A high-loop low-cut carpet pile inkjet customized color printing system and method

Abstract

The application discloses a high-loop low-cut carpet pile inkjet customized color printing system and a color printing method, and belongs to the technical field of digital printing of tufted carpets. The main control unit is connected with an illumination unit, an image acquisition unit, an inkjet printing unit and a cloth running unit. An encoder is arranged at the driving end of the cloth running unit. The color printing processing module is built in the main control unit. The color printing processing module performs regional segmentation and matches a printing image for each region. The high-loop low-cut carpet pile inkjet customized color printing system and the color printing method utilize the diffuse reflection texture characteristics of high-loop pile and the uniform mirror reflection characteristics of low-cut pile, enlarge the brightness and texture difference of the two types of regions through specific oblique illumination, realize accurate pixel-level segmentation of the regions, and customize printing of corresponding patterns for the two types of regions respectively.

Description

Technical Field

[0001] This invention specifically relates to a custom inkjet color printing system and method for high-loop, low-cut carpet pile, belonging to the field of digital printing technology for tufted carpets. Background Technology

[0002] With the upgrading of consumption in the home decoration market, consumers are constantly increasing their requirements for the decorative, tactile, and functional aspects of carpets. High-loop, low-cut tufted carpets, due to their unique three-dimensional pile structure, with the high loop pile being soft and comfortable and the low cut pile being wear-resistant and supportive, combine good foot feel, durability, and decoration, and are gradually becoming the mainstream product in the high-end carpet market. For example, Chinese Patent Announcement No. CN116876237B discloses a diacetate fiber no-steam washing digital direct-injection printing process, which includes the following steps: sequentially subjecting the diacetate fiber greige fabric to sizing, pre-drying, inkjet printing, and baking to obtain inkjet printed fabric. This invention only requires sizing, pre-drying, inkjet printing, and baking to complete the printing on acetate fiber fabric, without the need for steaming. Currently, the patterns for high-loop, low-cut carpets are mainly implemented in two ways: the first is by directly weaving the pattern on a tufting machine. This method requires pre-customized pattern templates, has a long changeover cycle, and can only be used for large-volume fixed orders. Small-volume personalized customization is extremely costly. At the same time, the color richness of the pattern is limited by the yarn color, making it impossible to achieve complex gradient colors and high-definition photo prints, resulting in severely insufficient pattern expressiveness. The second method is to perform full-surface digital printing on the woven high-loop, low-cut fabric. While it can achieve complex color patterns, it cannot perform customized printing for different areas of high and low pile. The pattern cannot be accurately matched with the three-dimensional structure of the carpet surface, which easily leads to problems such as pattern misalignment and chaotic layering. It cannot fully utilize the three-dimensional advantages of the high loop and low cut structure, and may even damage the texture of the original pile structure. In addition, the existing digital printing technology is not well adapted to high loop and low cut carpets. Due to the obvious difference in pile height between high and low pile areas, problems such as white showing, color bleeding, and large registration deviation are easy to occur during the printing process. It cannot achieve high-precision area printing, which seriously limits the product upgrade and application scenario expansion of high loop and low cut carpets. Summary of the Invention

[0003] To address the aforementioned issues, this invention proposes a custom inkjet color printing system and method for high-loop, low-cut carpet pile. By utilizing the diffuse reflection texture characteristics of high-loop pile and the specular reflection uniformity characteristics of low-cut pile, the brightness and texture differences between the two types of areas are amplified through specific oblique illumination, achieving pixel-level precise segmentation of the areas, and custom printing corresponding patterns on the two types of areas respectively.

[0004] The present invention discloses a custom inkjet color printing system for high-loop, low-cut carpet pile, comprising a main control unit for overall machine control. The main control unit is connected to an illumination unit, an image acquisition unit, an inkjet printing unit, and a fabric feeding unit. An encoder is installed at the drive end of the fabric feeding unit. The main control unit has a built-in color printing processing module, which performs area segmentation and matches a printing image to each area. During operation, the main control unit controls the fabric feeding unit to feed the carpet. Before feeding, the main control unit controls the illumination unit to illuminate the carpet surface. The encoder collects the step value of the fabric feeding unit in real time and triggers the image acquisition unit to synchronously acquire images. The image acquisition frequency is synchronized with the stepping unit. The images are sent to the main control unit and preprocessed by the color printing processing module. Then, the images are segmented into multiple high-loop and low-cut pile areas. After image segmentation, the color printing processing module matches a color printing image to each high-loop and low-cut pile area. Finally, the image data corresponding to the color printing image is sent to the inkjet printing unit for inkjet color printing on the carpet pile.

[0005] Furthermore, the image acquisition unit is a high-speed area array high-definition camera, installed above the feed end of the inkjet printing unit; the illumination unit is a high-intensity incandescent light source or a high color rendering index LED linear light source, arranged at an angle of 45° below the camera, with the illumination range covering the entire width of the printed carpet surface.

[0006] Furthermore, the image acquisition unit is a high-speed linear scan camera covering the entire printing width, with a linear scan resolution ≥12K and a scanning frequency ≥20kHz; the encoder outputs a fabric-moving pulse signal to trigger the high-speed linear scan camera to continuously scan the carpet surface at a fixed step size, with a single-line image exposure time ≤20μs; the color printing processing module uses a convolutional neural network model to perform real-time segmentation of the continuously scanned images line by line, with a single-frame image processing delay ≤10ms.

[0007] Furthermore, the inkjet printing unit includes a color dye ink channel and a functional special ink channel; during printing, while printing the foreground pattern in the low-pile L area, a hydrophobic and anti-fouling functional ink containing fluorocarbon hydrophobic additives is printed simultaneously; while printing the background pattern in the high-pile H area, a wear-resistant and anti-collapse functional ink containing nano-silica wear-resistant agent and polyurethane film-forming agent is printed simultaneously; the printed functional ink is cross-linked and formed into a film by UV curing or hot air curing.

