A glove size self-adaptive cutting method and system based on machine vision

Abstract

The present application relates to glove cutting processing technical field, specifically, the present application relates to a kind of glove size self-adapting cutting method and system based on machine vision, the front image of hand flat posture and bending posture is collected on calibration plate;Hand contour extraction algorithm and finger region segmentation technique are used to obtain finger edge point coordinate sequence;Finger root joint and fingertip center point positioning technology and finger palm midline fitting algorithm are used to generate finger palm midline curve in two postures respectively and calculate in-plane length;The ratio of bending in-plane length and flat in-plane length is used to determine the fingertip dynamic compensation coefficient;In glove cutting pattern, according to the fingertip dynamic compensation coefficient and material elastic correction factor, the fingertip offset increment is obtained, and the fingertip cutting line is independently offset compensated along the finger longitudinal direction;After compensation, the fingertip cutting line and the rest of the cutting line without adjustment form the final cutting path and execute cutting, to realize the adaptive accurate cutting of fingertip area.

Description

Technical Field

[0001] This invention relates to the field of glove cutting and processing technology, and more specifically, to a glove size adaptive cutting method and system based on machine vision. Background Technology

[0002] Glove cutting technology is an important technology. Currently, various vision- or sensor-based adaptive cutting solutions have emerged in customized glove production. Existing patent CN105615070A discloses a method for obtaining hand dimensions and manufacturing gloves. This method acquires front or back images of a consumer's hand, uses edge recognition technology to calculate dimensions such as finger joint width, finger length, palm length, and wrist width, and then manufactures glove pieces based on these dimensions. CN224089148U provides a glove cutting positioning mechanism that, through the cooperation of a feeding component and a cutting component, performs multiple positioning and guiding operations during the material transport and transfer process to accommodate materials of different widths. Furthermore, C... N201610256814.0 discloses an automatic glove cutting machine that uses a photoelectric sensor to determine the size of the raw material based on the brightness range of an LED light, and then a controller determines the cutting direction and position. However, the above-mentioned existing solutions only rely on a single-pose image of the hand in a flat, naturally extended state or the material size for cutting control, and fail to consider the difference in stretch allowance required by the bending movement of the fingers when the glove is worn after it is made. This leads to problems such as insufficient fit or excessive tightness at the fingertips, affecting wearing comfort and finger dexterity. To solve this technical problem, we provide a machine vision-based glove size adaptive cutting method and system. Summary of the Invention

[0003] The purpose of this invention is to provide a machine vision-based glove size adaptive cutting method and system to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, one objective of this invention is to provide a machine vision-based adaptive glove cutting method, comprising the following steps: S1. Sequentially acquire frontal images of the same hand in a flat position and a bent position on the calibration plate. The flat position is when each finger is naturally extended, and the bent position is when each finger is bent to its maximum natural bending state while keeping the palm position unchanged. S2. Perform hand contour extraction algorithms based on skin color model and edge detection on the flat frontal image and the curved frontal image respectively to obtain the flat hand contour and the curved hand contour. Then, segment each finger region from the flat hand contour and the curved hand contour to obtain the finger edge point coordinate sequence. S3. The center point of the finger root joint and the center point of the finger tip are determined by the finger root joint and the center point of the finger tip. Combined with the finger edge point coordinate sequence, the finger pad midline fitting algorithm is used to generate the finger pad midline curve in the flat posture and the finger pad midline curve in the bent posture, respectively. The length of the corresponding curve is calculated as the length in the flat surface and the length in the bent surface. S4. For each finger, the ratio of the length in the curved surface to the length in the flat surface is determined as the fingertip dynamic compensation coefficient for each finger. S5. In the glove pattern, for the fingertip area cutting line of each finger, according to the fingertip dynamic compensation coefficient and the pre-calibrated material elasticity correction factor, the fingertip offset increment is obtained through the fingertip offset increment calculation model. The fingertip cutting line is translated outward along the longitudinal direction of the finger by the fingertip offset increment to generate the compensated fingertip cutting line. The other cutting lines in the glove pattern, except for the fingertip area cutting line, remain in their original positions. S6. The final cutting path is formed by the compensated fingertip cutting line and the remaining unadjusted cutting lines, and the cutting tool is controlled to cut the glove piece.

[0005] The second objective of this invention is to provide a system for implementing a machine vision-based adaptive glove size cutting method as described in any one of the above-mentioned methods, comprising: The image acquisition and calibration unit is used to sequentially acquire frontal images of the same hand in a flat posture and a bent posture on the calibration plate. The hand contour processing and feature calculation unit is used to extract contours, segment finger regions, locate the center points of the finger roots and fingertips, and fit the midline curve of the finger pads on the tiling and bending posture images acquired by the image acquisition and calibration unit, and calculate the fingertip dynamic compensation coefficient for each finger. The adaptive cutting path generation unit is used to calculate the fingertip offset increment of each finger in the glove cutting pattern according to the dynamic compensation coefficient of each fingertip and the pre-stored material elasticity correction factor, and translate the corresponding original fingertip cutting line to generate the compensated fingertip cutting line. The cutting path fusion and execution control unit is used to combine the compensated fingertip cutting line with the remaining unadjusted cutting lines into the final cutting path, and control the cutting tool to perform the cutting.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires dual images of the hand in both a flat and bent posture, extracting a dynamic compensation coefficient for the fingertip area that reflects the material stretching requirements under bent conditions. This improves upon traditional fixed-amount compensation methods by providing dynamic, customized compensation based on the user's individual finger bending characteristics. This effectively solves the problems of insufficient fit or excessive tightness at the fingertips caused by bending and stretching. Furthermore, the dynamic compensation coefficient is used in conjunction with a material elasticity correction factor pre-calibrated through material mechanics experiments to calculate the fingertip offset increment. This ensures the compensation amount adapts to the differences in hand bending among users and matches the elasticity characteristics of different glove fabrics, achieving dual self-adaptation based on both hand features and material properties. In addition, only the fingertip cutting line is independently offset along the longitudinal direction of the finger, while the remaining cutting lines maintain their original positions. This precise and focused compensation method improves fingertip fit and wearing comfort while maintaining the integrity and stability of the overall glove structure. Attached Figure Description

