An asynchronous exposure dual-view field imaging system and imaging method

Abstract

The application relates to the technical field of industrial vision detection, and discloses an asynchronous exposure double-view field imaging system and an imaging method. Based on the asynchronous exposure double-view field imaging system, independent view field images collected by two parallel lenses can be combined into a single linear array camera through a mirror and a non-polarization beam splitting cube. Meanwhile, cooperating with precise timing control, the system drives two independent stroboscopic light sources and a camera line trigger signal to be synchronous, alternately illuminates different view fields in an odd-even line cycle, makes the camera asynchronously expose and collect two-way image information at different times, and finally reconstructs two complete, independent and seamlessly spliced detection images from the interwoven data stream through an odd-even line separation algorithm. The application realizes significant expansion of a detection view field with a single camera hardware cost, guarantees detection efficiency and image geometric accuracy under a high line scanning frequency, and realizes excellent balance of performance, cost and efficiency.

Description

Technical Field

[0001] This invention relates to the field of industrial visual inspection technology, and in particular to an asynchronous exposure dual-field imaging system and imaging method. Background Technology

[0002] In modern industrial production, visual inspection technology, with its advantages of being non-contact, highly precise, and efficient, has become a core means of product quality control. Taking OLED screen production as an example, indicators such as display uniformity and pixel defects directly determine product quality, thus requiring full-panel inspection for quality screening. As OLED screens develop towards larger sizes, the need for expanded inspection fields is becoming increasingly urgent, and traditional imaging technologies face a dilemma between cost, field of view, and efficiency.

[0003] Currently, there are two main technical solutions for large field-of-view inspection in the industrial field. One of them is multi-camera stitching scanning technology. The core of this solution is to arrange multiple line scan cameras side by side along the width of the workpiece. Each camera is responsible for capturing a sub-region of the area to be inspected. The sub-images captured by multiple cameras are then combined into a complete giant image using an image stitching algorithm. Its hardware components include: multiple line scan cameras, a corresponding number of image acquisition cards, a synchronization control module, and stitching processing software. The working principle is as follows: the synchronization control module ensures that the exposure sequence of all cameras is consistent. When the workpiece moves at a constant speed with the conveyor belt, multiple cameras simultaneously capture the row images of their respective responsible areas. Then, the stitching algorithm eliminates the field-of-view offset between cameras, forming a complete inspection image. The advantages of this solution are that it can achieve full-area coverage and has high inspection efficiency, requiring no additional movement of the workpiece or cameras. However, it has the following significant drawbacks: 1. High hardware cost: Each additional sub-field of view requires an extra camera, a data acquisition card, and a corresponding mounting bracket. For large workpieces, 2-4 cameras are usually needed, and the hardware cost increases linearly, which seriously conflicts with the cost control requirements of industrial production; 2. Complex synchronization control: The exposure sequence and acquisition rhythm of multiple cameras need to be strictly synchronized, otherwise it will lead to image stitching misalignment. Therefore, a high-precision synchronization control module is required, which increases the complexity and failure rate of the system; 3. Difficult installation and debugging: The installation positions of multiple cameras need to be precisely calibrated to ensure that the overlapping area of ​​the field of view meets the stitching requirements. The calibration process is time-consuming and laborious, and if the camera position is offset during later maintenance, complex recalibration work is required.

[0004] Another approach is multi-step scanning technology. This solution uses a combination of a single camera and a high-precision motion platform. The core principle is that the motion platform controls a single camera to move back and forth along the width of the workpiece, capturing multiple images of different sub-regions. These images are then stitched together using a stitching algorithm to create a complete image. The hardware components include: a single line scan camera, a high-precision linear motion platform, a motion control card, and an image acquisition and processing module. The working principle is as follows: the workpiece moves at a constant speed with a conveyor belt, and the motion platform drives the camera to move in a direction perpendicular to the workpiece's movement. Each time the camera moves across a field of view, it captures an image of a sub-region. This process is repeated until the entire workpiece width is covered. Finally, a stitching algorithm combines the images of multiple sub-regions into a complete image. The advantage of this solution is its lower hardware cost; a large field of view can be achieved with only a single camera. However, it has the following key drawbacks: 1. Extremely low detection efficiency: The camera needs to move back and forth with the motion platform, and multiple starts, stops, and movements are required to cover each complete field of view. The detection speed is limited by the moving speed and acceleration of the motion platform. For large-sized workpieces, the detection efficiency is reduced by more than 50% compared to the single-camera solution, which cannot meet the detection requirements of high-speed industrial production lines; 2. Imaging quality depends on the accuracy of the motion platform: The positioning accuracy and stability of the motion platform directly determine the stitching accuracy of the sub-region images. If the motion platform has problems such as vibration or positioning deviation, it will cause image stitching misalignment and blurring, affecting the detection accuracy; and the procurement cost of high-precision motion platforms is high, which partially offsets the cost advantage of the single-camera solution.

[0005] In summary, existing large field-of-view detection technologies are either prohibitively expensive in terms of hardware or suffer from low detection efficiency and poor stability, failing to simultaneously meet the comprehensive demands of industrial production for low cost, large field of view, high efficiency, and high precision. Therefore, there is an urgent need for an innovative imaging technology that can simultaneously guarantee a large field of view and high efficiency while controlling hardware costs, thus resolving the core contradictions of existing technologies. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, the present invention aims to provide an asynchronous exposure dual-field imaging system and imaging method, which aims to solve the problems of high hardware cost, low detection efficiency and image quality being limited by mechanical motion precision in existing large-size, high-resolution visual inspection schemes.

[0007] The first aspect of this invention provides an asynchronous exposure dual-field-of-view imaging system, comprising an imaging module, an acquisition module, an illumination control module, and a data processing module; the imaging module is connected to the acquisition module, the illumination control module is mounted in front of the imaging module, and the acquisition module is communicatively connected to the data processing module; the imaging module is used to merge optical information from two spatially separated fields of view on the object under test into a single coaxial outgoing beam, comprising a first lens and a second lens arranged side-by-side along a first direction, a first reflector and a second reflector arranged one-to-one with the first lens and the second lens, and a non-polarizing beam splitter cube; the illumination control module includes a first stroboscopic flash source, a second stroboscopic flash source, and a controller; the first stroboscopic flash source and the second stroboscopic flash source respectively illuminate the field of view areas of the first lens and the second lens; the controller is used to receive information from the object... The system generates a row trigger signal to synchronize body displacement, and based on the row trigger signal, generates a control timing sequence that alternately drives the first and second frequency flash sources to turn on and off, as well as a trigger signal to synchronously control the exposure of the line scan camera. This ensures that during odd-numbered row triggering periods, only the first frequency flash source is lit and the line scan camera acquires the first field-of-view image row, while during even-numbered row triggering periods, only the second frequency flash source is lit and the line scan camera acquires the second field-of-view image row. The acquisition module includes a single line scan camera, used to receive the emitted beam merged by the imaging module and convert it into a continuous image row data stream. The data processing module receives the image row data stream output by the line scan camera, and uses an odd-even row separation algorithm to divide the image row data stream into a first field-of-view image dataset and a second field-of-view image dataset, and performs image reconstruction on each dataset to obtain complete two-dimensional images with two independent fields of view.

[0008] Optionally, in a first implementation of the first aspect of the present invention, in the imaging module, the optical axes of the first lens and the second lens are parallel and perpendicular to the plane where the object being measured is located; the reflecting surfaces of the first mirror and the second mirror form a 45° angle with the horizontal plane, which is used to deflect the vertical incident beam from the corresponding lens by 90° into a horizontal beam; the non-polarized beam splitter is disposed at the center of the line connecting the first mirror and the second mirror, and the beam splitting surface inside it forms a 45° angle with the optical axes of the horizontal beams from both sides, which is used to reflect and transmit the two horizontal beams according to a preset beam splitting ratio, and to merge the reflected beams into one output beam coaxial with the receiving optical axis of the line scan camera.

[0009] Optionally, in a second implementation of the first aspect of the present invention, the imaging parameters of the first lens and the second lens are consistent, and their fields of view overlap in a region perpendicular to the direction of object movement, wherein the width of the overlapping region accounts for 10%-15% of the field of view of a single lens.

[0010] Optionally, in a third implementation of the first aspect of the present invention, the non-polarized beam splitter has a beam splitting ratio of 50:50 for the light beam in the wavelength range of 400nm to 700nm.

[0011] Optionally, in a fourth implementation of the first aspect of the present invention, the imaging module further includes an opaque housing that encloses the first lens, the second lens, the first reflector, the second reflector, and the optical path of the non-polarized beam splitter cube, forming an independent optical channel.

[0012] Optionally, in a fifth implementation of the first aspect of the present invention, the controller includes a high-frequency logic controller and a motion control card; the motion control card is connected to a servo driver that drives the motion platform and is used to control the motion platform to drive the object under test to move at a constant speed; the line trigger signal is provided by a rotary encoder installed on the motion platform; the high-frequency logic controller is used to receive the line trigger signal and perform frequency division, frequency multiplication and logic processing on the line trigger signal to generate two complementary, pulse width-adjustable light source drive pulse signals and one line scan camera exposure trigger signal synchronized with the light source drive pulse.

