth information threshold under current conditions based on the scene classification results, and adjusting the threshold parameters, wherein the threshold parameters include pixel level threshold and depth information threshold.

[0010] By adopting the above technical solution, image features including color distribution, texture information, edge contours and depth information are extracted from the real-time image set, and the current scene is classified using a pre-trained convolutional neural network to obtain accurate scene classification results. Based on these classification results, the system can predict the optimal pixel-level threshold and depth information threshold under the current conditions, and dynamically adjust the corresponding threshold parameters. This process ensures high accuracy and adaptability of occupancy status judgment, and maintains excellent performance even in the case of changes in lighting or diversified cargo types, thereby significantly improving the accuracy and reliability of truck loading rate recognition.

[0011] Optionally, in the process of performing occupancy status judgment on each adjusted cell based on the image set data using a pixel-level threshold and / or a depth information threshold, the method also includes: obtaining a multi-view image set from multiple different sources, and obtaining corresponding multi-view feature information from the multi-view image set through a feature extraction algorithm, the feature information including color distribution, texture information, edge contour, depth information and thermal imaging; performing a comprehensive judgment on the occupancy status of each cell based on the multi-view feature information, and predicting the movement trend of the cargo by tracking the position changes of the cargo in the truck based on the time series; judging whether there is any abnormality in the comprehensive judgment result of the occupancy status based on the movement trend, and if not, determining the comprehensive judgment result of the occupancy status as the final judgment result.

[0012] By adopting the above technical solution, a multi-view image set from multiple different sources is obtained, and a feature extraction algorithm is used to obtain rich multi-view feature information, including color distribution, texture information, edge contour, depth information and thermal imaging. Based on these multi-view feature information, the system can make a more comprehensive and accurate comprehensive judgment on the occupancy status of each cell. Combined with time series analysis, the system tracks the position changes of the goods in the truck and predicts its movement trend, further verifying the rationality of the occupancy status judgment result. If no abnormality is detected, the system determines the comprehensive judgment result as the final judgment result, thereby significantly improving the accuracy and reliability of the occupancy status judgment and ensuring the efficiency and accuracy of the truck loading rate recognition.

[0013] Optionally, the method also includes: retrieving the pre-set importance differences of cells in different positions, and determining the cell weight value corresponding to each cell based on the importance difference; calculating the proportion data of occupied state cells based on the cell weight values, wherein the occupancy state contribution of each cell is proportional to its weight value; and determining the truck loading rate based on the proportion data.

[0014] By adopting the above technical solution, the weight value corresponding to each cell is determined by calling the pre-set importance differences of cells in different locations, ensuring that key areas (such as areas near doors or areas with high loading and unloading frequency) are fully valued in the loading rate calculation. Based on these weight values, the system calculates the proportion data of occupied state cells so that the contribution of the occupied state of each cell is proportional to its importance. Finally, the truck loading rate is determined based on the weighted proportion data, thereby providing a more accurate and reasonable loading rate evaluation and optimizing cargo stacking and transportation planning.

[0015] Optionally, the method also includes: real-time monitoring of the cell distribution inside the truck, and detecting whether there is unreasonable distribution in a specific area based on the cell distribution; if so, determining the corresponding unreasonable area, and prioritizing fine-tuning the unreasonable area based on a preset fine-tuning strategy; after fine-tuning, monitoring the real-time feedback effect of the fine-tuning of the unreasonable area, and optimizing the fine-tuning strategy according to the real-time feedback effect.

[0016] By adopting the above technical solution, the distribution of cells inside the truck can be monitored in real time to automatically detect whether there is an unreasonable distribution problem in a specific area. Once an unreasonable area is found, the system will prioritize fine-tuning the area based on the preset fine-tuning strategy, and continue to monitor its effect after fine-tuning, and dynamically optimize the fine-tuning strategy based on real-time feedback. This process ensures the accuracy of local adjustments and the stability of global layout, avoids unnecessary interference, and significantly improves the rationality of grid division and the accuracy of occupancy status judgment.

[0017] Optionally, including a user interaction interface, the method also includes: obtaining a corresponding truck digital model based on the user inputting a specific truck model or selecting a predefined truck template in the user interaction interface; obtaining key dimension points of the user on the truck digital model, and obtaining context information input by the user in the user interaction interface; and optimizing the local grid of the truck digital model based on the key dimension points and the context information.