[0007] Figure 1 This is a flowchart illustrating the overall workflow of the present invention; Figure 2 This is a schematic diagram of the overall structure of the present invention; The meanings of the labels in the diagram are as follows: 1. Image acquisition and calibration unit; 2. Hand contour processing and feature calculation unit; 3. Adaptive cropping path generation unit; 4. Cropping path fusion and execution control unit. Detailed Implementation

[0008] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0009] Please see Figure 1 As shown, one of the objectives of this embodiment is to provide a machine vision-based adaptive glove cutting method, including the following steps: S1. Sequentially acquire frontal images of the same hand in a flat position and a bent position on the calibration plate. The flat position is when each finger is naturally extended, and the bent position is when each finger is bent to its maximum natural bending state while keeping the palm position unchanged. S2. Perform hand contour extraction algorithms based on skin color model and edge detection on the flat frontal image and the curved frontal image respectively to obtain the flat hand contour and the curved hand contour. Then, segment each finger region from the flat hand contour and the curved hand contour to obtain the finger edge point coordinate sequence. S3. The center point of the finger root joint and the center point of the finger tip are determined by the finger root joint and the center point of the finger tip. Combined with the finger edge point coordinate sequence, the finger pad midline fitting algorithm is used to generate the finger pad midline curve in the flat posture and the finger pad midline curve in the bent posture, respectively. The length of the corresponding curve is calculated as the length in the flat surface and the length in the bent surface. S4. For each finger, the ratio of the length in the curved surface to the length in the flat surface is determined as the fingertip dynamic compensation coefficient for each finger. S5. In the glove pattern, for the fingertip area cutting line of each finger, according to the fingertip dynamic compensation coefficient and the pre-calibrated material elasticity correction factor, the fingertip offset increment is obtained through the fingertip offset increment calculation model. The fingertip cutting line is translated outward along the longitudinal direction of the finger by the fingertip offset increment to generate the compensated fingertip cutting line. The other cutting lines in the glove pattern, except for the fingertip area cutting line, remain in their original positions. S6. The final cutting path is formed by the compensated fingertip cutting line and the remaining unadjusted cutting lines, and the cutting tool is controlled to cut the glove piece.

[0010] It needs further explanation that after the machine vision-based glove size adaptive cutting system completes initialization, the image acquisition and calibration unit starts a standardized acquisition process for dual-pose images of the hand. This process provides raw image data with a unified spatial reference for subsequent hand feature extraction.

[0011] The image acquisition and calibration unit uses a calibration plate printed with a grid coordinate system as the acquisition reference carrier. The grid coordinate system consists of orthogonal rigid grid lines uniformly printed on the surface of the calibration plate. The intersections of the grid lines form standard coordinate nodes. These coordinate nodes are used to establish the mapping relationship between image pixel coordinates and actual physical dimensions. The mathematical expression for this mapping relationship is as follows: In the formula This indicates the actual physical length on the calibration plate. This represents the calibration coefficient corresponding to the physical size of a single pixel. This indicates the number of pixels of a corresponding length in the image. The calibration board surface has preset contour position reference lines, which are divided into two categories: tiling posture reference lines and bending posture reference lines. Both types of reference lines are drawn based on the standard anatomical posture of the human hand and are used to constrain the hand's position to ensure complete consistency of the spatial reference for both acquisitions. When acquiring a tiling posture frontal image, the test subject keeps their hand in a naturally extended state, with the palm completely against the calibration board surface. Complete palm contact means that the palm plane and the calibration board plane have no gaps and no local lifting. The axes of each finger remain parallel to the tiling posture reference lines. The finger axis refers to the virtual line connecting the center point of the finger root joint to the center point of the fingertip. The wrist and the base of the palm are in contact with the preset fixed positioning area of ​​the calibration board, ensuring no translation or rotation of the hand. When acquiring a bending posture frontal image, the test subject keeps the wrist and the base of the palm completely fixed in the tiling posture position. Fixed state means that the contact position between the wrist and the base of the palm does not shift or deflect at any angle. Each finger is naturally bent to its ultimate comfortable position. The ultimate comfortable position means that the fingers, with the palm position fixed, naturally bend to their own... The flexor and extensor muscles are driven to the maximum physiological bending position without excessive muscle stretching or abnormal joint stress. This position represents the limit of natural finger bending without external forced force. The bent finger contour is aligned with the bending posture reference line on the calibration plate. The machine vision device is fixedly installed at a preset position directly above the calibration plate, with the device lens optical axis perpendicular to the calibration plate plane. The device's internal and external parameters have been pre-calibrated to eliminate imaging distortion and viewing angle deviation. After the hand is stable and without shaking in the corresponding posture, the machine vision device performs trigger shooting operations. During the shooting process, the calibration plate serves as a uniform background, and the grid coordinate system serves as a uniform spatial reference. Finally, a flat posture frontal image and a bent posture frontal image are acquired. The two types of images have consistent physical size mapping relationships and background features. After the image acquisition and calibration unit completes the dual-posture image acquisition, it transmits the image data to the hand contour processing and feature calculation unit in real time. Upon receiving the image data, the hand contour processing and feature calculation unit immediately performs preprocessing and hand contour extraction operations based on skin color model and edge detection on the two types of images.

