Multi-mode optical fiber probe, and method and device for identifying bending state of multi-mode optical fiber probe

Abstract

The invention provides a multimode optical fiber probe and a method and device for identifying the bending state of the multimode optical fiber probe, and relates to the technical field of optical fibers. The multimode optical fiber probe comprises a probe body, the probe body is twisted periodically at a preset angle in the optical axis direction, and the probe body is composed of a flexible multimode optical fiber; the side, close to the needle body, of the detection end can reflect detection light, and the detection end is connected with the needle body through a continuous multimode optical fiber; wherein after reaching the detection end along the needle body, the detection light carrying speckle light field information is reversely transmitted from the detection end along the needle body under the reflection effect of the detection end. According to the invention, real-time sensing of the bending form of the flexible optical fiber can be realized with high efficiency and high precision.

Description

Technical Field

[0001] This invention relates to the field of optical fiber technology, and in particular to a multimode optical fiber probe, a method and apparatus for identifying the bending state of a multimode optical fiber probe. Background Technology

[0002] In the fields of flexible robotic manipulation, minimally invasive medical devices, and industrial precision inspection, the need for real-time sensing of fiber optic bending shapes is becoming increasingly urgent. Traditional fiber optic bending state sensing technologies mainly rely on solutions such as Bragg gratings (FBGs), interferometers, or distributed sensing. While these can achieve a certain level of bending detection accuracy, they suffer from drawbacks such as limited bending degree, high system complexity, large computational delay, high cost, and reduced resolution due to error accumulation. For example, FBG-based sensors require inscribing periodic structures on the fiber, which is complex, easily damaged, and susceptible to environmental interference. Distributed sensing systems rely on long-distance fiber scanning, making it difficult to meet real-time requirements and suffer from error accumulation over length, thus failing to meet the high-precision bending state sensing needs in complex environments such as endoscopic procedures.

[0003] In recent years, sensing technologies based on speckle light fields in multimode optical fibers have attracted attention due to their simple structure and low cost. These technologies invert deformation information by analyzing mode coupling and speckle changes induced by fiber bending. However, existing methods are mainly limited to measuring the bending of the mid-segment of multimode fibers, and are mostly applicable to static or small-scale bending measurements, making it difficult to meet the high-precision, real-time requirements in dynamic environments. More importantly, traditional speckle calculation algorithms (such as the transfer matrix method) require pre-calibration of a large amount of data, cannot dynamically update model parameters, lack robustness under complex conditions, and have low computational efficiency, making it difficult to meet the requirements of high frame rate, real-time endoscopic, high-precision bending state sensing.

[0004] Therefore, how to achieve real-time sensing of the bending shape of flexible optical fibers efficiently and with high precision is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] This invention provides a multimode fiber probe, a method and apparatus for identifying the bending state of a multimode fiber probe, to solve the above-mentioned defects in the prior art, and to achieve real-time sensing of the bending shape of flexible optical fibers with high efficiency and high precision.

[0006] This invention provides a multimode fiber probe, comprising: The needle body is periodically twisted at a preset angle along the optical axis and is made of flexible multimode optical fiber. The detector end is located on the side of the needle body that can reflect the detector light and is connected to the needle body through continuous multimode optical fiber. After the detector light reaches the detector end along the needle body, it carries speckle light field information from the detector end and is transmitted in the reverse direction along the needle body under the reflection of the detector end.

[0007] The present invention also provides a multimode fiber optic sensor, comprising: A multimode fiber probe; a multimode fiber, the multimode fiber being seamlessly connected to the multimode fiber probe; wherein the multimode fiber probe includes: a needle body, the needle body being periodically twisted at a preset angle along the optical axis, the needle body being composed of flexible multimode fiber; a detection end, the side of the detection end near the needle body capable of reflecting detection light, the detection end being connected to the needle body via a continuous multimode fiber; wherein, after the detection light reaches the detection end along the multimode fiber and the needle body, under the reflection of the detection end, it carries speckle light field information from the detection end and propagates in the reverse direction along the needle body and the multimode fiber.

[0008] The present invention also provides a method for identifying the bending state of a multimode fiber optic probe, the method being applied to the multimode fiber optic sensor described above, the method comprising: In response to the probe tip of the multimode fiber probe reaching the probe point, a probe light is injected into the multimode fiber using a fiber coupler; an area array image sensor receives the outgoing light carrying speckle light field information from the multimode fiber to obtain speckle image data; wherein the outgoing light is generated by the probe light being transmitted to the probe tip via the multimode fiber and the probe body, and then reflected by the probe tip; based on the speckle image data, the bending state of the probe tip of the multimode fiber probe at the probe point is determined.

[0009] According to the multimode fiber probe bending state identification method provided by the present invention, determining the bending state of the probe end of the multimode fiber probe at the probe point based on the speckle image data includes: Based on the speckle image data, the bending state is obtained using a probe bending state recognition model; wherein, the probe bending state recognition model is obtained by training an initial machine learning model using a training dataset.

[0010] According to the multimode fiber probe bending state identification method provided by the present invention, the probe bending state identification model includes a first probe bending state identification model and a second probe bending state identification model, both of which are trained machine learning models. The step of obtaining the bending state based on the speckle image data using the probe bending state identification model includes: obtaining a first bending state based on the speckle image data using the first probe bending state identification model; fine-tuning the parameters of the second probe bending state identification model based on the speckle image data and the first bending state; obtaining a second bending state based on the speckle image data using the fine-tuned second probe bending state identification model; and determining the final bending state based on the first bending state and the second bending state.

[0011] According to the method for identifying the bending state of a multimode fiber probe provided by the present invention, the training dataset includes sample data and labels corresponding to the sample data; the sample data and its corresponding labels are obtained by: using a bending control device to place the multimode fiber probe into a preset bending state; using an optical fiber coupler to inject probe light into the multimode fiber; using an area array image sensor to receive the outgoing light carrying speckle light field information output from the multimode fiber to obtain speckle image data samples; wherein the outgoing light is generated by the probe light being transmitted to the probe end via the multimode fiber and the probe body, and then reflected by the probe end; the sample data is obtained based on the speckle image data; and the labels of the sample data are obtained based on the preset bending state.

[0012] According to the multimode fiber probe bending state recognition method provided by the present invention, after the multimode fiber probe is placed in a preset bending state, the method further includes: using a vision camera to capture images of the multimode fiber probe to obtain a morphological mask of the multimode fiber probe in the preset bending state; obtaining a label for the sample data based on the preset bending state includes: using the morphological mask of the multimode fiber probe in the preset bending state as a label for the sample data; obtaining the bending state based on the speckle image data using a probe bending state recognition model includes: obtaining a morphological mask of the multimode fiber probe based on the speckle image data using a probe bending state recognition model.

