CN-122023522-A - High-speed real-time visual navigation and image matching positioning method for micron-sized features of optical communication module

CN122023522ACN 122023522 ACN122023522 ACN 122023522ACN-122023522-A

Abstract

The invention belongs to the technical field of optical communication and machine vision intersection, in particular to a high-speed real-time vision navigation and image matching positioning method for micron-sized features of an optical communication module, which comprises constructing a system architecture integrating multi-scale feature guidance, dynamic disturbance modeling and closed-loop feedback correction, the hardware synchronous image acquisition, the multi-scale coarse-fine matching, the parameter self-adaptive adjustment driven by disturbance state estimation, the geometric consistency verification and the thermal and mechanical coupling compensation are realized, and the end-to-end deterministic time sequence control is realized. The method has the advantages that an invalid region is rapidly eliminated in a coarse positioning stage through a multi-scale feature guiding mechanism, calculation overhead caused by global high-dimensional matching is avoided, dynamic disturbance modeling and physical layer parameter adjustment are utilized, environmental interference is restrained at an image generation source instead of back-end algorithm compensation, weighted fusion of local phase consistency and gradient direction information is realized, and a matching strategy can adapt to feature reliability change under different noise levels.

Inventors

LI JIACHANG
XU HONGCHAO
Teng Baojian

Assignees

海克斯康制造智能技术(深圳)有限公司

Dates

Publication Date: 20260512
Application Date: 20260127

Claims (10)

1. The high-speed real-time visual navigation and image matching positioning method for the micron-scale features of the optical communication module is characterized by being executed based on a deterministic system architecture consisting of an image acquisition module, a preprocessing module, a multi-scale matching engine module, a disturbance state estimation module, a closed-loop feedback correction module and a motion control interface module, wherein the modules are coupled through a hardware-level synchronization and time sequence linkage mechanism, and the method comprises the following steps: The image acquisition module receives a hard trigger signal after the moving platform is in place, a camera is started to acquire an original image of the surface area of the optical communication module to be detected, and the camera is synchronized with a moving platform encoder through a hardware interface, so that the image acquisition time is aligned with the pose of the platform; The preprocessing module sequentially executes geometric correction based on a lens distortion parameter matrix, white balance adjustment based on feedback of an ambient light sensor and local contrast enhancement aiming at a preset characteristic region on an original image, and divides the processed image into subareas with overlapped boundaries, wherein each subarea corresponds to one type of target characteristic; the multi-scale matching engine module executes two-stage processing of a coarse positioning layer and a fine matching layer, namely, in the coarse positioning layer, projecting a sub-region image to a principal component space corresponding to a low-dimensional geometric template library constructed offline, calculating a confidence score, entering the fine matching layer if the score is higher than a safety threshold, otherwise, calling a disturbance state estimation module to start dynamic disturbance compensation; At the fine matching layer, adopting a local phase consistency and gradient direction histogram joint descriptor to perform feature matching, wherein the weight coefficient of the matching distance is dynamically adjusted by the disturbance state estimation value; The matching result is verified by geometric consistency, if convergence judging conditions are met, six-degree-of-freedom pose parameters are output, otherwise, a local resampling subprogram is triggered, a spiral scanning track is executed in a range, and multi-frame matching results are fused; The closed loop feedback correction module receives the effective pose sequence, detects systematic drift through time sequence trend analysis, generates thermal compensation quantity by combining clamp temperature data and a material thermal expansion model if the thermal deformation exists, and superimposes the thermal compensation quantity on the current pose parameter to form a final correction instruction; the motion control interface module receives the correction instruction, generates a control signal meeting the requirement of the servo driver, and drives the assembly executing mechanism to complete positioning operation.
2. The method of claim 1, wherein the dynamic disturbance compensation is performed by a disturbance state estimation module that uses the fluctuation of light source intensity, the micro-vibration amplitude of the platform, and the shift of the depth of field of the lens as state variables based on a three-state disturbance model, and uses a kalman filter to fuse the ambient light intensity signal, the accelerometer reading, and the auto-focus feedback value to estimate the current disturbance state in real time, and dynamically adjust the exposure time, the gain factor, and the focus position of the image acquisition module accordingly.
3. The method of claim 2, wherein the kalman filter is implemented with fixed point numbers, and the state transition matrix is obtained by offline training of production line history matching failure data, and the process noise covariance matrix and the observed noise covariance matrix are solidified in the nonvolatile memory in the system initialization stage.
4. The method of claim 1, wherein the local phase consistency in the fine matching layer is obtained by multi-scale Gabor filter response calculation, the gradient direction histogram descriptor is constructed by taking the local maximum point of the phase consistency as the center and only the partial component with the largest energy is reserved, and the weight coefficient alpha in the weighted Euclidean distance is calculated according to the formula Dynamic setting, wherein Is a state estimation value of the micro-vibration amplitude of the platform.
5. The method of claim 1, wherein the geometric consistency verification is based on affine transformation residual root mean square joint determination with scale factor deviation.
6. The method according to claim 1, wherein the local resampling subroutine controls the motion platform to perform spiral track on the X/Y plane with the current pose as the center, synchronously collect multi-frame images, and weight average the matching results of each frame according to the geometric consistency index.
7. The method of claim 1, wherein the closed loop feedback correction module is implemented by an FPGA logic unit and comprises a dual-port RAM buffer area for storing an effective pose sequence, and wherein the embedded motion controller periodically reads the sequence and activates the thermal and mechanical coupling compensation unit when detecting that the Z-axis displacement increment direction of the continuous set times is consistent and the amplitude is monotonically increased.
8. A high-speed real-time visual navigation and image matching positioning system for micron-sized features of an optical communication module, which is applicable to the high-speed real-time visual navigation and image matching positioning method for micron-sized features of an optical communication module according to any one of claims 1 to 7, and is characterized in that the system comprises: the image acquisition module is configured with a camera, a telecentric lens and an environment sensor array and is synchronous with the motion platform through hard wire triggering; The preprocessing module is integrated with the on-board image signal processor and used for executing geometric correction, white balance and local contrast enhancement; The multi-scale matching engine module adopts a pipelined hardware acceleration architecture and comprises a coarse positioning calculation unit, a fine matching calculation unit and a verification arbitration unit, wherein the coarse positioning calculation unit, the fine matching calculation unit and the verification arbitration unit are connected through an on-chip interconnection network and are synchronously controlled by a central scheduler frame; the disturbance state estimation module is integrated in an independent coprocessor mode and operates a fixed-point Kalman filter; the closed loop feedback correction module is realized by FPGA logic, the input end receives the pose flow, and the output end is connected with the motion control interface through an RS485 bus; And the motion control interface module analyzes the six-degree-of-freedom pose instruction and generates a servo pulse sequence.
9. The system of claim 8, wherein in the multi-scale matching engine module, the coarse positioning calculation unit is formed by a matrix multiplier array and is used for performing principal component projection operation, the fine matching calculation unit integrates a reconfigurable filter bank and a histogram accumulator, and the verification arbitration unit is internally provided with a geometric consistency judging logic circuit, wherein a threshold value register of the verification arbitration unit supports upper-level dynamic configuration.
10. The system of claim 8, wherein the image acquisition module is connected to the preprocessing module, the preprocessing module is connected to the multi-scale matching engine module, the disturbance state estimation module exchanges sensory data with the main processor through the shared memory, and the end-to-end delay of the whole system from image acquisition to correction instruction output is locked by the hardware clock divider within a certain time window.