[0012] After the image acquisition and calibration unit transmits the tiled frontal image and the bent frontal image with a unified spatial reference to the hand contour processing and feature calculation unit, the hand contour processing and feature calculation unit immediately initiates the image preprocessing and hand contour extraction process. This process provides accurate contour base data for subsequent finger region segmentation. The hand contour processing and feature calculation unit first performs preprocessing operations on the two types of images. The preprocessing includes three levels of processing: grayscale conversion, Gaussian filtering, and histogram equalization. Grayscale conversion converts the RGB color image into a single-channel grayscale image to remove redundant information in the color channels. Gaussian filtering uses a fixed-size convolution kernel to traverse the image pixels to suppress random Gaussian noise. Histogram equalization performs probability density normalization on the pixel grayscale values ​​of the grayscale image to stretch the grayscale distribution range. The three levels of processing work together to eliminate local brightness and darkness deviations in the image caused by uneven lighting in the acquisition environment. The problem is that this method uses the YCbCr color space as the carrier for skin color modeling. A color space is a numerical coordinate system that quantifies the color attributes of image pixels. The YCbCr color space decomposes RGB color information into independent luminance components (Y), blue chrominance components (Cb), and red chrominance components (Cr). The chrominance components Cb and Cr are unaffected by luminance changes, thus stably representing the core color features of skin color and adapting to skin color recognition needs under fluctuating lighting conditions. The skin color probability model is constructed based on the statistical analysis of the chrominance distribution of multiple sets of standard hand skin color samples in the YCbCr space. The construction process involves collecting hand skin color pixel samples under different lighting conditions, extracting the Cb and Cr chrominance values ​​of each sample pixel, statistically analyzing the two-dimensional joint distribution of Cb and Cr, and fitting the distribution characteristics using a double Gaussian mixture model to finally generate a standardized skin color probability model. The mathematical expression of the model is: , In the formula This represents the probability value of skin color for a single pixel. This indicates the sub-model number of the Gaussian mixture model. Indicates the first The weight coefficients of each sub-model Indicates the first The mean of the Cb components of each sub-model. This represents the mean of the Cr component in the k-th sub-model. Indicates the first The standard deviation of the Cb component of the sub-model Indicates the first The standard deviation of the Cr component in each sub-model Indicates the first The correlation coefficients between the Cb and Cr components in the sub-model Represents pi (π). The natural index calculation unit converts the pre-processed images from RGB space to YCbCr space, extracts the Cb and Cr chromaticity component values ​​of each pixel, and substitutes these component values ​​into the skin color probability model. The model calculates the skin color probability value for each pixel, iterates through all pixels in the image, and completes the skin color probability calculation for all pixels, calls a preset skin color probability threshold, and marks pixels with skin color probability values ​​higher than the threshold as skin color pixels. Consecutive distributions of skin color pixels form hand candidate regions, while pixels with skin color probability values ​​lower than the threshold are marked... To exclude background pixels from subsequent processing, the hand contour processing and feature calculation unit uses the Canny edge detection operator to extract edges from the candidate hand region. First, it calculates the horizontal and vertical gradient values ​​of pixels within the candidate region, synthesizes the gradient magnitude and direction, and then performs non-maximum suppression, retaining only pixels with the largest local gradient magnitude. Subsequently, it filters strong and weak edge pixels using high and low thresholds, connects strong edge pixels, and associates weak edge pixels connected to strong edges. Finally, it extracts all complete edges within the candidate hand region. Connectivity analysis is then performed on the extracted edge image; a connected component refers to the number of edge pixels that satisfy... The pixel set for the eight-neighbor connectivity rule is defined as follows: an eight-neighbor region refers to the eight adjacent pixel positions of a single pixel. All edge pixels are traversed, the pixel coordinates of each connected region are marked, and the area of ​​the connected region is calculated, which is the total number of pixels contained within the region. Then, a contour tracing operation is performed, traversing pixels clockwise along the outer edge of each connected region, recording the contour point coordinate sequence of each region, and simultaneously calculating the contour closure degree of each connected region as a contour integrity index. The contour integrity index and area values ​​of all connected regions are compared, and the connected region with the highest contour integrity and the largest area is selected. This connected region corresponds to a complete contour. For the hand contour region, a moving average smoothing algorithm is used to smooth and denoise the filtered connected component contour point set. For each target point in the contour point sequence and a fixed number of its neighboring points, the average horizontal and vertical coordinates of the neighboring points are calculated. The average value is used to replace the original coordinates of the target point. All contour points are processed in sequence to eliminate contour burrs and discrete noise points. After the above processing, the hand contour processing and feature calculation unit obtains the tiled hand contour corresponding to the tiled pose image and the curved hand contour corresponding to the curved pose image, respectively. Subsequently, finger region segmentation operation will be performed based on the two types of hand contours to obtain the edge point coordinate sequence of each finger.