[0018] By adopting the above technical solution, through the user interaction interface, users can enter the specific truck model or select a predefined template, and mark the key size points on the truck digital model and provide contextual information. Based on these user-entered data, the system optimizes the local grid of the truck digital model to ensure that each cell accurately fits the actual physical space, especially when dealing with special structures such as partitions and ramps. This process enhances the flexibility and customization capabilities of the system, making the grid division more in line with actual needs, thereby improving the accuracy of occupancy status judgment and the reliability of loading rate evaluation.

[0019] In a second aspect, the present application provides a truck loading rate identification device, which adopts the following technical solution:

[0020] A truck loading rate identification device, comprising:

[0021] A cell division module, which obtains three-dimensional data of the interior of the truck, establishes a corresponding digital model of the truck based on the three-dimensional data of the truck, and selects a regular gridding method or an adaptive gridding method based on the digital model of the truck to divide the interior of the truck into a plurality of cells;

[0022] A cell parameter adjustment module, which obtains real-time image set data of the interior of the truck, adjusts the cell parameters of the interior of the truck based on the real-time image set data, and performs occupancy status judgment on each adjusted cell using a pixel-level threshold and / or a depth information threshold based on the image set data;

[0023] The truck loading rate determination module determines the cell occupancy data based on the judgment result of each cell, calculates the proportion data of the occupied state cells based on the cell occupancy data, and uses the proportion data to determine the truck loading rate, and generates a corresponding visualization report based on the cell occupancy data.

[0024] In a third aspect, the present application provides an electronic device, which adopts the following technical solution:

[0025] An electronic device comprises a processor, wherein a program of any one of the above-mentioned truck loading rate identification methods is running on the processor.

[0026] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution:

[0027] A storage medium stores a program of any one of the above-mentioned methods for identifying a truck loading rate.

[0028] In summary, the present application includes at least one of the following beneficial technical effects:

[0029] By integrating high-precision 3D modeling, intelligent meshing and multi-source data fusion technology, the accuracy and reliability of occupancy status judgment have been significantly improved. The system can not only dynamically adapt to changes in the internal structure of the truck, but also eliminate visual blind spots through real-time image analysis and feature fusion, ensuring that the status judgment of each cell is comprehensive and accurate. Finally, the loading rate evaluation calculated based on the weighted average method is more in line with the actual loading situation, providing a detailed visual report to help users optimize cargo stacking, achieve efficient space utilization and transportation planning, and ensure that excellent recognition accuracy can be maintained in complex and changing loading environments.

[0030] To further enhance the accuracy of recognition, the system uses a pre-trained convolutional neural network to classify the image features extracted from the real-time image set, and predicts the optimal pixel-level and depth information thresholds based on the classification results, and dynamically adjusts the corresponding parameters. This process ensures that even in the case of changes in lighting or diversified cargo types, the occupancy status judgment remains highly accurate and adaptable, thereby significantly improving the accuracy and reliability of truck loading rate recognition. In addition, through multi-view image collection and time series analysis, the system can track the location changes of cargo and predict movement trends, further verifying the rationality of the occupancy status judgment results and ensuring the accuracy of the final judgment results.

[0031] By introducing a user interaction interface, the system allows users to enter specific truck models or select predefined templates, mark key dimension points and provide contextual information to optimize the local grid of the truck digital model to ensure that each cell accurately fits the actual physical space, especially when dealing with special structures such as partitions and slopes. At the same time, the system monitors the distribution of cells inside the truck in real time, automatically detects and prioritizes fine-tuning of unreasonable distribution areas, and optimizes the fine-tuning strategy based on real-time feedback to ensure the accuracy of local adjustments and the stability of the global layout. These functions enhance the flexibility and customization capabilities of the system, making the grid division more in line with actual needs, thereby greatly improving the accuracy of occupancy status judgment and the reliability of loading rate assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The present invention is a flowchart of a method for identifying a truck loading rate according to an exemplary embodiment.

[0033] Figure 2 The figure is a structural block diagram of a device for identifying a truck loading rate according to an exemplary embodiment. DETAILED DESCRIPTION

[0034] Embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings.

