Image detection and multi-stage anti-collision optimization method and system for baggage conveying

Abstract

The present application relates to the technical field of luggage conveying, in particular to a luggage conveying oriented image detection and multi-stage anti-collision optimization method and system. The method comprises obtaining luggage conveying equipment operation state data and real-time image data, obtaining luggage space motion features through feature extraction and semantic analysis, converting to a physical coordinate system through coordinate mapping to obtain a space distribution result, calculating a plurality of detection area collision probability values to generate a hierarchical risk assessment result and a multi-stage control instruction in combination with the operation state data, controlling the driving mechanism and collecting feedback information to update the assessment result. The present application realizes multi-stage active anti-collision of the luggage conveying process, and improves the anti-collision accuracy and real-time performance.

Description

Technical Field

[0001] This invention relates to the field of baggage handling technology, and in particular to an image detection and multi-level collision avoidance optimization method and system for baggage handling. Background Technology

[0002] In the field of baggage handling systems, especially in airports and logistics centers, automated baggage conveying and sorting equipment must ensure that baggage moves smoothly and without collisions along a predetermined path. Current conventional practices mainly rely on a combination of multiple sensors for collision detection. For example, photoelectric sensors, ultrasonic sensors, or lidar are installed at key nodes of the conveyor equipment to trigger stop or deceleration commands by monitoring the baggage's passage or the presence of obstacles. Some systems use mechanical anti-collision bars or buffer devices as passive protection, triggering emergency braking through physical contact. Furthermore, the control logic is usually based on preset fixed thresholds, immediately executing a single stop operation when sensors detect foreign objects or baggage spacing becomes too small.

[0003] These conventional practices have two significant drawbacks. Firstly, sensor placement is limited by physical installation locations, only covering local monitoring points and failing to capture the spatial distribution and movement trends of all luggage along the conveyor path in real time, leading to a severe lag in collision risk assessment. For example, when luggage slips or deviates on the conveyor belt, fixed-position sensors struggle to detect changes promptly, easily causing false alarms or missed alarms. Secondly, existing control strategies lack a tiered response mechanism; once a collision avoidance signal is triggered, the entire system is immediately shut down or given an emergency stop. This not only disrupts the normal transport process but also increases the time cost of manual intervention for cleaning and restarting, significantly reducing luggage handling efficiency during peak hours. Furthermore, frequent contact with mechanical collision avoidance devices accelerates equipment wear and tear, resulting in high maintenance costs. Therefore, the industry urgently needs a technological solution capable of real-time luggage dynamics sensing and intelligent tiered collision avoidance. Summary of the Invention

[0004] This invention provides an image detection and multi-level collision avoidance optimization method and system for baggage transportation, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides an image detection and multi-level collision avoidance optimization method for baggage transportation, comprising: Acquire operational status data of baggage conveyor equipment and real-time image data collected by image acquisition devices deployed along the conveyor path; Feature extraction and semantic parsing are performed on the real-time image data to obtain the spatial motion features of the luggage objects on the transport path; Based on the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, the spatial motion characteristics of the luggage object are mapped from the image coordinate system to the physical coordinate system of the conveying equipment to obtain the spatial distribution result of the luggage object; Based on the spatial distribution results and the operating status data, the collision probability values ​​in multiple preset detection areas are calculated to obtain the graded collision risk assessment results, and corresponding multi-level control commands are generated based on the graded collision risk assessment results. Based on the multi-level control commands, control operations are performed on the drive mechanism of the conveying equipment, and data on changes in the motion state of the luggage object after control execution are collected to obtain feedback information. The feedback information is then used to update the graded collision risk assessment results.

[0006] Feature extraction and semantic parsing are performed on the real-time image data to obtain the spatial motion features of luggage objects along the transport path, including: Perform multi-level feature extraction operations on the real-time image data to obtain visual representation information of the luggage object at different levels of abstraction; Based on the visual representation information, semantic classification and target localization are performed to identify the category attributes and spatial pose of the luggage object in the image coordinate system; Based on the spatial pose of the luggage object, a cross-frame association relationship of the luggage object is established between consecutive frames of image data, and a motion trajectory sequence of the luggage object is constructed. Based on the motion trajectory sequence, a time-series analysis is performed on the historical motion behavior of the luggage object to extract the motion pattern features of the luggage object, and the position change and posture change of the luggage object within a preset time window are calculated based on the motion pattern features. Based on the spatial pose, motion trajectory sequence, position change, and attitude change, the spatial motion features, including the current state description and the predicted state description, are determined.

[0007] Based on the motion trajectory sequence, a time-series analysis is performed on the historical motion behavior of the luggage object to extract its motion pattern features. Then, based on these features, the changes in position and posture of the luggage object within a preset time window are calculated, including: The motion trajectory sequence is segmented, and the motion trajectory sequence is divided into multiple trajectory segments according to the motion change characteristics of the luggage object; Statistical analysis is performed on each trajectory segment to extract the motion feature parameters of the trajectory segment, thereby obtaining a trajectory segment feature set; Based on the set of trajectory segment features, the motion behavior pattern of the luggage object on the transport path is determined, and the motion behavior pattern is associated and mapped with the topological structure information of the transport path to obtain motion pattern features; Based on the motion pattern characteristics, combined with the current location of the luggage object on the transport path and the topological information of the transport path, historical trajectory segments corresponding to the current transport path location are searched. Based on the duration of motion and average velocity amplitude of the historical trajectory segment, the change in position of the luggage object within a preset time window is calculated, and the change in attitude of the luggage object within the preset time window is calculated based on the directional deflection angle of the historical trajectory segment.

[0008] Based on the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, the spatial motion characteristics of the luggage object are mapped from the image coordinate system to the physical coordinate system of the conveying equipment to obtain the spatial distribution result of the luggage object, including: Based on the installation pose parameters of the image acquisition device, determine the coordinate transformation relationship between the image coordinate system and the physical coordinate system; The spatial motion features of the luggage object include position and posture descriptions in the image coordinate system, which are preprocessed by coordinate transformation to determine the projection area of ​​the luggage object on the image plane. The spatial distance between the luggage object and the image acquisition device is calculated based on the position of the projection area on the image plane and the observation angle. Based on the spatial distance and the coordinate transformation relationship, a coordinate mapping operation is performed on the position description and posture description to transform the position description and posture description from the image coordinate system to the physical coordinate system, thereby obtaining the spatial position and spatial posture of the luggage object in the physical coordinate system; The spatial position and orientation of the luggage object in the physical coordinate system are correlated with the structural layout information of the conveying equipment to determine the distribution position of the luggage object on the conveying equipment, thus obtaining the spatial distribution result.

[0009] Based on the spatial distribution results and the operational status data, collision probability values ​​within multiple preset detection areas are calculated to obtain a graded collision risk assessment result. Then, based on the graded collision risk assessment result, corresponding multi-level control commands are generated, including: Based on the structural layout information and topology information of the conveying equipment, multiple preset detection areas are defined on the conveying equipment, and a risk level identifier is assigned to each preset detection area. Extract the spatial position and spatial orientation of the luggage object in the physical coordinate system from the spatial distribution results, determine the preset detection area to which the luggage object belongs, and obtain the running status data corresponding to the preset detection area to which the luggage object belongs; Based on the spatial location, spatial posture, and operational status data of the luggage object, calculate the spatial distance and relative motion relationship between the luggage object and other luggage objects within the preset detection area; Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage object, the collision probability value in the preset detection area is calculated. The collision probability value is then associated with the risk level identifier corresponding to the preset detection area to obtain a graded collision risk assessment result. Based on the graded collision risk assessment results, a preset control strategy mapping table is queried to generate multi-level control commands corresponding to the risk level identifier.

[0010] Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage object, the collision probability value within the preset detection area is calculated. The collision probability value is then correlated with the risk level identifier corresponding to the preset detection area to obtain a graded collision risk assessment result, including: Extract positional and orientation changes from the spatial motion characteristics of the luggage object; The minimum safe distance between the luggage object and other luggage objects within the preset detection area is calculated based on the spatial distance. Based on the relative motion relationship and the position change, the movement trajectory of the luggage object within the preset time window is predicted. Based on the minimum safe distance and the movement trajectory, the intersection position of the trajectory of the luggage object with that of other luggage objects in the preset detection area is determined, and the degree of posture conflict at the trajectory intersection position is calculated according to the posture change amount; Based on the minimum safe distance, the trajectory intersection position, and the degree of attitude conflict, calculate the collision probability value within the preset detection area; Based on the collision probability value and the risk level identifier, the collision probability value is mapped to a risk level, and the collision probability value is divided into different risk level ranges. The risk level range is associated with the risk level identifier to obtain the graded collision risk assessment result.