[0013] According to the multimode fiber probe bending state identification method provided by the present invention, the probe end is an end ball, which is connected to a digital micromirror device (DMM). After the probe light reaches the end ball, a first part of the probe light, under the reflection of the side of the probe end near the needle body, carries speckle light field information and propagates in the reverse direction from the probe end along the needle body and the multimode fiber. A second part of the probe light passes through the probe end and reaches the DMM. Under the reflection of the DMM, it carries the image light field information and speckle light field information at the probe point, and propagates in the reverse direction through the probe end along the needle body and the multimode fiber. The step of receiving the outgoing light carrying speckle light field information from the multimode fiber using an area array image sensor to obtain speckle image data includes: receiving the outgoing light carrying speckle light field information and image light field information from the multimode fiber using an area array image sensor to obtain speckle image data carrying the image information of the probe point. The method further includes: using an image reconstruction model to obtain a reconstructed image of the probe point based on the speckle image data carrying the image information of the probe point.

[0014] The present invention also provides a multimode fiber probe bending state identification device, the device being installed on the multimode fiber sensor described above, the device comprising: An injection module is used to inject probe light into the multimode fiber optic cable using a fiber optic coupler in response to the probe end of the multimode fiber probe reaching the probe point; a receiving module is used to receive the outgoing light carrying speckle light field information from the multimode fiber using an area array image sensor to obtain speckle image data; wherein the outgoing light is generated by the probe light being transmitted to the probe end through the multimode fiber and the probe body, and then reflected by the probe end; a determining module is used to determine the bending state of the probe end of the multimode fiber optic probe at the probe point based on the speckle image data.

[0015] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multimode fiber probe bending state identification method as described above.

[0016] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multimode fiber probe bending state identification method as described above.

[0017] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multimode fiber probe bending state identification method as described above.

[0018] The multimode fiber probe provided by this invention, through its design of periodically twisting the probe body at a preset angle along the optical axis, combined with its flexible multimode fiber structure and a probe end that reflects the probe light, achieves modulation of the optical field transmitted by the bending deformation of the multimode fiber, encoding the bending state as characteristic changes in the speckle pattern. Specifically, when the fiber bends, it induces a mode coupling effect, causing the reflected light field transmitted in the reverse direction to carry speckle image data with a unique mapping relationship to the bending state of the multimode fiber probe. This design enhances the sensitivity of the speckle pattern to bending deformation, realizing a nonlinear mapping relationship between speckle pattern changes and the bending angle, direction, and curvature of the multimode fiber probe, thereby improving the resolution and accuracy of the bending state. It is suitable for various complex application scenarios such as medical endoscopic imaging, robot morphology estimation, and industrial sensing and detection. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the structure of the multimode fiber probe provided by the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of the multimode fiber optic sensor provided by the present invention.

[0022] Figure 3 This is a flowchart illustrating the multimode fiber probe bending state identification method provided by the present invention.

[0023] Figure 4 This is a flowchart illustrating the method for obtaining bending states using a probe bending state identification model provided by the present invention.

[0024] Figure 5 This is one of the structural schematic diagrams of the multimode fiber optic probe bending state detection system provided by the present invention.

[0025] Figure 6 This is one of the structural schematic diagrams of the training dataset acquisition system provided by the present invention.

[0026] Figure 7 This is one of the schematic diagrams illustrating the principle of the multimode fiber probe bending state identification method provided by the present invention.

[0027] Figure 8 This is the second schematic diagram of the multimode fiber optic probe bending state detection system provided by the present invention.

[0028] Figure 9 This is the second schematic diagram of the training dataset acquisition system provided by the present invention.

[0029] Figure 10 This is the second schematic diagram of the principle of the multimode fiber probe bending state identification method provided by the present invention.

[0030] Figure 11 This is a schematic diagram of the experimental results of the multimode fiber probe bending state identification method provided by the present invention.

[0031] Figure 12 This is a schematic diagram illustrating the experimental results of obtaining reconstructed images using the image reconstruction model provided by the present invention.

[0032] Figure 13 This is a schematic diagram of a mask for the bending state of a multimode fiber optic probe obtained using a vision camera, provided by the present invention.

[0033] Figure 14 This is a schematic diagram of the structure of the multimode fiber optic probe bending state identification device provided by the present invention.

[0034] Figure 15 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0035] Figure label: 10: Multimode fiber probe; 11: Probe end; 12: Needle body; 100: Multimode fiber sensor; 110: Multimode fiber probe; 120: Multimode fiber; 510: Coherent light source; 520: Fiber coupler; 530: Multimode fiber; 540: Multimode fiber probe; 550: Signal fiber; 560: Area array image sensor; 570: Data image processing module; 610: Bending control device; 620: Multimode fiber probe; 630: Multimode fiber; 640: Fiber coupler; 650: Coherent light source; 660: Area array image sensor; 670: Data image processing module; 700: Dynamically updated dual-model integrated network; 710: First probe bending state recognition model; 720: Second probe bending state recognition model; 810: Coherent light source; 820: Fiber optic coupler; 830: Multimode fiber; 840: Multimode fiber probe; 850: Signal fiber; 860: Area array image sensor; 870: Data image processing module; 880: Multimode fiber end ball; 890: Digital micromirror device; 910: Bending control device; 920: Multimode fiber probe; 930: Multimode fiber; 940: Fiber optic coupler; 950: Coherent light source; 960: Area array image sensor; 970: Data image processing module; 980: Digital micromirror device; 990-1: Top vision camera; 990-2: Front vision camera. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0037] The following is combined with Figure 1 The multimode fiber probe of the present invention is described as follows: Figure 1 As shown, the multimode fiber probe 10 includes a probe end 11 and a needle body 12.

[0038] The needle body 12 is periodically twisted at a preset angle along the optical axis, and the needle body is made of flexible multimode optical fiber.

[0039] The side of the probe end 11 closest to the needle body 12 can reflect the probe light, and the probe end 11 and the needle body 12 are connected by a continuous multimode optical fiber.

[0040] After the probe light reaches the probe end 11 along the needle body 12, it is reflected by the probe end 11 and carries the speckle light field information from the probe end 11 and is transmitted in the opposite direction along the needle body 12.

[0041] When the probe light propagates within the internal optical path of a multimode fiber, it undergoes physical phenomena such as intermodal dispersion and mode coupling. These phenomena cause changes in the distribution of the probe light's optical field, thereby enabling the probe light to carry speckle field information. This speckle field information accurately reflects the state within the fiber optic path.

[0042] The multimode fiber probe 10 is an optical sensor based on flexible multimode fiber provided by this invention, suitable for scenarios requiring high-precision bending detection. For example, in medical endoscopy applications, the multimode fiber probe 10 can be embedded in the endoscope tip to monitor the bending state of the endoscope inside the body in real time, assisting doctors in performing precise operations.

[0043] The needle body 12 is the main part of the multimode fiber probe 10, which is made of flexible multimode fiber and is periodically twisted at a preset angle along the optical axis. This design enhances the probe's sensitivity to micro-bending deformation and effectively improves sensing accuracy.

