Flexible cable surface defect full-circumferential detection method based on multi-dimensional visual fusion

Abstract

The present application relates to the technical field of flexible cable and protective pipe quality detection, and discloses a flexible cable surface defect full-circumferential detection method based on multi-dimensional visual fusion. The method aims to solve the problems of existing dynamic and static detection disconnection, easy missed detection of implicit defects, placement error affecting test reference, and single-point judgment being easily interfered and misjudged under extreme compression working conditions. By analyzing the outer diameter parameter to perform adaptive physical centering, and monitoring the micro contact force to establish the zero point under pressure, the initial feature matrix in the state of not being under pressure is constructed, the visual track oscillation type shooting and pixel level comparison are used in the synchronous pressing process to lock the implicit micro-cracks of the side wall caused by pressure, the asymmetric deformation characteristics are quantified by measuring the deformation gradient change rate, and the real cross-section closed jump state is established by combining the multi-point redundant logic to exclude the interference of falling debris. The synchronous extraction and comprehensive evaluation of the surface microstructure defects and compression resistance physical properties of the flexible pipe material under dynamic deformation working conditions are realized.

Description

Technical Field

[0001] This invention relates to the field of quality inspection technology for flexible cables and protective conduits, and in particular to a method for full-circumferential inspection of surface defects in flexible cables based on multi-dimensional visual fusion. Background Technology

[0002] In the fields of high-end manufacturing and new energy wiring harness processing, the surface integrity and compressive strength of flexible cables and their polymer protective tubes are directly related to the operational safety and service life of the entire electrical system. Currently, the industry's quality inspection of such flexible tubing often adopts a model that separates static surface visual screening from simple mechanical flattening testing. Conventional static visual equipment can usually only detect obvious damage or severe deformation on the surface of the tubing when it is placed in a natural, stress-free state, using a fixed light source and camera; while traditional mechanical compression testing mainly relies on a press to perform the pressing action, extracting macroscopic displacement and force curves through mechanical sensors, and then calculating macroscopic physical indicators such as the ring stiffness of the tubing.

[0003] However, under complex compression deformation processes, the latent quality defects of flexible polymer tubing often exhibit strong dynamic concealment. In actual testing, many micro-stress concentrations or slight material inhomogeneities caused by processing techniques are not obvious under normal conditions, only manifesting as minute tensile cracks or stress whitening under specific extreme compression conditions. Existing static visual inspection is disconnected from simple mechanical compression equipment, lacking highly sensitive internal visual synchronous tracking methods when performing dynamic compression tests. Conventional lighting and photography with fixed viewing angles easily miss these hidden defects caused by deformation. Furthermore, due to the inherent susceptibility of flexible tubing to deformation and its physical flexibility, placement misalignment or initial micro-deformation due to external forces is inevitable in the initial stages of automated feeding testing. This directly leads to distortion of subsequent testing benchmarks and visual measurement distortion. Furthermore, under complex conditions of extreme compression, the light transmittance inside the tube drops sharply and is easily accompanied by the shedding of debris from the inner wall. Traditional equipment, which relies on a single sensor or a local fixed vision judgment mechanism, is easily interfered with and produces misjudgments. It is difficult to achieve high-precision and high-reliability comprehensive online detection of micro-defects and asymmetric mechanical responses on the tube surface during the entire life cycle of compressive deformation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, in order to solve the problems of the disconnect between existing static visual inspection and mechanical compression testing, which leads to the easy omission of latent dynamic defects, the impact of initial placement error on the test benchmark, and the susceptibility of single-point judgment mechanism under extreme compression conditions to interference from debris inside the tube, this invention provides a full-circumferential detection method for surface defects of flexible cables based on multi-dimensional visual fusion.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion, which includes the following steps: S1. Read the material characteristics and nominal outer diameter of the protective tube to be tested, and control the feeding platform to push the protective tube to be tested to the axis centering position; S2. Drive the upper pressure plate to move downward, establish the zero point of pressure by collecting force data, and lock the spatial geometric relationship between the upper pressure plate and the protective tube to be tested. S3. Perform an initial full-circumferential rotational scan to obtain a panoramic view of the surface texture of the protective tube under test in an unpressurized state, and unfold the outer circumference of the protective tube under test into an initial feature matrix of a plane; S4. Control the upper pressure plate to continue pressing down, and at the same time perform trajectory oscillation shooting through the vision component to capture the characteristic changes caused by the pressure on the protective tube under test, and compare it with the initial feature matrix to identify surface defects; S5. Evaluate the mechanical response characteristics of the protective tube under test based on the deformation gradient change rate of the cross-sectional profile. When the deformation gradient change rate reaches a preset threshold and the cross-sectional profile undergoes a closed jump, it is determined that the flattening test state has been reached and a test report is output.

[0007] As a preferred embodiment of the multi-dimensional vision fusion-based circumferential surface defect detection method for flexible cables described in this invention, the following steps are taken: Reading the material characteristics and nominal outer diameter of the protective tube to be tested, and controlling the feeding platform to push the protective tube to be tested to the axis alignment position. The main control computer reads the identification information on the surface of the protective tube under test, and analyzes and obtains the outer diameter and wall thickness values ​​of the protective tube under test as the nominal outer diameter and material characteristics. The adaptive centering slider at the bottom of the feeding platform is controlled to shrink synchronously towards the center according to the outer diameter value. The mechanical thrust generated by the adaptive centering slider pushes the protective tube under test to the preset center position, ensuring that the central axis of the protective tube under test is on the same horizontal and vertical plane as the center of the lens of the vision component behind it.

[0008] As a preferred embodiment of the multi-dimensional vision fusion-based full-circumferential surface defect detection method for flexible cables described in this invention, the following steps are taken: The upper pressure plate is driven to move downwards; the zero-point of pressure is established by collecting force data; and the spatial geometric relationship between the upper pressure plate and the protective tube under test is locked. The control motor drives the upper pressure plate to move downward at a constant speed along the vertical axis, and the force sensor at the rear end of the upper pressure plate monitors the force data. If the force data reaches the preset initial contact force value limit, it is determined that the lower surface of the upper pressure plate has completed contact with the surface of the protective tube to be tested, and a braking command is sent to the motor to keep the upper pressure plate in a stationary state, and the current mechanical position is locked as the zero point of the flattening test.