[0013] A second aspect of the present invention provides an asynchronous exposure dual-field-of-view imaging method based on the imaging system described in the present invention, comprising the following steps: initializing the imaging system and configuring its parameters; placing the object under test on a motion platform and starting the motion platform and the imaging system; receiving a line trigger signal generated by a rotary encoder mounted on the motion platform, synchronized with the displacement of the object under test; performing asynchronous exposure control according to the parity of the line trigger signal: driving the first frequency flash source to light up and the second frequency flash source to turn off during odd-numbered line trigger cycles, and synchronously triggering the line scan camera to expose and acquire a first field-of-view image line; driving the second frequency flash source to light up and the first frequency flash source to turn off during even-numbered cycles, and synchronously triggering the line scan camera to expose and acquire a second field-of-view image line; outputting an image line data stream containing the first and second field-of-view image lines through the line scan camera; performing odd-even line separation processing on the image line data stream, extracting the odd-numbered line data into a first field-of-view image dataset and the even-numbered line data into a second field-of-view image dataset; and reconstructing two-dimensional images from the first and second field-of-view image datasets respectively to obtain two complete and independent first and second field-of-view images.

[0014] Optionally, in the first implementation of the second aspect of the present invention, the asynchronous exposure control is specifically implemented as follows: the row trigger signal is counted and judged by a high-frequency logic controller; when the count value is odd, a first light source driving pulse and a camera exposure trigger pulse are generated; when the count value is even, a second light source driving pulse and a camera exposure trigger pulse are generated; the width of the light source driving pulse is less than or equal to the width of the camera exposure trigger pulse.

[0015] Optionally, in a second implementation of the second aspect of the present invention, the odd-even row separation processing is implemented by the following algorithm: assigning a sequential row number index to each row in the image row data stream; performing a modulo-2 operation on each row index; classifying rows with a result of 1 as odd rows and storing them in the first storage area; classifying rows with a result of 0 as even rows and storing them in the second storage area.

[0016] Optionally, in a third implementation of the second aspect of the present invention, two-dimensional image reconstruction is performed on the first field-of-view image dataset and the second field-of-view image dataset, including the steps of: for the first field-of-view image dataset, its k-th row of data is used as the k-th row of the reconstructed image, where k is a natural number, and this row of data corresponds to the (2k-1)-th row in the original data stream; for the second field-of-view image dataset, its k-th row of data is used as the k-th row of the reconstructed image, and this row of data corresponds to the 2k-th row in the original data stream.

[0017] Beneficial effects: The asynchronous exposure dual-field imaging system and method provided by this invention achieves significant technical effects through innovative spatial folding optical path and time-segmented exposure collaborative design, systematically solving the core contradiction of cost, efficiency and accuracy in industrial large-size high-precision visual inspection.

[0018] First, this invention achieves nearly double the field of view coverage with a single linear array camera hardware configuration. Its core lies in employing a unique folded optical path composed of symmetrical lenses, a mirror group, and a non-polarizing beam splitter cube, successfully merging and guiding the image optical paths of two independent fields of view to a single camera sensor. This directly eliminates the need for an entire set of cameras, lenses, and acquisition channels. Compared to multi-camera stitching schemes that achieve the same field of view, the core imaging hardware cost is significantly reduced, while avoiding the complexity and additional costs associated with multi-camera calibration. This provides a cost-effective solution for large-scale industrial production line applications.

[0019] Secondly, this invention utilizes a precise timing control module to drive two independent strobe light sources that are strictly synchronized with the camera's line trigger signal. This alternately illuminates different fields of view within odd and even line cycles, enabling a single camera to acquire two image streams via asynchronous exposure and time-division multiplexing. This operating mode eliminates the need for frequent starting and stopping of the motion platform, unlike multi-step scanning schemes, achieving truly high-speed continuous scanning and significantly improving detection efficiency. Furthermore, because image acquisition is controlled entirely by a fixed optical path and electronic timing, the reliance on a high-precision stepping positioning platform is completely eliminated, fundamentally removing image blurring or stitching misalignment caused by mechanical motion errors and vibrations, resulting in stable image quality with high geometric fidelity.

[0020] Furthermore, this invention ensures excellent image consistency and crosstalk-free characteristics, specifically in the following ways: Optically, the symmetrical folded optical path design of this invention guarantees consistency in distortion and optical path length between the two imaging channels; in terms of control, the complementary switching strategy of the light source physically isolates light from the non-acquisition field of view from entering the sensor, achieving complete optical path isolation and avoiding the aliasing of dual-field-of-view image information. In back-end processing, based on a simple and efficient odd-even row separation algorithm, two independent and complete field-of-view images can be accurately and in real time reconstructed from the interleaved data stream, laying a solid foundation for subsequent seamless stitching.

[0021] Furthermore, the system of this invention has the advantages of compact structure, high reliability, and easy integration. Its folded optical path significantly compresses the volume of the imaging module, making its structure flat and easy to integrate into existing production lines with limited space. The core of the system consists of solid-state optical and electronic components, without high-speed reciprocating motion mechanisms, resulting in a low failure rate, simple maintenance, and stable and reliable operation.

[0022] With a simple and ingenious system architecture, this invention successfully achieves multiple technical effects, such as dual field-of-view coverage at the cost of a single camera, ensuring high-precision imaging in continuous scanning, and eliminating physical crosstalk with electronic synchronization control. It provides a new high-performance, low-cost, and highly reliable solution for large-size precision industrial inspection. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of an asynchronous exposure dual-field imaging system provided in an embodiment of the present invention.

[0026] Figure 2 This is a schematic diagram of the optical path structure in an asynchronous exposure dual-field imaging system according to the present invention.

[0027] Figure 3 The flowchart illustrates an asynchronous exposure dual-field imaging method provided by this invention. Detailed Implementation

[0028] This invention provides an asynchronous exposure dual-field imaging system and method. The terms "first," "second," "third," "fourth," etc. (if applicable) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] Please see Figure 1 and Figure 2 This invention provides an asynchronous exposure dual-field-of-view imaging system, as shown in the figure, which includes an imaging module 10, an acquisition module 20, an illumination control module 30, and a data processing module 40. The imaging module 10 is connected to the acquisition module 20, the illumination control module 30 is mounted in front of the imaging module 10, and the acquisition module 20 is communicatively connected to the data processing module 40. The imaging module 10 is used to merge optical information from two spatially separated fields of view on the object under test into a single coaxial outgoing beam. It includes a first lens 11 and a second lens 12 arranged side by side along a first direction, a first reflector 13 and a second reflector 14 corresponding one-to-one with the first lens 11 and the second lens 12, and a non-polarizing beam splitter cube 15. The illumination control module 30 includes a first stroboscopic flash source 31, a second stroboscopic flash source 32, and a controller. The first stroboscopic flash source 31 and the second stroboscopic flash source 32 respectively illuminate the fields of view of the first lens 11 and the second lens 12. The controller is used to receive a line trigger signal synchronized with the object's displacement, and based on the line trigger signal, generate a control timing sequence that alternately drives the first frequency flash source 31 and the second frequency flash source 32 to light up and turn off, and generate a trigger signal that synchronously controls the exposure of the line scan camera 21, so that in odd-numbered line trigger cycles, only the first frequency flash source 31 is lit and the line scan camera 21 acquires the first field of view image line, and in even-numbered line trigger cycles, only the second frequency flash source 32 is lit and the line scan camera 21 acquires the second field of view image line; the acquisition module 20 includes a single line scan camera 21, used to receive the emitted beam after being merged by the imaging module 10, and convert it into a continuous image line data stream; the data processing module 40 is used to receive the image line data stream output by the line scan camera 21, and divide the image line data stream into a first field of view image dataset and a second field of view image dataset through an odd-even line separation algorithm, and perform image reconstruction respectively to obtain complete two-dimensional images of two independent fields of view.

[0030] Specifically, in this embodiment, the imaging module 10 is the optical core of the entire system. It is responsible for merging the optical information of two spatially separated fields of view into one output port. The specific structural composition and connection relationships are as follows: Figure 2 As shown, the first lens 11 and the second lens 12 are arranged symmetrically side by side along a horizontal direction (i.e., the first direction) perpendicular to the direction of the object's movement. Their optical axes are parallel and perpendicular to the horizontal plane (i.e., the plane where the object being measured is located). The lenses need to be strictly selected or customized according to core parameters such as detection resolution, working distance, and pixel size to ensure that the optical performance of the two is highly consistent, such as focal length, distortion, and relative illumination. The image plane size error needs to be controlled within 1% to ensure the geometric accuracy of subsequent image stitching.

