Logistics navigation method and system based on visual identification and wireless communication

Abstract

The invention discloses a logistics navigation method and system based on visual identification and wireless communication, and the method comprises the following steps: collecting an environment image and an equipment state information set, and executing image preprocessing; the sending device generates a confidential dynamic two-dimensional code frame sequence according to the device state information set and an encryption algorithm; based on the confidential dynamic two-dimensional code frame sequence, calculating and generating a pose priori set and row time sequence parameters; executing an improved iNeRF algorithm, generating a positioning result set and outputting a feature weight map; carrying out structured encryption coding on the positioning result set and the feature weight map, and updating a confidential dynamic two-dimensional code frame sequence; a receiving device collects the confidential dynamic two-dimensional code frame sequence to complete analysis and encryption verification; executing path calculation and obstacle avoidance control, continuously updating the two-dimensional code, and periodically executing an improved iNeRF algorithm to complete navigation. According to the invention, the positioning precision of logistics navigation and the data interaction security are improved.

Description

Technical Field

[0001] This invention relates to the field of automated logistics equipment control, and in particular to a logistics navigation method and system based on visual recognition and wireless communication. Background Technology

[0002] Currently, automated logistics equipment is widely used in warehousing, transportation, and material handling. Traditional navigation methods mostly rely on lidar, ultrasonic sensors, or inertial measurement units for positioning and path planning. However, these sensor fusion-based navigation methods suffer from high costs, limited accuracy, and poor environmental adaptability, especially in complex lighting and dynamic scenarios, where positioning drift and path errors are prone to occur.

[0003] In terms of information exchange, existing wireless communication mechanisms mostly rely on radio frequency signals or network transmission methods, which are subject to signal interference and environmental obstruction, making it difficult to guarantee real-time performance and security. Although some communication methods based on QR codes or image tags reduce communication complexity, the static content and fixed refresh rate of QR codes cannot support dynamic task command transmission and status feedback, resulting in system interaction delays and insufficient synchronization accuracy.

[0004] Furthermore, existing vision-based positioning methods suffer from significant temporal mismatch and motion blur problems due to row-level temporal differences in images under rolling shutter exposure conditions. Traditional NeRF-like algorithms fail to effectively model row-level temporal and dynamic blur effects, making it difficult to meet the requirements of high-precision, real-time logistics navigation. Existing technologies have significant shortcomings in vision-driven high-precision positioning, dynamic QR code communication security, and real-time closed-loop navigation control. There is an urgent need for a technical solution that can achieve high-precision pose estimation, secure and reliable data interaction, and low-latency path control in complex environments.

[0005] Therefore, how to provide a logistics navigation method and system based on visual recognition and wireless communication is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a logistics navigation method and system based on visual recognition and wireless communication. By integrating an improved iNeRF algorithm with a confidential dynamic QR code interaction mechanism, it achieves high-precision pose estimation and secure data transmission for automated logistics equipment. This invention constructs a rolling shutter time modeling and feature adaptive sampling structure to solve the temporal mismatch and motion blur problems in traditional visual positioning. Simultaneously, it designs a layered hybrid encryption dynamic QR code communication mechanism to improve the security and real-time performance of task command interaction. This method achieves closed-loop collaboration between visual perception, encrypted communication, and path control, possessing advantages such as high positioning accuracy, secure data transmission, fast system response, and strong environmental adaptability.

[0007] A logistics navigation method based on visual recognition and wireless communication according to an embodiment of the present invention includes the following steps: Collect environmental images and equipment status information, perform image preprocessing to generate a standardized image sequence, and set dynamic QR code refresh cycle parameters; The sending device generates a confidential dynamic QR code frame sequence based on the device status information set and encryption algorithm, while simultaneously loading the device identifier, task instructions, timestamp and refresh cycle parameters; Based on the confidential dynamic QR code frame sequence, the QR code plane homography relation is generated, the initial pose estimate and covariance are calculated to form the pose prior set, and the row time series parameters are calculated. Based on standardized image sequences, pose prior sets, and row time parameters, an improved iNeRF algorithm is executed to generate a localization result set and output a feature weight map. The location result set, feature weight map and task instructions are structured and encrypted to generate confidential dynamic QR code update frames and appended to the confidential dynamic QR code frame sequence. The receiving device collects the confidential dynamic QR code frame sequence, completes the parsing and encryption verification, and outputs a set of pose data, a set of task instructions, uncertainty measure and solution delay parameters. Based on the pose data set and task instruction set, path calculation and obstacle avoidance control are performed. Low-confidence poses are suppressed with uncertainty metric. Based on refresh cycle parameter and solution delay parameter, images are continuously acquired and QR codes are updated. The improved iNeRF algorithm is periodically executed to complete the logistics navigation.

[0008] Optionally, the generation of the standardized image sequence and refresh period parameters includes: Each automated logistics device is equipped with a visual recognition module, a confidential dynamic QR code generation and parsing module, a visual recognition positioning module, and a data interaction control module, and the camera and QR code display device are activated. The camera captures environmental images and transmits them to the visual recognition module, where image preprocessing is performed, including brightness normalization, noise suppression, and edge enhancement, to generate a standardized image sequence. The data interaction control module collects a set of device status information and calculates dynamic QR code refresh cycle parameters based on device operating frequency, processing delay, and image frame rate. Perform consistency verification between the timestamp identifier of the standardized image sequence and the refresh cycle parameter of the dynamic QR code.