[0008] A method for custom inkjet printing on high-loop, low-cut carpet pile, employing a high-loop, low-cut carpet pile inkjet printing system, is described in detail below. S1 Carpet Surface Pre-fabrication: A double-needle tufting machine is used to prepare a high-loop, low-cut tufted carpet fabric. The front row of needle bars weaves the high-loop pile, and the rear row of needle bars weaves the low-cut pile, controlling the pile height difference between the high-loop pile and the low-cut pile. The prepared fabric is then sized and pre-dried before printing. S2 system calibration: The pixel and physical coordinates of the image acquisition unit and the inkjet printing unit are matched through the color printing processing module; S3 Image Acquisition and Preprocessing: When the fabric feeding unit moves, the encoder synchronously triggers the image acquisition unit and the lighting unit. The lighting unit provides supplementary lighting and simultaneously captures images of the carpet surface entering the printing area. The acquired raw images are preprocessed to amplify the differences in reflective characteristics between the high loop pile and low cut pile areas. S4 Region Intelligent Recognition and Segmentation: Based on the diffuse reflection texture features of high loop pile and the specular uniform reflection features of low loop pile, the preprocessed image is classified and segmented at the pixel level to identify the L region of low loop pile and the H region of high loop pile. The region contour is optimized through morphological operations to obtain the segmentation result of continuous closed loop. S5 Pattern Customization Matching and Color Separation Processing: Pre-design a foreground pattern that adapts to low-cut pile and a background pattern that adapts to high-loop pile. Match the foreground and background patterns with the identified L and H areas using coordinate matching and registration correction to generate inkjet printing data that is only valid in the corresponding areas. S6 Synchronous Linkage Inkjet Printing: The main control unit uses the real-time signal from the encoder to synchronously trigger the entire process of supplementary lighting, image acquisition, area recognition and inkjet printing, and controls the inkjet printing unit to print the foreground pattern in the L area and the background pattern in the H area, so as to realize customized color printing in different areas. S7 finishing and color fixing: The printed carpet surface is pre-baked, steamed for color fixing, washed, and dried to obtain the finished carpet.

[0009] Furthermore, during the prefabrication of the carpet surface, the pile height of the high loop pile is set to 5-9mm, the pile height of the low cut pile is set to 1.5-3mm, and the pile height difference is ≥2.5mm; the sizing and pre-drying treatment uses a polyacrylate thickener, the pre-drying temperature is 105-110℃, the treatment time is 3-5min, and the moisture regain rate of the fabric is controlled to be 6-8% after treatment.

[0010] Furthermore, when the image acquisition unit acquires images, the exposure time of a single frame image is ≤1ms; the preprocessing of the original image includes Gaussian denoising and histogram contrast enhancement processing.

[0011] Furthermore, the specific process of pixel-level classification and segmentation of the preprocessed image is as follows: pixel-level classification is performed using a pre-trained convolutional neural network binary classification model or a color library matching and binarization segmentation algorithm. The morphological operations include opening and closing operations, which are used to eliminate recognition noise and optimize the continuity of region contours.

[0012] Furthermore, both the foreground and background patterns are continuous loop patterns or dynamically generated patterns that adaptively match the contours of the L and H regions. The specific method for dynamically generating the patterns is as follows: contour extraction is performed on the segmented L and H regions to obtain the contour features of each closed loop region. The contour features include contour coordinates, size, and shape features. Then, a dynamic pattern that is completely adapted to the contour of the corresponding region is generated in real time based on the contour features. The pattern is automatically scaled and deformed to adapt to the size and shape of the region.

[0013] Furthermore, the prefabricated fabric of the carpet surface has a pile height gradient of 3 levels or more, including a first-level region of ultra-high loop pile, a second-level region of medium-high loop pile, and a third-level region of low cut pile, with a height difference of ≥2.5mm between adjacent loop piles; when the region is intelligently identified and segmented, a pre-trained multi-class convolutional neural network model is used to perform pixel-level multi-classification on the pre-processed image to identify multiple regions corresponding to different pile heights. Then, an appropriate graded pattern is designed for each identified region to generate inkjet printing data that is only valid in the corresponding region.

[0014] Compared with the prior art, the high-loop, low-cut carpet pile inkjet customized color printing system and method of the present invention have the following advantages: 1. High registration accuracy and pattern-structure matching: Utilizing the inherent reflective characteristics of high loop pile and low cut pile, pixel-level area segmentation is achieved through visual recognition. Combined with the synchronous linkage control of shooting, recognition and printing, the high loop low cut carpet print pattern and the three-dimensional structure of the carpet surface are accurately registered. The registration accuracy can reach within ±0.2mm, achieving precise matching between the three-dimensional structure of the carpet surface and the print pattern.

[0015] 2. Strong pattern expressiveness and rich three-dimensional layering: Different patterns are designed and printed for different areas of high and low pile. Through the natural light and shadow contrast of high and low pile and the color matching of the area printing, a three-dimensional layering effect far exceeding that of flat printing is achieved, resulting in richer pattern effects. It can achieve a high-end texture similar to hand-carved blankets and high-end Axminster loom blankets, greatly enhancing the decorativeness of the product.

[0016] 3. Flexible and efficient production: Using digital printing, there is no need to customize pattern plates. Changing styles only requires changing the pattern file, which can flexibly adapt to large-volume production orders and single-piece personalized customization needs.

[0017] 4. Excellent product performance and long service life: The pre-baking process solves the problems of color bleeding and white showing in high and low pile printing. At the same time, it can print with appropriate functional inks for different areas, which improves the wear resistance and anti-collapse performance of high loop pile areas and the stain resistance of low cut pile areas, thus greatly extending the service life of the carpet.

[0018] 5. Low production cost and wide adaptability: By modifying ordinary double-needle tufting machines and conventional digital printing machines, it has wide adaptability and significantly reduces the production threshold of high-end three-dimensional carpets. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the high-loop, low-cut carpet pile inkjet custom color printing system of the present invention.

[0020] Figure 2 This is a schematic diagram of another embodiment of the high-loop, low-cut carpet pile inkjet custom color printing system of the present invention.

[0021] Figure 3 This is a schematic diagram illustrating the workflow of the high-loop, low-cut carpet pile inkjet custom color printing system of the present invention.

[0022] Figure 4 This is a schematic diagram of the process of morphological optimization after region intelligent recognition and segmentation according to the present invention.

[0023] Figure 5 A schematic diagram of the coordinate matching and registration correction process for pattern customization matching in this invention.