[0044] For example, assuming a preset angle of 180 degrees and a cycle of one twist per centimeter, such a design can make the probe produce more obvious speckle pattern changes when it bends, making it easier to detect and analyze.

[0045] The probe end 11 is the end portion of the multimode fiber probe 10, and a reflective film is coated on the side facing the probe body 12. It is used to reflect the probe light, so that it carries the speckle light field information and transmits it back to the detection system.

[0046] In some embodiments, the probe end 11 is an end face with a reflective film coated on the side facing the needle body 12. After the probe light reaches the probe end 11 along the needle body 12, it carries speckle light field information from the probe end 11 and is transmitted in the opposite direction along the needle body 12 under the action of the reflective film.

[0047] In some embodiments, the detection end 11 is an end ball, fixed (e.g., by welding) to one side of the needle body 12. After the detection light reaches the detection end 11, due to the reflection effect of the end ball, a portion of the detection light, carrying speckle light field information, is reflected from the detection end 11 along the needle body 12 on the side of the detection end 11 near the needle body 12 and transmitted in the reverse direction.

[0048] In some embodiments, the detection end 11 is a ball-shaped end, and the detection end 11 is connected to a digital micromirror device. After the detection light reaches the detection end 11, the first part of the detection light, under the reflection of the side of the detection end 11 near the needle body 12, carries speckle light field information and is transmitted in reverse from the detection end 11 along the needle body 12. The second part of the detection light passes through the detection end 11 and reaches the digital micromirror device. Under the reflection of the digital micromirror device, it carries the image light field information and speckle light field information at the detection point, passes through the detection end 11 in reverse, and is transmitted in reverse along the needle body 12.

[0049] In some embodiments, the distance between the probe end 11 and the output end face of the multimode fiber probe 10 can be set to 55 mm. Furthermore, by twisting the probe body 12 180° along the optical axis, a nonlinear relationship between the bending amount and the speckle light field is successfully introduced. This nonlinear relationship makes the light field more sensitive to bending changes, thereby significantly improving the system's ability to sense bending states.

[0050] Speckle light field information refers to the light field distribution information about the bending state of the needle body 12 carried by the probe light after reflection at the probe end 11. By analyzing this information, the degree of bending or deformation state of the probe can be inferred. When the probe bends, the speckle light field information changes, and these changes have a non-linear mapping relationship with the bending angle, direction, and curvature of the probe. This information can be resolved through specific algorithms to achieve accurate identification of the bending state.

[0051] In some embodiments, a coherent light source can be used to emit a continuous light field as a probe light. A coherent light source can provide stable, continuous light waves for high-precision bending state sensing.

[0052] Coherent light sources can include, but are not limited to, fiber lasers, gas lasers, solid-state lasers, liquid lasers, and light emitted by non-laser sources with coherence.

[0053] In the specific implementation process, the multimode fiber probe 10 can be fabricated in the following way: A flexible multimode fiber is selected as the needle material, and its length and diameter are determined according to the specific application requirements. The multimode fiber is periodically twisted along the optical axis at a preset angle (e.g., 180 degrees) and period (e.g., twisted once per centimeter) to form the needle body 12. A high-reflectivity metal reflective film (e.g., silver film) is deposited on the end of the needle body 12 (i.e., the probe end 11) facing the needle body to ensure effective reflection of the probe light.

[0054] The multimode fiber probe 10 provided by this invention, through its periodic torsion design and a probe end that reflects the probe light, can generate reflected light carrying significant speckle light field information. By analyzing this speckle light field information, the degree of curvature or deformation state of the target region can be accurately determined. This multimode fiber probe has advantages such as high precision, low cost, and easy integration, and can be widely used in fields such as medical endoscopy, industrial inspection, and robot morphology estimation.

[0055] The following is combined with Figure 2 The multimode fiber optic sensor of the present invention is described as follows: Figure 2 As shown, the multimode fiber sensor 100 includes a multimode fiber probe 110 and a multimode fiber 120.

[0056] For a detailed description of the multimode fiber probe 110, please refer to [link / reference]. Figure 1 The relevant content will not be repeated here.

[0057] The multimode fiber is seamlessly connected to the multimode fiber probe.

[0058] In this process, after the probe light reaches the probe end along the multimode fiber and the needle body, it carries the speckle light field information from the probe end and is transmitted in reverse along the needle body and the multimode fiber under the reflection effect of the side end.

[0059] Multimode fiber 120 is another part of the multimode fiber sensor, used to transmit probe light to multimode fiber probe 110 and transmit the reflected speckle light field information to the detection device. Multimode fiber 120 can be a flexible multimode fiber, the length and diameter of which are determined according to specific application requirements.

[0060] Seamless connection refers to the absence of obvious interfaces or gaps at the connection point between the multimode fiber 120 and the multimode fiber probe 110, ensuring efficient and lossless transmission of optical signals. In practical implementation, seamless connection between the multimode fiber 120 and the multimode fiber probe 110 can be achieved through fiber optic fusion splicing technology.

[0061] In some embodiments, such as Figure 5 The high-precision bending detection system based on the multimode fiber optic sensor provided by this invention, as shown, enables high-precision real-time monitoring of the bending state of a multimode fiber optic probe. This system can be applied in fields such as medical endoscopic imaging, robot morphology estimation, and industrial sensing and detection, and has advantages such as high precision, wide deflection range, and strong real-time performance.

[0062] like Figure 5 As shown, the bending detection system includes a coherent light source 510, which uses a narrow linewidth fiber laser to emit a stable continuous light field, providing a detection light source for the bending detection system.

[0063] The fiber coupler 520, employing a (2+1)×1 side coupler, couples the probe light into the multimode fiber 530 (e.g., injects the probe light into the cladding of the multimode fiber 530), thereby integrating the illumination and probe optical paths.

[0064] Multimode fiber 530: Employs step-index double-clad fiber for transmitting signal optical fields.

[0065] Multimode fiber probe 540: fused with multimode fiber 530. See details for more information. Figure 1 The relevant content will not be repeated here.

[0066] The reflected light output from the multimode fiber 530 is output via the signal fiber 550. This reflected light carries information about the bending deformation of the multimode fiber probe. In a specific implementation, the signal fiber 550 can be the signal fiber carried by the fiber coupler 520 implemented with a (2+1)×1 side coupler.

[0067] The multimode fiber 530, the multimode fiber probe 540, and the fiber coupler 520 all use the same multimode step-index fiber to ensure mode field coupling efficiency.

[0068] Area array image sensor 560: Employs a camera CMOS sensor to capture reflected light emitted from the signal fiber end face of fiber coupler 520, and converts the reflected light into an electrical signal with a two-dimensional intensity distribution to obtain speckle image data.

[0069] The data image processing module 570 is implemented using a computing device (e.g., a computer) and includes a preprocessing unit and a bending recognition unit. The preprocessing unit performs region cropping, grayscale normalization, and dimensionality reduction sampling on the raw speckle image data, extracts effective feature regions, compresses the data volume, and generates a standardized input matrix. The bending recognition unit calculates the bending parameters of the multimode fiber probe based on a preset algorithm (see description of the preset algorithm). Figure 1 and Figure 2 The relevant content will not be repeated here.