[0009] Optionally, the generation of the confidential dynamic QR code frame sequence includes: The sending device is triggered by the data interaction control module to generate a confidential dynamic QR code, which calls the set of device status information registered in the previous period, loads the current encryption algorithm configuration and dynamic QR code refresh cycle parameters, and allocates the data load area. The confidential dynamic QR code generation module generates the initial frame of the confidential dynamic QR code based on the device status information set and encryption algorithm; The device identifier, task instruction, timestamp and refresh cycle parameters are loaded sequentially into the data payload area of ​​the confidential dynamic QR code initial frame to form a confidential dynamic QR code initial frame with complete parameter mapping, and a timestamp index field is established in the frame header area. The generated initial frame of the confidential dynamic QR code is appended to the confidential dynamic QR code frame sequence. The confidential dynamic QR code frame sequence is delivered to the visual recognition module, which drives the QR code display device to display it and performs inter-frame refresh and content replacement according to the dynamic QR code refresh cycle parameters.

[0010] Optionally, the generation of the pose prior set and the row timing parameters includes: The visual recognition module periodically collects dynamic QR code image frames within the dynamic QR code refresh cycle and extracts the timestamp information and pixel coordinate matrix corresponding to each collection. The confidential dynamic QR code parsing module parses the encoded data in the dynamic QR code image frame, reconstructs the coordinates of the QR code boundary vertices and geometric feature points, calculates the plane homography relationship of the QR code based on the pixel coordinate matrix, and outputs the homography matrix parameter set. The visual recognition and localization module calculates the initial pose estimate based on the homography relation of the QR code plane, and calculates the pose covariance matrix by using the pixel gradient distribution and feature point differences between periodically acquired image frames to form a pose prior set. The visual recognition module performs row-level scanning timing analysis in conjunction with standardized image sequences, calculates row timing parameters and frame-level sampling interval information, and calculates motion blur index based on image gradient change rate and optical flow characteristics.

[0011] Optionally, the generation of the localization result set and the feature weight map includes: The visual recognition and localization module receives standardized image sequences, pose prior sets, line time parameters, and motion blur indices. A rolling shutter time modeling structure is used to dynamically render the row-level time sequence. The exposure time term in the row time sequence parameters is introduced into the time dimension of each ray. The exposure time of each row pixel is weighted and modeled by the time function, so that the rendering result of each pixel strictly corresponds to its exposure time, forming a time-continuous row-level rendering result. Based on the feature adaptive sampling structure, the feature gradient magnitude distribution and texture saliency are calculated on the image plane to generate a feature weight map. High-density ray sampling is performed on the QR code region and high-texture region, and low-density ray sampling is performed on the low-texture region. Based on the motion blur index, the feature weight map is dynamically suppressed. The motion blur index is used as a reference for weight adjustment, and the feature weights in high motion regions are attenuated. In the multi-resolution space, the Anytime budget solution structure is used to perform optimization. The volume density and color field are backpropagated layer by layer within the time delay budget limited by the refresh cycle parameter. The intermediate optimization results are output in real time, and the localization result set is output immediately when the budget is reached. The localization result set includes an optimized pose vector, an uncertainty metric, and a solution delay parameter. The optimized pose vector represents the spatial pose estimation result after multi-resolution optimization. The uncertainty metric represents the solution confidence under constraints of optical flow, gradient, and sampling density. The solution delay parameter represents the termination time of the Anytime budget solution structure.

[0012] Optionally, updating the confidential dynamic QR code includes: The data interaction control module gathers the optimized pose vector, uncertainty measure and solution delay parameters from the positioning result set, and together with the feature weight map and task instructions, it forms a structured data load. The confidential dynamic QR code generation module performs encryption and integrity verification on the structured data payload based on the encryption algorithm, generates a confidential dynamic QR code update frame, and binds a timestamp and refresh cycle parameter to the confidential dynamic QR code update frame; Write the confidential dynamic QR code update frame to the next frame position in the confidential dynamic QR code frame sequence, set the valid marker and update the frame index; The updated QR code display device completes the display of confidential dynamic QR code update frames, and controls the display duration and frame replacement according to the dynamic QR code refresh cycle parameters; The data interaction control module records the interaction status of the current cycle. The recorded content includes frame index, timestamp, refresh cycle parameters, uncertainty measurement and solution delay parameters.

[0013] Optionally, the generation and data feedback of the pose data set, task instruction set, uncertainty metric, and solution delay parameters include: The data interaction control module of the receiving device starts the visual recognition module to periodically collect the confidential dynamic QR code frame sequence displayed by the sending device, extract the latest updated frame according to the refresh cycle parameter, and register the timestamp and frame index of the updated frame. The visual recognition module transmits the extracted confidential dynamic QR code update frame to the confidential dynamic QR code parsing module. The confidential dynamic QR code parsing module calls the decoding unit to parse the encrypted data payload in the update frame and recover the pose field, confidence field, delay field and task field in the structured data payload. The integrity verification unit is invoked to perform encryption verification. Based on the hash digest result, the integrity and temporal consistency of the structured data payload are compared, and a set of parsing results is generated, including the pose data set, the task instruction set, the uncertainty measure, and the solution delay parameter. The data interaction control module receives the parsing result set and registers the pose data set, task instruction set, uncertainty measure and solution delay parameter, and generates an interaction receipt message. The receiving device sends back the device status information set and the interactive receipt message to the sending device. After receiving the returned information, the sending device triggers the regeneration process of the initial frame of the confidential dynamic QR code.

[0014] Optionally, the execution of the logistics navigation includes: The data interaction control module of the receiving device reads the pose data set, task instruction set, uncertainty measurement and solution delay parameters, and sets the scheduling sequence according to the refresh cycle parameters; The execution path calculation and obstacle avoidance control uses uncertainty metric as a weighting factor to suppress low-confidence poses during the calculation process; Based on the refresh cycle parameter and the calculation delay parameter, the visual recognition module is scheduled to continue acquiring images to ensure that the image acquisition time is consistent with the path calculation execution time. The confidential dynamic QR code generation module is scheduled to continue updating the QR code, generating confidential dynamic QR code update frames and appending them to the confidential dynamic QR code frame sequence. The driver visual recognition and positioning module periodically executes the improved iNeRF algorithm to continuously update and optimize the pose vector, thus completing the closed-loop scheduling of logistics navigation execution.