[0035] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0036] The present application embodiment discloses a method for identifying a truck loading rate, referring to Figure 1 ,include:

[0037] S100, obtaining three-dimensional data of the interior of the truck, establishing a corresponding digital model of the truck based on the three-dimensional data of the truck, and selecting a regular gridding method or an adaptive gridding method based on the digital model of the truck to divide the interior of the truck into a plurality of cells.

[0038] Among them, high-precision laser radar (LiDAR), 3D cameras or structured light scanners are used. These devices can work under different lighting conditions. Multiple sensors are installed inside the truck to ensure that the entire compartment space is covered; through multiple scans and multi-view fusion, complete point cloud data is generated. Each scanning cycle should minimize motion artifacts to ensure data accuracy. The original point cloud data is denoised, spliced ​​and registered to eliminate errors caused by equipment movement or environmental changes.

[0039] Choose a 3D modeling tool such as AutoCAD, SolidWorks or a specialized one to convert the processed point cloud data into an accurate digital model of the truck, which contains not only the geometric shape but also information such as material properties and reflectivity.

[0040] Evaluate the complexity of the truck's internal structure and the characteristics of the load, and decide whether to use regular gridding or adaptive gridding. Regular gridding is suitable for areas with regular shapes and little change; adaptive gridding is for areas with complex shapes or requiring detailed analysis, and can dynamically adjust the cell size to optimize calculation efficiency and accuracy; then apply the selected gridding method to divide the interior of the truck into multiple cells. Each cell is assigned a unique identifier, and its coordinates, size, and other relevant attributes are recorded for subsequent processing.

[0041] S110, acquiring real-time image set data of the interior of the truck, adjusting the cell parameters of the interior of the truck based on the real-time image set data, and determining the occupancy status of each adjusted cell using a pixel-level threshold and / or a depth information threshold based on the image set data.

[0042] Multiple cameras are installed inside the truck to ensure full coverage without blind spots. The cameras can be ordinary RGB cameras or RGB-D cameras with depth perception function, the latter of which can provide additional distance information. The camera takes pictures at regular intervals to form an image sequence. The image set at each time point contains a snapshot of the status of the truck's interior for subsequent analysis. At the same time, the collected images need to be preprocessed, including color correction, distortion correction, resolution adjustment, etc., to ensure the consistency and comparability of all images.

[0043] Then, based on the real-time image, check whether the cells divided in the early stage accurately reflect the current cargo stacking situation; if it is found that the division of some cells is inaccurate, such as misclassification due to the special shape of the cargo, it is necessary to adjust the boundaries of these cells or other related parameters. At the same time, machine learning or computer vision algorithms can be used to automatically detect and adjust cell boundaries. For example, for irregularly shaped cargo, the system can perform adaptive segmentation based on the edge contour to ensure that each cell fits the actual physical space.

[0044] Extract features such as color distribution, texture information, edge contours, and depth information from real-time images. This step can be achieved through convolutional neural networks (CNNs) or other advanced image processing algorithms. Using pre-trained machine learning models (such as convolutional neural networks), the optimal pixel-level threshold or depth information threshold in the current scene is predicted based on the extracted features. These thresholds will take into account factors such as lighting conditions and cargo types to ensure the robustness of the judgment results.

[0045] Finally, based on the set threshold, the occupancy status of each cell is judged. For example, if the average depth value in a cell is lower than the set no-load threshold, it is considered unoccupied; otherwise, it is occupied. Combined with context perception and time series analysis, the rationality of the judgment result is further verified to avoid misjudgment.

[0046] S120, determining cell occupancy data based on the judgment result of each cell, calculating proportion data of occupied cells based on the cell occupancy data, determining a truck loading rate based on the proportion data, and generating a corresponding visualization report based on the cell occupancy data.

[0047] Among them, the information of all cells marked as occupied is sorted out, and their number and coverage area or volume are counted; during the summary process, the system can detect and correct possible anomalies, such as occlusion, reflection, etc., to further improve the robustness and accuracy of the judgment.

[0048] Compare the number of occupied cells or the space they occupy with the total number of cells or the total volume to get the loading rate percentage. Considering the differences in importance of cells in different positions, the final loading rate is calculated using the weighted average method to make the evaluation more reasonable. Combined with historical data, the changing trend of the loading rate is analyzed to help users understand the improvement of loading efficiency.