[0009] As a preferred embodiment of the multi-dimensional visual fusion-based circumferential surface defect detection method for flexible cables described in this invention, the method includes: performing an initial circumferential rotational scan to obtain a panoramic view of the surface texture of the protective tube under test in an unpressurized state, and unfolding the outer circumference of the protective tube under test into an initial feature matrix of a plane. The specific steps are as follows: After the upper pressure plate remains stationary, the vision component is controlled to automatically slide along the vertical guide rail to an observation height level with the cavity of the protective tube under test, based on the current vertical height coordinates of the upper pressure plate. The illumination lamp is then controlled to rotate and extend into the front opening of the protective tube under test to turn on the light source, thereby acquiring the full circumferential texture information of the surface of the protective tube under test. The outer circumferential surface of the protective tube under test is then unfolded through image stitching processing to construct the initial feature matrix representing the original morphological features.

[0010] As a preferred embodiment of the multi-dimensional vision fusion-based circumferential surface defect detection method for flexible cables described in this invention, the method involves controlling the upper pressure plate to continue pressing down while simultaneously using a vision component to perform trajectory oscillation imaging to capture the characteristic changes caused by the pressure on the protective tube under test, and comparing these changes with the initial feature matrix to identify surface defects. The specific steps are as follows: The upper pressure plate is controlled to continue pressing down at a preset speed, while the vision component is used to acquire dynamic images of the deformation inside the cavity of the protective tube under test at a fixed frame rate. The inner wall contour pixel set of each frame in the dynamic image is extracted, and the upper inner wall edge line and the lower inner wall edge line of the protective tube under test are located by establishing an image coordinate system. The height difference between the upper inner wall edge line and the lower inner wall edge line in the vertical direction is calculated to generate a vertical pixel distance sequence reflecting the degree of tube deformation. The trajectory oscillation shooting and defect identification process specifically includes: controlling the vision component to perform a reciprocating oscillation motion of the viewing angle during vertical pressing, and capturing the texture anomaly of the side wall of the protective tube under test by changing the light angle; projecting the dynamic image onto the local space corresponding to the initial feature matrix, and calculating the deviation value between the two in terms of pixel color grayscale and structural integrity; if the deviation value exceeds a preset safety threshold, the coordinates of the surface defect are automatically recorded, and the pressure data at this time is extracted simultaneously for stress defect correlation processing.

[0011] As a preferred embodiment of the multi-dimensional visual fusion-based circumferential surface defect detection method for flexible cables described in this invention, the mechanical response characteristics of the protective tube under test are evaluated based on the deformation gradient change rate of the cross-sectional profile. When the deformation gradient change rate reaches a preset threshold and the cross-sectional profile undergoes a closed jump, it is determined that the flattening test state has been reached and a test report is output. The specific steps are as follows: The motion vectors of the cross-sectional contour pixels in the dynamic image are tracked between adjacent frames. By identifying the displacement changes of the contour pixels as the compression process progresses, the deformation gradient change rate of the contour set composed of a preset number of discrete pixel sampling points over time is calculated. Ten detection positions are selected on average within the lateral inner diameter range of the protective tube under test. The data collected at each position is normalized, and the vertical pixel distance changes at multiple points are monitored in real time. A multi-point redundancy judgment logic is constructed. If the proportion of points where the vertical pixel distance is reduced to zero reaches 80%, the cross-sectional contour is determined to have undergone a closed jump, confirming the flattening test state. The mechanical response characteristic evaluation process includes: by recording the movement distance and force changes during the downward pressing process of the upper pressure plate in real time, and combining the acquired displacement data sequence and force data sequence, the compressive stiffness response of the protective tube under test at different deformation stages is calculated. The numerical dispersion of the deformation gradient change rate between discrete pixel sampling points is identified. By analyzing the numerical dispersion, the quantitative characteristics of the asymmetric deformation of the protective tube under test due to uneven material density and wall thickness tolerance are determined. The data processing and environmental adaptation process includes: analyzing the histogram brightness distribution of the dynamic image; dynamically adjusting the luminous power of the lighting lamps extending into the tube cavity to maintain the preset edge contrast based on the light attenuation after the tube cavity is compressed; after determining that the flattening test state has been reached, extracting the locked final displacement value and peak pressure data; combining the nominal outer diameter and material characteristics to calculate the initial inner diameter, and the preset tube sample length, calculating the ring stiffness value reflecting the deformation resistance of the protective tube under test; and fusing the defect image features detected by vision with the mechanical calculation results to generate a comprehensive test report including structural damage location and mechanical performance indicators.

[0012] Secondly, the present invention provides a circumferential detection system for surface defects of flexible cables based on multi-dimensional visual fusion, comprising: The feeding platform, upper pressure plate, vision assembly, lighting, motor, and pressure sensor; The detection system also includes a memory and a processor. The memory stores computer instructions, and the processor executes the following operations when executing the computer instructions: controlling the feeding platform to perform adaptive centering based on the material characteristics and nominal outer diameter of the protective tube under test; driving the upper pressure plate to move down and pausing to establish a zero point when the pressure data reaches a preset limit; controlling the vision component to perform full-circumferential scanning to construct an initial feature matrix, and identifying surface defects and deformation states through trajectory oscillation imaging and deformation gradient analysis during the pressing process; and locking the data and outputting a detection report when the cross-section is determined to be closed.