[0024] Figure 6 This is a schematic diagram of the dynamic deviation detection and real-time registration correction process of the present invention. Detailed Implementation

[0025] Example 1: like Figures 1 to 6 The high-loop, low-cut carpet pile inkjet custom color printing system shown includes a main control unit for overall machine control. The main control unit is connected to an illumination unit, an image acquisition unit, an inkjet printing unit, and a fabric feeding unit. The fabric feeding unit's drive end is equipped with an encoder. The main control unit has a built-in color printing processing module that performs area segmentation and matches a printing image to each area. During operation, the main control unit controls the fabric feeding unit to feed the carpet. Before feeding, the main control unit controls the illumination unit to illuminate the carpet surface. The encoder collects the step value of the fabric feeding unit in real time and triggers the image acquisition unit to synchronously acquire images. The image acquisition frequency is synchronized with the stepping unit. The images are sent to the main control unit and preprocessed by the color printing processing module. Then, the images are segmented into multiple high-loop and low-cut pile areas. After image segmentation, the color printing processing module matches a color printing image to each high-loop and low-cut pile area. Finally, the image data corresponding to the color printing images is sent to the inkjet printing unit for inkjet printing on the carpet pile.

[0026] The image acquisition unit is a high-speed area array high-definition camera, installed above the feed end of the inkjet printing unit; the illumination unit is a high-intensity incandescent light source or a high color rendering index LED linear light source, arranged at an angle of 45° below the camera, with the illumination range covering the entire width of the printed carpet surface.

[0027] The image acquisition unit is a high-speed linear scan camera covering the entire printing width, with a linear scan resolution ≥12K and a scanning frequency ≥20kHz; the encoder outputs a fabric-moving pulse signal to trigger the high-speed linear scan camera to continuously scan the carpet surface at a fixed step size, with a single-line image exposure time ≤20μs; the color printing processing module uses a convolutional neural network model to perform real-time segmentation of the continuously scanned images line by line, with a single-frame image processing delay ≤10ms.

[0028] The inkjet printing unit includes a color dye ink channel and a functional special ink channel. During printing, the inkjet printing unit prints the foreground pattern in the low-pile L area while simultaneously printing a hydrophobic and anti-fouling functional ink containing fluorocarbon hydrophobic additives. In the high-pile H area, the background pattern is printed while simultaneously printing a wear-resistant and anti-collapse functional ink containing nano-silica wear-resistant agent and polyurethane film-forming agent. The printed functional inks are cross-linked and formed into a film through UV curing or hot air curing.

[0029] A method for custom inkjet printing on high-loop, low-cut carpet pile, employing a high-loop, low-cut carpet pile inkjet printing system, is described in detail below. S1 Carpet Surface Pre-fabrication: A double-needle tufting machine is used to prepare a high-loop, low-cut tufted carpet fabric. The front row of needle bars weaves the high-loop pile, and the rear row of needle bars weaves the low-cut pile, controlling the pile height difference between the high-loop pile and the low-cut pile. The prepared fabric is then sized and pre-dried before printing. S2 system calibration: The pixel and physical coordinates of the image acquisition unit and the inkjet printing unit are matched through the color printing processing module; S3 Image Acquisition and Preprocessing: When the fabric feeding unit moves, the encoder synchronously triggers the image acquisition unit and the lighting unit. The lighting unit provides supplementary lighting and simultaneously captures images of the carpet surface entering the printing area. The acquired raw images are preprocessed to amplify the differences in reflective characteristics between the high loop pile and low cut pile areas. S4 Region Intelligent Recognition and Segmentation: Based on the diffuse reflection texture features of high loop pile and the specular uniform reflection features of low loop pile, the preprocessed image is classified and segmented at the pixel level to identify the L region of low loop pile and the H region of high loop pile. The region contour is optimized through morphological operations to obtain the segmentation result of continuous closed loop. S5 Pattern Customization Matching and Color Separation Processing: Pre-design a foreground pattern that adapts to low-cut pile and a background pattern that adapts to high-loop pile. Match the foreground and background patterns with the identified L and H areas using coordinate matching and registration correction to generate inkjet printing data that is only valid in the corresponding areas. S6 Synchronous Linkage Inkjet Printing: The main control unit uses the real-time signal from the encoder to synchronously trigger the entire process of supplementary lighting, image acquisition, area recognition and inkjet printing, and controls the inkjet printing unit to print the foreground pattern in the L area and the background pattern in the H area, so as to realize customized color printing in different areas. S7 finishing and color fixing: The printed carpet surface is pre-baked, steamed for color fixing, washed, and dried to obtain the finished carpet.

[0030] When the carpet is prefabricated, the pile height of the high loop pile is set to 5-9mm, the pile height of the low cut pile is set to 1.5-3mm, and the pile height difference is ≥2.5mm; the sizing and pre-drying treatment uses a polyacrylate thickener, the pre-drying temperature is 105-110℃, the treatment time is 3-5min, and the moisture regain of the fabric is controlled to be 6-8% after treatment.

[0031] When the image acquisition unit acquires images, the exposure time of a single frame image is ≤1ms; the preprocessing of the original image includes Gaussian denoising and histogram contrast enhancement processing.

[0032] The specific process of pixel-level classification and segmentation of the preprocessed image is as follows: a pre-trained convolutional neural network binary classification model or a color library matching and binarization segmentation algorithm is used for pixel-level classification. The morphological operations include opening and closing operations, which are used to eliminate recognition noise and optimize the continuity of regional contours.

[0033] Both the foreground and background patterns are continuous loop patterns or dynamically generated patterns that adaptively match the contours of regions L and H. The specific method of dynamically generating patterns is as follows: contour extraction is performed on the segmented regions L and H to obtain the contour features of each closed loop region. The contour features include contour coordinates, size and shape features. Then, a dynamic pattern that is completely adapted to the contour of the corresponding region is generated in real time based on the contour features. The pattern is automatically scaled and deformed to adapt to the size and shape of the region.

[0034] The prefabricated fabric of the carpet surface has a pile height gradient of 3 levels or more, including a first-level region of ultra-high loop pile, a second-level region of medium-high loop pile, and a third-level region of low cut pile, with a height difference of ≥2.5mm between adjacent loop piles; when the region is intelligently identified and segmented, a pre-trained multi-class convolutional neural network model is used to perform pixel-level multi-classification on the pre-processed image to identify multiple regions corresponding to different pile heights. Then, an appropriate graded pattern is designed for each identified region to generate inkjet printing data that is only valid in the corresponding region.