[0049] Use heat maps, bar charts, pie charts and other visualization tools to display the loading status inside the truck. For example, red represents high-density loading areas, green represents low-density areas, and blue represents vacant areas. Through intuitive color coding, users can quickly understand the loading status. By generating detailed digital reports, including but not limited to loading rates, underutilized space locations, historical loading record comparisons, etc., these reports provide comprehensive data support to help users make better decisions. Finally, a user-friendly interactive interface is provided, allowing users to view the loading status of different time periods, adjust parameters, and export reports.

[0050] In the process of determining the occupancy state of each adjusted cell by applying a pixel level threshold and / or a depth information threshold based on the image set data, the method further includes:

[0051] S131, obtaining a multi-view image set from multiple different sources, and obtaining corresponding multi-view feature information from the multi-view image set through a feature extraction algorithm.

[0052] Among them, multiple cameras are installed at different positions inside the truck, and each camera takes pictures at regular intervals to form an image sequence. These image sets not only provide a snapshot of the status of the truck's interior, but also eliminate the blind spots that may exist in a single perspective through multi-perspective fusion. Advanced feature extraction algorithms such as SIFT (Scale Invariant Feature Transform), SURF (Speeded Robust Features) or CNN (Convolutional Neural Network) can be used to extract rich feature information from multi-perspective images, including color distribution, texture information, edge contours, depth information, and thermal imaging.

[0053] The feature information extracted from different sources is fused to form a comprehensive feature representation. For example, color and texture features can help distinguish goods from the background, while depth information helps determine the three-dimensional shape and position of objects. Thermal imaging data can provide additional support in low-light conditions, especially when detecting temperature-sensitive items.

[0054] S132, making a comprehensive judgment on the occupancy status of each cell based on the multi-view feature information, determining the position change of the tracked goods in the truck based on the time series, and predicting the movement trend of the goods.

[0055] Among them, the extracted multi-view features are analyzed using a pre-trained machine learning model (such as a convolutional neural network) to determine whether each cell is occupied. The machine learning model will evaluate the current scene based on the feature information and output the occupancy state probability of each cell. Among them, the neighborhood consistency check is combined to ensure that the states between adjacent cells conform to the logical relationship and reduce the possibility of misjudgment of isolated points. For example, if a cell is judged to be occupied, but the surrounding cells are all unoccupied, the state of the cell needs to be further verified.

[0056] In addition, time series analysis is used to track the location changes of goods, and historical data is combined to predict the movement trend of goods. The system records the state changes of each cell over a period of time to confirm whether it is in a stable occupied state or has a movement trend.

[0057] As for the prediction of movement trends, based on time series data, the system can identify and track the movement path of goods and predict future movement directions. For example, during the loading and unloading process, the system can monitor the movement path of goods in real time to ensure that each update reflects the latest loading status. As more data accumulates, the machine learning model continuously optimizes itself, learns new scene features and threshold adjustment rules, and gradually improves the accuracy and generalization ability of occupancy status judgment. In particular, for complex scenes with frequent lighting changes or diverse cargo types, the system can quickly adapt and make accurate judgments.

[0058] S133, judging whether there is any abnormality in the comprehensive judgment result of the occupancy status based on the movement trend, and if not, determining the comprehensive judgment result of the occupancy status as the final judgment result.

[0059] Among them, the occupancy status is judged based on the movement trend to see if there is an abnormality in the comprehensive judgment result. For example, if the state of a cell is obviously inconsistent with other adjacent cells, or its movement trend is not as expected, it may indicate an abnormal situation. When an abnormal situation is detected, the system will automatically adjust the threshold parameters to avoid misjudgment. For example, when some areas cannot obtain clear images due to occlusion, the system can supplement them with data provided by other sensors (such as depth cameras); and for interference caused by reflective surfaces, the system will adjust the threshold parameters to ensure the accuracy of the judgment results.

[0060] In addition, if the system does not detect any abnormal situation, the comprehensive judgment result of the occupancy status is determined as the final judgment result. This means that the status of each cell has been fully verified and is consistent with the actual loading situation.

[0061] Furthermore, the method further comprises:

[0062] S134, extracting corresponding image features from the real-time image set, classifying the current scene corresponding to the image features through a pre-trained convolutional neural network, and obtaining a corresponding scene classification result.

[0063] Among them, image features include color distribution, texture information, edge contours and depth information; since the camera may be affected by changes in ambient lighting or other factors, resulting in noise in the image, denoising processing is required, such as applying a Gaussian filter or a bilateral filter; in order to improve computing efficiency and model inference speed, the original image may be adjusted to a specific resolution, such as the common 1024x768 or smaller size.