[0013] After obtaining the smoothed and denoised tiled and curved hand contours, the hand contour processing and feature calculation unit immediately initiates the finger region segmentation and edge point coordinate sequence generation process. This process achieves independent segmentation of the five fingers through convex hull calculation and concave point recognition, providing basic data for subsequent localization of the finger root joint center point and fingertip center point. The convex hull refers to the smallest convex polygon containing all discrete contour points in the plane. No interior angle of the convex polygon is greater than 180 degrees. The convex hull clearly presents the outer shape of the hand contour and highlights the concave features between the fingers. The hand contour processing and feature calculation unit uses the Graham scan algorithm to calculate the convex hull of the hand contour. First, the point with the smallest ordinate is selected from the hand contour point set as the initial pole. If multiple points have the same smallest ordinate, the point with the smallest abscissa is selected. The remaining contour points are arranged in ascending order according to the polar angle with the pole. The polar angle is the angle between the line connecting the contour point and the pole and the positive direction of the horizontal axis. The mathematical expression for polar angle calculation is: In the formula Indicates the first The polar angle of each contour point and Indicates the first The x and y coordinates of the contour points and The x and y coordinates of the poles are represented by the vector coordinates. Then, the sorted contour points are traversed sequentially. The orientation relationship between three consecutive points is determined by the cross product of vectors. The expression for the cross product is: , In the formula , , This represents three consecutive contour points. If the cross product of these three points is greater than 0, the current point is retained; if the cross product is less than or equal to 0, the current point is discarded. This process is repeated until all points satisfy the convex polygon constraint, ultimately generating the convex hull polygon point set corresponding to the hand contour. After the convex hull calculation is completed, the hand contour processing and feature calculation unit traverses the hand contour point set and the convex hull polygon point set, calculating the Euclidean distance from each hand contour point to the corresponding edge of the convex hull polygon. The Euclidean distance expression is... In the formula Indicates the first The distance from each contour point to the convex hull. , , The linear equation parameters representing the convex hull are used. Points with distance values ​​greater than a preset concave distance threshold are identified as concave points. These concave points are distributed in the gaps between adjacent fingers. These concave points are marked as the starting points for finger gap segmentation. A total of four finger gap segmentation starting points are identified in the hand contour, corresponding to the gaps between the thumb and index finger, index finger and middle finger, middle finger and ring finger, and ring finger and little finger. The hand contour processing and feature calculation unit uses the four finger gap segmentation starting points to complete the segmentation of the five-finger region. The four finger gap segmentation starting points are traversed in a clockwise order according to the hand contour. Marking sequentially, starting from the first finger gap segmentation point as the initial boundary, tracing clockwise along the hand contour to the second finger gap segmentation point, extracting the contour point set of this segment to form the thumb region sub-contour. Starting from the second finger gap segmentation point as the initial boundary, tracing clockwise to the third finger gap segmentation point, extracting the contour point set of this segment to form the index finger region sub-contour. Starting from the third finger gap segmentation point as the initial boundary, tracing clockwise to the fourth finger gap segmentation point, extracting the contour point set of this segment to form the middle finger region sub-contour. Starting from the fourth finger gap segmentation point as the initial boundary... Starting from the boundary, the hand contour processing and feature calculation unit traces clockwise back to the third finger gap segmentation starting point, extracts the contour point set of this segment to form the sub-contour of the ring finger region, and uses the fourth finger gap segmentation starting point as the starting boundary, tracing clockwise back to the first finger gap segmentation starting point, extracting the contour point set of this segment to form the sub-contour of the little finger region. The five independent contour point sets correspond to the finger region sub-contours of the thumb, index finger, middle finger, ring finger, and little finger, respectively. For each finger region sub-contour point set, the hand contour processing and feature calculation unit first locates the point in the sub-contour that is farthest from the corresponding finger gap segmentation starting point as the fingertip point. Using the fingertip point as the sorting starting point, it synchronously traverses along the sub-contour towards the two finger gap segmentation starting points, and stores the contour points obtained by the traversal in sequence according to the direction from the fingertip to the finger root. After removing duplicate points, an ordered point set is formed. This ordered point set is the finger edge point coordinate sequence of each finger. After obtaining the finger edge point coordinate sequence of all fingers, the hand contour processing and feature calculation unit immediately uses the finger root joint and fingertip center point positioning technology to perform the positioning operation of the finger root joint center point and fingertip center point for each finger region.

[0014] After the hand contour processing and feature calculation unit obtains the ordered coordinate sequence of finger edge points for each finger, it initiates a precise localization process for the center point of the finger root joint and the center point of the fingertip for a single finger region. This process is based on the contour width gradient change and the geometric extension rules of the fingertip point, providing key reference points for subsequent fingertip midline curve fitting. First, the coordinate sequence of finger edge points for a single finger is split into the upper edge point sequence and the lower edge point sequence according to the natural contour direction. The upper and lower edges are divided by the longitudinal axis of the finger. The number of points in the two edge point sequences is consistent and they are matched one-to-one in the order from fingertip to finger root. When determining the center point of the finger root joint, the finger edge point coordinate sequence is traversed position by position from fingertip to finger root. For each traversed position, the upper edge point and the corresponding lower edge point are taken, and the straight-line distance between the two points is calculated as the finger contour width at that position. The mathematical expression for the finger contour width is: , In the formula Indicates the first finger The width of the outline at each traversed position. and Indicates the first The x and y coordinates of the upper edge points at each location and Indicates the first The horizontal and vertical coordinates of the lower edge points at each location are calculated. After the unit completes the full contour width calculation in the order from fingertip to finger root, a continuous width change sequence is generated. Then, the contour width difference between adjacent traversed positions is calculated. The mathematical expression for the width difference is: In the formula Indicates the first The position relative to the first The contour width difference at each position is used to traverse the width difference sequence. The first traversed position with a width difference greater than zero is retrieved. This position is the first point where the contour width begins to increase. A sudden increase in contour width occurs at the base of the finger joint; this point is determined as the center point of the base of the finger joint. The center point of the base of the finger joint is the geometric center of the connection between the finger and the palm, and is the starting reference point for calculating finger length. When determining the fingertip center point, a fixed range of edge point subsequences is extracted from the fingertip portion of the finger contour. All matching point pairs on the upper and lower sides within this subsequence are traversed, and the straight-line distance between each pair of points is calculated. The mathematical expression for the point pair distance is: , In the formula Indicates the fingertip subsequence number Distance between pairs of points and Indicates the first Coordinates of the upper edge point of the group and Indicates the first The coordinates of the lower edge points of the group are used to filter out the point pairs with the smallest distance values. The geometric center position of this point pair with the smallest distance is calculated, and the mathematical expression for the center position coordinates is: , In the formula and This represents the x and y coordinates of the center of the pair of points with the minimum distance. , , , The coordinates of the minimum distance point pair are represented by the upper and lower sides respectively. The tangent of the finger contour at the fingertip is fitted to determine the extension direction of the fingertip. The center position is moved along the extension direction by a preset distance. The preset distance is a fixed physical length marked according to the anatomical features of the human finger. The final endpoint after extension is the center point of the fingertip. The center point of the fingertip is the geometric reference point of the outermost end of the finger and is the termination reference point for the calculation of finger length. After the hand contour processing and feature calculation unit completes the positioning of the center points of the root joints and the center points of the fingertips of all fingers, it combines the coordinate sequence of the edge points of each finger and starts the fingertip midline fitting algorithm to generate the fingertip midline curves in the flat posture and the bent posture respectively.