[0015] A logistics navigation system based on visual recognition and wireless communication according to an embodiment of the present invention includes: Automated logistics equipment is used to carry out various functional modules and operate in logistics scenarios; The camera is used to capture environmental images and confidential dynamic QR code images; A QR code display device for displaying a sequence of confidential dynamic QR code frames; The visual recognition module is used to perform image preprocessing to generate standardized image sequences, periodically acquire dynamic QR code image frames, extract timestamp information and pixel coordinate matrix, and calculate row time sequence parameters and motion blur index. The confidential dynamic QR code generation and parsing module is used to generate a confidential dynamic QR code frame sequence and parse and update the frame on the receiving device side to output a set of parsing results; The visual recognition and localization module is used to execute the improved iNeRF algorithm based on standardized image sequences, pose prior sets, and row time parameters, and output a localization result set and a feature weight map. The data interaction control module is used to set the dynamic QR code refresh cycle parameters, schedule each module to generate, parse and update confidential dynamic QR codes, perform path calculation and obstacle avoidance control, drive the visual recognition and positioning module to run periodically, and complete the closed-loop control of logistics navigation.

[0016] The beneficial effects of this invention are: First, it realizes row-level temporal rendering modeling under rolling shutter exposure conditions, effectively eliminating the pose drift and image blurring problems caused by row-level time differences in traditional visual positioning, and greatly improving the positioning accuracy and stability in complex dynamic environments.

[0017] Secondly, by utilizing the feature adaptive sampling structure and motion fuzzy constraint mechanism, the QR code region and the high-texture region are sampled and weighted for optimization, thereby improving the feature extraction efficiency and pose calculation accuracy.

[0018] Furthermore, by constructing a confidential dynamic QR code communication mechanism based on layered hybrid encryption, the security and synchronization of wireless data interaction are significantly enhanced, avoiding the information leakage and delay problems in task instruction transmission caused by traditional static QR codes.

[0019] In addition, the data interaction control module realizes the periodic scheduling of visual recognition, QR code updates and positioning algorithms, forming an efficient closed-loop logistics navigation control process. This enables the system to have the comprehensive advantages of high-precision positioning, secure communication and low-latency control, which significantly improves the intelligent operation level and task execution efficiency of automated logistics equipment. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0021] Figure 1 This is an overall flowchart of a logistics navigation method based on visual recognition and wireless communication proposed in this invention; Figure 2 This is a schematic diagram of the visual recognition and localization module structure based on the improved iNeRF algorithm proposed in this invention; Figure 3 This is a schematic diagram of the data interaction process of the confidential dynamic QR code generation and parsing module proposed in this invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0023] refer to Figure 1-3 A logistics navigation method based on visual recognition and wireless communication includes the following steps: In each automated logistics device, a visual recognition module, a confidential dynamic QR code generation and parsing module, a visual recognition positioning module and a data interaction control module are loaded. The camera and QR code display device are activated to collect environmental images and equipment status information. The visual recognition module performs image preprocessing to generate a standardized image sequence, and the data interaction control module sets the dynamic QR code refresh cycle parameters. The transmitting device is triggered by the data interaction control module to generate a confidential dynamic QR code. Based on the device status information set and encryption algorithm, it generates an initial frame of the confidential dynamic QR code, loads the device identifier, task instructions, timestamp and refresh cycle parameters, generates a confidential dynamic QR code frame sequence for wireless communication, and delivers it to the visual recognition module for display on the QR code display device. The visual recognition module acquires QR code images from the confidential dynamic QR code frame sequence. The confidential dynamic QR code generation module generates a QR code plane homography relation. The visual recognition positioning module calculates the initial pose estimate and covariance based on the homography relation to form a pose prior set. The visual recognition module also calculates the line time sequence parameters and motion blur index as inputs to the improved iNeRF algorithm. The visual recognition and localization module is based on standardized image sequences, pose prior sets, and row time parameters. It executes an improved iNeRF algorithm, uses a rolling shutter time modeling structure for row-level temporal rendering, and uses a feature adaptive sampling structure to generate a feature weight map. At the same time, it performs weight suppression on the feature weight map according to the motion blur index, performs high-density ray sampling on the QR code area and high-texture area, and uses an Anytime budget solution structure to perform multi-resolution optimization within the time delay budget limited by the refresh cycle parameter. It generates a localization result set, including optimized pose vector, uncertainty metric, and solution delay parameters, and outputs the feature weight map. The confidential dynamic QR code generation module performs structured encryption encoding on the location result set, feature weight map and task instructions, updates the QR code display device to generate confidential dynamic QR code update frames and appends them to the confidential dynamic QR code frame sequence, and the data interaction control module records the interaction status of this cycle. The receiving device's visual recognition module collects the confidential dynamic QR code frame sequence and extracts the latest updated frame. The confidential dynamic QR code parsing module completes parsing and encryption verification, and outputs a pose data set, a task instruction set, uncertainty measurement and solution delay parameters. These are then delivered to the data interaction control module for path calculation and obstacle avoidance control. At the same time, the data interaction control module sends the receiving device's device status information set back to the sending device to trigger the regeneration of the initial confidential dynamic QR code frame on the sending device. The data interaction control module of the receiving device performs path calculation and obstacle avoidance control based on the pose data set and task instruction set. During the path calculation process, the uncertainty metric is used as a weighting factor to suppress low-confidence poses. Based on the refresh cycle parameter and the solution delay parameter, the visual recognition module is scheduled to continue to acquire images, and the confidential dynamic QR code generation module is scheduled to continue to update the QR code. The visual recognition and positioning module is driven to periodically execute the improved iNeRF algorithm to continuously update and optimize the pose vector, thus completing the logistics navigation execution.