[0064] The features extracted from the real-time images are then input into another specially trained classification model, which is able to identify different loading states or cargo configuration modes, such as empty, partially loaded or fully loaded. Based on the results of the classification model, a specific scene label is assigned to each analyzed image frame, which helps to subsequently select appropriate processing strategies for different scenarios.

[0065] S135, predicting the optimal pixel level threshold and / or depth information threshold under the current conditions according to the scene classification result, and adjusting the threshold parameters.

[0066] Among them, the threshold parameters include pixel-level thresholds and depth information thresholds. Based on the above scene classification results, the system can predict the most suitable pixel-level thresholds and / or depth information thresholds under the current conditions. For example, in an unloaded state, a higher sensitivity may be required to detect small changes; while in a fully loaded state, the threshold can be appropriately lowered to reduce false alarms. As time and conditions change, the system will continuously collect new data samples and update its internal threshold adjustment rules accordingly, thereby achieving more intelligent and flexible occupancy status judgment. In addition, a feedback mechanism may be introduced so that the system can further optimize the threshold setting based on the experience or needs of actual operators.

[0067] Through real-time image processing technology, combined with advanced machine learning algorithms, it accurately understands and responds to dynamic changes inside the truck. Specifically, it uses a pre-trained convolutional neural network to extract features from real-time images and classify the current scene based on these features. Then, based on the classification results, it predicts the optimal pixel-level threshold and depth information threshold under the current conditions, and dynamically adjusts these threshold parameters. This method not only improves the accuracy of occupancy status judgment, but also automatically optimizes the judgment criteria according to different scenarios, enhancing the adaptability and reliability of the system. The ultimate goal is to ensure that the cargo status inside the truck can be efficiently and accurately monitored and managed in various complex environments.

[0068] In addition, the method includes:

[0069] S1356, real-time monitoring of the cell distribution inside the truck, and detecting whether there is unreasonable distribution in a specific area based on the cell distribution.

[0070] Among them, multiple sensor nodes are deployed inside the truck, including cameras, depth sensors and infrared sensors, etc., to obtain high-precision real-time images and distance data. The data collected by the sensors are transmitted to the central processing unit through a high-speed network. The system receives and processes new data packets regularly (such as every second or as needed) to ensure the real-time monitoring information; based on the rules or adaptive grid division in the digital model of the truck, the state changes of each cell are continuously tracked. The system will record information such as the occupancy of each cell, the type of cargo and its position changes.

[0071] The data from different sensors are integrated to eliminate the blind spots that may exist in a single perspective and ensure comprehensive monitoring of each cell. For example, combining RGB images, depth information and thermal imaging data can more accurately determine whether a cell is occupied.

[0072] S137: If so, determine the corresponding unreasonable area, and fine-tune the unreasonable area preferentially based on a preset fine-tuning strategy.

[0073] Among them, machine learning or statistical analysis methods are used to define a reasonable cell distribution pattern. For example, when unloaded, most cells should be empty, and when fully loaded, most cells should be occupied. For the cell distribution in each area, the deviation from the preset reasonable pattern is calculated. If the deviation exceeds the set threshold, it is considered that the area has an unreasonable distribution.

[0074] Once unreasonable distribution areas are detected, the system automatically locates and marks these areas. The marked information includes the area boundaries, the cell numbers involved, and the preliminary diagnosis results (such as overcrowded stacking, too many vacancies, etc.); according to the degree of unreasonableness and potential impact, each detected unreasonable area is assigned a priority, and priority is given to those areas that have a greater impact on the overall loading efficiency, such as those close to the door or areas with high loading and unloading frequency.

[0075] In the embodiments of the present application, a variety of fine-tuning strategies are pre-designed, covering different adjustment methods, such as changing the cell size, adjusting the boundary shape, redistributing the load, etc. These strategies can be flexibly combined according to actual conditions. According to the specific characteristics of the unreasonable area (such as distribution density, cargo type, etc.), the fine-tuning scheme that best suits the current scenario is selected from the strategy library. For example, for overly densely stacked areas, the method of expanding the cell size may be adopted; and for areas with too many vacancies, the method of reducing the cell size or increasing the number of divisions may be adopted.