[0011] Based on the multi-level control commands, control operations are performed on the drive mechanism of the conveying equipment, and data on changes in the motion state of the luggage object after control execution are collected to obtain feedback information. The feedback information is then used to update the graded collision risk assessment results, including: The control type identifier and control parameters contained in the multi-level control instructions are analyzed, and the target drive unit in the drive mechanism of the conveying equipment that needs to perform control operations is determined according to the control type identifier. The control parameters are converted into execution instructions for the target drive unit, thereby driving the target drive unit to perform the corresponding control operation; The image acquisition device acquires a sequence of images of the luggage object after the control is executed, and extracts data on the change in the motion state of the luggage object after the control is executed from the sequence of images of the luggage object after the control is executed. The motion state change data is correlated and compared with the control parameters to calculate the deviation between the actual effect and the expected effect of the control execution; Feedback information is generated based on the deviation amount. The collision probability value calculation method in the preset detection area is adjusted according to the feedback information. The collision probability value in the preset detection area is recalculated based on the adjusted collision probability value calculation method, and the graded collision risk assessment result is updated.

[0012] A second aspect of the present invention provides an image detection and multi-level collision avoidance optimization system for baggage transportation, comprising: The data acquisition unit is used to acquire the operating status data of the baggage conveyor equipment and the real-time image data collected by the image acquisition devices deployed along the conveyor path; The feature parsing unit is used to extract features and perform semantic parsing on the real-time image data to obtain the spatial motion features of the luggage objects on the transport path; The coordinate mapping unit is used to map the spatial motion characteristics of the luggage object from the image coordinate system to the physical coordinate system of the conveying equipment according to the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, so as to obtain the spatial distribution result of the luggage object; The risk assessment unit is used to calculate the collision probability values ​​in multiple preset detection areas based on the spatial distribution results and the operating status data, obtain the graded collision risk assessment results, and generate corresponding multi-level control commands based on the graded collision risk assessment results. The feedback update unit is used to perform control operations on the drive mechanism of the conveying equipment based on the multi-level control commands, collect data on the change in the motion state of the luggage object after the control is executed to obtain feedback information, and use the feedback information to update the graded collision risk assessment results.

[0013] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] This invention utilizes image acquisition and semantic parsing technology to extract the spatial motion features of luggage objects. Combined with equipment operating status data, it obtains accurate spatial distribution results in the physical coordinate system, which greatly improves the detection accuracy of luggage positions on the transport path. It avoids the blind spot problem caused by occlusion or interference of traditional sensors, and realizes continuous and dynamic tracking from image to physical space, effectively supporting the reliability of subsequent collision avoidance decisions.

[0016] This invention calculates the collision probability values ​​of multiple preset areas based on spatial distribution results, which can generate graded collision risk assessment results and output corresponding multi-level control commands. This eliminates the need for the conveying equipment to perform emergency braking every time a potential collision occurs. Instead, it takes differentiated actions such as deceleration, lane changing, or pausing according to the risk level, which significantly reduces the false trigger rate. Under the premise of ensuring safety, it maintains the continuity and efficiency of baggage transportation and avoids equipment wear and baggage damage caused by frequent emergency stops.

[0017] This invention generates feedback information by collecting data on changes in the motion state of luggage after control execution, and uses this information to update the graded collision risk assessment results in real time, forming a closed-loop optimization mechanism. This enables the system to autonomously adapt to dynamic factors such as different luggage sizes, changes in transport speed, and environmental interference, continuously enhancing the accuracy of risk assessment and the adaptability of control strategies. Under long-term operation, it can maintain stable anti-collision performance and reduce the need for manual intervention. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the image detection and multi-level collision avoidance optimization method for baggage transportation according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the process of determining spatial motion characteristics according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0020] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0021] Figure 1 This is a flowchart illustrating the image detection and multi-level collision avoidance optimization method for baggage transportation according to an embodiment of the present invention. Figure 1 As shown, the method includes: Acquire operational status data of baggage conveyor equipment and real-time image data collected by image acquisition devices deployed along the conveyor path; Feature extraction and semantic parsing are performed on the real-time image data to obtain the spatial motion features of the luggage objects on the transport path; Based on the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, the spatial motion characteristics of the luggage object are mapped from the image coordinate system to the physical coordinate system of the conveying equipment to obtain the spatial distribution result of the luggage object; Based on the spatial distribution results and the operating status data, the collision probability values ​​in multiple preset detection areas are calculated to obtain the graded collision risk assessment results, and corresponding multi-level control commands are generated based on the graded collision risk assessment results. Based on the multi-level control commands, control operations are performed on the drive mechanism of the conveying equipment, and data on changes in the motion state of the luggage object after control execution are collected to obtain feedback information. The feedback information is then used to update the graded collision risk assessment results.

[0022] Figure 2 This is a schematic diagram illustrating the process of determining spatial motion characteristics according to an embodiment of the present invention. Figure 2 As shown, feature extraction and semantic parsing are performed on the real-time image data to obtain the spatial motion features of luggage objects along the transport path, including: Perform multi-level feature extraction operations on the real-time image data to obtain visual representation information of the luggage object at different levels of abstraction; Based on the visual representation information, semantic classification and target localization are performed to identify the category attributes and spatial pose of the luggage object in the image coordinate system; Based on the spatial pose of the luggage object, a cross-frame association relationship of the luggage object is established between consecutive frames of image data, and a motion trajectory sequence of the luggage object is constructed. Based on the motion trajectory sequence, a time-series analysis is performed on the historical motion behavior of the luggage object to extract the motion pattern features of the luggage object, and the position change and posture change of the luggage object within a preset time window are calculated based on the motion pattern features. Based on the spatial pose, motion trajectory sequence, position change, and attitude change, the spatial motion features, including the current state description and the predicted state description, are determined.

[0023] After acquiring real-time image data generated by the baggage handling equipment during actual operation, the first step is to perform deep feature extraction on the image content. To achieve this, a convolutional neural network architecture can be used to perform forward propagation operations on the input image frames. In the shallow layers of the network, convolutional kernels filter the raw pixel values, extracting low-level visual features such as edge contours and texture distribution. As the network deepens, through the alternating effects of pooling and convolution, local features are gradually aggregated into higher-level semantic representations. During this process, feature output nodes can be set at different depths in the network to obtain visual representation information of baggage objects at multiple abstract levels, including edges, local textures, overall shape, and semantic categories. This multi-level visual representation information constitutes the foundational data for subsequent semantic parsing and spatial localization.

[0024] Based on the visual representation information obtained above, semantic classification and target localization operations are further performed on the luggage objects in the image. Specifically, the extracted high-level feature vectors can be input into the classifier module, and through nonlinear transformation of fully connected layers or convolutional layers, the confidence distribution of each category label is output, thereby identifying the category attribute of the current luggage object, such as hard-shell suitcase, soft backpack, tote bag, etc. At the same time, the position of the luggage object in the image coordinate system is accurately located through the regression branch of the object detection network. This regression branch outputs the coordinate parameters of the rectangular bounding box, including the pixel coordinates of the upper left and lower right corners, or the coordinates of the center point and the width and height dimensions. In some implementations, the orientation angle information of the luggage object can also be further output, thereby fully describing its spatial pose in the image coordinate system. The results of the above semantic classification and target localization operations provide a reliable single-frame detection foundation for subsequent cross-frame association in the temporal dimension.

[0025] When establishing cross-frame associations of luggage objects across multiple consecutive frames of image data, it is necessary to solve the matching problem between targets in different frames. An association strategy combining appearance features and motion prediction can be adopted. For each luggage object detected in the current frame, its visual representation vector obtained during the feature extraction stage is extracted as its appearance descriptor. For targets tracked in the previous frame, their possible positions in the current frame are predicted based on their historical motion states, and the spatial distance between the predicted position and the detection result in the current frame is calculated. Appearance similarity and spatial distance are weighted and fused to obtain a comprehensive association cost matrix. The optimal matching relationship is then solved using the Hungarian algorithm or a greedy matching strategy. Successfully matched objects are considered as observations of the same luggage at different times, thus establishing continuous cross-frame associations on the timeline. For luggage objects successfully associated in multiple consecutive frames, their spatial pose data in each frame are arranged chronologically to construct a motion trajectory sequence for the luggage object. This sequence records the evolution of the luggage object's position and posture over time during the transportation process.

[0026] Based on the constructed motion trajectory sequence, a temporal analysis is performed on the historical movement behavior of the luggage to extract its motion pattern features. The position coordinates in the trajectory sequence can be organized into a temporal vector in chronological order. This temporal vector is then encoded using a sliding window or recurrent neural network structure to capture the temporal dependencies of the luggage's movement. During this process, the statistical properties of the trajectory sequence can be calculated, such as the mean and variance of the position coordinates, and the statistical distribution of position increments between adjacent frames, thereby identifying the motion pattern features of the luggage, including uniform motion, accelerated motion, stationary states, or periodic oscillations. Furthermore, based on these motion pattern features, the changes in position and posture of the luggage within a preset time window are calculated.

[0027] The change in position can be calculated by performing a difference operation on the coordinate data in the trajectory sequence. Let the luggage object be at time... With time The position coordinates are respectively and Then the change in position can be expressed as To more comprehensively characterize motion behavior, the change in velocity vector, i.e., acceleration information, can also be calculated. Let the positions at three adjacent moments be... , , The corresponding time interval is Then the velocity vector is The acceleration vector is .

[0028] The calculation of attitude change measures focuses on the temporal evolution of the orientation angle of the luggage object, assuming that the luggage object at time... and The orientation angles are respectively and The change in attitude can be expressed as In practical applications, to avoid calculation errors caused by angle jumps, the angle difference can be normalized to map it to... Within the interval. For a time series of attitude angles, its rate of change, i.e., angular velocity, can also be calculated. and angular acceleration These time-series derivatives can more precisely characterize the rotational motion of luggage during transport.