[0024] In this embodiment, the generation of the standardized image sequence and refresh cycle parameters includes: In each automated logistics device, a visual recognition module, a confidential dynamic QR code generation and parsing module, a visual recognition positioning module, and a data interaction control module are loaded. The camera and QR code display device are activated, and the control binding of the camera and QR code display device is established in the data interaction control module. The camera acquires environmental images and transmits them to the visual recognition module. The visual recognition module performs image preprocessing, including brightness normalization, noise suppression, and edge enhancement, to generate a standardized image sequence and add a timestamp for subsequent reading by the visual recognition and positioning module. The data interaction control module collects a set of device status information, calculates the dynamic QR code refresh cycle parameters based on the device operating frequency, processing delay and image frame rate, writes the dynamic QR code refresh cycle parameters into the parameter cache area of ​​the data interaction control module, and simultaneously distributes them to the confidential dynamic QR code generation module and the visual recognition and positioning module. The data interaction control module performs a consistency check on the timestamp identifier of the standardized image sequence and the refresh cycle parameter of the dynamic QR code. After the check passes, the standardized image sequence and the refresh cycle parameter of the dynamic QR code are registered as inputs for image processing and positioning calculation, and preparation is complete.

[0025] In this embodiment, the generation of the confidential dynamic QR code frame sequence includes: The sending device is triggered by the data interaction control module to generate the confidential dynamic QR code, which calls the set of device status information registered in the previous period, loads the current encryption algorithm configuration and dynamic QR code refresh cycle parameters, initializes the initial frame structure of the confidential dynamic QR code and allocates the data payload area. The confidential dynamic QR code generation module generates an initial frame of confidential dynamic QR code based on the device status information set and encryption algorithm. The encryption algorithm performs data scrambling and key index matching to ensure frame-level identity consistency between devices. The device identifier, task instruction, timestamp and refresh cycle parameters are loaded sequentially into the data payload area of ​​the confidential dynamic QR code initial frame to form a confidential dynamic QR code initial frame with complete parameter mapping, and a timestamp index field is established in the frame header area to support subsequent refresh synchronization. The generated confidential dynamic QR code initial frame is appended to the confidential dynamic QR code frame sequence. The confidential dynamic QR code frame sequence is marked as a display sequence for wireless communication, and the frame index number, refresh cycle parameter and timing attribute are registered by the data interaction control module for use in the next cycle. The confidential dynamic QR code frame sequence is delivered to the visual recognition module, which drives the QR code display device to display it. The visual recognition module performs inter-frame refresh and content replacement according to the dynamic QR code refresh cycle parameters, so that the confidential dynamic QR code frame sequence is continuously displayed, ensuring that the receiving device can collect a valid QR code frame image in any refresh cycle.

[0026] In this embodiment, the generation of the pose prior set and the row timing parameters includes: The visual recognition module periodically collects dynamic QR code image frames within the dynamic QR code refresh cycle and extracts the timestamp information and pixel coordinate matrix corresponding to each collection. The confidential dynamic QR code parsing module parses the encoded data in the dynamic QR code image frame, reconstructs the coordinates of the QR code boundary vertices and geometric feature points, calculates the plane homography relationship of the QR code based on the pixel coordinate matrix, and outputs the homography matrix parameter set for the visual recognition and positioning module to perform geometric registration. The visual recognition and positioning module calculates the initial pose estimate based on the homography relation of the QR code plane, and calculates the pose covariance matrix by using the pixel gradient distribution and feature point differences between periodically acquired image frames, forming a pose prior set and registering it in the data interaction control module. The visual recognition module combines standardized image sequences to perform line-level scanning timing analysis, calculates line timing parameters and frame-level sampling interval information, and calculates motion blur index based on image gradient change rate and optical flow characteristics; Specifically, the visual recognition module first reads the timestamp information of adjacent image frames in the standardized image sequence, the total number of rows of the image sensor, the frame readout duration, and the exposure start time. The frame readout duration is divided by the total number of rows to obtain the single-row exposure time step, which is used to determine the sampling time interval of each row of images. Based on the exposure start time and the single-row exposure time step, the sampling time corresponding to each row is calculated to form a row timing parameter, which is used to characterize the imaging time sequence of each row during the rolling shutter exposure process. Subsequently, the visual recognition module calculates the frame-level sampling interval information based on the timestamp difference between two consecutive frames. Multiple inter-frame intervals are smoothed using a sliding window to obtain steady-state frame-level sampling interval information, which is used to constrain the time modeling input of the improved iNeRF algorithm, so that the row-level exposure time and the inter-frame sampling frequency remain consistent in the time domain. In the calculation of motion blur index, the visual recognition module first selects the QR code region as the region of interest (ROI), and then calculates the pixel-level optical flow vector in two adjacent normalized images using an optical flow algorithm. ,in For the pixel coordinates within the ROI, the visual recognition module calculates the average value of the optical flow amplitude: ; in, The average optical flow amplitude represents the motion intensity of a pixel, where M represents the number of pixels within the ROI. This average optical flow amplitude is used to characterize the overall motion speed of the QR code region. The visual recognition module further calculates the gradient energy distribution of the image, using the grayscale intensity of the k-th frame in the normalized image sequence. Calculate its gradient magnitude The gradient magnitudes within the ROI region are averaged to obtain the average gradient energy of the current frame. This value is then compared with the maximum average gradient energy within the time window to obtain the gradient energy ratio. The closer the value is to 1, the clearer the image; Based on this, the visual recognition module uses the average optical flow amplitude and gradient energy ratio as inputs to calculate the motion blur index: ; in, Indicates the motion fuzziness index, This represents the weighting coefficient, with values ​​ranging from 0 to 1, where higher values ​​indicate higher weighting. Indicates a relatively high or low speed of movement. This indicates a high degree of image blur, therefore The value comprehensively reflects the dual impact of motion speed and image sharpness; The row-time parameters, motion blur indices, and pose priors are registered together to form the input data set for the improved iNeRF algorithm, providing constraints for subsequent row-level temporal rendering based on time modeling.