[0013] The beneficial effects of this invention are: This invention establishes an objective zero-point of pressure, eliminating initial placement errors and initial deformation, by analyzing the outer diameter parameters of the protective tube under test to drive a slider for adaptive physical alignment and monitoring minute contact forces for braking and hovering. After extracting a panoramic image of the unpressurized surface to construct an initial feature matrix, the system controls the vision component to perform trajectory oscillation-style imaging during synchronous pressing. Through pixel-level comparison of the dynamic image with the reference matrix, it extracts and locks the hidden micro-cracks and stress discoloration areas on the sidewall caused by pressure. Simultaneously, the system calculates the deformation gradient change rate by tracking the contour motion vector, quantitatively assesses the asymmetric deformation characteristics caused by material inhomogeneity, and combines adaptive supplementary lighting and multi-point redundant detection logic to eliminate the obstruction interference of localized debris, establishing the true cross-sectional closed jump state. This invention achieves simultaneous extraction and objective comprehensive evaluation of the microscopic damage and compressive physical properties of the flexible tube surface structure under continuous dynamic deformation conditions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 For the control method flowchart; Figure 2 Flowchart for S1 parameter reading and adaptive centering; Figure 3 Flowchart for constructing the S3 baseline feature matrix; Figure 4 Flowchart for S4 dynamic compression and visual defect recognition; Figure 5 This is a flowchart for S5 multi-point redundancy determination and comprehensive evaluation. Detailed Implementation

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0018] Secondly, the term "one embodiment" or "example" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the invention. The appearance of an embodiment in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0019] Example 1 Reference Figures 1-5 This is the first embodiment of the present invention, which provides a method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion, including the following steps: S1. Read the material characteristics and nominal outer diameter of the protective tube to be tested, and control the feeding platform to push the protective tube to be tested to the axis alignment position. The specific operation steps are as follows: The main control computer reads the identification information on the surface of the protective tube under test, and analyzes and obtains the outer diameter and wall thickness values ​​of the protective tube under test as nominal outer diameter and material characteristics. The adaptive centering slider at the bottom of the feeding platform is controlled to shrink synchronously towards the center according to the outer diameter value. The mechanical thrust generated by the adaptive centering slider pushes the protective tube under test to the preset center position, ensuring that the central axis of the protective tube under test is on the same horizontal and vertical plane as the center of the lens of the vision component behind it.

[0020] It should be noted that in the initial preparation stage of automated flattening testing, the system first needs to eliminate physical placement errors caused by external handling, and perform precise dimensional reading and spatial alignment calibration of the protective tube under test. The specific operation steps are as follows: Scanning and reading the identification information and parsing the dimensional parameters of the protective tube under test: When the protective tube under test is placed on the feeding platform of the testing equipment by the external handling mechanism, the main control computer calls the associated scanning device to scan and read the identification information code on the surface of the protective tube under test. The main control computer automatically decodes and parses the read information to extract the outer diameter and wall thickness values ​​of the protective tube at the time of manufacture. These two real physical values ​​are not only recorded as the nominal outer diameter and material characteristics of this batch of tests, but are also directly used to guide the subsequent mechanical alignment and calibration.

[0021] Adaptive physical alignment and mechanical locking based on real-world dimensional data feedback: Due to the weight of the pipe itself, equipment vibration, and the inertia of the robotic arm during gripping and releasing, the landing point of the protective pipe under test often deviates from the center of the feeding platform, resulting in lateral offset and angular tilt. To correct this deviation, the main control computer converts the acquired outer diameter value into specific displacement stroke commands and sends them to the drive mechanism at the bottom of the feeding platform. Adaptive alignment sliders are configured on both sides of the bottom of the feeding platform. Under the control of the drive mechanism, these two sliders synchronously and uniformly retract towards the center point of the feeding platform according to the stroke set in the command. During synchronous retraction, the adaptive alignment sliders on both sides will adhere to and clamp the outer wall of the protective pipe under test. Using the mechanical thrust generated by the continuous inward retraction of the sliders, the originally misaligned protective pipe under test is forcibly pushed back and fixed in the center of the feeding platform. When the slider travels to the preset coordinate that perfectly matches the read outer diameter value, the slider automatically stops retracting and maintains a mechanically locked state.

[0022] Precise establishment and error elimination of the physical reference coordinates for visual imaging: If the downward pressure test is performed directly under skewed conditions, the cross-section of the tube captured by the vision component installed at the rear of the equipment will suffer from perspective distortion, leading to inaccurate subsequent pixel feature calculations. Through the aforementioned automated mechanical calibration action based on feedback from real-dimensional data, the system achieves positioning calibration in three-dimensional physical space, offsetting the initial material placement error and ensuring, geometrically, that the central axis of the protective tube under test is on the same horizontal and vertical plane as the center of the lens of the vision component installed at the rear of the equipment.

[0023] S2. Drive the upper pressure plate downwards, establish the zero pressure point by collecting force data, and lock the spatial geometric relationship between the upper pressure plate and the protective tube under test. The specific operation steps are as follows: The control motor drives the upper pressure plate to move downward at a constant speed along the vertical axis, and the force sensor at the rear end of the upper pressure plate monitors the force data. If the force data reaches the preset initial contact force value limit, it is determined that the lower surface of the upper pressure plate has completed contact with the surface of the protective tube to be tested, and a braking command is sent to the motor to keep the upper pressure plate in a stationary state and lock the current mechanical position as the zero point of the flattening test.

[0024] It should be noted that after the pipes are precisely aligned in physical position, the system needs to establish the starting benchmark for the test through force sensing monitoring to prevent excessive pressure from the upper pressure plate, which could cause initial deformation of the flexible pipes before the formal test. The specific operating steps are as follows: High-frequency pressure monitoring during the constant-speed motor drive and pressing process: After the test tube completes the centering operation, the main control computer controls the top motor to start, driving the upper pressure plate to move downwards at a constant speed along the vertical axis, gradually approaching the upper surface of the test tube. During the pressing process, in order to accurately capture the force jump at the moment of contact, the system continuously collects force data at a set high-frequency sampling rate through a pressure sensor installed at the rear end of the upper pressure plate, and monitors the dynamic changes of the pressure value in real time.