[0029] After calculating the position and pose changes, the spatial pose, motion trajectory sequence, position changes, and pose changes of the luggage object are integrated to form spatial motion features that include a current state description and a predicted state description. The current state description includes the luggage object's position coordinates, bounding box size, orientation angle, category label, and confidence level in the latest frame. The predicted state description, based on historical trajectory sequences and motion pattern features, uses kinematic models or machine learning prediction models to infer the luggage object's position and pose at several future moments. Commonly used prediction methods include Kalman filtering, particle filtering, or temporal prediction models based on recurrent neural networks.

[0030] For example, when using Kalman filtering for prediction, the state vector of the luggage object can be defined as... ,in , For position coordinates, , For velocity components, For the orientation angle, Let be the angular velocity. The state transition equation can be expressed as: ,in Here is the state transition matrix. This represents process noise. The observation equation is: ,in For the observation matrix, To mitigate observation noise, the Kalman filter alternates between prediction and update steps, allowing for the acquisition of the optimal estimate of the baggage object's state and its covariance matrix at each time step, thus enabling prediction of its future state.

[0031] In time-series prediction methods based on recurrent neural networks, historical trajectory sequences can be used as input. Long Short-Term Memory (LSTM) networks or gated recurrent units are used to model the temporal dependencies. The network's output layer generates predicted position coordinates and attitude angles for several future moments. During training, historical real trajectory data is used as a supervisory signal, and the network parameters are optimized through backpropagation to minimize the mean squared error between the predicted results and the actual values. After sufficient training, the network can accurately predict the future movement trend of luggage objects based on current and historical observation data.

[0032] The resulting spatial motion features are a structured dataset that includes both complete state information of the luggage object at the current moment and predictions of its future state based on historical motion patterns. These features provide comprehensive and accurate input data for subsequent coordinate system mapping, spatial distribution calculation, and collision risk assessment. Through the integrated processing flow of multi-level feature extraction, semantic parsing, cross-frame correlation, temporal analysis, and motion prediction, accurate extraction and effective description of the spatial motion features of luggage objects can be achieved in complex luggage handling scenarios, laying a solid foundation for the stable operation of the entire collision avoidance optimization system.

[0033] Based on the motion trajectory sequence, a time-series analysis is performed on the historical motion behavior of the luggage object to extract its motion pattern features. Then, based on these features, the changes in position and posture of the luggage object within a preset time window are calculated, including: The motion trajectory sequence is segmented, and the motion trajectory sequence is divided into multiple trajectory segments according to the motion change characteristics of the luggage object; Statistical analysis is performed on each trajectory segment to extract the motion feature parameters of the trajectory segment, thereby obtaining a trajectory segment feature set; Based on the set of trajectory segment features, the motion behavior pattern of the luggage object on the transport path is determined, and the motion behavior pattern is associated and mapped with the topological structure information of the transport path to obtain motion pattern features; Based on the motion pattern characteristics, combined with the current location of the luggage object on the transport path and the topological information of the transport path, historical trajectory segments corresponding to the current transport path location are searched. Based on the duration of motion and average velocity amplitude of the historical trajectory segment, the change in position of the luggage object within a preset time window is calculated, and the change in attitude of the luggage object within the preset time window is calculated based on the directional deflection angle of the historical trajectory segment.

[0034] After obtaining the motion trajectory sequence of the luggage, in-depth temporal analysis is needed to extract the motion pattern features of the luggage. To achieve this, the motion trajectory sequence is first segmented. Specifically, by detecting abrupt changes in velocity amplitude, inflection points in direction angle, and zero-crossing points in acceleration, the entire trajectory sequence is divided into multiple trajectory segments. In practice, a threshold for velocity amplitude change is set as follows: m / s, the threshold for change in direction angle is When the velocity amplitude changes more than 100 times in adjacent moments Or the change in orientation angle exceeds The moment in time is considered the trajectory segmentation point. This segmentation strategy can distinguish the movement behavior of luggage objects in straight transport sections, turning areas, merging areas, and diverging areas, providing a foundation for subsequent motion pattern recognition.

[0035] After completing trajectory segmentation, statistical analysis is performed on each trajectory segment to extract motion feature parameters. For the first segment... Each trajectory segment records its start time. and the end time Calculate the duration of the segment Simultaneously, the average speed amplitude of the luggage objects within this segment was statistically analyzed. Maximum speed amplitude and speed standard deviation The average velocity amplitude is obtained by calculating the arithmetic mean of the velocity amplitudes at all times within the segment, while the velocity standard deviation reflects the severity of velocity fluctuations within that segment. Furthermore, the directional characteristic parameters of the segment, including the segment's initial direction angle, need to be extracted. End direction angle and the angle of directional deflection To describe the spatial morphology of a trajectory segment, the curvature characteristic parameters of the segment are calculated by fitting the spatial path curve of the trajectory segment and determining the radius of curvature. This characterizes the degree of trajectory curvature. All the above feature parameters are organized into a feature vector. The feature vectors of all trajectory segments constitute the feature set of the trajectory segments. .

[0036] Based on the feature set of trajectory segments, cluster analysis is used to determine the movement behavior pattern of luggage objects along the transportation path. Specifically, the K-means clustering algorithm is used to divide all trajectory segment feature vectors into... There are several clusters, each representing a typical movement behavior pattern. In practical applications, it is usually set as follows: This corresponds to five basic motion modes: linear constant speed conveying, linear acceleration conveying, linear deceleration conveying, left-turn conveying, and right-turn conveying. After clustering, the cluster center is calculated for each cluster. The cluster center represents the first Typical characteristics of this motion behavior pattern. To associate and map the motion behavior pattern with the topological information of the delivery path, a topological map of the delivery path is first obtained. This topological map divides the delivery path into multiple segments, each with a unique segment identifier. and road segment type labels Road segment type labels include categories such as straight segments, left-turn segments, right-turn segments, merging segments, and diverging segments. By analyzing the road segment location of each trajectory segment, a correspondence between trajectory segments and road segment labels is established. Furthermore, the frequency of various motion behavior patterns occurring on each road segment is statistically analyzed to construct a correlation matrix between road segments and motion patterns. Matrix elements Indicates the section of road Movement mode appears on The frequency of motion patterns is determined. By normalizing the correlation matrix, the probability distribution of each motion pattern on each road segment is obtained, thus completing the correlation mapping between motion pattern characteristics and the topology of the transport path.

[0037] After obtaining the motion pattern features, based on the current location of the luggage object on the delivery path and the topological information of the delivery path, the historical trajectory segment corresponding to the current delivery path location is searched. Specifically, this is first achieved by using the current physical coordinates of the luggage object. Determine the road segment sign where it is located. Subsequently, all trajectory segments passing through this road segment are retrieved from the historical trajectory database to obtain a set of candidate trajectory segments. To improve matching accuracy, candidate trajectory segments are further filtered based on the current motion state characteristics of the luggage object. The current motion state feature vector is calculated. Euclidean distance between the feature vectors of the candidate trajectory segments The algorithm selects several trajectory segments with the smallest distances as matching historical trajectory segments. Typically, the segments with the smallest distances are selected from the previous segments. A trajectory segment, denoted as .

[0038] Based on the matched historical trajectory segments, the change in position of the luggage object within a preset time window is calculated. The preset time window is typically set to [value missing]. s represents the motion prediction window for the next 2 seconds. For each matched historical trajectory segment... Extract the duration of motion in the segment. and average velocity amplitude If the duration of a historical trajectory segment Greater than the preset time window Then extract the segment before that. Trajectory data within a time period; if Less than Then extrapolate based on the average velocity of that segment. Calculate the positional change of the historical trajectory segment within a preset time window. The positional changes of all matched historical trajectory segments are weighted and averaged, with weighting coefficients... It is inversely proportional to the Euclidean distance, that is Finally, the change in position of the luggage object within the preset time window is obtained. .

[0039] The calculation of attitude changes is also based on matching historical trajectory segments. For each historical trajectory segment... Extract the directional deflection angle of the segment. If the duration of a historical trajectory segment With preset time window The inconsistency necessitates time-scale normalization of the directional deflection angle. The normalized angular velocity is then calculated. Then, the corresponding attitude change is calculated according to the preset time window. Similarly, a weighted average method is used to calculate the change in posture of the luggage object within a preset time window. This change in attitude can reflect the orientation deflection trend of the luggage object over a period of time, providing important predictive information for subsequent collision risk assessment.

[0040] To further improve prediction accuracy, a dynamic adjustment mechanism can be introduced. When a significant deviation is detected between the actual movement state of the luggage object and the predicted results of historical trajectory segments, the weight coefficients of the historical trajectory segments are dynamically adjusted or the trajectory segments are re-matched. Specifically, the deviation between the predicted position and the actual observed position is calculated in real time. , when the deviation Exceeding the preset threshold At time m, the trajectory segment re-matching process is triggered. Simultaneously, for attitude changes, the deviation between the predicted orientation angle and the actual observed orientation angle is calculated. , when the deviation Exceeding the preset threshold At the same time, a re-matching process is triggered. This dynamic adjustment mechanism can adapt to abnormal movements of luggage during transportation, ensuring the accuracy and robustness of the prediction results.