[0070] In practical implementation, due to the small diameter and high flexibility of the multimode fiber probe, the multimode fiber probe 540 can be attached to the structure under test (e.g., a lesion in the human body). As the structure under test bends, the multimode fiber probe undergoes corresponding deformation. Using the above-mentioned bending detection system, the bending state of the structure under test can be accurately determined.

[0071] Finally, the calculated bending parameters are output in graphical or numerical form, allowing users to intuitively understand the bending state of the tested structure.

[0072] Alternatively, depending on the application scenario, the bending parameters can be further processed to determine the location information of the probe tip of the multimode fiber optic probe, and then the next operation can be performed based on this location information. For example, in endoscopic surgery, by acquiring the location information of the multimode fiber optic probe tip in real time, doctors can be assisted in precise navigation to ensure that surgical instruments accurately reach the lesion. Furthermore, by combining bending parameters and location information, the specific location of the lesion can be marked on medical images, providing accurate positioning data for subsequent treatment.

[0073] In some embodiments, the probe end is spherical and connected to a digital micromirror device, which can utilize, for example... Figure 8 The high-precision bending detection system based on the multimode fiber sensor provided by the present invention realizes high-precision real-time monitoring of the bending state of the multimode fiber probe.

[0074] like Figure 8 As shown, the system includes a coherent light source 810, which uses a narrow linewidth fiber laser to emit a stable continuous light field, providing a detection light source for the bending detection system.

[0075] The fiber coupler 820, employing a (2+1)×1 side coupler, couples the probe light into the multimode fiber 830 (e.g., injects the probe light into the cladding of the multimode fiber 830), thereby integrating the illumination and probe optical paths.

[0076] Multimode fiber 830: Employs step-index double-clad fiber for transmitting signal optical fields.

[0077] Multimode fiber probe 840: fused with multimode fiber 830. See details for more information. Figure 1 The relevant content will not be repeated here.

[0078] The multimode fiber end ball 880 is fixedly connected to the multimode fiber probe 840.

[0079] Digital micromirror device 890, multimode fiber end ball 880 fixed connection.

[0080] The reflected light output from the multimode fiber 830 is then output via the signal fiber 850. This reflected light carries information about the bending deformation of the multimode fiber probe and the image light field information at the detection point (see detailed description). Figure 1 (Related content in the document). In specific implementation, the signal fiber 850 can be the signal fiber carried by the fiber coupler 820 implemented by a (2+1)×1 side coupler.

[0081] The multimode fiber 830, multimode fiber probe 840, and fiber coupler 820 all use the same multimode step-index fiber to ensure mode field coupling efficiency.

[0082] The area array image sensor 860 uses a camera CMOS sensor to capture the reflected light emitted from the signal fiber end face of the fiber coupler 820 and convert the reflected light into an electrical signal with a two-dimensional intensity distribution to obtain speckle image data carrying image information at the detection point.

[0083] The data image processing module 870, implemented using a computing device (e.g., a computer), includes a preprocessing unit, a bending recognition unit, and an imaging unit. The preprocessing unit performs region cropping, grayscale normalization, and dimensionality reduction sampling on the raw speckle image data, extracts effective feature regions, compresses the data volume, and generates a standardized input matrix. The bending recognition unit calculates the bending parameters of the multimode fiber probe based on a preset algorithm (see description of the preset algorithm). Figure 1 and Figure 2 (The relevant content will not be repeated here.) The imaging unit reconstructs the image for the detection point based on the image information carried by the speckle image data.

[0084] The following is combined with Figures 3-8 The present invention describes a method for identifying the bending state of a multimode fiber optic probe.

[0085] Figure 3 This is a flowchart illustrating the multimode fiber probe bending state identification method provided by the present invention. This method is applied to applications such as... Figure 2 The multimode fiber optic sensor shown, such as Figure 3 As shown, the method includes the following steps.

[0086] Step 301: In response to the probe end of the multimode fiber probe reaching the probe point, the probe light is injected into the multimode fiber using a fiber coupler.

[0087] The detection point can be determined based on different application scenarios. For example, in medical endoscopy, the detection point is the specific location inside the human body reached by the probe end of a multimode fiber optic probe that is integrated with the endoscope (e.g., through embedding), such as the stomach, intestines, or respiratory tract—the area to be examined. When the endoscope is inserted into the human body and moved to the target location, the probe end of the multimode fiber optic probe reaches that detection point, at which point bending can be detected and imaged.

[0088] Step 302: Receive the outgoing light carrying speckle light field information from the multimode fiber using an area array image sensor to obtain speckle image data; wherein, the outgoing light is transmitted from the probe light through the multimode fiber and the needle to the probe end, and is generated after being reflected by the probe end.

[0089] Speckle image data can accurately reflect the bending state of a multimode fiber probe. By analyzing the speckle image, parameters such as the bending angle, direction, and curvature of the probe at the detection point can be calculated.

[0090] For the specific implementation process of steps 301 to 302, please refer to the above. Figure 5 The details of the bending detection system will not be elaborated here.

[0091] Step 303: Based on the speckle image data, determine the bending state of the probe end of the multimode fiber probe at the probe point.

[0092] In practice, the bending state of the probe end of the multimode fiber optic probe at the probe point can be determined in various ways based on speckle image data.

[0093] In the specific implementation process, the bending state can be obtained by using the probe bending state recognition model based on the speckle image data; the probe bending state recognition model is obtained by training the initial machine learning model using the training dataset.

[0094] In some embodiments, the input data for the probe bending state recognition model is a standardized input matrix obtained after preprocessing the original speckle image data.

[0095] The output of the probe bending state recognition model includes: Bending angle: indicates the degree of bending of the multimode fiber probe in a specific direction, usually in degrees (°).

[0096] Bending direction: Indicates the specific direction of the multimode fiber probe bending, such as up and down, left and right, or a specific angle.

[0097] Curvature: Describes the radius of curvature or rate of change of curvature of a multimode fiber probe, reflecting the smoothness or rate of change of the bend.

[0098] These output parameters provide doctors, engineers, or operators with intuitive, quantitative information that helps them accurately determine the position and status of multimode fiber optic probes, thereby enabling precise operation or diagnosis.

[0099] In some embodiments, the morphological mask of the multimode fiber probe can be obtained based on the speckle image data and using a probe bending state recognition model.

[0100] In this embodiment, the input data for the probe bending state recognition model is a standardized input matrix obtained after preprocessing the original speckle image data.

[0101] The output of the probe bending state recognition model is: the morphology mask of the multimode fiber probe.

[0102] For a detailed description of the training dataset used in this embodiment, please refer to [link / reference]. Figure 9 The relevant content will not be repeated here.