[0015] After locating the center points of the finger root joints and fingertips, the hand contour processing and feature calculation unit initiates the fingertip midline curve fitting and in-plane length calculation process based on the acquired finger edge point coordinate sequence. This process achieves accurate midline construction through edge partitioning, proportional resampling, center point fitting, and curve smoothing, providing length benchmark data for calculating the fingertip dynamic compensation coefficient. The hand contour processing and feature calculation unit first distinguishes between the finger ventral and dorsal edge point sequences. The longitudinal axis of the finger is constructed by connecting the center points of the finger root joints and the fingertips. The finger edge point coordinate sequence is segmented around this longitudinal axis as the core benchmark. All edge points are traversed, and the normal projection orientation of each edge point to the longitudinal axis is calculated. Edge points whose projection orientation falls on the longitudinal axis towards the palm are uniformly collected as the finger ventral edge point sequence, and edge points whose projection orientation falls on the longitudinal axis towards the back of the hand are uniformly collected as the finger dorsal edge point sequence. The unit sorts both types of edge point sequences in an ordered manner according to a single direction from fingertip to finger root, ensuring the ventral and dorsal order. The traversal direction of the columns is completely consistent. After the edge sequence is distinguished, a proportional resampling operation based on the longitudinal length of the finger is performed. First, the total physical length of the finger's longitudinal axis is calculated. This length is converted into the actual physical length by the pixel physical mapping relationship of the calibration plate grid coordinate system, converting the pixel distance from the center point of the finger root joint to the center point of the fingertip. A fixed total number of resampling points is set, which is a preset value to ensure sampling accuracy. The total longitudinal length of the finger is divided into equal longitudinal segments according to the total number of resampling points. Each segment corresponds to the same physical length increment. Sampling points are extracted sequentially from the position of the equal longitudinal segments along the direction from the fingertip to the finger root for the finger ventral edge point sequence and the finger dorsal edge point sequence, so that the ventral sampling points and the dorsal sampling points are completely corresponding in the longitudinal position of the finger. Finally, a set of ventral resampling points and a set of dorsal resampling points with the same number of points and one-to-one matching longitudinal positions are obtained. For each set of ventral and dorsal resampling points corresponding to the longitudinal position, the geometric center coordinates of the two sets of sampling points are calculated. The mathematical expression for the geometric center coordinates is: In the formula Indicates the first The ventral resampling point and the first The coordinates of the geometric center point corresponding to each backside resampling point and Indicates the ventral first The x and y coordinates of each resampling point and Indicates the dorsal side The x and y coordinates of each resampling point and Using the corresponding sampling point numbers at the same longitudinal position, all geometric center points are sequentially connected by straight lines from fingertip to finger root, forming a preliminary fingertip midline polygon connecting the finger root joint center point and the fingertip center point. A cubic spline smoothing algorithm is applied to this preliminary fingertip midline polygon, using all geometric center points on the polygon as interpolation nodes to construct a continuous and smooth curve fitting function. The mathematical expression of the cubic spline smoothing function is as follows: In the formula The coordinate function representing the smoothed fingertip midline curve. This represents the normalized positional parameter along the longitudinal direction of the finger. , , , The coefficients represent the fitting coefficients of the cubic spline function. These coefficients are obtained by solving the coordinate constraint equations of the interpolation nodes. The smoothed curve has no abrupt changes in line and completely covers the path from the center point of the finger root joint to the center point of the fingertip, forming a standard fingertip midline curve. The actual length of the fingertip midline curve is calculated using numerical integration. The fingertip midline curve is discretized into several small arc segments, and the arc length is calculated segment by segment and then summed to obtain the total length. The mathematical expression for calculating the curve length is: In the formula This represents the in-plane length of the finger in a given pose. This represents the total number of discrete, minute arc segments of the curve. and Indicates the first The x and y coordinates of the curve at the end of the arc segment. and Indicates the first The horizontal and vertical coordinates of the curve starting from the arc segment are used to perform the above calculations on the fingertip midline curves in both the flat and bent postures to obtain the flat in-plane length and the bent in-plane length of a single finger. After the hand contour processing and feature calculation unit completes the solution of the in-plane length of a single finger, the flat in-plane length and the bent in-plane length of the thumb, index finger, middle finger, ring finger, and little finger are calculated in sequence according to the same process, providing all the basic data for the subsequent calculation of the fingertip dynamic compensation coefficient.

[0016] After the hand contour processing and feature calculation unit completes the calculation of the length in the flat surface and the length in the curved surface of the five fingers, it performs a division operation on the length in the curved surface and the length in the flat surface for each finger to obtain the fingertip dynamic compensation coefficient. Then, the fingertip dynamic compensation coefficients of all fingers are transmitted to the cutting path adaptive generation unit. This unit immediately starts the standardized calculation process of fingertip offset increment. This process integrates the glove pattern design parameters and the physical properties of the glove material to output the fingertip cutting offset value adapted to a single finger. The adaptive cutting path generation unit retrieves the system's built-in standard glove pattern database. This database stores the original fingertip cutting line segment parameters for each finger. The longitudinal design length of the original fingertip cutting line segment refers to the baseline design length along the finger's longitudinal direction in the standard glove pattern. This length is a fixed physical parameter in the glove's industrial pattern design, and its direction is completely consistent with the finger's longitudinal axis. The dynamic compensation coefficient of a single fingertip is multiplied by the longitudinal design length of the corresponding original fingertip cutting line segment to obtain the basic compensation amount. The basic compensation amount characterizes the theoretical compensation length of the fingertip without considering material properties. The mathematical expression for the basic compensation amount is: In the formula This indicates the basic compensation amount for a single fingertip. This represents the dynamic compensation coefficient for a single fingertip. This represents the longitudinal design length of the original fingertip cut line segment for a single finger. The material elasticity correction factor is a dimensionless coefficient pre-calibrated through a material mechanics tensile test. The experiment applies an axial load to the target glove material, consistent with the physiological tensile force of human finger bending, while simultaneously detecting three core physical indicators: the material's elastic modulus, material thickness, and warp and weft elongation. Substituting these three measured indicators into the material elongation characteristic fitting formula yields a correction factor that comprehensively reflects the material's actual elongation capacity. This factor is used to eliminate the influence of elasticity differences between different glove materials on the fingertip compensation accuracy. The basic compensation amount is multiplied by the material elasticity correction factor. This multiplication calculation model ultimately yields the fingertip offset increment of the fingertip cut line for a single finger. The fingertip offset increment is the final fingertip cut offset value adapted to the dynamic bending of the hand and the material elasticity. The mathematical expression for the fingertip offset increment is... In the formula This represents the offset increment of a single fingertip. This indicates the basic compensation amount for a single fingertip. The material elasticity correction factor is represented by the cutting path adaptive generation unit. Following the above calculation process, the unit sequentially solves the fingertip offset increments of the thumb, index finger, middle finger, ring finger, and little finger. After obtaining the fingertip offset increments of all fingers, the unit immediately starts the geometric translation reconstruction operation of the fingertip cutting line. The original fingertip cutting line is translated outward along the longitudinal direction of the finger to correspond to the fingertip offset increment, generating the compensated fingertip cutting line.