[0025] Setting the Minimal Contact Force Limit and Deformation-Preventing Physical Adhesion Judgment: The compressive stiffness of flexible pipes is quite sensitive in the initial stage of stress. Even a small amount of excess mechanical pressure can cause initial deformation of the pipe wall, thus affecting the objectivity and accuracy of the test benchmark. Therefore, the system sets the preset initial contact force limit as a tiny force threshold (this threshold is typically set to a minimum value that only filters out background noise interference from the sensor system). When the system detects that the pressure data reaches this initial contact force limit, the main control computer determines that the lower surface of the upper pressure plate has made physical contact with the pipe wall of the protective pipe under test, confirming that both are in a non-destructive adhesion state.

[0026] Absolute locking of braking hovering and test space geometry: At the instant the fit is determined to be complete, to avoid continuous pressure deformation of the pipe, the main control computer immediately sends a braking command to the motor, controlling the upper pressure plate to stop moving downwards and maintain a stationary hovering state. This hovering action records and locks the current vertical mechanical height as the zero point of pressure for calculating displacement deformation in subsequent flattening tests; at the same time, this operation makes the upper pressure plate, the protective pipe under test, and the vision components form a relatively static and fixed geometrically locked state in three-dimensional space, effectively eliminating measurement hysteresis errors that may be caused by the initial mechanical clearance of the system, and providing a stable physical space reference for subsequent circumferential visual scanning.

[0027] S3. Perform an initial full-circumferential rotation scan to obtain a panoramic view of the surface texture of the protective tube under test in an unpressurized state, and unfold the outer circumference of the protective tube under test into an initial feature matrix of a plane. The specific operation steps are as follows: After the upper pressure plate remains stationary, the vision component automatically slides along the vertical guide rail to an observation height level with the cavity of the protective tube under test, based on the current vertical height coordinates of the upper pressure plate. It then controls the illumination lamp to rotate and extend into the front opening of the protective tube under test to turn on the light source, acquiring the full circumferential texture information of the surface of the protective tube under test. Through image stitching processing, the outer circumferential surface of the protective tube under test is unfolded to construct an initial feature matrix representing the original morphological features.

[0028] It should be noted that after establishing the zero-pressure point of the test, the system needs to provide an objective and accurate comparison benchmark for subsequent defect identification. The specific operation steps are as follows: Camera height adaptive adjustment and internal penetrating illumination construction: After the upper pressure plate comes to a stable stop, the main control computer obtains the current vertical height coordinates of the upper pressure plate. Based on these height coordinates, the control system drives the vision component to slide up and down along the vertical guide rail on the back, stopping at an observation height that allows for a level view of the internal cavity of the protective tube under test. After confirming that the camera is in place, the system controls the front lighting robotic arm to rotate, extending the light head into the front opening of the protective tube under test, and simultaneously turning on the light source. The internal light source works in conjunction with the external light to achieve uniform illumination of the inner and outer walls of the tube, eliminating blind spots in visual monitoring.

[0029] Full-circumferential rotational scanning and acquisition of baseline panoramic images: After the lighting environment is established, the system drives the vision component to perform a circumferential scan around the protective tube under test. During the circular motion, the vision component comprehensively acquires the texture features and color information of different areas on the outer surface of the tube, thereby obtaining the complete surface features of the tube in an unstressed state.

[0030] Image unfolding and stitching, and establishment of the initial feature matrix: The system stitches together multiple captured images of the local cylindrical surface into a continuous two-dimensional unfolded image through edge feature matching and projection mapping. The core purpose of constructing this unfolded image is that industrial pipes typically have mold scratches, color difference spots, and normal processing textures on their surfaces when they leave the factory. If these initial features are not recorded before testing, the system is very likely to misjudge them as structural cracks caused by pressure during subsequent pressure tests. Therefore, extracting the initial feature matrix under unpressured conditions establishes a comparative reference for subsequent defect judgment.

[0031] S4. Control the upper pressure plate to continue pressing down, and at the same time, perform trajectory oscillation imaging through the vision component to capture the characteristic changes caused by the pressure on the protective tube under test, and identify surface defects by comparing with the initial feature matrix. The specific operation steps are as follows: The upper pressure plate is controlled to continue pressing down at a preset speed. Simultaneously, a vision component is used to capture dynamic images of the deformation inside the cavity of the protective tube under test at a fixed frame rate. The inner wall contour pixel set of each frame in the dynamic image is extracted, and the upper and lower inner wall edges of the protective tube under test are located by establishing an image coordinate system. The height difference between the upper and lower inner wall edges in the vertical direction is calculated to generate a vertical pixel distance sequence reflecting the degree of tube deformation. The trajectory oscillation shooting and defect recognition process specifically includes: controlling the vision component to perform a reciprocating oscillation motion of the viewing angle during vertical pressing, capturing the texture anomalies of the sidewall of the protective tube under test by changing the light angle; projecting the dynamic image onto the local space corresponding to the initial feature matrix, calculating the deviation value between the two in pixel color grayscale and structural integrity; if the deviation value exceeds a preset safety threshold, the coordinates of the surface defect are automatically recorded, and the pressure data at this time is extracted simultaneously for stress defect correlation processing.

[0032] It should be noted that this step is the core of achieving dynamic latent defect identification. The system not only needs to monitor the degree of pressure deformation of the pipe in real time, but also needs to identify micro-cracks generated on the sidewall with high sensitivity. The specific operation steps are as follows: Dynamic synchronous compression and real-time extraction of internal contour deformation features: The main control computer controls the upper pressure plate to restart at a pre-set test speed, continuing to compress the protective tube under test downwards. During the downward compression process, the vision component continuously captures dynamic images of the tube's internal deformation at a set high fixed frame rate. The system removes the background from the images in real time, extracting a continuous set of pixels representing the upper and lower inner walls. To accurately quantify the degree of tube flattening, the system targets the selected... For each detection location, the vertical height difference between the bottommost pixel row of the upper inner wall and the topmost pixel row of the lower inner wall is calculated using the following formula:

[0033] : indicates the first The vertical pixel distance at each detection position (i.e., the remaining gap height in the cavity). Its value range is 0 ≤ ≤ ,in The maximum vertical resolution of industrial camera sensors (e.g., in high-definition images), The value can be either 1080 or 2160.