[0027] In this embodiment, the generation of the localization result set and the feature weight map includes: The visual recognition and localization module receives standardized image sequences, pose prior sets, line timing parameters, and motion blur indices. It sets a delay budget based on refresh cycle parameters, initializes the neural volume rendering network structure of the improved iNeRF algorithm, defines the ray direction set and sampling depth interval, and establishes a mapping relationship from pixel coordinates to world coordinates for subsequent volume rendering and optimization. The visual recognition and positioning module uses a rolling shutter time modeling structure to dynamically render and model the row-level time sequence. It introduces the exposure time term in the row time sequence parameters into the time dimension of each ray, and uses a time function to weight the exposure time of each row pixel to make the rendering result of each pixel strictly correspond to its exposure time, forming a time-continuous row-level rendering result. The time function is specifically: ; Where i represents the row index of the image, from 0 to... , This indicates the start time of the exposure for the k-th frame. Indicates the total number of rows of sensors. Indicates the frame readout duration. This indicates the exposure center time corresponding to the pixels in that row; The visual recognition and localization module is based on a feature adaptive sampling structure. It calculates the feature gradient magnitude distribution and texture saliency on the image plane, generates a feature weight map based on the feature gradient magnitude distribution and texture saliency, performs high-density ray sampling on the QR code area and high-texture area, and performs low-density ray sampling on low-texture area to achieve non-uniform sampling allocation. The visual recognition and positioning module dynamically suppresses the feature weight map based on the motion blur index. The motion blur index is used as a reference for weight adjustment. The feature weights in high motion areas are attenuated, so that the feature weight distribution is dynamically adjusted according to the local motion blur degree. In high motion areas, the feature response intensity is reduced and the ray sampling density is reduced, thereby reducing the impact of pixel blur caused by high-speed motion on the pose calculation accuracy. The visual recognition and localization module performs optimization in a multi-resolution space using an Anytime budget solution structure. Within a time delay budget limited by the refresh cycle parameter, it performs backpropagation optimization of volume density and color field layer by layer. In the low-resolution stage, it updates the coarse parameters of the network, and in the high-resolution stage, it refines the local structure. It outputs the intermediate optimization results in real time. When the budget is reached, it immediately outputs the localization result set, which includes the optimized pose vector, uncertainty metric, and solution delay parameter. The optimized pose vector represents the spatial pose estimation result after multi-resolution optimization, the uncertainty metric represents the solution confidence under the constraints of optical flow, gradient, and sampling density, and the solution delay parameter represents the termination time of the Anytime budget solution structure. The visual recognition and positioning module also outputs an updated feature weight map for subsequent encryption encoding and dynamic QR code update steps.

[0028] In this embodiment, the updating of the confidential dynamic QR code includes: The data interaction control module gathers the optimized pose vector, uncertainty measure and solution delay parameter from the positioning result set, and together with the feature weight map and task instructions, forms a structured data load. The structured data load includes pose field, confidence field, delay field, task field, timestamp, refresh cycle parameter and frame index. The confidential dynamic QR code generation module performs encryption and integrity verification on the structured data payload based on the encryption algorithm, generates a confidential dynamic QR code update frame, and binds a timestamp and refresh cycle parameter in the confidential dynamic QR code update frame for timing control. The encryption algorithm is specifically a layered hybrid encryption algorithm, which includes two sub-processes: symmetric key encryption and hash integrity verification. In the symmetric key encryption stage, a dynamic key derivation function is used to generate a periodically updated encryption key, and block encryption is performed on the pose field, confidence field, and task field of the structured data payload. In the hash integrity verification stage, a hash digest is calculated on the encrypted data payload and appended to the verification area of ​​the confidential dynamic QR code update frame to ensure the tamper resistance and consistency of the data during wireless transmission and visual recognition decoding. The confidential dynamic QR code generation module writes the confidential dynamic QR code update frame into the next frame position of the confidential dynamic QR code frame sequence, sets a valid marker and updates the frame index, and the confidential dynamic QR code frame sequence is maintained as a display sequence for wireless communication. The confidential dynamic QR code generation module updates the QR code display device to complete the display of the confidential dynamic QR code update frame, and controls the display duration and frame replacement according to the dynamic QR code refresh cycle parameter. The data interaction control module records the interaction status of the current cycle. The recorded content includes frame index, timestamp, refresh cycle parameter, uncertainty measure and solution delay parameter, which are used for alignment and triggering of subsequent parsing and path calculation.

[0029] In this embodiment, the generation and data feedback of the pose data set, task instruction set, uncertainty metric, and solution delay parameter include: The data interaction control module of the receiving device starts the visual recognition module to periodically collect the confidential dynamic QR code frame sequence displayed by the sending device, extract the latest updated frame according to the refresh cycle parameter, and register the timestamp and frame index of the updated frame. The visual recognition module transmits the extracted confidential dynamic QR code update frame to the confidential dynamic QR code parsing module. The confidential dynamic QR code parsing module calls the decoding unit to parse the encrypted data payload in the update frame and recover the pose field, confidence field, delay field and task field in the structured data payload. The confidential dynamic QR code parsing module calls the integrity verification unit to perform encryption verification. Based on the hash digest result, it compares the integrity and timing consistency of the structured data payload. If the verification result passes, it generates a parsing result set, including the pose data set, the task instruction set, uncertainty measure, and solution delay parameters. The data interaction control module receives the parsing result set and registers the pose data set, task instruction set, uncertainty measure and solution delay parameter as input data for path calculation and obstacle avoidance control, while generating an interaction receipt message. The data interaction control module sends the device status information set of the receiving device and the interaction receipt message back to the sending device. After receiving the back information, the sending device triggers the regeneration process of the initial frame of the confidential dynamic QR code, realizing a two-way closed-loop dynamic QR code update cycle.