[0103] Probe bending state recognition models can be built upon various machine learning models, such as convolutional neural network models. As an example only, a probe bending state recognition model may include: Convolutional layers extract local features from an image by performing convolution operations on the input matrix using convolution kernels. Multiple convolutional layers can be stacked together to form deep feature extraction networks to capture more complex feature patterns.

[0104] Pooling layers are used to downsample the feature maps output by convolutional layers, reducing data dimensionality and computational cost while preserving important feature information. Common pooling operations include max pooling and average pooling.

[0105] Fully connected layers flatten the feature maps output by pooling layers into one-dimensional vectors and associate the features with the output labels through fully connected operations. Fully connected layers can contain multiple hidden layers for further learning and extraction of high-level features.

[0106] The output layer, designed according to task requirements, may contain multiple neurons for tasks involving bending state recognition, corresponding to output parameters such as bending angle, bending direction, and curvature. The output layer typically uses activation functions (such as Softmax or linear activation functions) to convert the output of the fully connected layer into specific bending parameter values.

[0107] Regarding an example of obtaining the bending state using a probe bending state recognition model based on speckle image data, see [link to example]. Figure 4 The relevant content will not be repeated here.

[0108] For the training method of the probe bending state recognition model, please refer to [link / reference]. Figure 6 The relevant content will not be repeated here.

[0109] In some embodiments, the probe end is a ball-shaped device connected to a digital micromirror device. For a detailed description of this embodiment, see [link to relevant documentation]. Figure 1 The relevant content will not be repeated here.

[0110] In this embodiment, an area array image sensor is used to receive outgoing light carrying speckle light field information and image light field information from the multimode fiber to obtain speckle image data carrying the image information of the detection point; and an image reconstruction model is used to obtain a reconstructed image of the detection point based on the speckle image data carrying the image information of the detection point.

[0111] Image reconstruction models are machine learning models trained using training datasets. In practice, various machine learning models can be used to construct image reconstruction models, including but not limited to: The encoder-decoder network (U-Net structure) consists of an encoder that extracts multi-scale features (such as edges and textures) from speckle images through downsampling layers and progressively compresses the feature map size; a decoder that restores spatial resolution through deconvolutional and upsampling layers and combines low-level features from the encoder with skip connections to improve detail reconstruction capabilities; and a physical constraint layer that simulates the light field propagation process through differentiable rendering layers to enhance physical interpretability.

[0112] The Transformer model divides the speckle image into blocks into sequences and captures global correlations through a self-attention mechanism to reconstruct the probe point image.

[0113] The input data for the image reconstruction model is speckle image data carrying probe point image information, and the output data is the reconstructed image of the probe point, that is, the original scene image of the region where the probe point is located, recovered from the speckle. For example, in medical applications, the probe point is located in a region of interest (such as a lesion area) within biological tissue, and the method provided in this embodiment of the invention can obtain an image of this region of interest. For a detailed description of the training dataset for the image reconstruction model, see [link to relevant documentation]. Figure 9 The relevant content will not be repeated here.

[0114] Figure 4 This is a flowchart illustrating the method for obtaining bending states using a probe bending state identification model provided by the present invention.

[0115] In this embodiment, the probe bending state recognition model includes a first probe bending state recognition model and a second probe bending state recognition model, both of which are trained machine learning models.

[0116] like Figure 4 As shown, the method includes the following steps.

[0117] Step 401: Based on the speckle image data, the first bending state is obtained using the first probe bending state recognition model.

[0118] For a detailed description of speckle image data, see [link to relevant documentation]. Figure 2 and Figure 3 The relevant content will not be repeated here.

[0119] Step 402: Fine-tune the parameters of the second probe bending state recognition model based on the speckle image data and the first bending state.

[0120] Step 403: Based on the speckle image data, the second bending state is obtained using the finely tuned second probe bending state recognition model.

[0121] Step 404: Determine the final bending state based on the first bending state and the second bending state.

[0122] The first probe bending state recognition model and the second probe bending state recognition model can be the same probe bending state recognition model trained using the same dataset. For a detailed description of the probe bending state recognition model, see [link to relevant documentation]. Figure 3 The relevant content will not be repeated here.

[0123] The newly deployed first and second probe bending state recognition models exhibit high accuracy, but their robustness decreases after processing large amounts of speckle image data, affecting prediction accuracy. Therefore, speckle image data can be used as sample data, and the first bending state as a pseudo-label. The parameters of the second probe bending state recognition model can be fine-tuned in a real-time monitoring environment to enhance the model's adaptability to speckle image data and improve its robustness.

[0124] The first bending state is the estimated bending state value obtained by performing preliminary analysis of the speckle image data using the first probe bending state recognition model.

[0125] The second bending state is an estimate of the bending state obtained after analyzing the speckle image data using a finely tuned second probe bending state recognition model. For a detailed description of the bending state, see [link to relevant documentation]. Figure 2 and Figure 3 The relevant content will not be repeated here.

[0126] In the specific implementation process, the final bending state can be obtained in a variety of ways based on the first bending state and the second bending state.

[0127] For example, different weights can be assigned to the first and second bending states, and then their weighted average can be calculated as the final bending state. The weight assignment can be determined based on the accuracy, stability, or specific application requirements of the first and second probe bending state recognition models.

[0128] For example, the confidence levels of the first and second bending states can be evaluated, and then the bending state with the higher confidence level can be selected as the final bending state.

[0129] For example, a threshold can be set. When the difference between the first bending state and the second bending state is less than the threshold, they are considered to be consistent, and one of the states (such as the average value or any state) can be selected as the final bending state. When the difference is greater than the threshold, a weighted average algorithm can be used to obtain the final bending state based on the first bending state and the second bending state.

[0130] Figure 6 This is one of the structural schematic diagrams of the training dataset acquisition system provided by this invention. For example... Figure 6 As shown, the training data acquisition system includes: A bending control device 610, such as an electrically driven rotary stage, is used to precisely control the bending direction and curvature of the multimode fiber probe 620. The electrically controlled rotary stage, through its electrical rotation, can simulate the probe's behavior under different bending conditions, providing the system with diverse training data.

[0131] Multimode fiber probe 620, which is fused with multimode fiber 630 to obtain Figure 2 The multimode fiber sensor shown is used for bending state sensing. For a detailed description of the multimode fiber probe, see [link to documentation]. Figure 1 The relevant content will not be repeated here. In the specific implementation process, the near end of the multimode fiber probe 620 is fused and fixed to the multimode fiber 630, and the far end is fixed on the electric rotary displacement stage 610. The bending state is controlled by rotating the displacement stage.

[0132] Multimode fiber 630 is used to transmit signal optical fields.

[0133] Fiber optic coupler 640, such as a (2+1)×1 side coupler, is used to inject probe light into multimode fiber 220 to integrate the probe light with the probe optical path.

[0134] Coherent light sources 650, such as narrow-linewidth fiber lasers, are used to emit stable, continuous light fields to provide a detection light source for multimode fiber probes.