[0031] The first reflector 13 and the second reflector 14 are precisely fixed above the first lens 11 and the second lens 12, respectively. The reflective surface of the reflector is precisely at a 45° angle to the horizontal plane. Its function is to receive the vertically upward light beam from the lower lens and reflect it to bend it by 90°, changing it into a horizontally propagating light beam. The reflector is preferably a right-angle prism with high reflectivity (e.g., ≥99.5%) or a plane mirror with a high-efficiency anti-reflection coating on its surface to minimize light energy loss.

[0032] The unpolarized beam splitter cube 15 is located directly above the center point of the line connecting the first and second reflecting mirrors. Inside the beam splitter cube is a semi-transparent, semi-reflective beam splitting film (splitter surface), precisely set at a 45° angle to the optical axes of the horizontally incident beams from both the left and right sides. Its core function is as a beam combiner: for a beam incident horizontally from the left (corresponding to the first lens), the beam splitter surface reflects a portion (e.g., 50%) of its energy, causing it to bend upwards by 90°; simultaneously, for a beam incident horizontally from the right (corresponding to the second lens), the beam splitter surface transmits a portion (e.g., 50%) of its energy, also causing it to propagate upwards. Through careful design and adjustment of the optical path length, these two controlled beams (one from reflection, one from transmission) will be completely coaxial behind the beam splitter cube, merging into a single vertically upward-facing output beam. This embodiment chooses an unpolarized beam splitter cube to eliminate or greatly reduce the influence of the polarization state of light on the beam splitting ratio, ensuring the light intensity stability of the two channels and avoiding brightness differences introduced by different surface characteristics of the measured object.

[0033] This embodiment designs a double-folded optical path in the imaging module, characterized by vertical incidence, horizontal deflection, and vertical exit. Its advantages are threefold: First, it significantly reduces the system's lateral dimensions, allowing two lenses to be placed side-by-side below the object without requiring direct vertical space for the camera. This enables the imaging head to be made very flat (height not exceeding half the working distance), adapting to space-constrained industrial environments. Second, it achieves strict optical path symmetry; the optical path lengths and reflection counts of the two channels are designed to be essentially equal, ensuring consistency in optical distortion and aberration between the two field-of-view images. Third, through the spatial filtering effect of the beam-splitting cube, it ensures that only light from designated directions can enter the camera, effectively suppressing stray light crosstalk.

[0034] In this embodiment, the core of the acquisition module 20 is a high-frequency, high-resolution line scan camera 21, such as... Figure 2 As shown, it is precisely mounted directly above the unpolarized beam splitter cube 15, ensuring that its receiving optical axis is strictly coaxial with the merged outgoing beam. The linear scan camera is primarily responsible for converting the received optical signals into electrical signals line by line. Specifically, it receives the merged outgoing beam from the imaging module and converts it into a continuous stream of image line data. Its line frequency must be high enough to match the motion speed and accommodate alternating acquisition of two fields of view. For example, to achieve the same effective detection line frequency as a single field of view, the camera's highest line frequency must be at least twice the line frequency required for a single field of view.

[0035] In this embodiment, the lighting control module 30 is the time control center for realizing asynchronous exposure and dual field of view functions. It includes a first strobe light source 31 and a second strobe light source 32, which are two independent high-brightness line light sources, precisely aligned with the field of view illumination areas of the first lens 11 and the second lens 12, respectively. LED line light sources are preferred because they can strobe at high speed, have a long lifespan, and stable brightness. A strip-shaped light shield can be set at the front end of each light source, with a width slightly larger than the lens diameter, to strictly limit the illumination range and prevent light from overflowing into the field of view of adjacent lenses and causing optical crosstalk. It also includes a controller, which is the metronome of the system. It receives line trigger signals from the rotary encoder of the motion platform (each signal represents the distance the object has moved by one camera line resolution). The controller operates internally with a fixed cycle synchronized with the line trigger. Its core control logic is: when the Nth (odd number) line trigger signal is detected, the controller immediately generates a short pulse to drive the first strobe light source to light up, while ensuring that the second strobe light source is in the off state; at the same time, it sends an exposure trigger signal to the line scan camera, the width of which covers the flash time of the light source. At this moment, only the light rays from the line in the first field of view illuminated by the first light source can enter the camera and be captured through the reflection path of the first lens → first reflector → unpolarized beam splitter cube. When the trigger signal of the (N+1)th (even-numbered) line is detected, the controller drives the second flash source to light up and turns off the first light source, synchronously triggering the camera exposure, thereby capturing the image line of the second field of view; this process is repeated.

[0036] The controller in this embodiment serves three main purposes: First, it achieves physical-level optical path isolation. By alternating the on and off of the light source, only one field of view is activated and allowed to enter the optical path at any given exposure time, fundamentally eliminating aliasing (crosstalk) between the two field of view images at the sensor level. Second, it ensures accurate time synchronization. The strict synchronization between the light source flashing and the camera exposure ensures uniform brightness and controllable motion blur in the captured image. Third, it achieves efficient resource utilization. The time resources of a single camera are alternately allocated to the two fields of view, effectively doubling the data throughput without increasing the number of cameras.

[0037] In this embodiment, since the line scan camera outputs an interwoven data stream that alternately contains image rows of the first and second fields of view, the data processing module 40 first marks each row of data with a sequential index, and then extracts and arranges the odd-numbered rows of data sequentially using a simple odd-even row separation algorithm (such as row number modulo 2 operation) to reconstruct the complete two-dimensional image of the first field of view; similarly, it extracts and arranges the even-numbered rows of data sequentially to reconstruct the complete two-dimensional image of the second field of view. Afterwards, further processing such as image stitching (using overlapping areas) and defect analysis can be performed.

[0038] This embodiment achieves nearly double the field of view coverage with a single linear scan camera. Its core lies in a unique folded optical path composed of symmetrical lenses, a mirror group, and a non-polarizing beam splitter cube, successfully merging and guiding the image optical paths of two independent fields of view to a single camera sensor. This directly eliminates the need for a complete set of cameras, lenses, and acquisition channels. Compared to multi-camera stitching schemes that achieve the same field of view, the core imaging hardware cost is significantly reduced, while avoiding the complexity and additional costs associated with multi-camera calibration, providing a cost-effective solution for large-scale industrial production line applications. Secondly, this embodiment uses a precise timing control module to drive two independent stroboscopic light sources to be strictly synchronized with the camera's line trigger signal, alternately illuminating different fields of view within odd and even line cycles. This allows a single camera to acquire two image information streams in an asynchronous exposure and time-division multiplexing manner. This operating mode eliminates the need for frequent start-stop of the motion platform as in multi-step scanning schemes, achieving truly high-speed continuous scanning and significantly improving detection efficiency. Meanwhile, since the entire image acquisition process is controlled by a fixed optical path and electronic timing, the reliance on a high-precision stepping positioning platform is completely eliminated, fundamentally eliminating image blurring or stitching misalignment caused by mechanical motion errors and vibrations, resulting in stable image quality with high geometric fidelity. Furthermore, this embodiment ensures excellent image consistency and crosstalk-free characteristics, specifically: optically, the symmetrical folded optical path design of this invention ensures consistency in distortion and optical path between the two imaging channels; in terms of control, the complementary switching strategy of the light source physically isolates light from the non-acquisition field of view from entering the sensor, achieving complete optical path isolation and avoiding the aliasing of dual-field-of-view image information. In back-end processing, based on a simple and efficient odd-even row separation algorithm, two independent and complete field-of-view images can be accurately and in real-time reconstructed from the interleaved data stream, laying a solid foundation for subsequent seamless stitching. Furthermore, the system in this embodiment boasts advantages such as compact structure, high reliability, and ease of integration. Its folded optical path significantly compresses the size of the imaging module, resulting in a flat structure that is easily integrated into existing production lines with limited space. The core of the system consists of solid-state optical and electronic components, eliminating high-speed reciprocating motion mechanisms, resulting in a low failure rate, simple maintenance, and stable and reliable operation. This embodiment, with its simple and ingenious system architecture, successfully achieves multiple technical effects, including dual-field-of-view coverage at the cost of a single camera, ensuring high-precision imaging during continuous scanning, and eliminating physical crosstalk through electronic synchronization control. It provides a new high-performance, low-cost, and highly reliable solution for large-size precision industrial inspection.

[0039] In some embodiments, the imaging parameters of the first and second lenses are identical, and their fields of view overlap in a region perpendicular to the direction of object movement. The width of this overlapping region accounts for 10%-15% of the field of view of a single lens. In this embodiment, identical imaging parameters mean that the two lenses must be highly matched in key optical performance aspects such as focal length, aperture, optical distortion, relative illumination (edge ​​light reduction), and resolution (MTF). In practice, industrial lenses of the same model and batch are typically selected, and rigorous screening is conducted to ensure that their image plane size error is no greater than 1%, resulting in nearly identical imaging quality. This consistency is the optical basis for subsequent seamless image processing and precise stitching. Setting an overlapping region in the field of view is the key to the spatial design of this invention for achieving seamless large field of view detection. The two lenses are not simply arranged back-to-back; their respective rectangular fields of view intentionally share a common area perpendicular to the direction of object movement. The width of this overlapping region is specifically quantified as 10%-15% of the field of view of a single lens. For example, if the field of view of a single lens is 100mm, the width of the overlapping region should be 10mm to 15mm.