[0035] Example 2: The specific steps of the inkjet custom color printing method for high-loop, low-cut carpet pile in this embodiment are as follows: 2.1 Carpet Surface Pre-fabrication: A double-needle tufting machine is used to prepare a high-loop, low-cut tufted carpet fabric. The front row of needles weaves high-loop pile with a pile height of 5-8mm, and the rear row of needles weaves low-cut pile with a pile height of 2-3mm, with a pile height difference ≥3mm. The prepared fabric is then sized and pre-dried. Polyacrylate thickener is used as the sizing agent. The pre-drying temperature is 105-110℃, and the time is 3-5 minutes. The moisture regain of the fabric is controlled at 6-8% to avoid ink bleeding during the printing process. 2.2 System Setup and Calibration: A high-speed area array HD camera is installed above the feed end of the digital inkjet printer head. A high-intensity incandescent light source is arranged at a 45° angle below the high-speed area array HD camera, and the illumination range covers the carpet surface within the printing width. Camera distortion correction, pixel and physical coordinate matching, and synchronous calibration of camera and printhead stepping are completed, with a calibration accuracy of ≤±0.1mm.

[0036] 2.3 Image Acquisition and Preprocessing: A high-speed area array high-definition camera is triggered by an encoder linked to the fabric roller to capture the carpet surface entering the printing area frame by frame. The light source is synchronously flashed to ensure that the exposure time of a single frame is ≤1ms and to eliminate motion blur. The acquired raw images are preprocessed with Gaussian noise reduction and histogram contrast enhancement to amplify the reflective differences between high and low pile areas. 2.4 Intelligent Region Recognition and Segmentation: A pre-trained color library matching and binarization segmentation algorithm is employed. Based on the specular uniform reflection characteristics of the low-cut pile region and the diffuse reflection texture characteristics of the high-loop pile region, pixel-level classification is performed on the pre-processed image to accurately identify the L region of low-cut pile and the H region of high-loop pile. Morphological opening and closing operations are used to optimize the region contours, eliminate recognition noise, and obtain continuous closed-loop segmentation results. Morphological optimization of the segmented images for high-loop and low-cut pile regions is based on the binarized carpet surface feature images. The core operation combines opening and closing operations with small region removal and contour smoothing to eliminate recognition noise caused by pile texture interference, optimize the continuity and integrity of the region contours, and ensure that the segmentation results are continuous closed-loop high-loop pile B region and low-cut pile A region, adapting to the coordinate matching requirements of subsequent fixed-position printing. The optimization process uses 3×3 square structural elements (adapting to the pixel-level recognition accuracy requirements of the carpet surface; it can be adjusted to a 5×5 cross shape according to the actual pile texture thickness), implemented based on OpenCV image processing methods. The specific steps are as follows: 2.4.1 Perform image binarization calibration to convert the pixel-level classified textured feature image into a standard binary image, setting the high loop texture area as the foreground (pixel value 255, white) and the low loop texture area as the background (pixel value 0, black) (or the reverse setting); calibrate the image pixel threshold to eliminate false contours caused by grayscale blurring; 2.4.2 Perform opening operations to eliminate surface noise and burrs. Perform an opening operation of erosion followed by dilation on the binarized image to remove small noise in the foreground and edge burrs caused by carpet fiber reflection and uneven texture (such as isolated pixels misidentified as high loop pile in low loop pile areas, and small protrusions at the edges of high loop pile areas). The specific process is as follows: Erosion: Traverse the image using a 3×3 structuring element. Only when all pixels covered by the structuring element are foreground pixels are retained as foreground pixels; otherwise, they are set as background pixels, thus achieving preliminary removal of noise and burrs. Dilation: Traverse the eroded image again with the same structuring element. If there are any foreground pixels in the area covered by the structuring element, set the center pixel as the foreground to restore the original outline size of the high loop / low cut loop area and avoid the area shrinkage caused by erosion. 2.4.3 Perform a closing operation to fill the internal holes and discontinuous contours of the region. Then, perform a dilation followed by erosion closing operation on the image after the opening operation to resolve the issues of small holes and discrete discontinuous contours in the foreground caused by gaps in the yarn and lack of local reflections in the velvet region segmentation; specifically as follows: Dilation: Traverse the image using a 3×3 structuring element to fill tiny holes inside the foreground region, connect adjacent discrete foreground pixel blocks, and achieve preliminary connectivity of the region outline. Erosion: Traverse the dilated image using the same structuring element to restore the original shape and size of the region outline, avoiding excessive expansion of the region edges caused by dilation.

[0037] 2.4.4 Small connected region removal to filter out falsely identified invalid regions; Perform connected region analysis on the image after closing operation, calculate the pixel area of ​​all foreground connected regions, set an area threshold (e.g., ≤20 pixels), and remove connected regions with an area smaller than the threshold; This step mainly filters out small falsely identified regions in the carpet surface caused by two-color plying of yarn and uneven local pile height, avoiding such invalid regions from interfering with the coordinates of subsequent pattern positioning printing, and ensuring that the remaining foreground area is a real high loop pile / low cut pile effective area.

[0038] 2.4.5 Contour fitting and smoothing to generate continuous closed-loop region contours; contour extraction is performed on the filtered effective regions, and polygon approximation and Gaussian smoothing algorithms are used to optimize the contour curves, as detailed below: The original contour was approximated by a polygon using the Douglas-Puk algorithm to remove small inflection points and simplify the contour structure. Then, the approximated contour points were smoothed by Gaussian smoothing to make the contour curve continuous and smooth, and finally a continuous closed-loop contour without burrs, discontinuities and voids was generated, which is the final segmentation result of the high loop pile B region and the low cut pile A region.