[0135] An area array image sensor 660, such as a CMOS sensor, is used to capture reflected light generated at the probe end of a multimode fiber optic probe to obtain speckle image data.

[0136] The data image processing module 670, such as a computer, is used to receive speckle image data acquired by a CMOS sensor, perform data preprocessing, and obtain a training dataset.

[0137] The training dataset includes sample data and the corresponding labels for each sample. In practical implementation, methods such as... Figure 6 The training data acquisition system shown obtains sample data and its corresponding labels in the following ways: like Figure 6 As shown, the multimode fiber probe 620 is set to a preset bending state using the bending control device 610; the probe light emitted by the coherent light source 650 is injected into the multimode fiber 630 using the fiber coupler 640; and the outgoing light carrying speckle light field information from the multimode fiber 630 is received by the area array image sensor 660 to obtain speckle image data. Specifically, the received speckle light field information is converted into an electrical signal with a two-dimensional intensity distribution and recorded as two-dimensional image information, thus obtaining speckle image data samples. These sample data contain the speckle light field characteristics of the probe under specific bending conditions; wherein, the outgoing light is transmitted from the probe light through multimode fiber and the probe body to the probe end, and is generated after being reflected by the reflective film; using the data image processing module 670, sample data is obtained based on the speckle image data samples, for example, the speckle image data samples are preprocessed (such as region cropping, grayscale normalization, dimensionality reduction sampling, etc.) to extract effective feature regions and compress the data volume, generate a standardized input matrix suitable for model input, and use the obtained input matrix as sample data, and use the preset bending state as the label of the sample data.

[0138] To construct a training dataset with sufficient diversity and representativeness, the above steps need to be repeated. By setting the bending control device 610 (e.g., an electric rotary displacement stage) at different positions, the multimode fiber probe is placed at different bending angles and directions. In each specific bending state, multiple sets of speckle image data samples are acquired. This process involves multiple two-way acquisitions at each position to ensure the comprehensiveness and accuracy of the data.

[0139] After obtaining the training dataset using the above method, the initial machine learning model (e.g., convolutional neural network) can be trained in a supervised manner using the training dataset. By adjusting the parameters of the initial machine learning model (using a preset optimization algorithm, such as gradient descent optimization algorithm) to minimize the prediction error (determined using a loss function, such as squared difference loss function), a probe bending state recognition model that can accurately identify the bending state of multimode fiber probes can be obtained.

[0140] like Figure 7 As shown, the final probe bending state recognition model includes a first probe bending state recognition model 710 and a second probe bending state recognition model 720. Both have the same model structure and parameters, forming a dynamically updated dual-model integrated network 700. For an embodiment of using a dynamically updated dual-model integrated network to identify the bending state of a multimode fiber probe, see [link to example]. Figure 4 The relevant descriptions in the text will not be repeated here.

[0141] Figure 9 This is the second schematic diagram of the training dataset acquisition system provided by this invention. (See diagram below.) Figure 9 As shown, the training data acquisition system includes: A bending control device 910, such as an electrically driven rotary stage, is used to precisely control the bending direction and curvature of the multimode fiber probe 920. The electrically controlled rotary stage, through its electrical rotation, can simulate the probe's behavior under different bending conditions, providing the system with diverse training data.

[0142] Multimode fiber probe 920, which is fused with multimode fiber 930 to obtain Figure 2 The multimode fiber sensor shown is used for bending state sensing. The probe end of this multimode fiber probe is ball-shaped. For a detailed description of the multimode fiber probe, see [link to documentation]. Figure 1 The relevant content will not be repeated here. In the specific implementation process, the near end of the multimode fiber probe 920 is fused and fixed to the multimode fiber 930, and the far end is fixed on the electric rotary displacement stage 910. The bending state is controlled by the rotation of the displacement stage.

[0143] Multimode fiber 930 is used to transmit signal optical fields.

[0144] Fiber optic coupler 940, such as a (2+1)×1 side coupler, is used to inject probe light into multimode fiber 220 to integrate the probe light with the probe optical path.

[0145] Coherent light sources, such as narrow-linewidth fiber lasers, are used to emit stable, continuous light fields to provide a detection source for multimode fiber probes.

[0146] An area array image sensor 960, such as a CMOS sensor, is used to receive outgoing light carrying speckle light field information and image light field information from the multimode fiber, and to obtain speckle image data carrying the image information of the detection point.

[0147] The data image processing module 970, such as a computer, is used to receive speckle image data acquired by a CMOS sensor, perform data preprocessing, and obtain a training dataset.

[0148] The Digital Micromirror Device 980 images the area where the detection point is located by reflecting light.

[0149] 990-1 and 990-2 are the top vision camera and front vision camera in a binocular vision camera system, respectively. They are used to photograph the multimode fiber probe after it has been placed in a preset bent state, thereby obtaining a morphological mask of the multimode fiber probe in the preset bent state. Figure 13 As shown, for different bending states of the multimode fiber probe (corresponding to different speckle image data), the upper vision camera and the front vision camera in the binocular vision camera are used to capture images of the multimode fiber probe to obtain images of different bending states. The captured bending state images are then subjected to HSV threshold segmentation and morphological processing to obtain the morphological mask of the multimode fiber probe.

[0150] The training dataset includes sample data and the corresponding labels for each sample. In practical implementation, methods such as... Figure 9 The training data acquisition system shown obtains sample data and its corresponding labels in the following ways: like Figure 9 As shown, the multimode fiber probe 920 is placed in a preset bending state using the bending control device 910, and the multimode fiber probe is photographed using a vision camera. The photographed bending state image is then subjected to HSV threshold segmentation and morphological processing to obtain a morphological mask (binary image) of the multimode fiber probe in the preset bending state.

[0151] Using fiber coupler 940, the probe light emitted from coherent light source 950 is injected into multimode fiber 930. An area array image sensor 960 receives the outgoing light carrying speckle light field information and image light field information from multimode fiber 930, obtaining speckle image data carrying the image information of the probe points. Specifically, the received speckle light field information is converted into a two-dimensional intensity distribution electrical signal and recorded as two-dimensional image information, thus obtaining speckle image data samples. These sample data contain the speckle light field characteristics of the probe under specific bending states and the image information of the probe points. Using data image processing module 970, sample data is obtained based on the speckle image data samples carrying the image information of the probe points. For example, the speckle image data samples are preprocessed (e.g., region cropping, grayscale normalization, dimensionality reduction sampling, etc.) to extract effective feature regions and compress the data volume, generating a standardized input matrix suitable for model input. The obtained input matrix is ​​used as the sample data.

[0152] For the task of identifying the bending state of a multimode fiber optic probe, a morphological mask of the multimode fiber optic probe in a preset bending state can be used as a label for the sample data. The sample data and its corresponding label constitute a first training dataset. The initial machine learning model (e.g., a convolutional neural network model) is then trained in a supervised manner using the first training dataset to obtain a trained probe bending state identification model.