[0017] After the adaptive cutting path generation unit completes the calculation of the fingertip offset increment, it enters the geometric reconstruction process of the original fingertip cutting line. This process generates a compensated fingertip cutting line adapted to the dynamic characteristics of the hand through four steps: reference point identification, direction calibration, coordinate translation, and curve fitting. Throughout the process, the positions of the cutting lines on the glove piece, except for the fingertip area, remain unchanged. The adaptive cutting path generation unit first analyzes the geometric features of the original fingertip cutting line to be adjusted in the glove piece pattern. The original fingertip cutting line is the arc-shaped cutting line segment at the outermost end of the finger in the standard pattern. The line segment contains two endpoints connecting the cutting lines on the side of the finger and a continuous arc trajectory. The unit traverses all endpoint coordinates of the line segment and calculates the vertical distance from each endpoint to the corresponding fingertip midline curve fingertip segment. The mathematical expression for calculating the vertical distance is: In the formula It represents the vertical distance from the endpoint to the midline of the fingertip segment. , , The coefficients of the equation for the straight line running along the midline of the fingertip segment are represented. , The x and y coordinates of the original fingertip cut line endpoints are represented. The endpoint with the smallest vertical distance is selected; this endpoint is located at the extension of the finger's longitudinal central axis and is marked as the translation reference point. The translation reference point serves as the geometric datum for the fingertip cut line translation, ensuring that the translation trajectory perfectly coincides with the finger's central axis. The unit extracts the differential tangent vector of the fingertip midline curve at the fingertip center point. This differential tangent vector is obtained by solving for the points on the midline curve adjacent to the fingertip center point. The mathematical expression for calculating the tangent direction is: In the formula This represents the tangent direction vector at the fingertip. , This indicates the coordinates of the midline point in front of the center of the fingertip. , This represents the coordinates of the midline point behind the fingertip center point. The tangent direction is the longitudinal direction of the finger, pointing towards the outermost part of the finger and consistent with the natural extension of the fingertip midline curve. The unit moves the translation reference point along the longitudinal direction of the finger by a unit vector, the distance equal to the fingertip offset increment. The mathematical expression for calculating the new target point coordinates is: In the formula , Represents the x and y coordinates of the new target point. , Represents the x and y coordinates of the translation reference point. This represents the offset increment of a single fingertip. , Represents the horizontal and vertical components of the tangent direction vector. The modulus of the tangent direction vector is represented. The other endpoint of the original fingertip cutting line, except for the translation reference point, is a fixed endpoint, and its spatial position is not adjusted. The unit uses the new target point as the outer end reference point of the fingertip and the original fixed endpoint as the side connection point. It adopts a quadratic Bézier curve fitting algorithm to retain the curvature characteristics and connection angle of the original cutting line and reconstructs a continuous and smooth arc cutting trajectory. This reconstructed arc trajectory is the compensated fingertip cutting line. The cutting path adaptive generation unit completes the generation of the compensated fingertip cutting lines of the thumb, index finger, middle finger, ring finger, and little finger in sequence according to the above process. After the compensationd fingertip cutting lines of all fingers are constructed, the unit synchronously transmits the group of cutting lines and the remaining unadjusted cutting lines in the glove pattern to the cutting path fusion and execution control unit. After receiving the data, the cutting path fusion and execution control unit immediately starts the final cutting path fusion splicing and cutting execution control process.