[0039] After morphological optimization, the recognition accuracy of high loop pile and low cut pile areas is ≥99%. The segmentation contour is a continuous closed loop structure with no noise, no voids, and no discontinuities. It can be directly matched and registered with the printed pattern to avoid the problem of chaotic printing of the foreground and background patterns due to irregular recognition contours. 2.5 Pattern Customization Matching and Color Separation Processing: A foreground pattern adapted to the L region and a background pattern adapted to the H region are pre-designed. Both the foreground and background patterns are continuous loop patterns. The continuous loop patterns are then matched and registered with the identified regions (L and H regions) to generate inkjet data for the foreground pattern valid only in the L region and inkjet data for the background pattern valid only in the H region. The specific process of coordinate matching and registration correction is as follows: First, the calibration parameters fixed in the S2.2 system calibration process are called to complete the unified transformation of the three core coordinate systems and establish a one-to-one mapping relationship between pixels, physical coordinates, and patterns. The specific process is as follows: 2.5.1 Coordinate system definition: The physical coordinate system of the color printing equipment is established with the fabric edge positioning point on one side of the printer feed end as the origin of the physical coordinate system, the Y-axis along the fabric feeding direction as the longitudinal direction, and the X-axis perpendicular to the fabric feeding direction as the transverse direction; the image pixel coordinate system is established with the upper left corner of the camera image as the origin; and the pattern digital coordinate system is established with the upper left corner of the pattern file as the origin. 2.5.2 Conversion Coefficient Matching: Call the pre-calibrated pixel and physical conversion coefficients (e.g., a 12K line scan camera corresponds to a 3600mm printing width, with a conversion coefficient of 3.33 pixels / mm; a 5-megapixel area scan camera corresponds to a 1600mm printing width, with a conversion coefficient of 3.125 pixels / mm), physical and pattern conversion coefficients (fixed at 1mm physical length corresponding to 10 pixels of pattern to ensure 1:1 size matching), and the longitudinal spacing compensation value between the camera and the print head (e.g., a longitudinal spacing of 200mm between the camera and the print head corresponds to 20,000 encoder pulses and an encoder resolution of 100 pulses / mm). 2.5.3 Alignment of the loop reference: According to the preset longitudinal loop length of the pattern (usually 500mm / 1000mm), the pattern is divided into continuous units that match the fabric step length according to the loop length. Each loop unit corresponds to a fixed number of encoder pulses (e.g., 50000 pulses for a 500mm loop) to ensure that the pattern loop is completely aligned with the high loop low cut weaving loop reference of the greige fabric. 2.5.4 Regional Pattern Pre-mapping and Dedicated Printing Mask Generation: For the pre-designed low-cut pile area to adapt to the foreground pattern and high-loop pile area to adapt to the background pattern, physical coordinate system pre-mapping and regional printing permission control are completed respectively. The specific process is as follows: 2.5.4.1 Pattern Physical Pre-mapping: Map the foreground and background patterns to the equipment physical coordinate system according to the actual printing width and cycle length, and generate a pattern coordinate matrix that corresponds to the physical size of the carpet surface in a 1:1 ratio, ensuring that the horizontal width and vertical cycle length of the pattern are completely matched with the fabric parameters; 2.5.4.2 Regional Printing Mask Generation: A dedicated printing mask M1 for the high-ring area of ​​the foreground pattern is generated, and a dedicated printing mask M2 for the low-cut texture area of ​​the background pattern is generated. The mask is a binary matrix with the same size as the pattern coordinate matrix. In mask M1, only the pixel value corresponding to the high-ring area is 1 (allowing inkjet printing), and the rest are 0 (disallowing inkjet printing). In mask M2, only the pixel value corresponding to the low-cut texture area is 1, and the rest are 0. This avoids the problem of cross-region inkjet printing and chaotic printing of foreground and background patterns from the source.

[0040] 2.5.5 Real-time region contour coordinate transformation and initial matching: The real-time output region segmentation result completes the transformation from pixel coordinates to physical coordinates and achieves initial alignment with the pre-mapped pattern. 2.5.5.1 Contour Coordinate Transformation: Through pre-calibrated pixel-to-physical conversion coefficients, the closed-loop contour coordinates of regions A and B in the image pixel coordinate system are converted into actual physical coordinates in the device physical coordinate system; at the same time, according to the longitudinal distance between the camera and the print head, the pulse compensation of the fabric encoder is superimposed to ensure that the contour coordinates and the current printing position of the print head are completely synchronized in the longitudinal direction. 2.5.5.2 Cyclic reference matching: Based on the number of fabric pulses fed back by the encoder in real time, the pattern cycle unit corresponding to the current carpet surface is locked, and the converted area contour is initially aligned with the pattern coordinate matrix of the corresponding cycle unit to eliminate the cycle misalignment caused by the cumulative deviation of fabric movement; 2.5.5.3 Lateral width alignment: Using the positioning marks on both sides of the blanket as a reference, correct the lateral coordinates of the area outline to eliminate the left and right misalignment caused by lateral offset of the blanket and fabric deviation. 2.5.6 Dynamic Deviation Detection and Real-time Registration Correction: For dynamic errors occurring during production, such as fabric slippage, blanket stretching, and weft skew, real-time detection and corresponding corrections are performed to ensure a registration accuracy of ≤±0.2mm throughout the entire process. 2.5.6.1 Longitudinal fabric deviation correction: The actual number of fabric pulses is compared with the theoretical number of pulses in real time by the encoder. When the longitudinal position deviation exceeds ±0.1mm, the ink jet triggering timing of the print head is dynamically adjusted. At the same time, the longitudinal Y-axis of the pattern coordinate matrix is ​​linearly scaled and corrected to compensate for the longitudinal misalignment caused by fabric slippage or carpet stretching. 2.5.6.2 Lateral Skew and Offset Correction: By identifying the selvage marks or pre-printed registration crosshairs on both sides of the carpet surface, the lateral offset and weft skew angle of the carpet surface are detected in real time. When the lateral offset exceeds ±0.1mm and the weft skew exceeds 0.15mm / m, an affine transformation is performed on the pattern coordinate matrix. Through translation, rotation, and skewing operations, the pattern misalignment caused by the lateral offset and weft skew is eliminated.

[0041] 2.5.6.3 Local Deformation Correction: For the regional contour deformation caused by local stretching / shrinkage of the carpet surface, a thin plate spline interpolation algorithm is adopted. The feature inflection points of the regional contour are used as anchor points to perform local mesh deformation correction on the pattern coordinates within the contour, ensuring that the pattern edge is completely aligned with the regional contour. After correction, the local registration accuracy is ≤ ±0.2mm. 2.5.7 Final Regional Inkjet Printing Data Generation: After coordinate matching and registration correction, the final inkjet control commands are generated through mask calculations. 2.5.7.1 Perform a pixel-by-pixel AND operation between the corrected pattern X pixel matrix and the mask M1, retaining only the pattern X inkjet data in region A, and setting the rest of the region to zero; 2.5.7.2 Perform a pixel-by-pixel AND operation between the corrected pattern Y pixel matrix and the mask M2, retaining only the pattern Y inkjet data in region B, and setting the rest of the region to zero; 2.5.7.3 The two sets of area printing data are merged into a complete inkjet control command according to the printhead channel allocation rules and output to the inkjet printing unit to achieve precise binding printing of the pattern and the high and low textured areas.