[0041] In practical applications, to reduce computational overhead, the historical trajectory database can be preprocessed and indexed for optimization. An index structure can be built based on road segment identifiers, enabling rapid location of historical trajectory segments corresponding to the current road segment. Simultaneously, for abnormal trajectory segments with low occurrence frequency, a minimum frequency threshold can be set for filtering to prevent isolated samples from interfering with motion pattern recognition results. Furthermore, as the system runs, the historical trajectory database will continuously grow, requiring periodic data cleaning and compression to retain representative trajectory segments and delete redundant and outdated data, ensuring system efficiency.

[0042] Through the complete time-series analysis process described above, motion pattern features can be extracted from the historical movement behavior of luggage objects. Based on these features, the changes in position and attitude of the luggage objects within a preset time window can be accurately calculated. This predictive information provides crucial input data for subsequent collision risk assessment and multi-level collision avoidance control, significantly improving the safety and reliability of the luggage handling system.

[0043] Based on the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, the spatial motion characteristics of the luggage object are mapped from the image coordinate system to the physical coordinate system of the conveying equipment to obtain the spatial distribution result of the luggage object, including: Based on the installation pose parameters of the image acquisition device, determine the coordinate transformation relationship between the image coordinate system and the physical coordinate system; The spatial motion features of the luggage object include position and posture descriptions in the image coordinate system, which are preprocessed by coordinate transformation to determine the projection area of ​​the luggage object on the image plane. The spatial distance between the luggage object and the image acquisition device is calculated based on the position of the projection area on the image plane and the observation angle. Based on the spatial distance and the coordinate transformation relationship, a coordinate mapping operation is performed on the position description and posture description to transform the position description and posture description from the image coordinate system to the physical coordinate system, thereby obtaining the spatial position and spatial posture of the luggage object in the physical coordinate system; The spatial position and orientation of the luggage object in the physical coordinate system are correlated with the structural layout information of the conveying equipment to determine the distribution position of the luggage object on the conveying equipment, thus obtaining the spatial distribution result.

[0044] In baggage handling systems, image acquisition devices are often distributed in different spatial locations and at different installation angles. The baggage data they observe needs to be uniformly converted to the physical coordinate system of the handling equipment in order to provide a reliable basis for subsequent collision risk assessment and control decisions. Therefore, a complete mapping process from the image coordinate system to the physical coordinate system needs to be established.

[0045] First, obtain the pose parameters recorded during installation of each image acquisition device. These parameters include the camera's three-dimensional installation position in the physical coordinate system and the rotation angle of the camera's optical axis relative to the reference plane of the conveying equipment. Let the camera's installation position in the physical coordinate system be denoted as... The pitch angle of the camera's optical axis relative to the horizontal reference plane of the conveying equipment is... The yaw angle relative to the direction of transport is The roll angle about the optical axis is Based on these installation pose parameters, a rotation matrix from the camera coordinate system to the physical coordinate system can be constructed. Translation vector The rotation matrix is ​​obtained by applying rotation transformations about each axis in sequence, and the translation vector is the coordinate of the camera's mounting position.

[0046] On the image plane, the spatial motion features of the luggage object are expressed in pixel coordinates. Let the center pixel coordinates of the bounding box of the luggage object on the image plane be denoted as... The bounding box has a pixel width and height of 100 pixels. and To convert this two-dimensional image information into its position in three-dimensional physical space, it is necessary to first determine the projection area of ​​the luggage object on the image plane. This is then combined with the camera intrinsic parameter matrix. This can convert pixel coordinates to normalized camera coordinates. Let the camera's focal length be denoted as . The principal point coordinates are Then the normalized camera coordinates are: ; Normalized coordinates This describes the ray direction from the camera's optical center to the center of the luggage object's projection. However, this ray direction alone is insufficient to determine the luggage object's absolute position in three-dimensional space; the spatial distance between the luggage object and the image acquisition device also needs to be calculated. .

[0047] The calculation of spatial distance relies on estimating the actual size of the luggage object and measuring its projected size on the image plane. In luggage transport scenarios, luggage objects are typically placed on the conveyor belt surface, with their bottoms in contact with the conveyor belt plane. Let the height of the conveyor belt surface in the physical coordinate system be denoted as . Assuming the height of the luggage object aligns with the vertical direction of the physical coordinate system, the physical height of the luggage object's bottom is... The physical height of the top is ,in The physical height of the luggage object.

[0048] By analyzing the projection area of ​​the luggage object on the image plane, the ordinates of its bottom and top edges in the image coordinate system can be obtained. and Combined with camera installation height and pitch angle This establishes a mapping from the image's vertical coordinate to the physical height. Based on perspective projection geometry, the intersection of the ray corresponding to the bottom of the luggage object and the conveyor belt plane represents the position of the luggage object's bottom in the physical coordinate system. Let the direction vector of this ray in the camera coordinate system be denoted as... Then the ray equation can be expressed as: ; in For camera position, Let be the ray parameter. Let the intersection of the ray and the conveyor belt plane satisfy... Solve for the parameters This allows us to obtain the position of the bottom of the luggage object in the physical coordinate system. Similarly, the ray corresponding to the top of the luggage object and its height can be calculated. The intersection of the planes determines the three-dimensional extent of the luggage object in physical space.

[0049] After obtaining the bottom position of the luggage object, it can be used as the representative position of the luggage object in the physical coordinate system. Let this position be denoted as... ,in Meanwhile, the projection area of ​​the luggage object onto the image plane also carries pose information. Let the orientation angle of the luggage object in the image coordinate system be denoted as . This angle describes the angle between the projection of the luggage object's principal axis direction onto the image plane and the image's horizontal axis. To convert this image orientation angle to an orientation angle in the physical coordinate system... The camera mounting yaw angle needs to be considered. The influence of this. The transformation relationship is: ; This conversion assumes that the main axis of the luggage object is parallel to the plane of the conveyor belt, and the camera roll angle is... Approaching zero. When the roll angle is not negligible, additional rotational correction is required for the image orientation angle.

[0050] At this point, the coordinate mapping operation from the image coordinate system to the physical coordinate system has been completed, and the spatial position of the luggage object in the physical coordinate system has been obtained. and spatial attitude However, this global coordinate information alone is insufficient to directly guide collision risk assessment and control decisions; it is also necessary to correlate the physical location of the luggage with the structural layout information of the transport equipment.

[0051] The structural layout information of the conveyor equipment includes the start and end positions of each conveyor belt segment, the center position and radius of curvature of turning sections, the coordinates of merging and diverging points, and the physical boundaries of various mechanical structural components (such as support frames, guide plates, and baffles). This structural information is loaded as a configuration file during system initialization and represented as a series of polygonal regions or curve segments in the physical coordinate system. Let's denote the... The centerline of the conveyor belt in the physical coordinate system is a parametric curve. ,in The parameter is used. For straight segments, the curve expression is a linear interpolation; for curved segments, the curve expression is a circular arc equation or a Bézier curve.

[0052] Physical location of the luggage Calculate the distance to the centerline of each conveyor belt segment and find the conveyor belt segment with the smallest distance. This allows us to determine the current section of the conveyor belt where the luggage is located. Let this minimum distance be denoted as . ,like Less than the preset conveyor belt width threshold If the condition is met, the luggage is considered to be on the conveyor belt segment; otherwise, it is considered to be off the conveyor belt or located in the structural component area. Simultaneously, the parameters corresponding to the projection point of the luggage object on the center line of the conveyor belt are calculated. This allows us to determine the longitudinal position of the luggage item on that section of the conveyor belt.

[0053] After determining the section of the conveyor belt where the luggage is located, the lateral offset of the luggage relative to the centerline of the conveyor belt is further calculated. The unit vector of the tangent direction of the centerline of the conveyor belt at the projection point is denoted as... The unit vector of the normal direction is The lateral offset is the projection of the luggage object's position vector relative to the offset vector of the projection point onto the normal direction. This lateral offset reflects the distribution of luggage objects on the cross-section of the conveyor belt and is an important basis for subsequent collision risk assessment. For example, when the lateral offset is close to the edge of the conveyor belt, the luggage objects collide with the guide plates on both sides of the conveyor belt; when the lateral positions of multiple luggage objects are close, collisions between luggage objects occur.

[0054] In addition to location information, the spatial orientation of the luggage object It also needs to be correlated with the direction of the conveyor belt. Let the tangent angle of the conveyor belt centerline at the projection point be denoted as... Then the orientation deviation of the luggage object relative to the direction of the conveyor belt is: ; This attitude deviation describes the angle between the main axis direction of the luggage object and the conveyor direction. When the attitude deviation is large, the luggage object will move on the conveyor belt in an inclined or horizontal posture, increasing the risk of collision with adjacent luggage or structural components.