[0018] After the cutting path fusion and execution control unit acquires all compensated fingertip cutting segments and the remaining unadjusted cutting segments in the glove pattern, it immediately initiates the integrated fusion and splicing process for the final digital cutting path. This process achieves seamless combination of line segments based on a unified coordinate system and topological connection rules, providing continuous closed trajectory data for subsequent cutting execution. First, a unified physical coordinate system for the glove pattern is constructed, completely consistent with the previous image acquisition and pattern design. The origin of this coordinate system coincides with the reference origin of the glove pattern, and the horizontal and vertical axes are aligned with the calibration plate grid coordinate system. The system is then configured according to the topological structure of the complete glove shape. The connection order of the clipping segments is determined as follows: thumb-compensated fingertip clipping segment, thumb side original clipping segment, palm outer original clipping segment, little finger side original clipping segment, little finger-compensated fingertip clipping segment, ring finger-compensated fingertip clipping segment, middle finger-compensated fingertip clipping segment, index finger-compensated fingertip clipping segment, index finger side original clipping segment, palm inner original clipping segment, and thumb root connection original clipping segment. Finally, the connection returns to the starting endpoint to form a closed clipping contour. The unit iterates through the connection endpoints of adjacent clipping segments, calculating the coordinate spacing between adjacent endpoints. The mathematical expression for calculating the coordinate spacing is: In the formula This indicates the coordinate spacing between the endpoints of adjacent cut line segments. , This indicates the coordinates of the end point of the previous cut line segment. , This indicates the starting coordinates of the next cutting segment. If the coordinate spacing is greater than zero, the unit uses a short arc or straight line segment with the same tangent direction as the adjacent segment to complete the transition, eliminating breakpoints and misalignments between segments. After all segments are connected, the cutting trajectory is unified in a clockwise direction. The coordinate points of all connected segments are sequentially integrated into a continuous sequence of coordinate points, which is the final digital cutting path. The physical coordinate data of the final digital cutting path is converted into motion control commands that the cutting equipment can recognize. First, a coordinate mapping transformation is performed to convert the coordinate values ​​in the physical coordinate system of the fabric piece into mechanical displacement coordinates of the X and Y axes of the cutting platform. The mathematical expression for coordinate mapping is: , In the formula , These represent the mechanical displacement coordinates of the cutting platform along the X and Y axes, respectively. This represents the conversion ratio from physical coordinates to mechanical coordinates. , This represents the physical coordinate values ​​in the final digital cutting path. Based on the displacement coordinates, motion control commands are generated, including interpolation mode, feed speed, and start / stop instructions. The interpolation modes include linear interpolation and circular interpolation, adapting to the motion control requirements of both straight and curved paths. The glove material to be cut is laid flat on the cutting platform surface. Positioning markers aligned with a unified cutting coordinate system are pre-set on the material surface. The cutting tool head is initially positioned at the starting coordinate point of the final digital cutting path. The unit drives the X-axis and Y-axis servo motors of the cutting platform through a motion controller, causing the cutting tool head to move continuously according to the coordinate sequence of the final digital cutting path. The cutting tool head completes the cutting operation of the glove material during its movement along the path. During the cutting process, a machine vision device fixedly installed above the cutting platform acquires real-time images of the cutting tool head and the material positioning markers at a fixed frame rate. Feature extraction is performed on the real-time images to obtain the actual mechanical coordinates of the cutting tool head center and the reference coordinates of the positioning markers. The positional deviation between the actual coordinates and the theoretical coordinates of the final digital cutting path is calculated. The mathematical expression for the positional deviation calculation is as follows: In the formula This indicates the real-time position deviation of the cutting head. , This represents the actual mechanical coordinates of the cutter head center. , The theoretical target coordinates of the final digital cutting path are represented. The position deviation data is fed back to the motion controller in real time. The motion controller adjusts the output speed and displacement of the servo motor according to the deviation value and corrects the movement trajectory of the cutter head in real time. If the position deviation value exceeds the preset allowable deviation range, the unit immediately outputs a pause command to stop the cutting action and trigger an abnormal prompt until the deviation returns to the allowable range and cutting resumes. Through real-time monitoring and dynamic correction, the cutting process is fully closed-loop controlled to ensure the cutting accuracy and size adaptability of the glove pieces.

[0019] Please see Figure 2 As shown, a second objective of this invention is to provide a system for implementing a machine vision-based adaptive glove size cutting method, including any one of the above-mentioned methods, comprising: The image acquisition and calibration unit is used to sequentially acquire frontal images of the same hand in a flat posture and a bent posture on the calibration plate. The hand contour processing and feature calculation unit is used to extract contours, segment finger regions, locate the center points of the finger roots and fingertips, and fit the midline curve of the finger pads on the tiling and bending posture images acquired by the image acquisition and calibration unit, and calculate the fingertip dynamic compensation coefficient for each finger. The adaptive cutting path generation unit is used to calculate the fingertip offset increment of each finger in the glove cutting pattern according to the dynamic compensation coefficient of each fingertip and the pre-stored material elasticity correction factor, and translate the corresponding original fingertip cutting line to generate the compensated fingertip cutting line. The cutting path fusion and execution control unit is used to combine the compensated fingertip cutting line with the remaining unadjusted cutting lines into the final cutting path, and control the cutting tool to perform the cutting.

[0020] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A machine vision-based adaptive glove size cutting method, characterized in that: Includes the following steps: S1. Sequentially acquire frontal images of the same hand in a flat position and a bent position on the calibration plate. The flat position is when each finger is naturally extended, and the bent position is when each finger is bent to its maximum natural bending state while keeping the palm position unchanged. S2. Perform hand contour extraction algorithms based on skin color model and edge detection on the flat frontal image and the curved frontal image respectively to obtain the flat hand contour and the curved hand contour. Then, segment each finger region from the flat hand contour and the curved hand contour to obtain the finger edge point coordinate sequence. S3. The center point of the finger root joint and the center point of the finger tip are determined by the finger root joint and the center point of the finger tip. Combined with the finger edge point coordinate sequence, the finger pad midline fitting algorithm is used to generate the finger pad midline curve in the flat posture and the finger pad midline curve in the bent posture, respectively. The length of the corresponding curve is calculated as the length in the flat surface and the length in the bent surface. S4. For each finger, the ratio of the length in the curved surface to the length in the flat surface is determined as the fingertip dynamic compensation coefficient for each finger. S5. In the glove pattern, for the fingertip area cutting line of each finger, according to the fingertip dynamic compensation coefficient and the pre-calibrated material elasticity correction factor, the fingertip offset increment is obtained through the fingertip offset increment calculation model. The fingertip cutting line is translated outward along the longitudinal direction of the finger by the fingertip offset increment to generate the compensated fingertip cutting line. The other cutting lines in the glove pattern, except for the fingertip area cutting line, remain in their original positions. S6. The final cutting path is formed by the compensated fingertip cutting line and the remaining unadjusted cutting lines, and the cutting tool is controlled to cut the glove piece.

2. The machine vision-based adaptive glove size cutting method according to claim 1, characterized in that: On the calibration board, frontal images of the same hand in a flat position and a bent position were acquired sequentially, specifically including: Use a calibration plate printed with a grid coordinate system, and pre-set reference lines on the calibration plate corresponding to the flat and bent postures of the human hand; When acquiring a flat, frontal image, the hand is naturally extended and the palm is completely pressed against the surface of the calibration plate, with the axis of each finger parallel to the contour position reference line. When acquiring a frontal image of the bending posture, while keeping the wrist and the base of the palm fixed on the calibration plate, bend each finger naturally to the extreme comfortable position. Simultaneously, using a machine vision device set at a fixed position above the calibration plate, images are captured after the hand position is stabilized in each posture, obtaining frontal images of the flat posture and the curved posture with a unified spatial reference and background.