[0153] For the image reconstruction task of the probe point, the image of the area where the probe point is located (e.g., an image obtained through optical imaging technology) can be used as the label of the sample data. The sample data and its corresponding label constitute a second training dataset. The initial machine learning model (e.g., a Transformer model) is trained in a supervised manner using the second training dataset to obtain the trained image reconstruction model.

[0154] like Figure 12 As shown, using the image reconstruction model, the reconstructed image of the input image can be accurately obtained based on the speckle image corresponding to the input image.

[0155] To construct a training dataset with sufficient diversity and representativeness, the above steps need to be repeated. By setting the bending control device 910 (e.g., an electric rotary displacement stage) at different positions, the multimode fiber probe is placed at different bending angles and directions. Multiple sets of speckle image data samples are acquired at each specific bending state. This process involves multiple two-way acquisitions at each location to ensure the comprehensiveness and accuracy of the data.

[0156] like Figure 10As shown, based on the speckle image data, the continuous bending state represented by the morphological mask of the multimode fiber probe can be obtained using the probe bending state recognition model. Furthermore, based on the speckle image data, a reconstructed image of the region where the probe point is located can be obtained using the image reconstruction model.

[0157] Figure 11 The diagram illustrates the experimental results of the multimode fiber probe bending state identification method provided by this invention. To verify the practical application value of this invention, systematic verification experiments were conducted on its core performance characteristics, including high precision, full-range bending, and high robustness.

[0158] The experiment employed a two-way data acquisition mode (i.e., each bending state included a forward bending and a reverse reset process), precisely controlling the bending angle of the multimode fiber probe using an electrically driven rotary stage. The resulting training dataset was divided into training and test sets in a 7:1 ratio, with each two-way acquisition containing 10 speckle images to ensure data diversity and temporal correlation. The specific experimental design and results are as follows: In the high-precision experiment (0.025° resolution), 20 bending positions were set at 0.025° intervals within a bending range of ±0.25°. A total of 7 round trips (1400 images) were collected as the training set, and 1 round trip (200 images) was collected as the test set. Figure 8 The confusion matrix results shown in the left subfigure demonstrate that the multimode fiber probe bending state identification method provided by this invention can accurately identify angle differences of 0.025° even under micro-bending conditions, with a test set classification accuracy of 95.63%. This verifies the high-resolution sensing capability of the multimode fiber probe provided by this invention for minute bending amounts.

[0159] In the full-area bending experiment (±90° coverage), 180 locations were uniformly selected within the ±90° bending range, and 10 images were acquired at each location. The training set (7 round trips, 12600 images) and the test set (1 round trip, 1800 images) maintained a spatially uniform distribution. Figure 8 The confusion matrix results shown in the right-hand subplot demonstrate that the test set classification accuracy reaches 100%, indicating that the system can maintain a stable mapping relationship under large-scale bending conditions, breaking through the bending range limitation of traditional speckle sensors.

[0160] In the high robustness experiment (time-varying disturbance compensation), with 20 bending positions at fixed 1° intervals, the training set used 7 two-way (1400 images) static data, while the test set simulated time-varying disturbance by continuously collecting 30 dynamic two-way (6000 images) data over 100 seconds. As shown in accuracy curve 530, the traditional VGG network experienced accuracy fluctuations and declines due to environmental disturbances, while the dynamically updated dual-model ensemble network (consisting of a first probe bending state recognition model and a second probe bending state recognition model) achieved 100% stable classification through online pseudo-label generation and model fine-tuning, highlighting the strong adaptability of the self-supervised mechanism to complex working conditions. Furthermore, the test time for a single test set was approximately 2.3ms, with a frame rate greater than 400Hz.

[0161] The combined results of the three experiments above demonstrate that the multimode fiber probe bending state identification method provided by this invention effectively overcomes the bottleneck of traditional sensing technology in balancing accuracy, range, and robustness.

[0162] The multimode fiber probe provided by this invention significantly enhances the sensitivity of speckle to bending through a periodically twisted fiber design. It also innovatively combines a dynamically updated dual-model integrated network to achieve intelligent analysis of the speckle-bending state mapping relationship. The multimode fiber probe employs a periodically twisted structure; this unique design not only enhances the sensitivity of intermodal interference to micro-bending deformation but also establishes a nonlinear amplified response relationship between speckle modes and bending parameters. Based on this, and combined with the powerful representation capability of neural networks for higher-order features, the system successfully achieves an angular resolution at the 0.025° level, ensuring accurate capture of minute angular changes even in complex bending environments, providing a solid guarantee for precise bending state monitoring.

[0163] Compared to traditional FBG (Fiber Bragg Grating) or distributed fiber optic sensing solutions, the multimode fiber optic probe provided by this invention exhibits comprehensive advantages such as sub-millimeter-level flexible structure, 0.025°-level angular resolution, ±90° full-range coverage, and 100Hz-level real-time response. This perfect combination of high precision and real-time performance benefits from the application of a self-supervised learning technique to generate pseudo-labels for online optimization. This technique establishes a dynamically updated dual-model integrated network (a first probe bending state recognition model and a second probe bending state recognition model), ensuring that the system can quickly respond to bending changes in high-speed dynamic environments and fully meet the stringent requirements of real-time monitoring.

[0164] Furthermore, the sub-millimeter diameter and unique twisted structure design of the multimode fiber probe provided by this invention endow it with single-end reflection detection characteristics, enabling the system to achieve bending monitoring of up to 180° within cavities of a specific length. This precise adaptability makes the system particularly suitable for bending state sensing in confined spaces and complex structures, such as medical endoscopes and industrial inspection applications. It provides an innovative, all-fiber, and highly reliable solution for minimally invasive surgical navigation, flexible robot morphology estimation, and industrial equipment deformation monitoring, demonstrating significant technological breakthroughs and broad applicability.

[0165] The following describes the multimode fiber optic probe bending state identification device provided by the present invention. The multimode fiber optic probe bending state identification device described below can be referred to in correspondence with the multimode fiber optic probe bending state identification method described above.

[0166] Figure 14 This is a schematic diagram of the multimode fiber probe bending state identification device provided by the present invention. The device 1400 includes: The injection module 1410 is used to inject probe light into the multimode fiber optic cable using a fiber optic coupler in response to the probe end of the multimode fiber probe reaching the probe point.

[0167] The receiving module 1420 is used to receive the outgoing light carrying speckle light field information output from the multimode fiber using an area array image sensor, and obtain speckle image data; wherein the outgoing light is transmitted from the probe light through the multimode fiber and the needle body to the probe end, and is generated after being reflected by the probe end.

[0168] The determination module 1430 is used to determine the bending state of the probe end of the multimode fiber probe at the probe point based on the speckle image data.