[0055] By combining the spatial location of each piece of luggage in the physical coordinate system, the identification of its conveyor belt segment, longitudinal position parameters, lateral offset, and attitude deviation, a complete spatial distribution result is constructed. This result is output in structured data format, including the fields: luggage object identifier, physical coordinates, etc. Conveyor belt section Longitudinal position parameters Lateral offset Spatial attitude Posture deviation This information provides precise spatial positioning data for subsequent multi-level collision risk assessment, enabling the calculation of collision probabilities within each preset detection area based on the specific distribution of luggage objects on the conveyor equipment, and generating corresponding multi-level control commands.

[0056] Based on the spatial distribution results and the operational status data, collision probability values ​​within multiple preset detection areas are calculated to obtain a graded collision risk assessment result. Then, based on the graded collision risk assessment result, corresponding multi-level control commands are generated, including: Based on the structural layout information and topology information of the conveying equipment, multiple preset detection areas are defined on the conveying equipment, and a risk level identifier is assigned to each preset detection area. Extract the spatial position and spatial orientation of the luggage object in the physical coordinate system from the spatial distribution results, determine the preset detection area to which the luggage object belongs, and obtain the running status data corresponding to the preset detection area to which the luggage object belongs; Based on the spatial location, spatial posture, and operational status data of the luggage object, calculate the spatial distance and relative motion relationship between the luggage object and other luggage objects within the preset detection area; Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage object, the collision probability value in the preset detection area is calculated. The collision probability value is then associated with the risk level identifier corresponding to the preset detection area to obtain a graded collision risk assessment result. Based on the graded collision risk assessment results, a preset control strategy mapping table is queried to generate multi-level control commands corresponding to the risk level identifier.

[0057] In the actual operation of baggage conveyor systems, differentiated monitoring of multiple key areas along the conveying path is required. Based on the mechanical structure layout of the conveyor equipment, including different functional sections such as straight sections, turning sections, merging sections, diverging sections, and lifting sections, and considering the topological connections of the conveying path, multiple pre-defined detection zones are delineated in the physical coordinate system. Each pre-defined detection zone corresponds to a segment of physical space along the conveying path and is described by its boundary coordinates in the physical coordinate system. To achieve refined risk management, a risk level label is assigned to each pre-defined detection zone, reflecting the potential danger of a collision in that area. Areas with small turning radii, merging areas where multiple conveyor lines converge, and buffer areas with dense baggage accumulation are typically marked as high-risk; while areas in straight conveyor sections with low baggage flow are marked as low-risk.

[0058] After obtaining the spatial distribution of luggage objects in the physical coordinate system, the three-dimensional spatial position and spatial orientation information of each luggage object are extracted. Spatial position is typically represented by a set of coordinate triplets in the physical coordinate system, while spatial orientation is characterized by the orientation angle of the luggage object relative to the conveyor direction and its tilt angle relative to the conveyor belt plane. The preset detection area to which the luggage object belongs is determined by judging whether its spatial position falls within the boundary of a preset detection area. If a luggage object spatially overlaps with multiple preset detection areas, its affiliation must be determined based on the area where its center point or main body is located. After determining the preset detection area to which the luggage object belongs, the corresponding operating status data for that area is obtained, including the current operating speed of the conveyor belt within that area, the motor load power, the conveyor belt tension, and the total number of luggage objects detected within that area.

[0059] For multiple luggage objects within the same pre-defined detection area, it is necessary to calculate their spatial distance and relative motion relationships. For any two luggage objects within the area, the Euclidean distance is calculated based on their respective spatial coordinates. This distance reflects the proximity of the two luggage objects in physical space. When two luggage objects are arranged one behind the other in the conveying direction, their longitudinal distance in the conveying direction is calculated; when two luggage objects are placed side by side in the transverse direction perpendicular to the conveying direction, their transverse distance is calculated. In addition to the static spatial distance, it is also necessary to analyze the relative motion relationships between the luggage objects. By combining the velocity vectors of each luggage object, the relative velocity, i.e., the velocity of one luggage object relative to another, is calculated. If two luggage objects move in the same direction but at different speeds, the luggage object with the faster speed behind has a tendency to catch up with the luggage object with the slower speed in front; if two luggage objects move in different directions, a collision will occur at the intersection point. The relative motion relationship also includes the relative acceleration between the luggage objects, used to determine whether the distance between two luggage objects is decreasing or increasing.

[0060] Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage objects, the collision probability value within a preset detection area is calculated. The spatial motion characteristics include parameters such as the current velocity amplitude, motion direction, acceleration, and attitude stability of the luggage objects. For any two luggage objects within the area, it is first determined whether their motion trajectories might intersect or approach each other. By extrapolating the current velocity and acceleration vectors of the luggage objects, their positional change trajectories over a future period are predicted. If the predicted trajectories intersect in the spatiotemporal domain, a collision risk is considered. The collision probability value is quantified by comprehensively considering multiple factors, including the current spatial distance between the two luggage objects, the amplitude of their relative velocity, the angle between the relative velocity direction and the direction of the connecting line, the size of the luggage objects, and their motion stability. When the spatial distance is small, the relative velocity is large, and the relative velocity direction is pointing towards each other, the collision probability value is high; conversely, when the spatial distance is large or the relative velocity is small, the collision probability value is low.

[0061] To calculate the collision probability more accurately, a time-based collision prediction method can be used. Assuming two pieces of luggage move at constant speeds along their respective directions, the time it takes for them to reach their nearest approach point is calculated and denoted as the estimated collision time. If the estimated collision time is less than a preset safe time threshold, and the spatial distance at the nearest approach point is less than a safe distance threshold, the collision risk is considered high. The collision probability can be expressed as a function of the estimated collision time and the nearest approach distance, typically modeled using a negative exponential or logistic function, so that the collision probability increases rapidly as the estimated collision time and the nearest approach distance decrease. Furthermore, the impact of the luggage's posture on the collision risk must be considered. If the luggage has a large posture deviation relative to the transport direction, or if the luggage tilts or flips on the conveyor belt, the uncertainty of its trajectory increases, and the collision probability increases accordingly.

[0062] After calculating the collision probability values ​​between all baggage pairs within a preset detection area, these collision probability values ​​are correlated with the risk level identifier corresponding to that preset detection area. The purpose of this correlation is to combine local collision probability information with the overall risk level of the area to form a graded collision risk assessment result. For preset detection areas marked as high-risk, even if the collision probability value is at a medium level, higher attention and stricter control measures should be given; while for areas marked as low-risk, the control threshold can be appropriately relaxed. The correlation is achieved by introducing a risk level weighting coefficient, with a higher weighting coefficient for high-risk areas than for low-risk areas. The final graded collision risk assessment result is organized by region, with each preset detection area corresponding to a comprehensive risk score. This score comprehensively considers factors such as the collision probability values ​​of all baggage pairs within the area, the inherent risk level identifier of the area, and the total number of baggage pairs within the area.

[0063] Based on the graded collision risk assessment results, corresponding multi-level control commands need to be generated to regulate the conveyor equipment. A control strategy mapping table predefines the correspondence between different risk assessment results and control commands, and this mapping table is configured according to the actual control capabilities and safety specifications of the conveyor equipment. The multi-level control commands manifest as control measures of varying intensities and ranges of action. When the comprehensive risk score of a preset detection area is below the first threshold, it is considered a safe state, and the generated control command is to maintain the current operating parameters unchanged, allowing the conveyor equipment to operate at normal speed. When the comprehensive risk score is between the first and second thresholds, it is considered a low-risk state, and the generated control command is to reduce the operating speed of the conveyor belt within that detection area, typically reducing the speed to a certain percentage of the rated speed, while simultaneously activating a warning signal to alert operators to the area. When the comprehensive risk score is between the second and third thresholds, it is considered a medium-risk state, and the generated control command is to further reduce the operating speed and activate auxiliary adjustment mechanisms to fine-tune the position and posture of the luggage, for example, by adjusting the lateral position of the luggage through pneumatic push rods or mechanical guide devices to maintain a greater safe distance between luggage items. When the comprehensive risk score exceeds the third threshold, it is determined to be a high-risk state. The generated control command is to apply emergency braking to the conveyor belt in the detection area, activate the audible and visual alarm device, and send an abnormal event report to the central control system.

[0064] Multi-level control commands not only target individual preset detection areas but also consider coordinated control between adjacent areas. When an area needs to reduce its operating speed or undergo emergency braking, its upstream area should correspondingly slow down the transport speed of luggage to prevent luggage from accumulating in front of the deceleration area; its downstream area should accelerate the evacuation speed of luggage to reduce the density of luggage within the area. This cross-area coordinated control is achieved by defining cascaded control rules in the control strategy mapping table. Furthermore, multi-level control commands also include personalized control measures for specific luggage items. For abnormal luggage items identified in the spatial distribution results, such as oversized, irregularly shaped, or severely deviated luggage, targeted control commands can be generated, such as activating a dedicated transport channel to divert it to a manual processing area, or using a robotic arm to adjust its posture.

[0065] In practical applications, the generation frequency of graded collision risk assessment results is synchronized with the frame rate of the image acquisition device, typically updating every tens of milliseconds. The execution response time of multi-level control commands needs to meet real-time requirements, ensuring sufficient time to complete control actions before a collision occurs. To improve control effectiveness, predictive control strategies can be introduced. This involves predicting the collision risk of luggage in a pre-defined detection area based on its motion state and historical trajectory in the upstream area before it enters that area, and generating control commands in advance for intervention. This predictive control strategy effectively avoids collisions caused by control response delays, further enhancing the safety and reliability of the luggage conveyor system.