[0034] : Indicates the number of the selected detection location. The value range is 1 to 1. The positive integer, in this embodiment the detection points are evenly divided horizontally, therefore =10.

[0035] : indicates the first At each detection location, the coordinates of the lowest pixel row in the upper inner wall contour pixel set extracted by the system on the vertical axis of the image. The value ranges from 0 to... < .

[0036] : indicates the first At each detection location, the coordinates of the topmost pixel row in the pixel set of the lower inner wall contour extracted by the system on the vertical axis of the image. The value ranges from 0 to... < .

[0037] By comparing the vertical height difference between the bottommost pixel row of the upper inner wall and the topmost pixel row of the lower inner wall, the system continuously outputs a vertical pixel distance sequence that reflects the real-time deformation of the pipe.

[0038] Trajectory-based oscillating imaging and the exposure of latent defects through light and shadow: In the dynamic compression stage, the system introduces a trajectory-based oscillating imaging mechanism. When flexible polymer pipes are subjected to extreme compression, the areas of highest stress concentration on their sidewalls often exhibit microscopic stress whitening and minute tensile cracks. These early latent defects are difficult to effectively extract under single, fixed-angle lighting conditions. During imaging, the vision component performs high-frequency reciprocating oscillations within a small space, continuously changing the refraction and reflection angles of light on the pipe surface. Once the originally smooth pipe wall develops rough surfaces due to stress tearing, it will exhibit high-contrast abnormal bright spots and shadow lines under specific reflection angles.

[0039] Dynamic image projection comparison and precise binding of stress defects: While performing vibration imaging, the system aligns the real-time captured local images with the initial feature matrix constructed in step S3 using coordinates and performs pixel subtraction. Through this comparison process, the system filters out the original processing marks on the pipe and extracts and locks newly added stress discoloration areas and structural cracks. Once the deviation value exceeds the preset safety threshold, the system immediately records the coordinates of the defect location and retrieves the stress data transmitted back by the pressure sensor at the same time stamp. This operation synchronously maps and binds the structural defect with the critical pressure value that caused the defect in spatiotemporal data, achieving highly sensitive online defect detection.

[0040] S5. Evaluate the mechanical response characteristics of the protective tube under test based on the deformation gradient change rate of the cross-sectional profile. When the deformation gradient change rate reaches the preset threshold and the cross-sectional profile undergoes a closed jump, it is determined that the flattening test state has been reached and a test report is output. The specific operation steps are as follows: The system tracks the motion vectors of cross-sectional contour pixels in dynamic images between adjacent frames. By identifying the displacement changes of contour pixels as the compression process progresses, it calculates the deformation gradient rate of the contour set composed of a preset number of discrete pixel sampling points over time. Ten detection positions are selected on average within the transverse inner diameter range of the protective tube under test. Data collected at each position is normalized, and the vertical pixel distance changes at multiple points are monitored in real time. A multi-point redundancy judgment logic is constructed. If the proportion of points where the vertical pixel distance shrinks to zero reaches 80%, the cross-sectional contour is determined to have undergone a closed jump, confirming the flattening test state. The mechanical response characteristic evaluation process includes: by recording the movement distance and force changes during the downward pressing process of the upper pressure plate in real time, and combining the acquired displacement data sequence and force data sequence, the compressive stiffness response of the protective tube under test at different deformation stages is calculated. The numerical dispersion of the deformation gradient rate of change between discrete pixel sampling points is identified. By analyzing the numerical dispersion, the quantitative characteristics of the asymmetric deformation of the protective tube under test due to uneven material density and wall thickness tolerance are determined. The data processing and environmental adaptation process includes: analyzing the histogram brightness distribution of the dynamic image; dynamically adjusting the luminous power of the lamps inserted into the tube to maintain the preset edge contrast based on the light attenuation after pressure inside the tube; extracting the locked final displacement value and peak pressure data after determining that the flattening test state has been reached; calculating the ring stiffness value reflecting the deformation resistance of the protective tube under test by combining the nominal outer diameter and material characteristics with the preset tube sample length; and integrating the defect image features detected by vision with the mechanical calculation results to generate a comprehensive test report that includes structural damage location and mechanical performance indicators.

[0041] It should be noted that as the upper pressure plate continues to move downwards, the pipe approaches its limit deformation. The system needs to intelligently determine the material uniformity and output a braking control signal in a timely manner when the cross-sectional profile is completely closed. The specific operating steps are as follows: Motion vector tracking and quantitative evaluation of asymmetric deformation: The system not only evaluates the overall compression of the cross-section, but also tracks the rate of deformation (i.e., the deformation gradient rate) of multiple discrete pixels on the cross-sectional contour in the image. Specifically, for the j-th discrete pixel sampling point, its deformation gradient rate... The calculation formula is as follows:

[0042] : indicates the first The deformation gradient change rate (physically equivalent to the acceleration of deformation at that point) of each discrete pixel sampling point. The value range is... ≥0. When this value shows an abnormal peak, it indicates that local asymmetric deformation has occurred or that a rupture is imminent.

[0043] : Indicates the number of the selected discrete pixel sampling point. It is a positive integer greater than 1.

[0044] : indicates the first The number of pixels at the current moment The lateral (or radial) displacement coordinate values.

[0045] and : These represent the position of the pixel in the previous frame ( (moment) and the first two frames ( The displacement coordinate values ​​extracted at time (time).

[0046] This represents the time interval between two consecutive frames of motion captured by the vision component. The value range is usually determined by the camera's frame rate; for example, when the frame rate is 60 frames per second... It takes approximately 0.016 seconds.

[0047] For pipes of uniform material, the deformation rate in all directions should remain smooth and symmetrical when subjected to pressure. The system compares the differences in deformation rate (i.e., numerical dispersion) between sampling points. If a significant deviation in deformation rate is detected in different directions, it is determined that the protective pipe under test has uneven thickness or local density defects, and the quantitative characteristics of asymmetric deformation are output simultaneously.