[0040] Setting a 10%-15% overlap area in this embodiment is a necessary condition for achieving high-precision and robust image stitching. Its technical effects are reflected in multiple aspects: The overlap area is a part shared by two independent images, providing a natural anchor point or feature source for automatic image registration algorithms. The algorithm can extract features such as edges and corners within this area for matching, accurately calculate the relative positional relationship between the two images (such as translation and slight rotation), and thus guide the stitching; Through the overlap area, image fusion techniques (such as fade-in / fade-out and multi-band fusion) can be used to smooth the stitching boundary, making the final synthesized panoramic image completely seamless; Although the optical path is symmetrical in the design, there will always be micron-level tolerances in machining and assembly. The 10%-15% overlap area provides a safe tolerance space, ensuring that even if there are slight optical axis non-parallelism or positional deviations, the two images still have enough common area for registration, greatly enhancing the engineering practicality and robustness of the system.

[0041] In this embodiment, the ratio of 10%-15% is the preferred range after careful consideration. If the ratio is too small (e.g., <5%), the effective feature information provided by the overlapping area may be insufficient, which may lead to registration failure in low-texture areas and poor fault tolerance. If the ratio is too large (e.g., >20%), it means that the utilization rate of the system's effective field of view is reduced, wasting the field of view that should be expanded, which contradicts the core purpose of this invention to expand the field of view. This range achieves the best balance between ensuring reliable stitching and maximizing the benefits of the field of view.

[0042] In some embodiments, the unpolarized beam splitter cube has a 50:50 splitting ratio for beams in the wavelength range of 400nm to 700nm. This means that for any beam incident at a 45-degree angle within this wavelength range, regardless of its polarization state, the beam splitter cube will distribute its energy approximately evenly into reflected and transmitted light (each accounting for approximately 50%). This limitation directly determines the signal balance and stability of the two imaging channels of the system, which is crucial to ensuring consistent final image quality. It has the following functions: ensuring balanced light energy in both channels: In the optical path design of this invention, the beam from the first lens mainly enters the camera through the reflected portion of the beam splitter cube, while the beam from the second lens mainly enters through the transmitted portion. If the splitting ratio is strictly 50:50, then theoretically, the light energy reaching the camera sensor from both channels is equal. This ensures that, under the same lighting conditions, the average grayscale level of the image signals output by the camera to the two fields of view is basically consistent, avoiding systematic deviations caused by hardware limitations that result in one field of view being brighter and the other darker. This greatly simplifies subsequent image processing, eliminating the need for complex inter-channel brightness correction that may introduce noise. Maintaining a stable splitting ratio, unaffected by polarization: In industrial production scenarios, the surfaces of the objects being measured are diverse (metal, glass, plastic, coatings, etc.), and the polarization state of their reflected or transmitted light can be complex and unpredictable. Ordinary beam-splitting elements (such as polarizing beam splitters or certain coated beam splitters) have a splitting ratio that strongly depends on the polarization state of the incident light. Using a non-polarized beam-splitting cube means that its 50:50 splitting ratio characteristic is insensitive to the polarization state of light throughout the specified wavelength band. Regardless of the polarization of the light generated by the object surface, the ratio of light intensity received by the two channels remains stable. This fundamentally eliminates the risk of image contrast fluctuations introduced by changes in the measured object, making the system applicable to a wider range of industrial material inspections and improving its versatility and reliability. Band limitation ensures actual performance: The applicable range of 400nm-700nm is clearly defined because this is the most sensitive visible spectrum range for mainstream industrial camera sensors (CMOS / CCD) and LED lighting. This limitation ensures that the beam splitter cube performs optimally within the core band of the system's actual operation, avoiding excessive costs or sacrifice of core band uniformity in pursuit of unnecessary (such as ultraviolet or far-infrared) wide-band performance.

[0043] In some embodiments, the imaging module further includes an opaque housing that encloses the optical path of the first lens, second lens, first reflector, second reflector, and unpolarized beam splitter cube, forming independent optical channels. This embodiment adds a crucial opaque housing to the imaging module. This housing is not a simple protective cover, but a precisely designed, completely sealed cavity that encloses the optical components and their optical paths, forming two independent optical channels (from the lens to the beam splitter cube) that are isolated from each other and from the external environment. The introduction of this housing solves a critical problem in complex optical path systems used in industrial applications: stray light interference. Specifically, industrial lighting conditions are complex and variable, including natural light from windows and illumination from other equipment. This housing acts as a physical barrier, completely blocking the direct illumination or reflection of these uncontrolled ambient lights onto the internal optical components (especially the reflectors and beam splitter cube surfaces) and camera sensors. This ensures that the image signal acquired by the camera originates entirely from a controlled, synchronously flickering dedicated light source, significantly improving the signal-to-noise ratio and contrast of the image, and allowing even weak defect signals to be clearly displayed. Furthermore, in the optical path of the invention, the beam splitter cube, in addition to combining the desired light beams, also generates unwanted transmitted or reflected light beams (for example, a portion of the light from lens A transmitted through the beam splitter cube will be directed towards lens B). Without proper handling, this "waste light" may reflect multiple times on the inner wall of the housing, eventually mixing into another channel and causing crosstalk between the two fields of view image information. The inner wall of the opaque housing is typically treated with a special matte black finish (such as coating with light-absorbing paint or attaching light-absorbing velvet) to quickly absorb this stray light. The formation of independent optical channels means that dedicated optical tunnels from the lens entrance to the beam splitter cube are established for both A and B light paths, ensuring, from a physical structure perspective, that the core advantages of asynchronous exposure and crosstalk-free acquisition of this invention are realized.

[0044] In some embodiments, the controller includes a high-frequency logic controller and a motion control card; the motion control card is connected to a servo driver that drives the motion platform and is used to control the motion platform to drive the object under test to move at a constant speed; the line trigger signal is provided by a rotary encoder installed on the motion platform; the high-frequency logic controller is used to receive the line trigger signal and perform frequency division, frequency multiplication and logic processing on the line trigger signal to generate two complementary, pulse width adjustable light source drive pulse signals and one line scan camera exposure trigger signal synchronized with the light source drive pulse.

[0045] Specifically, the motion control card is responsible for generating macroscopic motion. It connects to the servo driver of the motion platform, forming a closed-loop motion control system. Its core task is to precisely control the motion platform, ensuring the object under test passes through the detection area at a constant speed V. Line scan imaging essentially uses time integration to obtain a spatial image. The stability of the object's velocity V directly determines the accuracy of the image's scale in the direction of motion and whether distortion will occur. The motion control card actively maintains the constancy of V through closed-loop control, which is fundamentally different from the passive approach in the background technology that relies on the platform's own precision. It ensures that the line trigger signal generated by the rotary encoder is uniform in time, thus establishing a stable and reliable spatiotemporal coordinate system for the entire imaging system. This is the fundamental prerequisite for the correct arrangement of all subsequent image lines and the precise alignment of two field-of-view images.

[0046] The high-frequency logic controller is responsible for generating micro-time sequences. It receives line trigger signals from the rotary encoder mounted on the motion platform, which are strictly synchronized with the displacement (each pulse corresponds to the object moving a fixed distance, such as the ground resolution of one pixel). Its core task is to perform high-speed and precise electronic processing (frequency division, frequency multiplication, and logical judgment) on this signal, and finally generate a complex pulse sequence that drives the two light sources to flash alternately and the line scan camera to expose synchronously.

[0047] This embodiment utilizes a high-frequency logic controller to achieve nanosecond-level precision synchronization and isolation, which is the core technology enabling asynchronous exposure. The high-frequency logic controller (typically an FPGA) plays the following role:

[0048] Strict synchronization: It ensures that the pulses driving the light source a / b and the pulses triggering the camera are synchronized with nanosecond-level precision in both rise edge and pulse width. This allows the camera sensor's integration window to perfectly match the light source's flash window, guaranteeing the capture of the brightest and clearest instantaneous image while avoiding integration noise when the light source is off.

[0049] Precise alternation (complementary): The pulses it generates to drive the two light sources are complementary; that is, when one is high, the other must be low, and the switching is smooth and without overlap. This physically ensures from the source of the electrical signal that the two light sources will not light up simultaneously, making it the first and most reliable electronic barrier for achieving optical channel isolation and eliminating crosstalk.

[0050] Flexible and adjustable: Adjustable pulse width means that the flash and exposure time can be flexibly adjusted according to the reflectivity and speed of the object being measured to optimize image quality, reflecting the adaptability and intelligence of the system.

[0051] This embodiment connects motion control (uniform speed), position feedback (encoder), timing control (FPGA), and image acquisition (camera) into a complete closed loop. The rotary encoder, as the key sensor in this closed loop, converts actual physical displacement into electronic clock signals. This design tightly couples mechanical motion with electronic timing, ensuring that the image acquisition cycle is entirely determined by the actual displacement of the object. This completely eliminates image misalignment or overlap caused by slippage in the transmission system, error accumulation, etc., achieving true synchronous position scanning and obtaining extremely high geometric fidelity.