Description

High-speed real-time visual navigation and image matching positioning method for micron-sized features of optical communication module Technical Field The invention belongs to the field of intersection of optical communication and machine vision, and particularly relates to a high-speed real-time visual navigation and image matching positioning method for micron-sized features of an optical communication module. Background With the rapid development of mobile communication, data center interconnection and high-speed optical fiber access network, the optical communication module is used as an optical-electrical signal conversion core device, the manufacturing and assembling precision requirements of the optical communication module are up to the micrometer to submicron level, and in the automatic production and precise assembling process, the rapid and accurate visual identification and spatial positioning of the microstructures (such as optical fiber end faces, coupling lenses and the like) in the module are key for guaranteeing the product performance consistency and the yield, and the navigation positioning technology based on machine vision is widely introduced into the intelligent manufacturing process of the optical communication module because of the advantages of non-contact, high resolution and integrability, so that the optical communication module becomes a core enabling means for supporting high-precision assembling, automatic alignment and online detection. The current mainstream technical scheme adopts a high-resolution industrial camera to match with a precision motion platform, combines a template matching or feature point extraction algorithm (such as SIFT, SURF and edge contour matching) to realize target positioning, can meet basic requirements under static or low-speed working conditions, relies on a pre-acquired standard template image, completes matching by calculating similarity or geometric transformation parameters of an image to be detected and a template, has good engineering practicability in early-stage scenes with simple optical module structure and larger feature scale (more than tens of micrometers), and is deployed as an industry standard solution for a long time because the algorithm has clear logic and definite hardware dependence. The method is oriented to a new generation of high-speed optical communication module (400G/800G coherent module and multichannel silicon optical integrated device), the internal micron-level features of the method are densely distributed, the optical contrast is low, the method is easy to be interfered by illumination and lens distortion, the traditional method faces fundamental challenges, the core contradiction is irreconcilable conflict of instantaneity and matching robustness, high-dimensional features or global optimization are required for improving accuracy, so that the calculation complexity is too high to meet the production line beat, the algorithm is simplified to pursue high speed, mismatching is easily caused by tiny deformation, shielding and the like, the contradiction is caused by strong assumption of the traditional method on the consistency of image content, the assumption is frequently failed due to factors such as light source fluctuation, reflection and the like in an actual production line, the existing improvement means can only solve the problem, and the difficulty that the error accumulation is easy to be caused by closed loop feedback correction capability of micron-level positioning errors, the error is not accurate due to the trapping, and the criterion is not fast is severely restricted, and the automatic production efficiency and the process stability of the high-end optical communication module are severely limited. Therefore, the invention provides a high-speed real-time visual navigation and image matching positioning method for micron-sized features of an optical communication module. Disclosure of Invention In order to achieve the above-mentioned purpose, the present invention provides a method for high-speed real-time visual navigation and image matching positioning of micron-sized features of an optical communication module, which solves the above-mentioned technical problems. The method comprises the following steps of acquiring an original image of a surface area of an optical communication module to be detected under the control of a synchronous trigger signal through a high-frame-rate global shutter industrial camera arranged on a precision motion platform, connecting the camera with an encoder signal of the motion platform through a hardware-level synchronous interface to ensure that the image acquisition time is strictly aligned with the pose state of the platform, performing preprocessing operation on the original image through an on-board image signal processor, wherein the preprocessing operation comprises geometric correction based on a lens distortion parameter matrix, self-adaptive white balance a