[0048] Closed-loop adaptive adjustment of internal lighting environment: In the final stage of flattening, as the upper and lower inner walls of the tube move closer together, the light transmission channels inside the tube are gradually blocked, causing the captured image to darken rapidly. The system analyzes the histogram of brightness and darkness in the image to perceive this light attenuation trend in real time and actively increases the luminous power of the lamps inserted into the tube. This adaptive lighting mechanism ensures that the contrast of the upper and lower inner wall edges remains within a preset range of clarity even under the highly compressed state of the tube.

[0049] Multi-point redundancy anti-interference mechanism and closure jump determination: When determining whether the pipe has reached a completely flattened state, relying solely on the numerical jump of the pressure sensor and the overlap of a single visual point is highly susceptible to interference from local debris inside the pipe wall that may obstruct the lens, leading to system misjudgment and premature termination of the test. Therefore, the system evenly distributes multiple (e.g., ten) independently monitored detection points along the transverse direction of the pipe cavity. This multi-point redundancy determination logic satisfies the following mathematical formula:

[0050] This indicates the total number of detection locations selected in the image. The value range is... Positive integers greater than 1, in this embodiment of the invention =10.

[0051] : Indicates the threshold number of qualifying points required to determine the flattened state. The value range is 1 < ≤ Positive integers, in this embodiment of the invention, are set to account for 80%, that is... =8.

[0052] : indicates the first A state determination function for each detection location. It is a Boolean value (0 or 1). When the vertical pixel distance at that location... ( When the pixel error tolerance value is extremely small (typically 0 or 1 pixel), it is determined that the point has been touched. =1; when season =0.

[0053] This indicates the total number of valid contact points on the upper and lower inner walls of the current screen.

[0054] : Indicates the final system judgment result. When the above inequality holds, the system judges that the cross-sectional profile has undergone a closed jump, confirming that the flattening test state has been reached.

[0055] Only when more than 80% of the points (e.g., eight) simultaneously detect that the pixel distance between the upper and lower inner walls has returned to zero, does the system recognize that the pipe has undergone real large-area inner wall adhesion and confirm that the deformation data has generated a closed jump characteristic.

[0056] The closed-loop output of multi-dimensional data fusion and comprehensive testing report: Once the closed transition is confirmed, the main control computer immediately sends a highest-priority braking command to lock the final downward displacement value and ultimate bearing peak pressure of the pressure plate. Subsequently, the pressure plate is raised, and the lamp arm and vision component perform a reset action. The system backend integrates the recorded displacement data, pressure data, and physical parameters such as the initial inner diameter and preset cutting length of the pipe to calculate the true ring stiffness value of the pipe. The physical calculation model for this ring stiffness value satisfies the following formula:

[0057] This indicates the calculated ring stiffness value of the pipe, usually expressed in kN / m², and its range is [not specified]. >0.

[0058] This indicates the peak force data locked by the system when a specified deformation rate is reached (e.g., the pipe diameter is flattened by 3%). The value range is... >0.

[0059] This represents the actual vertical displacement of the upper pressure plate (i.e., the absolute deformation of the pipe). The value range is 0 < < .

[0060] This indicates the standard cut length of the protective tube sample to be tested, placed on the feeding platform. The value range is... >0.

[0061] This indicates the initial inner diameter of the protective tube to be tested. This value is calculated by subtracting twice the wall thickness from the outer diameter value read in system step S1. The value range is... >0.

[0062] The system summarizes the surface microcrack image features extracted throughout the process, the asymmetric deformation quantification data, and the calculated ring stiffness values ​​to generate a comprehensive mechanical and visual inspection report.

[0063] Example 2 This embodiment provides a full-circumferential inspection system for surface defects of flexible cables based on multi-dimensional vision fusion. The inspection system includes the following hardware physical architecture: feeding table, upper pressure plate, vision component, lighting lamp, motor and pressure sensor.

[0064] The detection system also includes a memory and a processor. The memory stores computer instructions, and the processor executes the following operations when executing the computer instructions: controlling the feeding table to perform adaptive centering based on the material characteristics and nominal outer diameter of the protective tube under test; driving the upper pressure plate to move down and pausing to establish the zero point when the pressure data reaches the preset limit; controlling the vision component to perform full-circumferential scanning to construct the initial feature matrix, and identifying surface defects and deformation states through trajectory oscillation imaging and deformation gradient analysis during the pressing process; and locking the data and outputting the detection report when the cross-section is determined to be closed.

[0065] Example 3 To further clarify the technical solution of the present invention, the following is a third embodiment of the present invention, using a specific scenario of detecting hidden quality defects in polymer flexible pipes on a high-end manufacturing production line as an example, to provide a more detailed description of the technical solution of the present invention.

[0066] In this embodiment, the multi-dimensional vision fusion-based circumferential surface defect detection system for flexible cables of this invention is deployed in the quality inspection stage of a new energy vehicle wiring harness manufacturing plant. One day, the production line was conducting random inspections on a batch of corrugated flexible tubes used to protect high-voltage wiring harnesses. One of the tubes had uneven wall thickness and microscopic stress concentration on its sidewalls due to fluctuations in the extrusion molding process; however, these defects were undetectable by the naked eye and conventional inspection equipment under normal conditions. The tube was randomly placed on the unloading platform of the inspection system by a robotic arm.

[0067] Step S1: Accurate reading of actual size parameters and precise execution of physical alignment When the protective tube under test fell onto the feeding platform, due to the inertia of the robotic arm, the tube's landing point deviated from the centerline by approximately 5 millimeters and exhibited a slight tilt. The system's main control computer immediately invoked the barcode scanner to read the identification information code on the tube's surface, deciphering it to have a nominal outer diameter of 40 millimeters and a wall thickness of 2 millimeters. The system then converted this 40-millimeter outer diameter value into a displacement stroke command. Upon receiving the command, the adaptive centering sliders on both sides of the bottom of the feeding platform synchronously and uniformly contracted towards the center. During the contraction process, the sliders contacted and clamped the tilted tube wall, using mechanical thrust to force the tube back to the center. When the distance between the two sliders precisely reached 40 millimeters, automatic mechanical locking occurred, completely correcting the 5-millimeter placement offset and ensuring absolute alignment between the tube's central axis and the center of the rear vision component's lens.