[0076] S138, after fine-tuning, monitoring the real-time feedback effect of the unreasonable area after fine-tuning, and optimizing the fine-tuning strategy according to the real-time feedback effect.

[0077] During the fine-tuning process, the system continuously monitors the adjusted effects and further optimizes the adjustment strategy based on real-time feedback. For example, if the fine-tuning of a certain area fails to achieve the expected effect, the system will automatically try other fine-tuning solutions until the optimal solution is found.

[0078] It should be pointed out here that a series of performance indicators, such as occupancy, uniformity, stability, etc., are defined to evaluate the actual effect after fine-tuning. The system will regularly collect data on these indicators and compare and analyze them with the results before adjustment. As more data accumulates, the system can continuously optimize itself, learn new scene features and adjustment rules, and gradually improve the effectiveness and generalization ability of the fine-tuning strategy.

[0079] By monitoring the distribution of cells inside the truck in real time, it automatically detects whether there is unreasonable distribution in a specific area. Once an unreasonable area is found, the system will prioritize fine-tuning the area based on the preset fine-tuning strategy, and continuously monitor its real-time feedback effect after fine-tuning, and dynamically optimize the fine-tuning strategy based on the feedback. This method not only ensures the accuracy of local adjustments and the stability of global layout, but also avoids unnecessary interference, significantly improving the rationality of grid division and the accuracy of occupancy status judgment.

[0080] It should be noted here that the method also includes:

[0081] S141, retrieve the pre-set importance differences of cells at different positions, and determine the cell weight value corresponding to each cell based on the importance differences.

[0082] Among them, the importance of each area is defined according to the internal structure of the truck and the actual usage scenario. For example, the area near the door is usually easier to pick up goods and has a high frequency of loading and unloading, so it has a higher importance; while the area in the middle or corner of the car may be relatively unimportant. According to the above evaluation results, a weight value is set for each cell, and the weight value reflects the relative importance of the cell in the overall loading efficiency. For example, cells near the door may be assigned a higher weight value (such as 1.5), while cells in the middle of the car may maintain the default weight value (such as 1.0). In addition, users are allowed to make manual adjustments according to specific needs or special circumstances to ensure that the weight value can flexibly adapt to different application scenarios. For example, in some cases, users may want to prioritize the use of space at the rear of the car. At this time, the weight value of the cell in this area can be temporarily increased.

[0083] S142, calculating the proportion data of occupied cells based on the cell weight values.

[0084] For each cell, the occupancy status is determined based on the pixel-level threshold and depth information threshold set in advance. If the average depth value in a cell is lower than the set idle threshold, it is considered unoccupied; otherwise, it is occupied.

[0085] For the calculation of the proportion data, the weight values ​​of all cells marked as occupied are added together to obtain the total occupied weight. At the same time, the sum of the weight values ​​of all cells is calculated as the benchmark value, and the proportion data of occupied cells is calculated using the following formula: The contribution of each cell's occupancy state is proportional to its weight value, which means that cells with high importance have a greater impact on the final loading rate, making the calculation result closer to the actual loading situation.

[0086] S143, determining the truck loading rate based on the ratio data.

[0087] Among them, the overall loading rate of the truck is determined based on the calculated proportion of occupied cells. This loading rate not only reflects the proportion of cargo occupying the compartment space, but also takes into account the importance differences of cells in different positions, providing a more reasonable and accurate evaluation.

[0088] By calling up the pre-set importance differences of cells in different locations, corresponding weight values ​​are set for each cell, and the proportion data of occupied cells is calculated based on these weight values. The contribution of the occupancy status of each cell is proportional to its weight value, thereby ensuring that key areas (such as areas near doors or areas with high loading and unloading frequency) are fully valued in the loading rate calculation. Finally, the truck loading rate is determined based on the weighted proportion data, providing a more accurate and reasonable loading rate evaluation. This method not only improves the accuracy and practicality of loading rate assessment, but also optimizes cargo stacking and transportation planning, ensuring maximum space utilization.

[0089] A corresponding user interaction interface is provided in the implementation scheme of the present application, so the method also includes:

[0090] S144 obtains a corresponding truck digital model based on the user inputting a specific truck model or selecting a predefined truck template in the user interaction interface.