[0052] In some embodiments, an asynchronous exposure dual-field imaging method based on the imaging system described in this invention is also provided, such as... Figure 3 As shown, it includes the following steps:

[0053] S10. Initialize the imaging system and configure its parameters.

[0054] Specifically, this step involves zeroing and parameter loading of the entire imaging system before any inspection task begins. It is not a single action but a series of configuration activities, including: Optical configuration: Calculating and setting the system's working distance (WD) based on the size of the object under test (determining the required field of view) and the minimum size of the defect to be inspected (determining the required resolution); adjusting the focal length and aperture of the first lens (A) and the second lens (B) to ensure consistent image sharpness on the focal plane and that the depth of field meets requirements; verifying the coplanarity of the two lenses, ensuring their optical axes are strictly parallel and perpendicular to the plane of the motion platform. Illumination configuration: Setting the drive current of the first stroboscopic flash source (a) and the second stroboscopic flash source (b) to ensure their output light intensity meets the signal-to-noise ratio required for camera imaging, and ensuring that the brightness output of the two light sources is highly consistent to avoid inherent brightness differences in dual-channel images; it may also be necessary to adjust the installation angle and position of the light sources to ensure uniform illumination covering their respective field of view lines. Controller timing configuration: This is the core of initialization, where key timing parameters are preset in the controller (especially its internal high-frequency logic controller).

[0055] Light source driving pulse width (T) flash T is determined based on the object's velocity (V) and the allowable amount of motion ambiguity; the faster the velocity, the greater the ambiguity. flash The shorter the length, the better to freeze the image;

[0056] Camera exposure trigger pulse width (T) exp ): Set to slightly greater than or equal to T flash ;

[0057] Synchronization relationship: Establish a definite mapping relationship in the controller logic between row trigger signal → parity judgment → corresponding light source drive pulse and camera exposure trigger pulse generation;

[0058] Motion configuration: Set the target motion speed (V) and acceleration parameters in the motion control card;

[0059] This step precisely calibrates and configures the imaging system from a general-purpose device state to a dedicated state adapted to a specific detection task. It establishes the baseline parameters for all subsequent steps and is a fundamental prerequisite for ensuring the repeatability, comparability, and accuracy of detection results.

[0060] S20. Place the object to be measured on the motion platform and start the motion platform and imaging system;

[0061] Specifically, this step is a startup process. First, the motion control card sends a command to the servo driver, driving the motion platform (conveyor belt) to accelerate the object under test from rest until it reaches and stabilizes at a preset constant speed (V). Then, the imaging system is started: the line scan camera (F), high-frequency logic controller, and stroboscopic light source driver are powered on; the camera begins to wait for an external trigger signal, the light source is in a ready-to-trigger state, and the controller program begins to run and listen for the line trigger signal.

[0062] S30. Receive a line trigger signal generated by a rotary encoder mounted on a motion platform, which is synchronized with the displacement of the object being measured.

[0063] Specifically, when an object moves at a constant speed, a high-precision rotary encoder (usually directly mounted on the drive shaft or driven wheel) monitors the platform displacement in real time. Each time the encoder's internal grating detects a fixed, minute movement of the platform (this distance corresponds to the size of one pixel in the object plane, i.e., the ground sampling distance GSD), it generates a standard digital pulse (such as a TTL rising edge). This pulse sequence is the line trigger signal. This step achieves a precise conversion of physical displacement into an electronic clock. The arrival of each line trigger pulse not only signifies the start of acquiring a new line of images but also precisely defines the absolute spatial position of the object corresponding to that line of images. It serves as a bridge connecting the continuous physical world and discrete digital images, and is the foundation for the spatiotemporal synchronization of the entire line scanning method.

[0064] S40. Based on the parity of the row trigger signal, perform asynchronous exposure control: in odd-numbered row trigger cycles, drive the first frequency flash source to light up and the second frequency flash source to turn off, and simultaneously trigger the line scan camera to expose, acquiring the first field of view image row; in even-numbered cycles, drive the second frequency flash source to light up and the first frequency flash source to turn off, and simultaneously trigger the line scan camera to expose, acquiring the second field of view image row.

[0065] Specifically, this step is the core innovation of the imaging method. The controller (specifically executed by a high-frequency logic controller) responds instantly to each received line trigger signal:

[0066] Signal determination: The internal counter is incremented by 1, and the parity of the current count value is immediately determined.

[0067] Control Decision-Making and Execution (Taking Odd-Numbered Periods as an Example):

[0068] Light source control: Generate a pulse with a width of T within a nanosecond delay. flash A square wave pulse is sent to the driver of the first frequency flash source (a) to illuminate it instantly. At the same time, the signal sent to the driver of the second frequency flash source (b) is ensured to be low (forced off).

[0069] Camera control: Almost synchronously, a pulse width of T is generated. exp The square wave pulse is sent to the exposure trigger port of the line scan camera (F). After receiving the rising edge, the camera immediately begins the exposure integration of a new line.

[0070] Optical effects: During this odd-numbered period, only the light from the first field of view illuminated by the light source (a) travels through the optical path of lens A → mirror C → beam splitter E (reflection path) to reach the camera sensor and be recorded. The optical path of the second field of view is dark because the light source (b) is off, and therefore does not contribute to the sensor.

[0071] Alternating Cycle: When the trigger signal for the next (even-numbered) row arrives, the controller performs a symmetrical operation: turns on the light source (b), turns off the light source (a), synchronously triggers the camera, and acquires the second field-of-view image row.

[0072] This step enables electronically responsive switching of the optical field of view. It utilizes time-division to dynamically allocate the sensor resources of a single camera to two physically separate fields of view within odd and even row cycles. Its innovation lies in achieving dual-field coverage through electronically synchronized light source alternation, rather than mechanically moving the camera or objects, thus fundamentally achieving the goals of high efficiency (continuous scanning) and zero mechanical error.

[0073] S50, Output an image line data stream containing the first field-of-view image line and the second field-of-view image line through the line scan camera;

[0074] Specifically, after each exposure, the line scan camera converts the accumulated charge on each pixel of the sensor into voltage, which is then converted into a digital value by an analog-to-digital converter. Once a line of data is converted, it is immediately streamed out via a high-speed interface. Due to the alternating control of step S40, the data stream output by the camera naturally forms a structure in time series where odd and even rows alternate, carrying different field-of-view information: [Row 1 data (from field of view A)], [Row 2 data (from field of view B)], [Row 3 data (from field of view A)], [Row 4 data (from field of view B)]... The purpose of this step is to generate raw interleaved data containing dual-channel information. This is the initial form after all information has been digitized and serves as the input source for subsequent software processing.

[0075] S60. Perform odd-even row separation processing on the image row data stream, extract the odd-numbered row data into a first field-of-view image dataset, and extract the even-numbered row data into a second field-of-view image dataset.

[0076] In this embodiment, the odd-even row separation processing of the image row data stream is implemented by the following algorithm: assign a sequential row number index to each row in the image row data stream; perform a modulo-2 operation on each row index; classify rows with an operation result of 1 as odd rows and store them in the first storage area; classify rows with an operation result of 0 as even rows and store them in the second storage area.

[0077] Specifically, this process is usually completed automatically at the hardware level by the image acquisition card. The acquisition card has a counter inside. Whenever it receives a complete row of pixel data from the camera interface and packages it, it assigns the current counter value to the data packet as the row number, and then increments the counter by 1. This row number corresponds strictly one-to-one with the camera's exposure trigger signal, thus originating from the "odd-even cycle" logic in the above steps.

[0078] In the software processing thread, for each incoming data row, its row number n is read, and a modulo 2 operation is performed on each row index. The operation logic is: row index ÷ 2, take the remainder; if the remainder is 1, it is determined to be an odd row, and if the remainder is 0, it is determined to be an even row.

[0079] If the remainder is 1, the memory copy function is called to copy all pixel data of the row completely to the end of the pre-allocated contiguous memory block (first storage area) used to store the first field of view image; if the remainder is 0, it is copied to the end of another independent contiguous memory block (second storage area); the first storage area and the second storage area are two independent dynamic arrays or buffers in memory, and their growth is only related to the arrival of the corresponding field of view data and do not interfere with each other.

[0080] The algorithm in this embodiment provides a simple solution for mapping time-interleaved order to spatially separated storage. Its core is to utilize row numbers, which are natural, reliable, and already contained in the data as metadata.

[0081] Because the generation of row number indices is strictly synchronized and continuous with the physical acquisition process, this separation logic is mathematically absolutely correct. As long as the data stream is uninterrupted and not out of order, the separation result is 100% accurate. It does not depend on the image content itself (such as features or textures), so it is effective for any object being measured (even solid colors and textureless objects), and is extremely robust. The modulo operation is a simple operation that can be completed in a single clock cycle of a modern CPU. The decision overhead for separating tens of thousands of pixels in a row is negligible. The main overhead is memory copying, which can be further accelerated by technologies such as DMA (Direct Memory Access). The high efficiency of this algorithm ensures that the separation processing will never become the bottleneck of the entire system's real-time processing pipeline.