[0042] 2.6 Synchronous Linkage Inkjet Printing: The main control unit controls the carpet to move continuously at a speed of 5-10m / min. The entire process of shooting, recognition and printing is triggered synchronously through the encoder signal in real time. Each round trip of the print head completes the printing of the pattern in the corresponding area, realizing the composite overlay printing of the L area and the foreground pattern, and the H area and the background pattern; the fabric deviation is corrected in real time throughout the process, and the registration accuracy is ≤±0.2mm. 2.7 Finishing and color fixing: The printed carpet surface is pre-baked at 110℃ for 5 minutes, then steamed with saturated steam at 102-105℃ for 8-10 minutes to fix the color. After that, it is washed with water to remove the floating color and paste, and finally dried at 120℃ to fix the shape, thus obtaining the finished carpet.

[0043] Example 3: The specific steps of the inkjet custom color printing method for high-loop, low-cut carpet pile in this embodiment are as follows: 3.1 Multi-gradient blanket fabric prefabrication: A three-station double-needle tufting machine is used to weave a high-loop, low-cut greige fabric with three levels of pile height gradient: the first level is the C region with a pile height of 7-9mm (super high loop pile), the second level is the B region with a pile height of 4-6mm (medium-high loop pile), and the third level is the A region with a pile height of 1.5-2.5mm (low-cut pile), with an adjacent pile height difference ≥2.5mm; the greige fabric is sized and pre-dried, with the process being the same as in Example 1; 3.2 System Setup and Calibration: Two high-speed area array cameras are installed above the printer feed end, arranged obliquely and symmetrically. Two sets of symmetrical high color rendering index LED linear light sources are configured below them, with incident angles of 30° and 60° respectively. The difference in reflectivity between different pile heights is amplified by dual-angle illumination. The joint calibration of the two high-speed area array cameras, pixel and physical coordinate matching, and synchronous calibration with the print head are completed. The calibration accuracy is ≤ ±0.1mm. 3.3 Image Acquisition and Preprocessing: The encoder triggers the simultaneous shooting of dual cameras (high-speed area array cameras), and the two sets of light sources flash synchronously, with a single frame exposure time ≤0.5ms; the images acquired by the dual cameras are fused and combined with the reflection data from different angles to generate a textured image, which is then subjected to noise reduction and contrast enhancement preprocessing. 3.4 Multi-region intelligent recognition and segmentation: A pre-trained multi-class convolutional neural network model is adopted. Based on the reflectivity and texture density features of different texture height regions, pixel-level multi-classification is performed on the pre-processed image to accurately identify three types of regions: A, B, and C. Multi-level morphological optimization of the contours of each region is used to eliminate recognition noise in the texture height transition area. 3.5 Customized matching and color separation of hierarchical patterns: Three-level patterns are pre-designed to adapt to three types of areas respectively: area A prints a high-saturation main pattern, area B prints a transition texture pattern, and area C prints a low-saturation background pattern. The three-level patterns are matched with the corresponding identified areas for coordinate matching and registration correction to generate hierarchical inkjet printing data that is only valid in the corresponding areas. 3.6 Synchronous and Linked Graded Inkjet Printing: The main control unit controls the carpet to move continuously at a speed of 3-8m / min. The encoder signals in real time synchronously trigger dual-camera shooting, multi-class recognition, and multi-channel printhead graded inkjet printing to achieve seamless connection between shooting, recognition, and graded printing. The system detects the fabric deviation and pile height fluctuation in real time and dynamically adjusts the coordinate mapping of the printing data. The registration accuracy is ≤±0.25mm throughout the process.

[0044] 3.7 Finishing and color fixing: Using the same pre-drying, steaming, washing and setting process as in Example 1, a finished carpet with a three-dimensional structure and graded color pattern is obtained.

[0045] Example 4: The specific steps of the inkjet custom color printing method for high-loop, low-cut carpet pile in this embodiment are as follows: This embodiment, based on Embodiment 1, replaces area array acquisition with linear array continuous scanning, significantly improving production speed. The specific steps are as follows: 4.1 Blanket fabrication: High loop low cut fabric was prepared using the same double needle bar tufting process as in Example 1, and sizing and pre-drying treatment was completed; 4.2 System Setup and Calibration: A high-speed linear scan camera covering the entire printing width is installed above the printer feed end, with a linear scan resolution ≥12K and a scanning frequency ≥20kHz; a linear focused incandescent light source matching the width is configured below the camera, with the light source incident at a 45° angle along the fabric feeding direction to form a uniform linear illumination band; the line frequency-fabric feed speed matching calibration, pixel-physical coordinate matching, and longitudinal synchronization calibration with the print head are completed, with a calibration accuracy ≤±0.08mm; 4.3 Image Acquisition and Preprocessing: A high-precision encoder coaxial with the fabric roller outputs a fabric pulse signal in real time, triggering a linear scan camera to continuously scan the carpet surface at a fixed step size. The light source synchronously flashes at high frequency to ensure that the exposure time of each scanned image is ≤20μs, completely eliminating the ghosting and uneven brightness under high-speed fabric movement. The continuously scanned images are preprocessed with row alignment, noise reduction, and contrast enhancement to generate continuous full-width carpet surface feature images. 4.4 Intelligent Region Recognition and Segmentation: A pre-trained convolutional neural network model is used to perform real-time, row-by-row pixel-level binary classification on continuously scanned images, quickly identifying low-cut pile areas and high-loop pile areas; the region contour is optimized through pipelined morphological operations, realizing simultaneous scanning, recognition, and segmentation result output, with a single-frame image processing latency of ≤10ms and a recognition accuracy of ≥99.2%; 4.5 Pattern Customization Matching and Color Separation Processing: A foreground pattern adapted to area A and a background pattern adapted to area B are pre-designed. The patterns are then split into continuous rows of print data according to the fabric feeding direction. Based on the real-time output of the area segmentation results, dynamic coordinate matching and registration correction are performed on the row print data to generate inkjet data for the foreground pattern in area A and the background pattern in area B for the corresponding row. During dynamic coordinate matching and registration correction, the line scan camera performs coordinate transformation, deviation correction, and mask matching simultaneously for each row of images after scanning and recognition, generating the area-specific print data for that row. A pipelined processing architecture is adopted, with a single-row processing delay ≤10ms, matching the high-speed fabric feeding requirements of 15-30m / min and ensuring continuous registration between rows. 4.6 High-speed synchronous inkjet printing: The main control unit controls the carpet to move continuously at a high speed of 15-30m / min. Through the real-time pulse signal of the encoder, it synchronously triggers the entire process of linear scan, real-time recognition, and inkjet printing, realizing a seamless connection between scanning, recognition, and printing in an assembly line, and uninterrupted continuous production; the closed-loop feedback system corrects the fluctuation of the fabric speed and the line alignment deviation in real time, and the registration accuracy is ≤±0.2mm throughout the process; 4.7 Finishing and color fixing: The same finishing process as in Example 1 is used to obtain the finished carpet.