[0066] Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage object, the collision probability value within the preset detection area is calculated. The collision probability value is then correlated with the risk level identifier corresponding to the preset detection area to obtain a graded collision risk assessment result, including: Extract positional and orientation changes from the spatial motion characteristics of the luggage object; The minimum safe distance between the luggage object and other luggage objects within the preset detection area is calculated based on the spatial distance. Based on the relative motion relationship and the position change, the movement trajectory of the luggage object within the preset time window is predicted. Based on the minimum safe distance and the movement trajectory, the intersection position of the trajectory of the luggage object with that of other luggage objects in the preset detection area is determined, and the degree of posture conflict at the trajectory intersection position is calculated according to the posture change amount; Based on the minimum safe distance, the trajectory intersection position, and the degree of attitude conflict, calculate the collision probability value within the preset detection area; Based on the collision probability value and the risk level identifier, the collision probability value is mapped to a risk level, and the collision probability value is divided into different risk level ranges. The risk level range is associated with the risk level identifier to obtain the graded collision risk assessment result.

[0067] The spatial motion characteristics of the luggage object include multi-dimensional dynamic parameters. A three-dimensional coordinate system is used to record the real-time position information of the luggage object within a preset detection area. The coordinate system uses Cartesian coordinates, with the origin set at the geometric center of the detection area. The positive X-axis points in the direction of the conveyor belt movement, the positive Y-axis points in the lateral expansion direction, and the positive Z-axis points vertically upward. Position changes are extracted through continuous sampling at a frequency adjustable between 20 and 50 times per second. Each sample records the coordinates of the luggage object's center of mass in three-dimensional space. The difference between the coordinates of two adjacent samples constitutes the instantaneous position change. The position changes from 10 consecutive samples are averaged using a sliding window to obtain the smoothed position change rate. Attitude changes are extracted based on the luggage object's three-axis rotation angles. Euler angles are used to record the pitch angle around the X-axis, the roll angle around the Y-axis, and the yaw angle around the Z-axis. Attitude angle sampling is performed synchronously with position sampling. The angle difference between adjacent sampling times is the attitude change. A significant attitude change is indicated when the angular velocity in any axis exceeds 15 degrees per second.

[0068] Spatial distance is calculated using a rectangular prism envelope model circumscribed in the luggage object. The six faces of the envelope are parallel to the three reference planes of the coordinate system. The shortest surface distance between the target luggage object's envelope and the envelopes of other luggage objects within the preset detection area is calculated as the spatial distance measurement. The minimum safe distance is determined by comprehensively considering the luggage object's size and motion state. The baseline minimum safe distance is set to 0.3 times the longest side length of the larger object's envelope. When the relative speed exceeds 0.5 meters per second, the safe distance coefficient increases linearly with the speed, with the increment being the speed value multiplied by a correction factor of 0.2. For luggage objects with irregular shapes or made of soft materials, an additional 20% redundancy margin is added to the safe distance coefficient to ensure the conservatism of collision prediction.

[0069] The motion trajectory prediction unfolds within a preset time window, the length of which is dynamically adjusted based on the conveyor system's operating speed. The standard setting is between 1.5 and 3 seconds, shortened to 1 second in fast operation mode to improve response speed. The trajectory prediction employs a hybrid strategy of a constant-velocity linear model and an acceleration correction model. When the standard deviation of the position change is less than 10% of the mean change within five consecutive samples, it is considered uniform motion. The current velocity vector is extrapolated to the future position sequence, with an extrapolation time step of 0.1 seconds. The acceleration correction model is suitable for scenarios where significant acceleration changes are detected. The average acceleration is calculated using the three most recent velocity changes, and the trajectory points are recursively derived from the current velocity and acceleration according to their kinematic relationship. During the recursion, an attenuation factor is applied to the acceleration, with the attenuation coefficient decreasing by 5% per step to simulate the effects of friction and resistance. The relative motion relationship is represented by the difference in velocity vectors between the target luggage and surrounding luggage. A relative velocity field is constructed to assess potential approach trends; when the relative velocities point towards each other and the angle is less than 30 degrees, an approach situation is identified.

[0070] The determination of trajectory intersection positions is achieved by traversing all time step combinations of the predicted trajectory of the target baggage object and the predicted trajectories of other baggage objects. The three-dimensional Euclidean distance between the corresponding points of each pair of time steps is calculated. When the distance is less than the sum of the minimum safe distances of both parties, the spatiotemporal point is recorded as a candidate intersection position. Multiple candidate positions are sorted chronologically, and the point with the earliest time and smallest distance is selected as the primary trajectory intersection position. The remaining five candidate positions are retained for risk accumulation assessment. The calculation of the attitude conflict level is based on the attitude changes of both baggage objects at the intersection position. The attitude angular velocity vectors of the two objects at the intersection moment are multiplied. The absolute value of the multiplication result reflects the degree of cooperation or antagonism in attitude changes. An absolute value greater than 50 degrees squared per second is considered high attitude conflict, between 20 and 50 degrees squared per second is considered medium conflict, and below 20 degrees squared per second is considered low conflict. The envelope box volume change rate is introduced as an auxiliary indicator for the attitude conflict level. When the volume expansion rate exceeds 15% per second, the conflict level is increased by one level.

[0071] The collision probability calculation integrates three factors: minimum safe distance, trajectory intersection location, and attitude conflict level, employing a weighted product model to construct the probability estimation framework. The distance factor is defined as the reciprocal of the ratio of the actual distance to the minimum safe distance. A distance factor of 1 is assigned when the ratio is less than 0.5; a linear interpolation is used when the ratio is between 0.5 and 1.5; and a zero distance factor indicates no collision risk when the ratio is greater than 1.5. The trajectory factor is assigned based on the temporal urgency and spatial compactness of the intersection location. A trajectory factor of 0.9 is assigned when the intersection time is less than 0.5 seconds from the current time and the distance is less than 0.3 times the safe distance. The trajectory factor decreases by 0.15 for every 0.2-second increase in time interval and by 0.1 for every 0.2 times the safe distance. The attitude factor is assigned according to the conflict level: 0.8 for high conflict, 0.5 for medium conflict, 0.2 for low conflict, and 0.1 for no significant attitude change. The weighting coefficients for the three factors are set to 0.4, 0.4, and 0.2, respectively. The equal weights for distance and trajectory emphasize the importance of the spatial and temporal dimensions, while the lower weight for attitude reflects its auxiliary role. The collision probability value is obtained by multiplying the distance factor by 0.4, the trajectory factor by 0.4, and the attitude factor by 0.2. The calculation result is rounded to two decimal places and ranges from 0 to 1.

[0072] Risk level identifiers are predefined into four discrete levels: low risk, medium risk, high risk, and emergency risk. Each level is associated with a specific response strategy and processing priority. Risk level mapping is achieved by setting collision probability threshold ranges: collision probability values ​​less than 0.25 are classified as low risk, 0.25 to 0.55 as medium risk, 0.55 to 0.8 as high risk, and greater than 0.8 as emergency risk. The threshold boundary settings consider a balance between system responsiveness and false alarm tolerance; the wider threshold in the low-risk range allows for fluctuations during normal operation, while the narrower threshold in the emergency risk range ensures rapid triggering in critical situations. A hysteresis mechanism is introduced in the mapping process to prevent frequent jumps across levels. When the probability value fluctuates near the threshold boundary, it must remain within the new range for three consecutive sampling periods before a level change is confirmed; a single out-of-bounds error does not trigger a level switch.

[0073] The correlation processing binds the risk level range determination results with the risk level identifiers to generate structured, graded collision risk assessment results. The assessment results include the unique identifier of the target baggage object, the current collision probability value, the corresponding risk level identifier, the duration of the risk level, the 3D coordinates and estimated arrival time of the primary trajectory intersection, a list of other baggage object identifiers involved, and suggested response measure codes. Response measure codes correspond to risk levels: 0 for low risk (maintain monitoring), 1 for medium risk (increase tracking frequency), 2 for high risk (early warning and preparation for intervention), and 3 for emergency risk (immediately activate avoidance or braking mechanisms). The assessment results are pushed to the execution control module in real time via a message queue. Message priority is positively correlated with risk level, with emergency risk messages having the highest priority and preempting processing resources.

[0074] Based on the multi-level control commands, control operations are performed on the drive mechanism of the conveying equipment, and data on changes in the motion state of the luggage object after control execution are collected to obtain feedback information. The feedback information is then used to update the graded collision risk assessment results, including: The control type identifier and control parameters contained in the multi-level control instructions are analyzed, and the target drive unit in the drive mechanism of the conveying equipment that needs to perform control operations is determined according to the control type identifier. The control parameters are converted into execution instructions for the target drive unit, thereby driving the target drive unit to perform the corresponding control operation; The image acquisition device acquires a sequence of images of the luggage object after the control is executed, and extracts data on the change in the motion state of the luggage object after the control is executed from the sequence of images of the luggage object after the control is executed. The motion state change data is correlated and compared with the control parameters to calculate the deviation between the actual effect and the expected effect of the control execution; Feedback information is generated based on the deviation amount. The collision probability value calculation method in the preset detection area is adjusted according to the feedback information. The collision probability value in the preset detection area is recalculated based on the adjusted collision probability value calculation method, and the graded collision risk assessment result is updated.