[0082] S70. Perform two-dimensional image reconstruction on the first field-of-view image dataset and the second field-of-view image dataset respectively to obtain two complete and independent first field-of-view images and second field-of-view images.

[0083] Specifically, when the number of rows in a separated dataset reaches the expected value (corresponding to the length of the measured object in the direction of motion), reconstruction begins. For the first field-of-view dataset, which is a list containing all odd-numbered rows of data, the reconstruction algorithm simply follows the storage order of this list (i.e., the order of acquisition time), taking the first element as the first row of the image, the second element as the second row, ..., and the kth element as the kth row. Mathematically, the kth row of this image corresponds to the (2k-1)th row in the original data stream. For the second field-of-view dataset, following the same rule, its kth element is taken as the kth row of the image, corresponding to the 2kth row in the original data stream. This step restores the one-dimensional, time-ordered data sequence to a two-dimensional digital image with the correct spatial topology. This step restores the geometric meaning of the image, allowing the row indices in the pixel matrix to re-establish a linear proportional relationship with the actual movement distance of the object.

[0084] This invention's imaging method integrates complex multidisciplinary technologies (optics, mechanics, electronics, and software) into a standardized operating procedure. Users or host computer systems only need to call the commands sequentially to automatically complete the entire process from environmental preparation to image output, greatly simplifying operation, lowering the technical threshold, and enabling the rapid deployment and replication of high-end detection technologies. The method constructs a closed loop from motion control → displacement sensing → synchronous acquisition → sequential output → sequential reconstruction. This closed loop ensures that the position of each pixel in the final digital image can be found in a unique and precisely corresponding physical point in the real world, providing a methodological foundation for high-precision measurement and positioning. Through continuous motion in S20 and electronic switching in S40, this invention achieves uninterrupted high-speed scanning, which is inherently efficient. Furthermore, since the switching in S40 is based on electronic timing rather than mechanical motion, its accuracy and speed far exceed those of mechanical methods, thus ensuring high-quality image acquisition (no vibration, no positioning error) while achieving high efficiency.

[0085] In some implementations, the asynchronous exposure control is specifically implemented as follows: the line trigger signal is counted and judged by a high-frequency logic controller; when the count value is odd, a first light source drive pulse and a camera exposure trigger pulse are generated; when the count value is even, a second light source drive pulse and a camera exposure trigger pulse are generated; the width of the light source drive pulse is less than or equal to the width of the camera exposure trigger pulse.

[0086] Specifically, the high-frequency logic controller has a dedicated register as a row counter. Whenever it detects a low-to-high transition (rising edge) on the row trigger signal input pin, the value of the counter is automatically incremented by 1. This process is directly implemented by the hardware logic, with a response delay in the nanosecond range. Parity determination is also completed instantaneously by the hardware logic. The most direct way is to bring out the least significant bit (LSB) of the counter value. In digital circuits, LSB=1 indicates an odd number, and LSB=0 indicates an even number. No arithmetic operations are required. This is a simple wiring logic.

[0087] When LSB=1, a set of preset gate circuits or state machines are activated, which immediately control the two output pins:

[0088] Pin 1 (light source a driver): The pin transitions from low to high, starting a hardware timer. The timer operates at a preset time T. flash After a certain time, pull the pin back to low level;

[0089] Pin 2 (Camera Exposure Trigger): It transitions from low to high almost simultaneously (possibly within a few clock cycles), starting another hardware timer at T... exp It will be lowered over time.

[0090] Pin 3 (light source b driver): Forced to remain low.

[0091] When LSB=0, another set of symmetrical gate circuits is activated, controlling pin 2 (camera trigger) to operate in the same way, but pin 1 is set low, and pin 3 outputs a signal with a width of T. flash High pulse.

[0092] Pulse width relationship maintenance: When designing the controller, it is essential to ensure that T... exp The timer value is set to be greater than or equal to T. flash The set value. T exp ≥T flash This ensures that the camera's image sensor integrates throughout the entire effective light emission period, capturing all the effective light signal energy, which is the physical basis for obtaining high signal-to-noise ratio images.

[0093] This step anchors the abstract concept of asynchronous exposure control in high-speed, deterministic operations implemented using hardware logic circuits. It eliminates the possibility of implementing this function through non-real-time methods such as software interrupts or operating system threads, because the latter cannot guarantee that the judgment and response can be completed within tens of microseconds of line cycle time, and have poor timing accuracy and large jitter.

[0094] In some implementations, two-dimensional image reconstruction is performed on the first field-of-view image dataset and the second field-of-view image dataset, including the steps of: for the first field-of-view image dataset, its k-th row of data is used as the k-th row of the reconstructed image, where k is a natural number, and this row of data corresponds to the (2k-1)-th row in the original data stream; for the second field-of-view image dataset, its k-th row of data is used as the k-th row of the reconstructed image, and this row of data corresponds to the 2k-th row in the original data stream.

[0095] Specifically, assuming that after separation in step S60, the first field-of-view dataset is a list List_A = [a1,a2, a3, ..., aM], where a1 is the first odd-numbered row of data stored (corresponding to the original row number 1), a2 is the second odd-numbered row of data stored (corresponding to the original row number 3), and so on. The second field-of-view dataset is List_B = [b1, b2,b3, ..., bM] (assuming the objects have the same length, and M is the height of the two field-of-view images).

[0096] The reconstruction process is independent for each field of view:

[0097] Reconstructing the first field-of-view image Image_A:

[0098] Create a blank two-dimensional array Image_A with a width of W (camera resolution) and a height of M.

[0099] For k = 1 to M:

[0100] Copy all pixel values ​​of the kth element a_k (i.e., the kth row of the list) in List_A to the kth row of Image_A.

[0101] At this point, Image_A(k, :) = a_k = Raw(2k-1, :), where Raw(r, :) represents the r-th row of the original data stream.

[0102] Reconstructing the second field-of-view image Image_B:

[0103] Create a blank two-dimensional array Image_B of the same size.

[0104] For k = 1 to M:

[0105] Copy the k-th element b_k from List_B to the k-th row of Image_B.

[0106] At this point, Image_B(k, :) = b_k = Raw(2k, :).

[0107] This embodiment defines an equidistant mapping relationship from a one-dimensional acquisition sequence to a two-dimensional spatial image. It is not just a simple data arrangement, but a mathematical expression of the reconstructed image spatial geometric coordinate system.

[0108] This rule guarantees the absolute accuracy of geometric measurements. The rule that the k-th row of Image_A equals the (2k-1)-th row of the original image means that the physical distance between adjacent rows in the final image is constant and known. This distance is not equal to twice the ground sampling distance (GSD) because adjacent rows in the original data stream (rows 1 and 2) correspond to different spatial locations of the object (one in field of view A, and one in field of view B). However, adjacent rows in Image_A (rows k and k+1) correspond to data separated by one row in the original data stream (rows (2k-1) and (2k+1)), which correspond to two adjacent positions on the object in the direction of motion, with a physical distance of exactly one GSD. Therefore, this reconstruction rule restores the correct spatial sampling relationship within each independent field of view image, and any size measurement based on this image is an accurate reflection of the true physical size.

[0109] This process is a pure data rearrangement, without any pixel value interpolation, averaging, or transformation. Every digital quantized value captured by the original sensor is placed unchanged into a specific position in the output image matrix. This maximizes the preservation of the original information's fidelity and avoids additional noise or distortion introduced by the reconstruction algorithm. The reconstructed Image_A and Image_B in this embodiment are independent images conforming to standard image formats (such as two-dimensional matrices) and possessing correct geometric meaning. They can be directly input into any existing or custom image processing library for filtering, enhancement, feature extraction, defect segmentation, OCR, and other operations. Their standard format and correct geometry make subsequent algorithm development and application integration simple and direct.

[0110] The present invention will be further explained and illustrated below through specific embodiments:

[0111] Example 1: Online detection of surface defects in OLED screen modules

[0112] This embodiment simulates a real-world application scenario of high-speed, high-precision surface defect detection on a 300mm wide OLED screen module.

[0113] A dual-field-of-view asynchronous exposure imaging system, its detailed configuration and parameter selection criteria:

[0114] Testing requirements:

[0115] Test object (x): Rigid OLED module, approximately 300 mm wide.

[0116] Inspection speed: The production line cycle requires a conveyor belt speed (V) of 500 mm / s.

[0117] Detection resolution: It is required to be able to identify defects ≥20μm, so the spatial resolution must be ≤10μm.

[0118] Field of view requirement: It needs to cover the entire width of the product.