[0046] Example 5: The specific steps of the inkjet custom color printing method for high-loop, low-cut carpet pile in this embodiment are as follows: This embodiment, based on embodiment 1, realizes adaptive dynamic pattern generation based on region contour, solving the problem of misregistration caused by fabric error. The specific steps are as follows: 5.1 Blanket prefabrication: High loop low cut fabric is prepared using the same double needle bar tufting process as in Example 1. Random high and low pile pattern fabric without fixed cycle can be used to complete the sizing and pre-drying treatment.

[0047] 5.2 System Setup and Calibration: Using the same area array camera, light source, and printer architecture as in Example 1, the camera and printhead were calibrated synchronously. At the same time, a main control unit with real-time pattern generation capability was set up and an edge computing unit was configured. 5.3 Image Acquisition and Preprocessing: Using the same method as in Example 1, the carpet surface image was acquired and preprocessed frame by frame to amplify the reflectivity differences between high and low pile areas; 5.4 Intelligent Region Recognition and Contour Extraction: Using the same binary classification segmentation algorithm as in Example 1, the low-cut pile region A and the high-loop pile region B are identified; simultaneously, contour extraction is performed on the segmented regions A and B to obtain the contour coordinates, size, and shape features of each closed-loop region, generating a region contour vector file. 5.5 Real-time Adaptive Pattern Generation and Matching: Based on the extracted region contour features, the edge computing unit calls a pre-defined pattern algorithm library to generate dynamic patterns that perfectly match the corresponding region contours in real time. For each closed-loop contour in region A, a foreground pattern X that fits the contour is automatically generated, and the pattern is automatically scaled and deformed to adapt to the region size and shape. For the remaining contours in region B, a background pattern Y that corresponds to the X pattern is automatically generated. No pre-design of the entire loop pattern is required, achieving dynamic customization of a set of adaptive patterns for each region, with a single-frame pattern generation latency of ≤20ms. Based on the real-time extracted closed-loop contour features of regions A and B, a pattern that perfectly matches the contours is generated in real time. The adapted dynamic patterns directly match the coordinates of the dynamic patterns with the coordinates of the area outline in a 1:1 ratio, without the need for loop reference alignment; contour fitting correction is performed separately for each closed loop area to ensure that the pattern edge and the area outline are completely coincident; the pattern algorithm library is adapted to the set of pre-made pattern generation algorithms for high-loop and low-cut carpets with area-specific fixed printing, and has built-in contour fitting deformation, size adaptive scaling, color style matching, and intelligent texture filling modules. Based on the real-time recognition of high and low pile area outlines, it can automatically generate printed patterns that are completely adapted to the area shape, without the need for manual pre-design of the entire loop pattern, quickly achieving accurate fitting of the pattern and the pile structure, and adapting to personalized customization needs; 5.6 Synchronous Linkage Inkjet Printing: The main control unit controls the continuous fabric feeding of the carpet, and triggers the entire process of shooting, recognition, pattern generation and printing in real time through encoder signals; for regional misalignment and deformation that occur during the fabric feeding process, the pattern data is dynamically adjusted in real time to ensure that the pattern and the outline of the corresponding area are completely aligned, with a registration accuracy of ≤±0.15mm.

[0048] 5.7 Finishing and color fixing: The same finishing process as in Example 1 is used to obtain the finished carpet.

[0049] Example 6: The specific steps of the inkjet custom color printing method for high-loop, low-cut carpet pile in this embodiment are as follows: This embodiment expands upon the previous embodiment by adding regional functional ink composite printing to improve product performance. The specific steps are as follows: 6.1 Blanket fabrication: High loop low cut fabric was prepared using the same double needle bar tufting process as in Example 1, and sizing and pre-drying treatment was completed; 6.2 System Setup and Calibration: Based on the equipment architecture of Example 1, expand the number of channels in the inkjet printhead, of which 4-6 channels are color dye ink channels and the remaining channels are functional special ink channels; complete the synchronous calibration of the camera and all printing channels, with a calibration accuracy of ≤±0.1mm; 6.3 Image Acquisition and Region Recognition: Using the same method as in Example 1, the carpet surface image was acquired and preprocessed to accurately identify the low-cut pile area A and the high-loop pile area B; 6.4 Pattern and Functional Ink Regional Matching Process: A color pattern X adapted to area A and a color pattern Y adapted to area B are pre-designed, and corresponding functional inks are matched: For the high-loop pile area B (high frequency of trampling, prone to flattening and wear), abrasion-resistant and anti-flattening functional ink with added nano-silica abrasion-resistant agent and polyurethane film-forming agent is matched; For the low-cut pile area A (low-lying areas of the carpet surface, prone to dirt accumulation and color bleeding), a hydrophobic and stain-resistant functional ink with added fluorocarbon hydrophobic additives is matched; Printing data for the corresponding color pattern and functional ink is generated. 6.5 Synchronous Linkage and Regional Composite Printing: The main control unit controls the continuous fabric movement of the carpet surface, and synchronously triggers the entire process of shooting, recognition, color printing and functional ink printing through real-time encoder signals. Color pattern X and hydrophobic and stain-resistant functional ink are printed simultaneously in area A, and color pattern Y and wear-resistant and anti-collapse functional ink are printed simultaneously in area B; achieving precise composite of color pattern and functional coating in the same area, with a registration accuracy of ≤±0.2mm.

[0050] 6.6 Post-processing, color fixing, and functional curing: The printed carpet surface is pre-baked and then fixed with saturated steam to ensure that the color dyes are fully applied; then, UV curing or hot air curing is used to fully cross-link and cure the film-forming agent of the functional ink, ensuring the adhesion and durability of the functional coating; finally, it is washed, dried, and shaped to obtain the finished carpet.