[0030] In this embodiment, the execution of the logistics navigation includes: The data interaction control module of the receiving device reads the pose data set, task instruction set, uncertainty measurement and solution delay parameters, and sets the scheduling sequence according to the refresh cycle parameter for time alignment of path calculation and obstacle avoidance control; The data interaction control module performs path calculation and obstacle avoidance control. During the calculation process, it uses uncertainty metric as a weighting factor to suppress low-confidence poses, and completes motion planning and obstacle avoidance constraint application based on pose data set and task instruction set. Based on the refresh cycle parameter and the calculation delay parameter, the visual recognition module is scheduled to continue acquiring images to ensure that the image acquisition time is consistent with the path calculation execution time. The confidential dynamic QR code generation module continues to update the QR code, generating confidential dynamic QR code update frames and appending them to the confidential dynamic QR code frame sequence for use by the visual recognition module in the next cycle and the confidential dynamic QR code parsing module. The driving vision recognition and positioning module periodically executes the improved iNeRF algorithm to continuously update and optimize the pose vector, and registers the updated optimized pose vector, uncertainty metric and solution delay parameter for path calculation and obstacle avoidance control in the next cycle, thus completing the closed-loop scheduling of logistics navigation execution.

[0031] A logistics navigation system based on visual recognition and wireless communication includes: Automated logistics equipment is used to carry out various functional modules and operate in warehousing scenarios; The camera, electrically connected to the visual recognition module, is used to capture environmental images and confidential dynamic QR code images; A QR code display device, electrically connected to a confidential dynamic QR code generation and parsing module, is used to display a confidential dynamic QR code frame sequence; The visual recognition module communicates bidirectionally with the camera and data interaction control module. It is used to perform image preprocessing to generate standardized image sequences, periodically acquire dynamic QR code image frames under the constraint of dynamic QR code refresh cycle parameters, extract timestamp information and pixel coordinate matrix, calculate row time sequence parameters and motion blur index and register them as inputs for the improved iNeRF algorithm. The confidential dynamic QR code generation and parsing module communicates bidirectionally with the data interaction control module, the visual recognition module, and the QR code display device. It is used to generate the initial frame and the update frame of the confidential dynamic QR code through encryption algorithm on the sending device side and form a confidential dynamic QR code frame sequence. On the receiving device side, it parses the update frame of the confidential dynamic QR code and completes encryption verification to output the pose data set, the task instruction set, the uncertainty measure, and the solution delay parameter. The visual recognition and localization module communicates bidirectionally with the visual recognition module and the data interaction control module. It is used to execute the improved iNeRF algorithm based on the standardized image sequence, pose prior set and row time sequence parameters, perform row-level temporal rendering for rolling shutter time modeling, generate feature weight map according to feature adaptive sampling structure and implement weight suppression under motion blur index constraint, and perform high-density ray sampling of QR code area and high texture area. Within the time delay budget limited by refresh cycle parameter, it uses Anytime budget solution structure to perform multi-resolution optimization to output localization result set and feature weight map. The data interaction control module communicates bidirectionally with the visual recognition module, the confidential dynamic QR code generation and parsing module, and the visual recognition positioning module. It is used to set the dynamic QR code refresh cycle parameters, trigger the generation of the confidential dynamic QR code initial frame and confidential dynamic QR code update frame of the transmitting device, register and schedule standardized image sequences and time parameters, read the pose data set and task instruction set on the receiving device side and execute path calculation and obstacle avoidance control, suppress low confidence poses with uncertainty metric as a weight factor, schedule the next cycle of the visual recognition module and the confidential dynamic QR code generation and parsing module according to the solution delay parameters and refresh cycle parameters, and drive the visual recognition positioning module to periodically execute the improved iNeRF algorithm to complete closed-loop navigation control.

[0032] Example 1: To verify the feasibility of this invention in practice, it was applied to an automated warehousing and logistics scenario for path coordination and task scheduling among multiple mobile logistics devices. Each automated logistics device is equipped with a visual recognition module, a confidential dynamic QR code generation and parsing module, a visual recognition positioning module, a data interaction control module, a camera, and a QR code display device. The system achieves wireless information interaction and pose synchronization between devices through confidential dynamic QR codes. In the experiment, 10 logistics vehicles were arranged in the same warehouse aisle, with approximately 3 meters between each vehicle, all performing collaborative handling and obstacle avoidance tasks. The ambient light ranged from 200 to 800 lux, the device movement speed ranged from 0.3 to 1.2 m / s, and the test lasted for 8 consecutive hours.

[0033] During operation, each device acquires images in real time through a visual recognition module and generates standardized image sequences, with a refresh cycle parameter set to 120ms. The confidential dynamic QR code generation module performs structured encryption encoding on the positioning result set, task instructions, and feature weight map, generating a QR code with a resolution of 256×256 pixels and an update frame rate of 8 frames per second. The visual recognition positioning module uses an improved iNeRF algorithm for row-level temporal rendering, combined with motion blur indicators to dynamically suppress sampling weights in high-motion regions, and finally outputs the pose calculation results under a multi-resolution optimization mechanism.

[0034] To verify the performance of this invention in wireless communication synchronization and secure transmission, a comparison was made between the traditional method (ordinary static QR code communication + PnP positioning) and the method of this invention, using data from the same task and environment. Test metrics included positioning accuracy, frame decoding success rate, communication latency, anti-interference rate, and system stable operation time. The results are shown in the table below.

[0035] Table 1. Performance comparison results between traditional methods and the method of this invention.