[0119] System Design and Selection:

[0120] (1) Imaging module:

[0121] Lens A and B selection: The same model of telecentric linear scanning lens is selected, with a focal length f=60mm, magnification ratio 1:1 (1x), and working distance WD=300mm. The line field of view (corresponding to object width) of a single lens on the image side is 81mm. The center distance between the two lenses is precisely adjusted to ensure a 12mm overlap in their object-side fields of view perpendicular to the direction of motion (approximately 15% of the width of a single field of view), used for registration during subsequent image stitching.

[0122] Reflectors C and D: Right-angle prism reflectors with a wide-band anti-reflection coating (400-700nm) on the reflective surface, reflectivity R>99.5%, and installation angle calibrated by an autocollimator to ensure that the angle with the horizontal plane is 45.00°±0.02°.

[0123] The unpolarized beam-splitting cube E has dimensions of 50.8mm x 50.8mm x 50.8mm (2-inch cube), a splitting ratio of 50:50 (at 530nm), and a polarization-dependent loss (PDL) of <0.5%. Its position is precisely adjusted using a six-dimensional adjustment frame to ensure that the two horizontal beams from C and D are incident at the same height, equidistant, and symmetrically.

[0124] Housing: All optical components are housed within a mechanical structure made of black anodized aluminum, with internal channels coated in matte black paint to form sealed, opaque conduits. The overall module height is H=110mm.

[0125] (2) Data Acquisition Module:

[0126] Line scan camera F selection: Select a 16K resolution CMOS line scan camera with 16384 effective pixels, a pixel size of 5μm x 5μm, and a maximum line frequency of 70kHz; output data through the Camera Link Full interface.

[0127] (3) Lighting control module:

[0128] Strobe light sources a and b: High-brightness white LED line light sources with a light emission length of 85mm, forming a uniform bright line through a lens. The light source driving current is adjustable, supports external TTL triggering, and the shortest flash pulse width can reach 5μs.

[0129] Controller: An FPGA-based high-frequency logic controller. A rotary encoder (mounted on the conveyor belt drive shaft) generates a pulse (row trigger signal) every 5μm movement of the conveyor belt (i.e., the object size corresponding to one pixel). The FPGA controller performs phase-locked loop (PLL) frequency multiplication and logic processing on this pulse signal to generate two complementary TTL pulse signals to control light sources a and b respectively. Simultaneously, it generates an exposure trigger signal strictly synchronized with the light source pulses for the camera F. The timing relationship is as follows: when odd-numbered row pulses arrive, the drive pulse of light source a is active (e.g., 10μs wide), while light source b is low; the camera exposure gate signal is high during the light source pulse period. The opposite is true for even-numbered rows.

[0130] (4) Data processing module:

[0131] An industrial control computer (IPC) equipped with a high-performance image acquisition card and GPU, running custom image acquisition and processing software.

[0132] An asynchronous exposure dual-field imaging method based on the imaging system includes the following steps:

[0133] Step S1: System startup and initialization.

[0134] Operation: After power-on, the computer software loads the parameter file and initializes the camera, acquisition card, and controller. Parameters include: camera gain, exposure time (software set to be slightly greater than the light source pulse width, such as 15μs), and line frequency upper limit; the controller sets the odd-even line judgment logic and the light source drive pulse width; the motion platform starts.

[0135] Function and Effect: Ensures all hardware is ready and parameters are set correctly. Setting the exposure time slightly longer than the light source pulse width ensures the camera sensor's effective integration time fully covers the light source's emission period, thereby capturing the brightest effective signal. It also avoids integrating stray light from the environment when the light source is off, improving the signal-to-noise ratio.

[0136] Step S2: Trigger signal generation and reception.

[0137] Operation: The conveyor belt moves at a constant speed of 200 mm / s. The rotary encoder generates a standard TTL pulse (period T = 5 μm / 200 mm / s = 25 μs) every time it detects a displacement of 5 μm. This pulse serves as a line trigger signal and is simultaneously transmitted to the FPGA controller and the image acquisition card via cable.

[0138] Function and Effect: The line trigger signal establishes a strict synchronization relationship between image line acquisition and the physical displacement of the object. This is the foundation of line scan imaging, ensuring the dimensional accuracy of the image in the direction of motion. Its period (25μs) determines the theoretical maximum line acquisition frequency of the system to be 40kHz.

[0139] Step S3: Controller internal timing logic processing (odd-number cycle example).

[0140] Operation: The FPGA has an internal counter that counts the row trigger signals. When the 1st, 3rd, 5th... (odd number) pulse is detected, the following sequence of actions is triggered (completed within a nanosecond delay):

[0141] Light source control: Immediately set the light source a drive pin to high level and output a square wave pulse with a width of 10μs; at the same time, ensure that the light source b drive pin remains at low level.

[0142] Camera trigger: Almost synchronously, the camera exposure trigger pin is set to a high level, and its rising edge triggers the camera to begin a new line of exposure. This high level lasts for 10μs or slightly longer, aligned with the end time of the light source pulse.

[0143] Function and Effect: This is the core step in achieving asynchronous and field-of-view control. Function 1 (Asynchronous): The controller decouples the unified displacement trigger signal into two independent, alternating action sequences. Function 2 (Field-of-View): Through complementary switching of the light sources (one on, the other off), physical isolation of the two fields of view is achieved at the optical source. Temporal isolation allocates different time slices to the signals of the two fields of view.

[0144] Step S4: Optical imaging and signal acquisition (corresponding to odd-numbered cycles).

[0145] Operation: Light source a is turned on instantly, illuminating an extremely narrow line area on the object that corresponds to the field of view of lens A, with a width of 81mm.

[0146] The reflected light (carrying surface information) from the illuminated object's area enters lens A vertically upwards, forming an image beam.

[0147] The beam continues vertically upward, then is deflected by mirror C by 90°, becoming a beam that propagates horizontally to the left.

[0148] A horizontal beam enters the left side of the beam-splitting cube E. Due to the characteristics of the beam-splitting cube, the incident light is reflected and transmitted through the beam-splitting membrane. Key point: Through precise optical path calibration, this embodiment ensures that the light entering the camera is primarily the portion reflected by the beam-splitting membrane (approximately 50% energy). The transmitted portion (the remaining 50%) passes horizontally through the cube and points to the right. However, since there is no light source b illuminating the right side, this portion of light is unavailable and is absorbed by the outer casing, causing no interference.

[0149] The reflected portion of the beam of light is then deflected 90° again, becoming a vertically upward beam, with its optical axis precisely aligned with the photosensitive array of camera F.

[0150] During the effective period of the exposure trigger signal (10μs), the sensor of camera F performs photoelectric conversion, converting the optical information of the first field of view carried by this beam of light into an analog electrical signal of 16384 pixels, and then converting it into a digital signal through internal A / D conversion.

[0151] Function and Effect: This step completes the physical conversion from a line in object space to a line of electrical signal data. The purpose of the folded optical path design is to allow two side-by-side lenses to share the space directly below the camera, and to fold the optical path to a vertical exit through a horizontal bend. Light from the left and right sides, after reflection and transmission through the beam splitter cube, exits from the same surface (the top surface) and is strictly coaxial. This is superior to using a tilted beam combiner to direct two beams of light to different areas of the camera sensor because it avoids problems such as vignetting and aberration differences that may exist at the edges of the camera sensor, ensuring that both fields of view enjoy completely consistent optical performance on the camera image plane.

[0152] Step S5: Data output.

[0153] Operation: After camera F completes the A / D conversion of the data for that line, it packages and transmits the digital values ​​(e.g., 12-bit depth) of one line of 16384 pixels to the image acquisition card via the Camera Link interface. The acquisition card then writes this data into a buffer in the computer's memory. The data stream in the buffer is continuous: line 1 (field of view A), line 2 (field of view B), line 3 (field of view A), line 4 (field of view B)...

[0154] Function and effect: It forms a one-dimensional interwoven data stream, which is the inevitable form of a single camera outputting dual field of view information.

[0155] Step S6: Parity row separation of the image data stream (software processing).

[0156] In the software processing thread, for each incoming data row, its row number n is read, and a modulo-2 operation is performed on each row index. The operation logic is: row index ÷ 2, take the remainder; if the remainder is 1, it is determined to be an odd row, and if the remainder is 0, it is determined to be an even row. If the remainder is 1, the memory copy function is called to completely copy all pixel data of the row to the end of the pre-allocated contiguous memory block (first storage area) used to store the first field of view image; if the remainder is 0, it is copied to the end of another independent contiguous memory block (second storage area).

[0157] Algorithm details:

[0158] Python

[0159] # Code Example

[0160] line_counter = 0

[0161] buffer_A = [] # A list storing rows of the image in field of view A.

[0162] buffer_B = [] # A list storing rows of the image in field B.

[0163] while True:

[0164] one_line_data = read_from_camera_buffer() # Read a new line of data

[0165] line_counter += 1

[0166] if line_counter % 2 == 1: # Odd-numbered rows

[0167] buffer_A.append(one_line_data)

[0168] else:# Even-numbered rows

[0169] buffer_B.append(one_line_data)

[0170] # When buffer_A or buffer_B accumulates to a certain number of rows (such as the height of a frame of image), subsequent processing is performed.