[0091] Among them, an intuitive user interface is provided, allowing users to enter specific truck models through a drop-down menu, text box or search bar. Users can select common truck templates predefined by the system, or manually enter a specific model. For common truck types, the system provides predefined truck templates, and users can directly select the appropriate template through a graphical interface. Each template contains the standard size and internal structure information of the model.

[0092] When the user selects or enters a specific truck model, the system retrieves the corresponding truck 3D data from the database, which may include CAD files, point cloud data or other forms of geometric descriptions; the retrieved data is loaded into memory to generate a 3D digital model of the truck's interior. The 3D digital model not only contains the spatial geometry of the compartment, but may also include key structural information such as the location of the partitions and the size of the door frames.

[0093] S145, obtaining key dimension points of the user on the digital model of the truck, and obtaining context information input by the user in the user interaction interface.

[0094] Among them, a set of easy-to-use annotation tools are provided on the user interaction interface, such as clicking, dragging, rectangular box selection, etc., so that users can easily mark key size points on the digital model of the truck. For example, users can mark the location of the partition, the starting point of the slope, special loading area, etc. When the user marks, the system will display the marking results in real time and provide instant feedback to ensure that the user's operation is accurate. For example, when the user clicks on a certain location, the system will place an obvious marking point at that location and display its coordinate information.

[0095] In addition to marking key dimension points, users can also enter additional contextual information through text boxes, check boxes, etc. For example, users can indicate whether certain areas require special handling (such as fragile storage areas), whether there are special loading and unloading requirements, etc. The system will associate the contextual information entered by the user with the corresponding cells or areas to ensure that these special requirements can be fully considered during subsequent processing.

[0096] S146, Optimizing the local mesh of the truck digital model based on key dimension points and context information.

[0097] Among them, the system automatically adjusts the local meshing rules according to the key dimension points annotated by the user and the context information entered. For example, for the location where the partition is annotated, the system will use a finer mesh in this area to capture subtle changes; while for ordinary areas, it will maintain a coarser mesh division to improve calculation efficiency. It is necessary to ensure that the adjusted mesh can accurately reflect the key dimension points and context information annotated by the user. For example, if the user specifies a special loading area, the system will ensure that the mesh division of this area can meet the specific loading requirements.

[0098] In addition, by combining multiple optimization objectives, such as mesh accuracy, computational efficiency, and user experience, the local mesh is optimized. For example, genetic algorithms, simulated annealing, and other methods are used to minimize unnecessary complexity while ensuring mesh accuracy. As users continue to add new annotations or modify existing annotations, the system can dynamically adjust the meshing to ensure that it always meets the latest user needs. This flexibility enables the system to maintain efficient and accurate performance in different scenarios.

[0099] Through the user interface, users can enter specific truck models or select predefined templates, and mark key dimension points on the truck digital model and provide contextual information. Based on these user-entered data, the system optimizes the local grid of the truck digital model to ensure that each cell accurately fits the actual physical space, especially when dealing with special structures such as partitions and ramps. This process enhances the flexibility and customization of the system, making the grid division more in line with actual needs, thereby improving the accuracy of occupancy status judgment and the reliability of loading rate assessment.

[0100] The present application embodiment discloses a truck loading rate identification device, referring to Figure 2 , the devices include but are not limited to:

[0101] The cell division module 200 obtains the three-dimensional data of the truck inside the truck, establishes a corresponding digital model of the truck based on the three-dimensional data of the truck, and selects a regular gridding method or an adaptive gridding method based on the digital model of the truck to divide the inside of the truck into a plurality of cells;

[0102] The cell parameter adjustment module 210 obtains real-time image set data of the interior of the truck, adjusts the cell parameters of the interior of the truck based on the real-time image set data, and performs occupancy status judgment on each adjusted cell using a pixel level threshold and / or a depth information threshold based on the image set data;

[0103] The truck loading rate determination module 220 determines the cell occupancy data based on the judgment result of each cell, calculates the proportion data of the occupied state cells based on the cell occupancy data, determines the truck loading rate based on the proportion data, and generates a corresponding visualization report based on the cell occupancy data.

[0104] An embodiment of the present application further discloses an electronic device, comprising a processor, wherein a program of any one of the above-mentioned methods for identifying a truck loading rate is running on the processor.

[0105] An embodiment of the present application further discloses a storage medium storing a program of any one of the above-mentioned methods for identifying a truck loading rate.