[0169] Figure 15 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 15As shown, the electronic device may include: a processor 1510, a communication interface 1520, a memory 1530, and a communication bus 1540, wherein the processor 1510, the communication interface 1520, and the memory 1530 communicate with each other through the communication bus 1540. The processor 1510 can call logic instructions in the memory 1530 to execute a multimode fiber probe bending state identification method, which includes: in response to the probe end of the multimode fiber probe reaching the probe point, injecting probe light into the multimode fiber using a fiber coupler; receiving the outgoing light carrying speckle light field information output from the multimode fiber using an area array image sensor to obtain speckle image data; wherein the outgoing light is generated by the probe light being transmitted to the probe end through the multimode fiber and the probe body, and then reflected by the probe end; and determining the bending state of the probe end of the multimode fiber probe at the probe point based on the speckle image data.

[0170] Furthermore, the logical instructions in the aforementioned memory 1530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0171] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multimode fiber probe bending state identification method provided by the above methods. The method includes: in response to the probe end of the multimode fiber probe reaching the probe point, injecting probe light into the multimode fiber using a fiber coupler; receiving outgoing light carrying speckle light field information output from the multimode fiber using an area array image sensor to obtain speckle image data; wherein the outgoing light is generated by the probe light being transmitted to the probe end via the multimode fiber and the probe body, and then reflected by the probe end; and determining the bending state of the probe end of the multimode fiber probe at the probe point based on the speckle image data.

[0172] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for identifying the bending state of a multimode fiber probe provided by the methods described above. The method includes: in response to the probe end of the multimode fiber probe reaching a probe point, injecting probe light into the multimode fiber using a fiber coupler; receiving outgoing light carrying speckle light field information output from the multimode fiber using an area array image sensor to obtain speckle image data; wherein the outgoing light is generated by the probe light being transmitted to the probe end via the multimode fiber and the probe body, and then reflected by the probe end; and determining the bending state of the probe end of the multimode fiber probe at the probe point based on the speckle image data.

[0173] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0174] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0175] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these 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. A multimode fiber optic probe, characterized in that, include: The needle body is periodically twisted at a preset angle along the optical axis, and the needle body is made of flexible multimode optical fiber. The detection end has a side near the needle body that can reflect detection light, and the detection end is connected to the needle body through a continuous multimode optical fiber. Wherein, after the probe light reaches the probe end along the needle body, it carries the speckle light field information from the probe end and is transmitted in reverse along the needle body under the reflection of the probe end.

2. A multimode fiber optic sensor, characterized in that, include: Multimode fiber probe; A multimode optical fiber, which is seamlessly connected to the multimode optical fiber probe; The multimode fiber probe includes: The needle body is periodically twisted at a preset angle along the optical axis, and the needle body is made of flexible multimode optical fiber. The detection end has a side near the needle body that can reflect detection light, and the detection end is connected to the needle body through a continuous multimode optical fiber. Wherein, after the probe light reaches the probe end along the multimode fiber and the needle body, it carries the speckle light field information from the probe end and is transmitted in reverse along the needle body and the multimode fiber under the effect of reflection at the probe end.

3. A method for identifying the bending state of a multimode fiber optic probe, characterized in that, The method is applied to the multimode fiber optic sensor as described in claim 2, and the method includes: In response to the probe end of the multimode fiber probe reaching the probe point, the probe light is injected into the multimode fiber using a fiber coupler; A speckle image data is obtained by receiving outgoing light carrying speckle light field information from the multimode fiber using an area array image sensor; wherein the outgoing light is generated by the detection light being transmitted to the detection end via the multimode fiber and the needle body, and then reflected by the detection end. Based on the speckle image data, the bending state of the probe end of the multimode fiber probe at the probe point is determined.

4. The method for identifying the bending state of a multimode fiber optic probe according to claim 3, characterized in that, Determining the bending state of the probe tip at the detection point based on the speckle image data includes: Based on the speckle image data, the bending state is obtained using a probe bending state recognition model; wherein, the probe bending state recognition model is obtained by training an initial machine learning model using a training dataset.

5. The method for identifying the bending state of a multimode fiber optic probe according to claim 4, characterized in that, The training dataset includes sample data and the labels corresponding to the sample data; The sample data and its corresponding labels are obtained using the following methods: The multimode fiber probe is placed in a preset bending state using a bending control device. The probe light is injected into the multimode fiber using an optical fiber coupler; A speckle image data sample is obtained by receiving the outgoing light carrying speckle light field information from the multimode fiber using an area array image sensor; wherein the outgoing light is generated by the detection light being transmitted to the detection end via the multimode fiber and the needle body, and then reflected by the detection end. The sample data is obtained based on the speckle image data; The label of the sample data is obtained based on the preset bending state.

6. The method for identifying the bending state of a multimode fiber optic probe according to claim 5, characterized in that, After placing the multimode fiber probe into a preset bent state, the method further includes: The multimode fiber probe is photographed using a vision camera to obtain a morphological mask of the multimode fiber probe under the preset bending shape. The step of obtaining the label of the sample data based on the preset bending state includes: The morphological mask of the multimode fiber probe in the preset bending state is used as a label for the sample data. The step of obtaining the bending state based on the speckle image data using a probe bending state recognition model includes: Based on the speckle image data, the morphological mask of the multimode fiber probe is obtained using the probe bending state recognition model.

7. The method for identifying the bending state of a multimode fiber optic probe according to any one of claims 3 to 6, characterized in that, The probe end is a ball, and the ball is connected to the digital micromirror device. After the probe light reaches the end ball, the first part of the probe light, under the reflection of the side of the probe end near the needle body, carries the speckle light field information from the probe end and propagates in reverse along the needle body and the multimode fiber. The second part of the probe light passes through the probe end and reaches the digital micromirror device. Under the reflection of the digital micromirror device, it carries the image light field information and speckle light field information at the probe point, and propagates in reverse along the needle body and the multimode fiber. The step of using an area array image sensor to receive the outgoing light carrying speckle light field information from the multimode fiber to obtain speckle image data includes: The area array image sensor receives the outgoing light carrying speckle light field information and image light field information from the multimode fiber, and obtains speckle image data carrying the image information of the detection point. The method further includes: using an image reconstruction model to obtain a reconstructed image of the detection point based on speckle image data carrying the image information of the detection point.

8. A multimode fiber optic probe bending state identification device, characterized in that, The device is mounted on the multimode fiber optic sensor as described in claim 2, and the device comprises: An injection module is used to inject probe light into the multimode fiber optic cable in response to the probe end of the multimode fiber probe reaching the probe point, using a fiber optic coupler. The receiving module is used to receive the outgoing light carrying speckle light field information from the multimode fiber using an area array image sensor, and obtain speckle image data; wherein the outgoing light is generated by the probe light being transmitted to the probe end through the multimode fiber and the needle body, and then reflected by the probe end. The determination module is used to determine the bending state of the probe end of the multimode fiber probe at the detection point based on the speckle image data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the multimode fiber probe bending state identification method as described in any one of claims 3 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multimode fiber probe bending state identification method as described in any one of claims 3 to 7.

Images