[0171] This step precisely segments the temporally interwoven mixed data stream into two independent data sets based on the key attribute of their generated timestamps (parity). This is a crucial step in recovering the original two streams of information from the mixed signal.

[0172] Step S7: Two-dimensional image reconstruction.

[0173] Operation: For buffer_A, arrange each row of data from top to bottom according to the acquisition order (i.e., list order) to form a two-dimensional array. Assuming buffer_A collects N rows, the reconstructed image size is 16384 (width) x N (height); in memory or in the file, the k-th row of the image (k starts from 1) corresponds to the (2k-1)-th row of the original camera output.

[0174] Similarly, for buffer_B, its k-th row corresponds to the 2k-th row of the original camera output.

[0175] Let the original interleaved data stream be I_original(row), where row = 1, 2, 3, ... Then the reconstructed field-of-view image A, I_A, satisfies: I_A(k) = I_original(2k - 1); the reconstructed field-of-view image B, I_B, satisfies: I_B(k) = I_original(2k). This step restores the one-dimensional row data sequence to a two-dimensional image with spatial meaning.

[0176] Step S8: Post-processing (image stitching and detection).

[0177] Procedure: After obtaining I_A and I_B, image registration and stitching are performed using the known 12mm overlap area between them. Feature-based registration algorithms (such as SIFT, ORB) or stitching based on known displacements can be used. The stitched image yields a complete field-of-view image with a width of approximately 81mm + 81mm - 12mm = 150mm. Then, a defect detection algorithm (such as thresholding, blob analysis, deep learning classification, etc.) is run on this stitched image. This step ultimately enables seamless and complete imaging and analysis of large objects (150mm in width in this example).

[0178] Verification of the effect of this embodiment:

[0179] Field of view extension: The single camera system achieved an imaging width of 150mm, compared to 81mm for a single lens, representing an 85% increase in field of view, which is largely in line with the design target (>90%). Minor differences stem from adjustments to the overlap area.

[0180] Detection efficiency: The system's effective line frequency is 40kHz (each field of view actually occupies 20kHz). For a conveyor belt speed of 200mm / s, the image resolution in the direction of motion is 200mm / s / 40kHz = 5μm, meeting the requirements for high-precision detection. A single scan can cover an area of ​​150mm x [required length], which is far more efficient than step scans of the same precision.

[0181] Cost Comparison: Compared to the stitching solution using two cameras and lenses of the same specifications, this system saves one camera, one lens, and some acquisition and processing resources, resulting in a significant cost reduction.

[0182] Image quality: The acquired images are clear, with uniform brightness and contrast in both fields of view. The overlapping area features are well matched, and there are no visible seams or misalignments after stitching, proving the effectiveness of optical path consistency and timing control.

[0183] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An asynchronous exposure dual-field-of-view imaging system, characterized in that, It includes an imaging module, an acquisition module, an illumination control module, and a data processing module; The imaging module is connected to the acquisition module, the lighting control module is installed in front of the imaging module, and the acquisition module is communicatively connected to the data processing module. The imaging module is used to combine optical information from two spatially separated fields of view on the object under test into a single coaxial outgoing beam. It includes a first lens and a second lens arranged side by side along a first direction, a first mirror and a second mirror arranged in a one-to-one correspondence with the first lens and the second lens, and a non-polarized beam splitter cube. The lighting control module includes a first strobe light source, a second strobe light source, and a controller; the first strobe light source and the second strobe light source respectively illuminate the field of view areas of the first lens and the second lens. The controller is used to receive a row trigger signal synchronized with the object's displacement, and based on the row trigger signal, generate a control timing sequence that alternately drives the first strobe source and the second strobe source to turn on and off, and generate a trigger signal that synchronously controls the exposure of the line scan camera, so that in odd-numbered row trigger cycles, only the first strobe source is lit and the line scan camera acquires the first field of view image row, and in even-numbered row trigger cycles, only the second strobe source is lit and the line scan camera acquires the second field of view image row; The acquisition module includes a single linear array camera, used to receive the emitted beam after being combined by the imaging module and convert it into a continuous image line data stream; The data processing module is used to receive the image row data stream output by the line scan camera, and to divide the image row data stream into a first field of view image dataset and a second field of view image dataset through an odd-even row separation algorithm, and to reconstruct the images in each dataset to obtain complete two-dimensional images of two independent fields of view.

2. The asynchronous exposure dual-field imaging system according to claim 1, characterized in that, In the imaging module, the optical axes of the first lens and the second lens are parallel and perpendicular to the plane where the object being measured is located; the reflecting surfaces of the first mirror and the second mirror form a 45° angle with the horizontal plane, which is used to deflect the vertical incident beam from the corresponding lens by 90° into a horizontal beam; the non-polarized beam splitter cube is set at the center of the line connecting the first mirror and the second mirror, and its internal beam splitting surface forms a 45° angle with the optical axes of the horizontal beams from both sides, which is used to reflect and transmit the two horizontal beams according to a preset beam splitting ratio, and to merge the reflected beams into one output beam coaxial with the receiving optical axis of the line scan camera.

3. The asynchronous exposure dual-field imaging system according to claim 2, characterized in that, The imaging parameters of the first lens and the second lens are the same, and their fields of view overlap in a region perpendicular to the direction of object movement. The width of the overlapping region accounts for 10%-15% of the field of view of a single lens.

4. The asynchronous exposure dual-field imaging system according to claim 2, characterized in that, The non-polarized beam splitter has a splitting ratio of 50:50 for beams in the wavelength range of 400nm to 700nm.

5. The asynchronous exposure dual-field imaging system according to claim 1, characterized in that, The imaging module also includes an opaque housing that encloses the first lens, the second lens, the first reflector, the second reflector, and the optical path of the non-polarized beam splitter cube, forming an independent optical channel.

6. The asynchronous exposure dual-field imaging system according to claim 1, characterized in that, The controller includes a high-frequency logic controller and a motion control card; the motion control card is connected to a servo driver that drives the motion platform and is used to control the motion platform to drive the object under test to move at a constant speed; the line trigger signal is provided by a rotary encoder installed on the motion platform; the high-frequency logic controller is used to receive the line trigger signal and perform frequency division, frequency multiplication and logic processing on the line trigger signal to generate two complementary light source drive pulse signals with adjustable pulse width and one line scan camera exposure trigger signal synchronized with the light source drive pulse.

7. An asynchronous exposure dual-field imaging method based on the imaging system according to any one of claims 1-6, characterized in that, Including the following steps: Initialize the imaging system and configure its parameters; Place the object to be measured on the motion platform and start the motion platform and imaging system; Receives a line trigger signal generated by a rotary encoder mounted on a motion platform, which is synchronized with the displacement of the object being measured; Asynchronous exposure control is performed based on the parity of the row trigger signal: in odd-numbered row trigger cycles, the first frequency flash source is driven to light up and the second frequency flash source is driven to turn off, and the line scan camera is triggered to expose simultaneously to acquire the first field of view image row; in even-numbered cycles, the second frequency flash source is driven to light up and the first frequency flash source is driven to turn off, and the line scan camera is triggered to expose simultaneously to acquire the second field of view image row. The line scan camera outputs an image row data stream containing the first field-of-view image row and the second field-of-view image row. The image row data stream is subjected to odd-even row separation processing, and the odd-numbered rows are extracted into a first field-of-view image dataset, and the even-numbered rows are extracted into a second field-of-view image dataset. Two-dimensional image reconstruction was performed on the first field-of-view image dataset and the second field-of-view image dataset respectively to obtain two complete and independent first field-of-view images and second field-of-view images.

8. The asynchronous exposure dual-field imaging method according to claim 7, characterized in that, The asynchronous exposure control is implemented as follows: the line trigger signal is counted and judged by a high-frequency logic controller; when the count value is odd, a first light source drive pulse and a camera exposure trigger pulse are generated; when the count value is even, a second light source drive pulse and a camera exposure trigger pulse are generated. The width of the light source driving pulse is less than or equal to the width of the camera exposure trigger pulse.

9. The asynchronous exposure dual-field imaging method according to claim 7, characterized in that, The odd-even row separation process is implemented by the following algorithm: assign a sequential row number index to each row in the image row data stream; perform a modulo-2 operation on each row index; classify rows with a result of 1 as odd rows and store them in the first storage area; classify rows with a result of 0 as even rows and store them in the second storage area.

10. The asynchronous exposure dual-field imaging method according to claim 7, characterized in that, Two-dimensional image reconstruction is performed on the first field-of-view image dataset and the second field-of-view image dataset respectively, including the following steps: For the first field-of-view image dataset, its k-th row of data is used as the k-th row of the reconstructed image, where k is a natural number, and this row of data corresponds to the (2k-1)-th row in the original data stream; For the second field-of-view image dataset, its k-th row of data is used as the k-th row of the reconstructed image, and this row of data corresponds to the 2k-th row in the original data stream.

Images