[0036] As shown in Table 1, the present invention outperforms traditional methods in terms of positioning accuracy, communication reliability, and system stability. The average positioning accuracy improved from 2.16 cm to 1.80 cm, indicating that the improved iNeRF algorithm, by introducing row-time modeling and motion blur constraints, effectively reduced image distortion errors in dynamic scenes, resulting in more accurate pose calculation. The frame decoding success rate increased from 82.2% to 89.4%, mainly due to the enhanced noise resistance of the layered hybrid encryption algorithm during QR code encoding and parsing, reducing recognition failures and data corruption. The communication latency decreased from 118.3 ms to 95.2 ms, indicating that the Anytime budget solution structure improved the computational and transmission coordination efficiency under dynamic refresh cycles, making wireless communication more real-time. The anti-interference rate increased from 84.5% to 91.8%, demonstrating a significant improvement in the system's robustness to data interaction under complex lighting and occlusion environments. The stable operating time increased from 6.9 h to 8.1 h, indicating that the overall architecture can maintain high synchronization and security in long-term continuous tasks.

[0037] The performance improvement of this invention lies in the deep integration of visual recognition and wireless communication, which constructs a multi-module collaborative system centered on temporal continuity and secure interaction. The improved iNeRF algorithm introduces rolling shutter time modeling and motion blur constraints, enabling pose calculation to move beyond single-frame static features and achieve dynamic compensation by combining row-level temporal information, fundamentally reducing positioning errors caused by motion blur and illumination changes. The feature adaptive sampling structure performs high-density sampling on high-texture and QR code regions, improving image feature resolution and optimizing convergence speed within a finite time delay budget. The confidential dynamic QR code employs a layered hybrid encryption algorithm, combined with dynamic refresh and hash integrity verification mechanisms, effectively improving the anti-interference and anti-tampering capabilities of data transmission. The Anytime budget solution structure further optimizes real-time performance, enabling path calculation, QR code updates, and visual recognition to operate collaboratively under unified temporal constraints. This allows the system to maintain high positioning accuracy and communication stability even in multi-source interference and highly dynamic environments, significantly improving overall navigation performance and reliability.