[0051] The above embodiments are merely preferred embodiments of the present invention. Therefore, all equivalent changes or modifications made to the structure, features and principles described in the claims of the present invention are included within the scope of the present invention.

Claims

1. A custom inkjet color printing system for high-loop, low-cut carpet pile, characterized in that: The system includes a main control unit for overall machine control. The main control unit is connected to the lighting unit, image acquisition unit, inkjet printing unit, and fabric feeding unit. The drive end of the fabric feeding unit is equipped with an encoder. The main control unit has a built-in color printing processing module. The color printing processing module performs area segmentation and matches the printed image to each area.

2. The high-loop, low-cut carpet pile inkjet custom color printing system according to claim 1, characterized in that: The image acquisition unit is a high-speed area array high-definition camera, installed above the feed end of the inkjet printing unit; the illumination unit is a high-intensity incandescent light source or a high color rendering index LED linear light source, arranged at an angle of 45° below the camera, with the illumination range covering the entire width of the printed carpet surface.

3. The high-loop, low-cut carpet pile inkjet custom color printing system according to claim 1, characterized in that: The image acquisition unit is a high-speed linear scan camera covering the entire printing width. The encoder outputs a fabric-moving pulse signal to trigger the high-speed linear scan camera to continuously scan the carpet surface at a fixed step size. The exposure time of a single line image is ≤20μs. The color printing processing module uses a convolutional neural network model to perform real-time segmentation of the continuously scanned images line by line. The processing delay of a single frame image is ≤10ms.

4. The high-loop, low-cut carpet pile inkjet customized color printing system according to claim 1, characterized in that: The inkjet printing unit includes a color dye ink channel and a functional special ink channel.

5. A method for custom inkjet color printing on high-loop, low-cut carpet pile, employing the high-loop, low-cut carpet pile inkjet custom color printing system as described in claim 1, characterized in that... The method is as follows: S1 Carpet Surface Pre-fabrication: A double-needle tufting machine is used to prepare a high-loop, low-cut tufted carpet fabric. The front row of needle bars weaves the high-loop pile, and the rear row of needle bars weaves the low-cut pile, controlling the pile height difference between the high-loop pile and the low-cut pile. The prepared fabric is then sized and pre-dried before printing. S2 system calibration: The pixel and physical coordinates of the image acquisition unit and the inkjet printing unit are matched through the color printing processing module; S3 Image Acquisition and Preprocessing: When the fabric feeding unit moves, the encoder synchronously triggers the image acquisition unit and the lighting unit. The lighting unit provides supplementary lighting and simultaneously captures images of the carpet surface entering the printing area. The acquired raw images are preprocessed to amplify the differences in reflective characteristics between the high loop pile and low cut pile areas. S4 Region Intelligent Recognition and Segmentation: Based on the diffuse reflection texture features of high loop pile and the specular uniform reflection features of low loop pile, the preprocessed image is classified and segmented at the pixel level to identify the L region of low loop pile and the H region of high loop pile. The region contour is optimized through morphological operations to obtain the segmentation result of continuous closed loop. S5 Pattern Customization Matching and Color Separation Processing: Pre-design a foreground pattern that adapts to low-cut pile and a background pattern that adapts to high-loop pile. Match the foreground and background patterns with the identified L and H areas using coordinate matching and registration correction to generate inkjet printing data that is only valid in the corresponding areas. S6 Synchronous Linkage Inkjet Printing: The main control unit uses the real-time signal from the encoder to synchronously trigger the entire process of supplementary lighting, image acquisition, area recognition and inkjet printing, and controls the inkjet printing unit to print the foreground pattern in the L area and the background pattern in the H area, so as to realize customized color printing in different areas. S7 finishing and color fixing: The printed carpet surface is pre-baked, steamed for color fixing, washed, and dried to obtain the finished carpet.

6. The method for customized inkjet color printing on high-loop, low-cut carpet pile as described in claim 5, characterized in that: When the carpet is prefabricated, the pile height of the high loop pile is set to 5-9mm, the pile height of the low cut pile is set to 1.5-3mm, and the pile height difference is ≥2.5mm; the sizing and pre-drying treatment uses a polyacrylate thickener, the pre-drying temperature is 105-110℃, the treatment time is 3-5min, and the moisture regain of the fabric is controlled to be 6-8% after treatment.

7. The method for customized inkjet color printing on high-loop, low-cut carpet pile as described in claim 5, characterized in that: When the image acquisition unit acquires images, the exposure time of a single frame image is ≤1ms; the preprocessing of the original image includes Gaussian denoising and histogram contrast enhancement processing.

8. The method for customized inkjet color printing on high-loop, low-cut carpet pile as described in claim 5, characterized in that: The specific process of pixel-level classification and segmentation of the preprocessed image is as follows: a pre-trained convolutional neural network binary classification model or a color library matching and binarization segmentation algorithm is used for pixel-level classification. The morphological operations include opening and closing operations, which are used to eliminate recognition noise and optimize the continuity of regional contours.

9. The method for customized inkjet color printing on high-loop, low-cut carpet pile as described in claim 5, characterized in that: Both the foreground and background patterns are continuous loop patterns or dynamically generated patterns that adaptively match the contours of regions L and H. The specific method of dynamically generating patterns is as follows: contour extraction is performed on the segmented regions L and H to obtain the contour features of each closed loop region. The contour features include contour coordinates, size and shape features. Then, a dynamic pattern that is completely adapted to the contour of the corresponding region is generated in real time based on the contour features. The pattern is automatically scaled and deformed to adapt to the size and shape of the region.

10. The method for customized inkjet color printing on high-loop, low-cut carpet pile according to claim 5, characterized in that: The prefabricated fabric of the carpet surface has a pile height gradient of 3 levels or more, including a first-level region of ultra-high loop pile, a second-level region of medium-high loop pile, and a third-level region of low cut pile, with a height difference of ≥2.5mm between adjacent loop piles; when the region is intelligently identified and segmented, a pre-trained multi-class convolutional neural network model is used to perform pixel-level multi-classification on the pre-processed image to identify multiple regions corresponding to different pile heights. Then, an appropriate graded pattern is designed for each identified region to generate inkjet printing data that is only valid in the corresponding region.

Images