[0068] Step S2: Sensitive capture of minute contact forces and locking of spatial reference After centering, the top motor drives the upper pressure plate to move downwards at a constant speed. During this process, the pressure sensor at the rear of the upper pressure plate continuously samples at a high frequency of 1000Hz. To prevent the flexible tube, with a wall thickness of only 2 mm, from being prematurely crushed, the system sets the initial contact force limit to a very small 0.5 Newtons (only slightly higher than the sensor system's noise floor). When the pressure reaches a certain coordinate, the sensor collects a force feedback of 0.55 Newtons, and the system determines that the metal bottom surface of the upper pressure plate has achieved physical contact with the top of the tube. The main control computer immediately sends a braking command to control the upper pressure plate to hover. At this point, the mechanical height is precisely locked by the system as the zero point of pressure for displacement calculation, and the geometric reference in the three-dimensional detection space is established.

[0069] Step S3: Construction of the baseline feature matrix and filing of initial surface features With the pressure plate hovering, the vision unit automatically slides down to the observation height of the tube cavity, and the illumination lamp extends into the opening at the front end of the tube to provide internal illumination. The vision unit performs a full circumferential scan around the tube. The system seamlessly stitches the captured images to construct the initial feature matrix of the tube. In the generated planar unfolded image, the system extracts and records a very shallow scratch on the tube surface caused by the extrusion die. This record establishes an objective comparison benchmark for subsequent defect determination, effectively avoiding misjudging the initial factory feature as a newly added crack under pressure.

[0070] Step S4: Trajectory Oscillation Photography and Latent Defect Location under Dynamic Pressure The upper pressure plate continues to press down at a preset speed of 5 millimeters per second. The vision component acquires dynamic images of the inside of the tube at a rate of 60 frames per second and outputs the vertical pixel distance sequence in real time. As the deformation of the tube under pressure increases, the stress on the sidewall of the tube continues to rise. During the capture, the vision component performs high-frequency reciprocating oscillation motion with a small amplitude, changing the angle of light refraction. When the pressure plate moves down to the point where the tube diameter is compressed by about 30%, the system extracts an abnormally high-contrast reflective area on the right sidewall of the tube at a specific reflection angle, which appears as stress whitening and micro-tension cracks. The system performs pixel-level subtraction comparison between this real-time image and the initial feature matrix, eliminating the initial shallow scratches recorded in step S3 and confirming the presence of a new structural fracture on the right sidewall. The system records the coordinates of the crack and simultaneously extracts the pressure data of 120 Newtons at this time, realizing the identification of latent defects and spatiotemporal data binding.

[0071] Step S5: In-depth evaluation of mechanical response and multi-point establishment of closed state During continuous pressure application to the pipe, the system tracks the motion vectors of the contour pixels. Calculations show that the deformation gradient rate of the right side of the pipe contour, i.e., the radial expansion rate, is 15% higher than that of the left side. After analyzing the dispersion of this value, the system determines that the pipe material exhibits an asymmetric deformation defect due to a thinner wall on one side. As the upper and lower inner walls gradually approach each other, the light transmittance inside the pipe cavity decreases, and the system adaptively increases the internal illumination power to maintain the clarity of edge features. Among the 10 detection points arranged laterally in the pipe cavity, the system detects in real-time that internal debris is obscuring one point. The system continues monitoring based on multi-point redundancy logic until the vertical pixel distance of the remaining 8 points synchronously reduces to zero, meeting the 80% threshold. The system confirms a closed transition in the cross-sectional contour and immediately triggers a braking command, locking the ultimate peak pressure at 350 Newtons. Substituting the above parameters along with the initial inner diameter of 36 mm into the physical calculation model, the system calculates the actual ring stiffness of the pipe to be 6.8 kN / m², lower than the preset acceptable standard. Ultimately, the system generated a comprehensive inspection report including anomaly warnings. The report contained substandard ring stiffness values, image features of the hidden microcracks on the right side, and a quantitative assessment of uneven wall thickness. This inspection report guided the production line's rejection device to precisely remove the defective pipe from the good-product queue.

[0072] In summary, this invention establishes an objective zero-point of pressure that eliminates initial placement errors and initial deformation by analyzing the outer diameter parameters of the protective tube under test to drive the slider to perform adaptive physical alignment and monitoring minute contact forces for braking and hovering. After extracting a panoramic image of the unpressurized surface to construct an initial feature matrix, the system controls the vision component to perform trajectory oscillation-style imaging during synchronous pressing. Through pixel-level comparison of the dynamic image with the reference matrix, the system extracts and locks the hidden micro-cracks and stress discoloration areas on the sidewalls caused by pressure. Simultaneously, the system calculates the deformation gradient change rate by tracking the contour motion vector, quantitatively assesses the asymmetric deformation characteristics caused by material inhomogeneity, and combines adaptive supplementary lighting and multi-point redundant detection logic to eliminate the occlusion interference of localized debris, establishing the true cross-sectional closed jump state. This invention achieves the simultaneous extraction and objective comprehensive evaluation of microscopic damage and compressive physical properties of the surface structure of flexible tubes under continuous dynamic deformation conditions.

[0073] It should be noted that the above 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion, characterized in that, Includes the following steps: S1. Read the material characteristics and nominal outer diameter of the protective tube to be tested, and control the feeding platform to push the protective tube to be tested to the axis centering position; S2. Drive the upper pressure plate to move downward, establish the zero point of pressure by collecting force data, and lock the spatial geometric relationship between the upper pressure plate and the protective tube to be tested; S3. Perform an initial full-circumferential rotational scan to obtain a panoramic view of the surface texture of the protective tube under test in an unpressurized state, and unfold the outer circumference of the protective tube under test into an initial feature matrix of a plane; S4. Control the upper pressure plate to continue pressing down, and at the same time perform trajectory oscillation shooting through the vision component to capture the characteristic changes caused by the pressure on the protective tube under test, and compare it with the initial feature matrix to identify surface defects; S5. Evaluate the mechanical response characteristics of the protective tube under test based on the deformation gradient change rate of the cross-sectional profile. When the deformation gradient change rate reaches a preset threshold and the cross-sectional profile undergoes a closed jump, it is determined that the flattening test state has been reached and a test report is output.

2. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 1, characterized in that, Step S1 specifically includes: The main control computer reads the identification information on the surface of the protective tube under test, and analyzes and obtains the outer diameter and wall thickness values ​​of the protective tube under test as the nominal outer diameter and material characteristics. The adaptive centering slider at the bottom of the feeding platform is controlled to shrink synchronously towards the center according to the outer diameter value. The mechanical thrust generated by the adaptive centering slider pushes the protective tube under test to the preset center position, ensuring that the central axis of the protective tube under test is on the same horizontal and vertical plane as the center of the lens of the vision component behind it.

3. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 1, characterized in that, Step S2 specifically includes: The control motor drives the upper pressure plate to move downward at a constant speed along the vertical axis, and the force sensor at the rear end of the upper pressure plate monitors the force data. If the force data reaches the preset initial contact force value limit, it is determined that the lower surface of the upper pressure plate has completed contact with the surface of the protective tube to be tested, and a braking command is sent to the motor to keep the upper pressure plate in a stationary state, and the current mechanical position is locked as the zero point of the flattening test.

4. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 1, characterized in that, Step S3 specifically includes: After the upper pressure plate remains stationary, the vision component is controlled to automatically slide along the vertical guide rail to an observation height level with the cavity of the protective tube under test, based on the current vertical height coordinates of the upper pressure plate. The illumination lamp is then controlled to rotate and extend into the front opening of the protective tube under test to turn on the light source, thereby acquiring the full circumferential texture information of the surface of the protective tube under test. The outer circumferential surface of the protective tube under test is then unfolded through image stitching processing to construct the initial feature matrix representing the original morphological features.

5. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 1, characterized in that, The specific process of step S4 includes: The upper pressure plate is controlled to continue pressing down at a preset speed. Simultaneously, the vision component is used to acquire dynamic images of the deformation inside the cavity of the protective pipe under test at a fixed frame rate. The inner wall contour pixel set of each frame in the dynamic image is extracted, and the upper inner wall edge line and the lower inner wall edge line of the protective pipe under test are located by establishing an image coordinate system. The height difference between the upper inner wall edge line and the lower inner wall edge line in the vertical direction is calculated to generate a vertical pixel distance sequence reflecting the degree of pipe deformation.

6. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 5, characterized in that, The trajectory oscillation-based imaging and defect identification process specifically includes: The vision component is controlled to perform a reciprocating oscillation motion of the viewing angle during vertical downward pressure, and the texture anomaly of the side wall of the protective tube under test is captured by changing the light receiving angle; the dynamic image is projected onto the local space corresponding to the initial feature matrix, and the deviation value between the two in terms of pixel color grayscale and structural integrity is calculated; if the deviation value exceeds the preset safety threshold, the coordinates of the surface defect are automatically recorded, and the pressure data at this time is extracted simultaneously for stress defect correlation processing.

7. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 5, characterized in that, The specific process of step S5 includes: The motion vectors of the cross-sectional contour pixels in the dynamic image are tracked between adjacent frames. By identifying the displacement changes of the contour pixels as the compression process progresses, the deformation gradient change rate of the contour set composed of a preset number of discrete pixel sampling points over time is calculated. Ten detection positions are selected on average within the lateral inner diameter range of the protective tube under test. The data collected at each position is normalized, and the vertical pixel distance changes at multiple points are monitored in real time. A multi-point redundancy judgment logic is constructed. If the proportion of points where the vertical pixel distance is reduced to zero reaches 80%, it is determined that the cross-sectional contour has undergone a closed jump, confirming the flattening test state.

8. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 7, characterized in that, The mechanical response characteristic evaluation process includes: By recording the moving distance and force changes of the upper pressure plate during the pressing process in real time, and combining the obtained displacement data sequence and force data sequence, the compressive stiffness response of the protective tube under test at different deformation stages is calculated, the numerical dispersion of the deformation gradient change rate between discrete pixel sampling points is identified, and the quantitative characteristics of the asymmetric deformation of the protective tube under test due to uneven material density and wall thickness tolerance are determined by analyzing the numerical dispersion.

9. The method for full-circumferential detection of surface defects in flexible cables based on multi-dimensional visual fusion as described in claim 8, characterized in that, The data processing and environment adaptation process includes: The histogram of the dynamic image is analyzed for light and dark distribution. Based on the light attenuation after the tube is compressed, the luminous power of the lamps extending into the tube is dynamically adjusted to maintain the preset edge contrast. After determining that the flattening test state has been reached, the locked final displacement value and peak pressure data are extracted. Combined with the nominal outer diameter and the initial inner diameter calculated from the material characteristics, as well as the preset tube sample length, the ring stiffness value reflecting the deformation resistance of the protective tube under test is calculated. The defect image features detected by vision and the mechanical calculation results are fused to generate a comprehensive test report that includes structural damage location and mechanical performance indicators.

10. A circumferential detection system for surface defects of flexible cables based on multi-dimensional visual fusion, used to implement the method described in claims 1 to 9, characterized in that, include: The feeding platform, upper pressure plate, vision assembly, lighting, motor, and pressure sensor; The detection system also includes a memory and a processor. The memory stores computer instructions, and the processor executes the computer instructions to perform the following operations: control the feeding platform to perform adaptive centering based on the material characteristics and nominal outer diameter of the protective tube under test; drive the upper pressure plate to move down and hover when the pressure data reaches a preset limit to establish a zero point; control the vision component to perform full-circumferential scanning to construct an initial feature matrix, and identify surface defects and deformation states through trajectory oscillation imaging and deformation gradient analysis during the pressing process. When the cross-section is determined to be closed, the data is locked and a test report is output.

Images