[0038] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A logistics navigation method based on visual recognition and wireless communication, characterized in that, The method comprises the following steps: Collecting environmental images and device state information sets, performing image preprocessing to generate standardized image sequences, and setting dynamic two-dimensional code refresh cycle parameters; The sending device generates a confidential dynamic two-dimensional code frame sequence based on the device state information set and the encryption algorithm, simultaneously loads the device identifier, task instruction, timestamp, and refresh cycle parameter, and generates a confidential dynamic two-dimensional code frame sequence; Based on the confidential dynamic two-dimensional code frame sequence, a two-dimensional code plane homography relationship is generated, an initial pose estimate value and a covariance are calculated to form a pose prior set, and a line timing parameter is calculated; Based on the standardized image sequence, the pose prior set, and the line timing parameter, an improved iNeRF algorithm is executed to generate a positioning result set and output a feature weight map; The positioning result set, the feature weight map, and the task instruction are structured and encrypted to generate a confidential dynamic two-dimensional code update frame, which is appended to the confidential dynamic two-dimensional code frame sequence; The receiving device collects the confidential dynamic two-dimensional code frame sequence to complete analysis and encryption verification, and outputs a pose data set, a task instruction set, an uncertainty measure, and a calculation delay parameter; Based on the pose data set and the task instruction set, path calculation and obstacle avoidance control are performed, low-confidence poses are suppressed according to the uncertainty measure, and based on the refresh cycle parameter and the calculation delay parameter, images are continuously collected and the two-dimensional code is updated, the improved iNeRF algorithm is periodically executed, and logistics navigation is completed. 2.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The generation of the standardized image sequence and the refresh cycle parameter comprises: In each automated logistics device, load a visual recognition module, a confidential dynamic two-dimensional code generation and analysis module, a visual recognition positioning module, and a data interaction control module, start a camera and a two-dimensional code display device; The camera obtains environmental images and transmits them to the visual recognition module, performs image preprocessing, including brightness normalization, noise suppression, and edge enhancement, and generates a standardized image sequence; The data interaction control module collects a device state information set, calculates a dynamic two-dimensional code refresh cycle parameter based on device operating frequency, processing delay, and image frame rate; The timestamp identifier of the standardized image sequence and the dynamic two-dimensional code refresh cycle parameter are subjected to consistency verification. 3.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The generation of the confidential dynamic two-dimensional code frame sequence comprises: The sending device triggers the confidential dynamic two-dimensional code generation module from the data interaction control module, calls the device state information set registered in the previous period, loads the current encryption algorithm configuration and the dynamic two-dimensional code refresh cycle parameter, and allocates a data load area; The confidential dynamic two-dimensional code generation module generates a confidential dynamic two-dimensional code initial frame based on the device state information set and the encryption algorithm; In the data load area of the confidential dynamic two-dimensional code initial frame, the device identifier, task instruction, timestamp, and refresh cycle parameter are loaded in sequence to form a confidential dynamic two-dimensional code initial frame with complete parameter mapping, and a timestamp index field is established in the frame header area; The generated confidential dynamic two-dimensional code initial frame is appended to the confidential dynamic two-dimensional code frame sequence; The confidential dynamic two-dimensional code frame sequence is delivered to the visual recognition module, the two-dimensional code display device is driven by the visual recognition module for display, and frame refresh and content replacement are performed based on the dynamic two-dimensional code refresh cycle parameter. 4.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The generation of the pose prior set and the line timing parameter comprises: The visual recognition module periodically collects dynamic two-dimensional code image frames within a dynamic two-dimensional code refresh cycle, and extracts the corresponding timestamp information and pixel coordinate matrix for each collection; The confidential dynamic two-dimensional code analysis module analyzes the encoded data in the dynamic two-dimensional code image frame, reconstructs the two-dimensional code boundary vertex coordinates and geometric feature points, calculates the two-dimensional code plane homography relationship based on the pixel coordinate matrix, and outputs the homography matrix parameter set; The visual recognition positioning module calculates the initial pose estimate value based on the two-dimensional code plane homography relationship, calculates the pose covariance matrix using the pixel gradient distribution and feature point difference between the periodically collected image frames, and forms the pose prior set; The visual recognition module performs line-level scan timing analysis in combination with the standardized image sequence, calculates the line timing parameters and frame-level sampling interval information, and calculates the motion blur index based on the image gradient change rate and optical flow features. 5.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The generation of the positioning result set and the feature weight map includes: The visual recognition positioning module receives the standardized image sequence, the pose prior set, the line timing parameters, and the motion blur index; A rolling shutter time modeling structure is used to dynamically render and model the line-level timing. An exposure time item in the line timing parameters is introduced in the time dimension of each ray. The exposure time of each row of pixels is modeled by a time function, so that the rendering result of each pixel point strictly corresponds to its exposure time, forming a time-continuous line-level rendering result. Based on the feature adaptive sampling structure, the feature gradient amplitude distribution and texture saliency are calculated on the image plane to generate a feature weight map. High-density ray sampling is performed on the two-dimensional code region and high-texture region, and low-density ray sampling is performed on the low-texture region. According to the motion blur index, the feature weight map is dynamically suppressed. The motion blur index is used as a weight adjustment reference to attenuate the feature weight of the high-motion region. In the multi-resolution space, an Anytime budget solving structure is used to perform optimization. The inverse propagation optimization of volume density and color field is performed layer by layer within the time delay budget limited by the refresh cycle parameters, and the optimization intermediate result is output in real time. When the budget is reached, the positioning result set is immediately output. The positioning result set includes the optimized pose vector, the uncertainty measure, and the solving time delay parameter. 6.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The update of the confidential dynamic two-dimensional code includes: The data interaction control module collects the optimized pose vector, the uncertainty measure, and the solving time delay parameter in the positioning result set, together with the feature weight map and the task instruction to form a structured data load; The confidential dynamic two-dimensional code generation module performs encryption and integrity verification on the structured data load based on the encryption algorithm, generates a confidential dynamic two-dimensional code update frame, binds the timestamp and refresh cycle parameters in the confidential dynamic two-dimensional code update frame; The confidential dynamic two-dimensional code update frame is written into the next frame position of the confidential dynamic two-dimensional code frame sequence, the effective flag is set, and the frame index is updated; The updated two-dimensional code display device completes the display of the confidential dynamic two-dimensional code update frame, and controls the display time and frame replacement according to the dynamic two-dimensional code refresh cycle parameters; The data interaction control module records the current cycle interaction state, and the recording content includes the frame index, the timestamp, the refresh cycle parameter, the uncertainty measure, and the solving time delay parameter. 7.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The generation of the pose data set, the task instruction set, the uncertainty metric and the calculation time delay parameter and data return include: The data interaction control module of the receiving device starts the visual recognition module, performs periodic collection on the confidential dynamic two-dimensional code frame sequence displayed by the sending device, extracts the latest updated frame according to the refresh period parameter, and registers the timestamp and frame index of the updated frame; The visual recognition module transmits the extracted confidential dynamic two-dimensional code updated frame to the confidential dynamic two-dimensional code analysis module, and the confidential dynamic two-dimensional code analysis module calls the decoding unit to analyze the encrypted data load in the updated frame, and restores the pose field, confidence field, time delay field and task field in the structured data load; Call the integrity checking unit to perform encryption checking, compare the integrity and time sequence consistency of the structured data load according to the hash digest result, generate an analysis result set including the pose data set, the task instruction set, the uncertainty metric and the calculation time delay parameter; The data interaction control module receives the analysis result set, and registers the pose data set, the task instruction set, the uncertainty metric and the calculation time delay parameter, and generates an interaction return message; The device state information set of the receiving device and the interaction return message are returned to the sending device together, and the sending device triggers the regeneration process of the confidential dynamic two-dimensional code initial frame after receiving the return information. 8.The logistics navigation method based on visual recognition and wireless communication of claim 1, wherein, The execution of the logistics navigation includes: The data interaction control module of the receiving device reads the pose data set, the task instruction set, the uncertainty metric and the calculation time delay parameter, and sets the scheduling time sequence according to the refresh period parameter; Perform path calculation and obstacle avoidance control, and in the calculation process, the low confidence pose is suppressed by taking the uncertainty metric as a weight factor; According to the refresh period parameter and the calculation time delay parameter, the visual recognition module continues to collect images to ensure that the image collection time and the path calculation execution time are consistent; The confidential dynamic two-dimensional code generation module continues to update the two-dimensional code, generates a confidential dynamic two-dimensional code updated frame and appends it to the confidential dynamic two-dimensional code frame sequence; Drive the visual recognition positioning module to periodically execute the improved iNeRF algorithm to continuously update the optimized pose vector, complete the closed-loop scheduling of the logistics navigation execution.

9. A logistics navigation system based on visual identification and wireless communication, performing a logistics navigation method based on visual identification and wireless communication according to any one of claims 1 to 8. It includes: An automated logistics device for carrying various functional modules and running in a logistics scene; A camera for collecting environment images and confidential dynamic two-dimensional code images; A two-dimensional code display device for displaying a confidential dynamic two-dimensional code frame sequence; A visual recognition module for performing image preprocessing to generate a standardized image sequence, periodically collecting dynamic two-dimensional code image frames, extracting timestamp information and pixel coordinate matrices, and calculating line timing parameters and motion blur indicators; A confidential dynamic two-dimensional code generation and analysis module for generating a confidential dynamic two-dimensional code frame sequence and analyzing updated frames on the receiving device side to output an analysis result set; A visual recognition positioning module for executing an improved iNeRF algorithm based on a standardized image sequence, a pose prior set and line timing parameters to output a positioning result set and a feature weight map; The data interaction control module is used for setting a dynamic two-dimensional code refreshing period parameter, scheduling modules to execute confidential dynamic two-dimensional code generation, analysis and updating, executing path calculation and obstacle avoidance control, driving the visual recognition positioning module to run periodically, and completing the logistics navigation closed-loop control.

Images