[0106] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for identifying a truck loading rate, characterized in that: include: Acquire three-dimensional data of the interior of the truck, establish a corresponding digital model of the truck based on the three-dimensional data of the truck, and select a regular gridding method or an adaptive gridding method based on the digital model of the truck to divide the interior of the truck into a plurality of cells; Acquire real-time image set data of the interior of the truck, adjust the cell parameters of the interior of the truck based on the real-time image set data, and perform occupancy status judgment on each adjusted cell using a pixel-level threshold and / or a depth information threshold based on the image set data; Cell occupancy data is determined based on the judgment result of each cell, proportion data of occupied cells is calculated based on the cell occupancy data, a truck loading rate is determined based on the proportion data, and a corresponding visualization report is generated based on the cell occupancy data.

2. The method for identifying a truck loading rate according to claim 1, characterized in that: The method also includes: Extract corresponding image features from the real-time image set, classify the current scene corresponding to the image features through a pre-trained convolutional neural network, and obtain the corresponding scene classification results, wherein the image features include color distribution, texture information, edge contour and depth information; The optimal pixel level threshold and / or depth information threshold under the current conditions is predicted according to the scene classification result, and the threshold parameters are adjusted, wherein the threshold parameters include the pixel level threshold and the depth information threshold.

3. The truck loading rate identification method according to claim 1, characterized in that: In the process of determining the occupancy state of each adjusted cell by performing pixel level threshold and / or depth information threshold based on the image set data, the method further includes: Acquire a multi-view image set from multiple different sources, and obtain corresponding multi-view feature information from the multi-view image set through a feature extraction algorithm, the feature information including color distribution, texture information, edge contour, depth information and thermal imaging; Based on the multi-view feature information, a comprehensive judgment is made on the occupancy status of each cell, and based on the time series, the position change of the tracked goods in the truck is determined to predict the movement trend of the goods; Based on the movement trend, it is determined whether there is any abnormality in the comprehensive judgment result of the occupancy status. If not, the comprehensive judgment result of the occupancy status is determined as the final judgment result.

4. The method for identifying a truck loading rate according to claim 1, characterized in that: The method also includes: Retrieving the pre-set importance differences of cells at different positions, and determining the cell weight value corresponding to each cell based on the importance differences; Calculating the proportion data of occupied state cells based on the cell weight values, wherein the occupied state contribution of each cell is proportional to its weight value; A truck loading factor is determined based on the ratio data.

5. The method for identifying a truck loading rate according to claim 1, characterized in that: The method also includes: Monitor the cell distribution inside the truck in real time, and detect whether there is unreasonable distribution in a specific area based on the cell distribution; If there is, determine the corresponding unreasonable area, and fine-tune the unreasonable area first based on the preset fine-tuning strategy; After fine-tuning, monitor the real-time feedback effect of fine-tuning in unreasonable areas and optimize the fine-tuning strategy based on the real-time feedback effect.

6. The method for identifying a truck loading rate according to claim 1, characterized in that: Including a user interaction interface, the method also includes: Acquiring a corresponding truck digital model based on an operation in which a user inputs a specific truck model or selects a predefined truck template in a user interaction interface; Obtain key dimension points of the user on the digital model of the truck and obtain context information input by the user in the user interaction interface; The local mesh of the digital model of the truck is optimized based on the key dimension points and the context information.

7. A truck loading rate identification device, characterized in that: include: A cell division module, which obtains three-dimensional data of the interior of the truck, establishes a corresponding digital model of the truck based on the three-dimensional data of the truck, and selects a regular gridding method or an adaptive gridding method based on the digital model of the truck to divide the interior of the truck into a plurality of cells; A cell parameter adjustment module, which obtains real-time image set data of the interior of the truck, adjusts the cell parameters of the interior of the truck based on the real-time image set data, and performs occupancy status judgment on each adjusted cell using a pixel-level threshold and / or a depth information threshold based on the image set data; The truck loading rate determination module determines the cell occupancy data based on the judgment result of each cell, calculates the proportion data of the occupied state cells based on the cell occupancy data, and uses the proportion data to determine the truck loading rate, and generates a corresponding visualization report based on the cell occupancy data.

8. An electronic device, characterized in that: It comprises a processor, in which a program of the truck loading rate identification method as described in any one of claims 1 to 6 is running.

9. A storage medium, characterized in that: A program storing the truck loading rate identification method according to any one of claims 1 to 6.

Images