[0075] After acquiring the spatial motion characteristics of the luggage object, the positional and orientation changes are extracted. The positional change reflects the spatial displacement of the luggage object within a preset time window, while the orientation change describes the deflection of the luggage object's orientation angle. Together, they constitute a complete representation of the luggage object's motion state, providing basic data for subsequent collision probability calculations.

[0076] Based on the spatial distance obtained from the preceding steps, calculate the minimum safe distance between the luggage object and other luggage objects within the preset detection area. This minimum safe distance must consider both the physical dimensions and current positions of the two luggage objects. Assume that luggage object A and luggage object B are located in the physical coordinate system as follows: and The envelope radii of the two are respectively and The center distance between the two is The minimum safe distance is defined as the center distance minus the sum of the two envelope radii, i.e.: ; when Values ​​close to zero or negative indicate that the envelopes of the two luggage objects have already overlapped or are about to overlap, posing a collision risk. To ensure the robustness of the calculation, a safety margin needs to be introduced in practical applications. That is, it is believed that when A collision warning should be triggered immediately.

[0077] By combining relative motion and positional change, the trajectory of luggage objects within a preset time window is predicted. The relative motion includes information such as the speed difference between the luggage object and the conveyor belt, and the relative speeds between luggage objects. Assume the velocity vector of luggage object A at the current moment is... The acceleration vector is Then within the preset time window Within this context, the predicted trajectory of luggage object A can be described by kinematic equations. On a discrete time series, the preset time window is divided into... There are 1 time steps, each time step having a length of 1. Then, luggage object A is in the first... Predicted location at each time step It can be represented as: ; ; in , velocity vector Horizontal and vertical components, , acceleration vector The horizontal and vertical components are used to predict the motion trajectory of all luggage objects within the preset detection area in the future time window using the above method, resulting in a series of position prediction sequences at discrete time steps.

[0078] Based on the minimum safe distance and movement trajectory, the intersection position of the trajectories of the baggage object and other baggage objects within the preset detection area is determined. The trajectory intersection position refers to the point in space where the predicted trajectories of the two baggage objects are closest, and the corresponding time. The predicted positions of baggage object A and baggage object B at each time step within the preset time window are traversed, and any combination of time steps is calculated. The distance between the two: ; Iterate through all possible combinations of time steps to find the one that makes The time step to obtain the minimum value The location corresponding to this time step is the trajectory intersection point. Record the minimum distance corresponding to this trajectory intersection point. and the corresponding time and .

[0079] The degree of attitude conflict at the trajectory intersection is calculated based on the attitude change. This degree of conflict reflects the impact of the orientation difference between the two luggage objects at the trajectory intersection on the collision risk. Assume that the orientation angles of luggage object A and luggage object B at the corresponding moments at the trajectory intersection are... and Then the angle of difference in orientation between the two Defined as: ; When the orientation difference angle is close to 0 degrees or 180 degrees, it indicates that the two pieces of luggage are moving parallel or towards each other, with a high risk of collision. When the orientation difference angle is close to 90 degrees, it indicates that the two pieces of luggage are moving in opposite directions, with a relatively low risk of collision. An attitude conflict coefficient is introduced. Mapping the orientation difference angle using a cosine function: ; This coefficient ranges from 0 to 2, reaching a minimum of 0 when the orientation difference angle is zero degrees, a maximum of 2 when the orientation difference angle is 180 degrees, and a value of 1 when the orientation difference angle is 90 degrees. A higher attitude conflict coefficient indicates a higher degree of attitude conflict and a greater risk of collision.

[0080] Based on the minimum safe distance, trajectory intersection location, and attitude conflict degree, a collision probability value is calculated within a preset detection area. This collision probability value comprehensively considers risk factors from three dimensions: spatial distance, trajectory intersection, and attitude conflict. A collision probability calculation function is introduced to map these factors to a probability range of zero to one. It can be represented as: ; in , , For the weighting coefficients, satisfying , The scale parameter for trajectory intersection distance. The scale parameter for the minimum safe distance. This represents the maximum value of the attitude conflict coefficient. The first term reflects the spatial proximity at the trajectory intersection, the second term reflects the contribution of attitude conflict to collision risk, and the third term reflects the spatial distance risk between the two luggage objects at the current moment. By adjusting the weighting coefficients and scale parameters, the risk assessment needs of different application scenarios can be adapted.

[0081] Based on the collision probability value and risk level identifier, a risk level mapping is performed on the collision probability value, dividing it into different risk level ranges. These risk level ranges are set according to the numerical range of the collision probability value, typically divided into four levels: low risk, medium risk, high risk, and extremely high risk. Collision probability thresholds corresponding to the four levels are set as follows: , , The risk level mapping rule is as follows: when When the risk level is low; when At that time, the risk level was medium risk; when At that time, the risk level was high risk; when At that time, the risk level is extremely high. Different risk levels correspond to different risk level labels, such as label one for low risk, label two for medium risk, label three for high risk, and label four for extremely high risk.

[0082] By associating risk level ranges with risk level identifiers, a graded collision risk assessment result is obtained. This association process establishes a mapping relationship between collision probability values, risk level ranges, and risk level identifiers, forming structured risk assessment data. The graded collision risk assessment result includes the collision probability values, corresponding risk level ranges, and risk level identifiers for all pairs of luggage within the preset detection area, providing a decision-making basis for the generation of subsequent multi-level control commands.

[0083] The process involves parsing the control type identifier and control parameters contained in multi-level control instructions. The control type identifier indicates the type of control operation to be performed, including deceleration control, stop control, and path change control. Control parameters include specific values ​​such as target speed, deceleration rate, stop duration, and path offset. Based on the control type identifier, the target drive unit in the conveyor's drive mechanism that needs to perform the control operation is determined. The conveyor's drive mechanism contains multiple drive units, each controlling the operating state of a specific conveyor belt segment. By parsing the device address information in the control type identifier, the corresponding target drive unit is located.

[0084] The control parameters are converted into execution instructions for the target drive unit, which then performs the corresponding control operations. The control parameters are expressed in standardized numerical form and need to be converted according to the communication protocol and instruction format of the target drive unit. For example, the target speed parameter for deceleration control needs to be converted into the drive unit's frequency control instruction or pulse width modulation duty cycle, and the stop duration parameter for stop control needs to be converted into the drive unit's hold state instruction. After conversion, the execution instructions are sent to the target drive unit via an industrial bus or dedicated communication interface. Upon receiving the instructions, the target drive unit executes the corresponding control operations, adjusting the conveyor belt's running speed or state.

[0085] A sequence of images of the luggage object after control execution is acquired using an image acquisition device. This sequence covers a period of time before and after the control operation, allowing observation of the actual changes in the luggage object's motion state. Motion state change data is extracted from this sequence of images. This data includes changes in the luggage object's position coordinates, velocity vector, and attitude angle. Using the same feature extraction and coordinate transformation methods as described above, the features of the luggage object in each frame of the image sequence are mapped to a physical coordinate system, forming time-series motion state change data.

[0086] The data on changes in motion state are correlated and compared with control parameters to calculate the deviation between the actual and expected effects of the control execution. The expected effect is calculated based on the control parameters and the kinematic model of the luggage object. For example, when the control type is deceleration control and the target speed is... At that time, the expected effect is that the speed of the luggage object is close to that after the control is executed. The actual effect is obtained by directly measuring the change in motion state data. The deviation is defined as the absolute value of the difference between the actual speed and the target speed. ; in To control the actual measured speed of the baggage object after execution, similarly, for stop control, the deviation is the distance between the actual stop position and the expected stop position; for path control, the deviation is the difference between the actual path offset and the expected path offset.

[0087] Feedback information is generated based on the deviation, quantifying the difference between the control execution effect and the expected target. This feedback includes the value and direction of the deviation, as well as the correlation between control parameters and the actual effect. The calculation method for collision probability values ​​within the preset detection area is adjusted based on the feedback information. When the deviation exceeds a preset threshold, it indicates that the current collision probability calculation method fails to accurately reflect the actual risk, requiring adjustment of the parameters in the calculation method. Adjustment strategies include correcting scale parameters. and To make it more consistent with actual control response characteristics; adjust the weighting coefficients. , , This enhances sensitivity to key risk factors. Based on the adjusted collision probability calculation method, the collision probability values ​​within the preset detection area are recalculated, updating the graded collision risk assessment results. The updated risk assessment results reflect the latest risk status after control execution, providing a more accurate basis for subsequent control decisions and forming a closed-loop optimization mechanism.