3. The machine vision-based adaptive glove size cutting method according to claim 1, characterized in that: The process of performing hand contour extraction algorithms based on skin color models and edge detection on both the flat-pose frontal image and the curved-pose frontal image specifically includes: Preprocessing is performed on the flat and curved frontal images to eliminate the effects of uneven illumination. A skin color probability model pre-established in the color space is used to calculate the skin color probability of each pixel in the flat and curved frontal images. Regions with skin color probabilities higher than a set threshold are initially identified as candidate hand regions. An edge detection operator is applied to the candidate hand regions to extract all edges in the image. Then, connected component analysis and contour tracking are performed on the edges to select the connected components with the highest contour integrity and the largest area. Finally, the contour point set of the connected components is smoothed and denoised to obtain the contours of the flat and curved hands, respectively.

4. The machine vision-based adaptive glove size cutting method according to claim 3, characterized in that: The process of segmenting each finger region from both the flat and curved hand contours specifically includes: The convex hull of the hand contour is calculated, and the indentation points located between adjacent fingers on the hand contour are identified as the starting points for finger gap segmentation. The overall hand contour is segmented into five independent sub-contours corresponding to the thumb, index finger, middle finger, ring finger, and little finger, respectively. Each sub-contour represents a finger region. The sub-contour point set of each finger region is reordered according to the direction from the fingertip to the finger root to obtain the finger edge point coordinate sequence of each finger.

5. The machine vision-based adaptive glove size cutting method according to claim 4, characterized in that: The center points of the base of the finger joint and the fingertip are determined using a technique that uses the finger root joint and fingertip center point. Specifically, this includes: For each finger region, along the coordinate sequence of finger edge points, the width change of the contour is calculated from the fingertip to the base of the finger. The first point where the width begins to increase is determined as the center point of the base of the finger joint. For the fingertip center point, the center position of the fingertip is calculated by finding the pair of points with the smallest distance between them in the contour point sequence on the upper and lower sides of the finger. The center position is then extended outward from the fingertip by a preset distance along the direction of the finger contour to locate the outermost point of the fingertip as the fingertip center point.

6. The machine vision-based adaptive glove size cutting method according to claim 5, characterized in that: By combining the finger edge point coordinate sequence, a fingertip midline curve is generated for both flat and bent fingertip midline postures using a fingertip midline fitting algorithm. Specifically, this includes: Each finger edge point coordinate sequence is divided into a finger ventral edge point sequence and a finger dorsal edge point sequence. The finger ventral edge point sequence and the finger dorsal edge point sequence are resampled proportionally based on the finger longitudinal length, so that the sampling points on the finger ventral side and the finger dorsal side correspond one-to-one. The geometric center point of each pair of corresponding sampling points is calculated. The geometric center points are connected sequentially to form a preliminary fingertip midline broken line. Finally, a curve smoothing algorithm is applied to the preliminary fingertip midline broken line to obtain the fingertip midline curve extending from the center point of the finger root joint to the center point of the fingertip. The length of the fingertip midline curve is calculated as the in-plane length of the finger in the corresponding posture.

7. The machine vision-based adaptive glove size cutting method according to claim 1, characterized in that: Based on the aforementioned fingertip dynamic compensation coefficient and the pre-calibrated material elasticity correction factor, the fingertip offset increment is obtained through a fingertip offset increment calculation model, specifically including: The basic compensation amount is obtained by multiplying the dynamic compensation coefficient of each fingertip by the longitudinal design length of the original fingertip cutting line segment corresponding to that finger in the glove pattern. The basic compensation amount is then multiplied by the material elasticity correction factor, which is pre-calibrated through material experiments. The material elasticity correction factor comprehensively reflects the actual elongation characteristics of a specific glove material under the combined action of its elastic modulus, material thickness, and warp and weft stretch rate when subjected to the same tensile force as finger bending. The fingertip offset increment for each fingertip cutting line is finally obtained through the multiplication calculation model.

8. The machine vision-based adaptive glove size cutting method according to claim 7, characterized in that: The fingertip trimming line is shifted outward along the longitudinal direction of the finger by the fingertip offset increment to generate a compensated fingertip trimming line, specifically including: The geometric features of the original fingertip cut line to be adjusted in the glove pattern are determined. The point located at the center of the fingertip among the endpoints of the original fingertip cut line is identified as the translation reference point. Based on the tangent direction of the fingertip midline curve at the fingertip, the longitudinal direction of the finger is determined. The translation reference point is moved outward along this longitudinal direction by a distance equal to the fingertip offset increment to obtain a new target point. Finally, based on the new target point and other fixed points of the original fingertip cut line, the compensated fingertip cut line is refitted and generated.

9. The machine vision-based adaptive glove size cutting method according to claim 8, characterized in that: The final cutting path is formed by the compensated fingertip cutting line and the remaining unadjusted cutting lines. This path controls the cutting tool to cut the glove pieces, specifically including: In the computer, all compensated fingertip cutting lines are seamlessly connected and combined with all other cutting lines in the glove pattern that retain their original positions, according to the complete shape of the glove, to form the final digital cutting path. The coordinate data of the final digital cutting path is converted into control commands to drive the cutting tool head to move along the final cutting path on the cutting platform covered with glove material. During the cutting process, the relative position of the tool and the material marking points is monitored in real time by machine vision equipment and compared with the final digital cutting path to achieve closed-loop control.

10. A system for implementing a machine vision-based adaptive glove size cutting method according to any one of claims 1-9, characterized in that, include: The image acquisition and calibration unit (1) is used to sequentially acquire frontal images of the same hand in a flat posture and a bent posture on the calibration plate; The hand contour processing and feature calculation unit (2) is used to extract contours, segment finger regions, locate the center points of the finger roots and fingertips, and fit the midline curve of the finger pads of the tiling and bending posture images obtained by the image acquisition and calibration unit, and calculate the dynamic compensation coefficient of the fingertip of each finger. The cutting path adaptive generation unit (3) is used to calculate the fingertip offset increment of each finger in the glove cutting pattern according to the dynamic compensation coefficient of each fingertip and the pre-stored material elasticity correction factor, and translate the corresponding original fingertip cutting line to generate the compensated fingertip cutting line. The cutting path fusion and execution control unit (4) is used to combine the compensated fingertip cutting line with the remaining unadjusted cutting lines into the final cutting path and control the cutting tool to perform cutting.

Images