[0088] A second aspect of the present invention provides an image detection and multi-level collision avoidance optimization system for baggage transportation, comprising: The data acquisition unit is used to acquire the operating status data of the baggage conveyor equipment and the real-time image data collected by the image acquisition devices deployed along the conveyor path; The feature parsing unit is used to extract features and perform semantic parsing on the real-time image data to obtain the spatial motion features of the luggage objects on the transport path; The coordinate mapping unit is used to map the spatial motion characteristics of the luggage object from the image coordinate system to the physical coordinate system of the conveying equipment according to the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, so as to obtain the spatial distribution result of the luggage object; The risk assessment unit is used to calculate the collision probability values ​​in multiple preset detection areas based on the spatial distribution results and the operating status data, obtain the graded collision risk assessment results, and generate corresponding multi-level control commands based on the graded collision risk assessment results. The feedback update unit is used to perform control operations on the drive mechanism of the conveying equipment based on the multi-level control commands, collect data on the change in the motion state of the luggage object after the control is executed to obtain feedback information, and use the feedback information to update the graded collision risk assessment results.

[0089] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0090] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0091] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An image detection and multi-level collision avoidance optimization method for baggage transportation, characterized in that, include: Acquire operational status data of baggage conveyor equipment and real-time image data collected by image acquisition devices deployed along the conveyor path; Feature extraction and semantic parsing are performed on the real-time image data to obtain the spatial motion features of the luggage objects on the transport path; Based on the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, the spatial motion characteristics of the luggage object are mapped from the image coordinate system to the physical coordinate system of the conveying equipment to obtain the spatial distribution result of the luggage object; Based on the spatial distribution results and the operating status data, the collision probability values ​​in multiple preset detection areas are calculated to obtain the graded collision risk assessment results, and corresponding multi-level control commands are generated based on the graded collision risk assessment results. Based on the multi-level control commands, control operations are performed on the drive mechanism of the conveying equipment, and data on changes in the motion state of the luggage object after the control is executed are collected to obtain feedback information. The feedback information is then used to update the graded collision risk assessment results.

2. The method according to claim 1, characterized in that, Feature extraction and semantic parsing are performed on the real-time image data to obtain the spatial motion features of luggage objects along the transport path, including: Perform multi-level feature extraction operations on the real-time image data to obtain visual representation information of the luggage object at different levels of abstraction; Based on the visual representation information, semantic classification and target localization are performed to identify the category attributes and spatial pose of the luggage object in the image coordinate system; Based on the spatial pose of the luggage object, a cross-frame association relationship of the luggage object is established between consecutive frames of image data, and a motion trajectory sequence of the luggage object is constructed. Based on the motion trajectory sequence, a time-series analysis is performed on the historical motion behavior of the luggage object to extract the motion pattern features of the luggage object, and the position change and posture change of the luggage object within a preset time window are calculated based on the motion pattern features. Based on the spatial pose, motion trajectory sequence, position change, and attitude change, the spatial motion features, including the current state description and the predicted state description, are determined.

3. The method according to claim 2, characterized in that, Based on the motion trajectory sequence, a time-series analysis is performed on the historical motion behavior of the luggage object to extract its motion pattern features. Then, based on these features, the changes in position and posture of the luggage object within a preset time window are calculated, including: The motion trajectory sequence is segmented, and the motion trajectory sequence is divided into multiple trajectory segments according to the motion change characteristics of the luggage object; Statistical analysis is performed on each trajectory segment to extract the motion feature parameters of the trajectory segment, thereby obtaining a trajectory segment feature set; Based on the set of trajectory segment features, the motion behavior pattern of the luggage object on the transport path is determined, and the motion behavior pattern is associated and mapped with the topological structure information of the transport path to obtain motion pattern features; Based on the motion pattern characteristics, combined with the current location of the luggage object on the transport path and the topological information of the transport path, historical trajectory segments corresponding to the current transport path location are searched. Based on the duration of motion and average velocity amplitude of the historical trajectory segment, the change in position of the luggage object within a preset time window is calculated, and the change in attitude of the luggage object within the preset time window is calculated based on the directional deflection angle of the historical trajectory segment.

4. The method according to claim 1, characterized in that, Based on the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, the spatial motion characteristics of the luggage object are mapped from the image coordinate system to the physical coordinate system of the conveying equipment to obtain the spatial distribution result of the luggage object, including: Based on the installation pose parameters of the image acquisition device, determine the coordinate transformation relationship between the image coordinate system and the physical coordinate system; The spatial motion features of the luggage object include position and posture descriptions in the image coordinate system, which are preprocessed by coordinate transformation to determine the projection area of ​​the luggage object on the image plane. The spatial distance between the luggage object and the image acquisition device is calculated based on the position of the projection area on the image plane and the observation angle. Based on the spatial distance and the coordinate transformation relationship, a coordinate mapping operation is performed on the position description and posture description to transform the position description and posture description from the image coordinate system to the physical coordinate system, thereby obtaining the spatial position and spatial posture of the luggage object in the physical coordinate system; The spatial position and orientation of the luggage object in the physical coordinate system are correlated with the structural layout information of the conveying equipment to determine the distribution position of the luggage object on the conveying equipment, thus obtaining the spatial distribution result.

5. The method according to claim 1, characterized in that, Based on the spatial distribution results and the operational status data, collision probability values ​​within multiple preset detection areas are calculated to obtain a graded collision risk assessment result. Then, based on the graded collision risk assessment result, corresponding multi-level control commands are generated, including: Based on the structural layout information and topology information of the conveying equipment, multiple preset detection areas are defined on the conveying equipment, and a risk level identifier is assigned to each preset detection area. Extract the spatial position and spatial orientation of the luggage object in the physical coordinate system from the spatial distribution results, determine the preset detection area to which the luggage object belongs, and obtain the running status data corresponding to the preset detection area to which the luggage object belongs; Based on the spatial location, spatial posture, and operational status data of the luggage object, calculate the spatial distance and relative motion relationship between the luggage object and other luggage objects within the preset detection area; Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage object, the collision probability value in the preset detection area is calculated. The collision probability value is then associated with the risk level identifier corresponding to the preset detection area to obtain a graded collision risk assessment result. Based on the graded collision risk assessment results, a preset control strategy mapping table is queried to generate multi-level control commands corresponding to the risk level identifier.

6. The method according to claim 5, characterized in that, Based on the spatial distance and relative motion relationship, and combined with the spatial motion characteristics of the luggage object, the collision probability value within the preset detection area is calculated. The collision probability value is then correlated with the risk level identifier corresponding to the preset detection area to obtain a graded collision risk assessment result, including: Extract positional and orientation changes from the spatial motion characteristics of the luggage object; The minimum safe distance between the luggage object and other luggage objects within the preset detection area is calculated based on the spatial distance. Based on the relative motion relationship and the position change, the movement trajectory of the luggage object within the preset time window is predicted. Based on the minimum safe distance and the movement trajectory, the intersection position of the trajectory of the luggage object with that of other luggage objects in the preset detection area is determined, and the degree of posture conflict at the trajectory intersection position is calculated according to the posture change amount; Based on the minimum safe distance, the trajectory intersection position, and the degree of attitude conflict, calculate the collision probability value within the preset detection area; Based on the collision probability value and the risk level identifier, the collision probability value is mapped to a risk level, and the collision probability value is divided into different risk level ranges. The risk level range is associated with the risk level identifier to obtain the graded collision risk assessment result.

7. The method according to claim 1, characterized in that, Based on the multi-level control commands, control operations are performed on the drive mechanism of the conveying equipment, and data on changes in the motion state of the luggage object after control execution are collected to obtain feedback information. The feedback information is then used to update the graded collision risk assessment results, including: The control type identifier and control parameters contained in the multi-level control instructions are analyzed, and the target drive unit in the drive mechanism of the conveying equipment that needs to perform control operations is determined according to the control type identifier. The control parameters are converted into execution instructions for the target drive unit, thereby driving the target drive unit to perform the corresponding control operation; The image acquisition device acquires a sequence of images of the luggage object after the control is executed, and extracts data on the change in the motion state of the luggage object after the control is executed from the sequence of images of the luggage object after the control is executed; The motion state change data is correlated and compared with the control parameters to calculate the deviation between the actual effect and the expected effect of the control execution; Feedback information is generated based on the deviation amount. The collision probability value calculation method in the preset detection area is adjusted according to the feedback information. The collision probability value in the preset detection area is recalculated based on the adjusted collision probability value calculation method, and the graded collision risk assessment result is updated.

8. An image detection and multi-level collision avoidance optimization system for baggage transportation, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire the operating status data of the baggage conveyor equipment and the real-time image data collected by the image acquisition devices deployed along the conveyor path; The feature parsing unit is used to extract features and perform semantic parsing on the real-time image data to obtain the spatial motion features of the luggage objects on the transport path; The coordinate mapping unit is used to map the spatial motion characteristics of the luggage object from the image coordinate system to the physical coordinate system of the conveying equipment according to the installation pose parameters of the image acquisition device and the preset coordinate transformation relationship, so as to obtain the spatial distribution result of the luggage object; The risk assessment unit is used to calculate the collision probability values ​​in multiple preset detection areas based on the spatial distribution results and the operating status data, obtain the graded collision risk assessment results, and generate corresponding multi-level control commands based on the graded collision risk assessment results. The feedback update unit is used to perform control operations on the drive mechanism of the conveying equipment based on the multi-level control commands, collect data on the change in the motion state of the luggage object after the control is executed to obtain feedback information, and use the feedback information to update the graded collision risk assessment results.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images