Online inspection system for cloth label washing mark defects

Abstract

This invention discloses an online defect detection system for fabric labels and care labels, belonging to the technical field of detection systems. The invention includes an image acquisition module for acquiring images of fabric labels and care labels to be inspected, and a sample management module for pre-acquiring standard product samples and setting exposure parameters. The images to be inspected acquired by the image acquisition module are transmitted to an image comparison module, which compares the images to be inspected with the standard product samples and sends the comparison results to a defect judgment module, which identifies defect information in the images to be inspected. This invention achieves fully automated online inspection of fabric labels and care labels, replacing inefficient and unreliable manual visual inspection, significantly improving inspection efficiency and accuracy. Through the combination of image comparison and a multi-dimensional defect judgment module, it can accurately identify various defects such as typos, numerical errors, graphic defects, and dimensional deviations, providing comprehensive detection capabilities.

Description

Technical Field

[0001] This invention relates to the field of detection system technology, specifically to an online defect detection system for fabric label washing marks. Background Technology

[0002] Care labels (also known as washing labels or fabric labels) are indispensable identification components on clothing, home textiles, toys, and other fabric products. They typically contain information such as product specifications, washing instructions, ingredient content, and production batch. With increasingly stringent product quality traceability and brand management, the accuracy and completeness of the characters and barcode data on care labels have become one of the core indicators for customer inspection. Large purchasers, in particular, have extremely high requirements for the matching degree between label data and the source data in their Manufacturing Execution System (MES). Any label defects caused by printing errors, abnormal data conversion, or printing quality issues can lead to the return of the entire batch of products or even damage to brand reputation.

[0003] Currently, the industry primarily relies on manual visual inspection or offline sampling for the inspection of wash labeling. The specific procedure involves operators sampling products from the same batch and visually inspecting the labels or using handheld barcode scanners to determine for defects such as typos, duplicate or missing numbers, missing graphics, or font size discrepancies. However, this inspection method has significant shortcomings: firstly, manual inspection is heavily influenced by subjective factors, and prolonged operation can easily lead to visual fatigue, resulting in a higher rate of missed inspections; secondly, the sampling method cannot cover all products, making it difficult to detect occasional printing failures or batch-wide data errors.

[0004] Therefore, an online defect detection system for cloth label washing marks is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide an online defect detection system for fabric label washing marks in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention specifically adopts the following technical solution: The online defect detection system for fabric label care labels includes an image acquisition module for acquiring images of fabric label care labels to be inspected, and a sample management module for pre-acquiring standard product label samples and setting exposure parameters. The images to be inspected acquired by the image acquisition module are transmitted to an image comparison module, which compares the images to be inspected with the standard product label samples and sends the comparison results to a defect judgment module. The defect judgment module identifies defect information in the images to be inspected, and the defect information includes at least one or more of the following: typos, numerical printing errors, graphic defects, or font size deviations.

[0007] Furthermore, the sample management module includes a sample sampling unit and a parameter setting unit. The sample sampling unit is used to sample the standard product to be tested, and the parameter setting unit is used to set the exposure parameters of the sample to serve as a reference sample for comparison during the testing process.

[0008] Furthermore, the image comparison module performs pixel-level or feature-level comparison between the image to be detected and the standard product label sample to identify the difference regions between the image to be detected and the standard sample.

[0009] Furthermore, the defect judgment module further includes a character recognition unit, a graphic recognition unit, and a size detection unit. The character recognition unit is used to recognize the character content in the image to be detected and compare it with standard character data to determine whether there are typos or numerical errors. The graphic recognition unit is used to recognize the graphic elements in the image to be detected and compare them with standard graphic data to determine whether there are graphic defects. The size detection unit is used to detect the size of the characters or graphics in the image to be detected and compare it with standard size data to determine whether there are font size deviations.

[0010] Furthermore, it also includes a data verification module connected to the Manufacturing Execution System (MES) data. This data verification module is used to verify the correctness of the label data read from the image to be inspected against the source data of the MES, in order to identify data printing errors, duplications or confusion caused by human error, software failure, data conversion errors, network failures or printer failures.

[0011] Furthermore, the image acquisition module includes an industrial camera and a light source device that provides uniform illumination for the fabric label / washing label to be inspected. The industrial camera acquires the image to be inspected under preset exposure parameters.

[0012] The beneficial effects of this invention are as follows: This invention enables fully automated online inspection of fabric labels and care labels, replacing inefficient and unreliable manual visual inspection and significantly improving inspection efficiency and accuracy. By combining image comparison with a multi-dimensional defect judgment module, it can accurately identify various defects such as typos, numerical errors, graphic defects, and dimensional deviations, providing comprehensive inspection capabilities.

[0013] By introducing a data verification module that works in conjunction with the MES system, the detection dimension is extended from appearance to the source of the data, effectively intercepting data errors caused by software, network, or printer failures, and meeting the ultimate requirements of the high-end market for quality traceability. Attached Figure Description

[0014] Figure 1 This is a system structure block diagram of the present invention; Detailed Implementation To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0015] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0016] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0017] All electrical components mentioned in this article are connected to an external main controller and 220V AC mains power, and the main controller can be a conventional known device such as a computer that can control it.

[0018] In the description of the embodiments of the present invention, it should be noted that the terms "inner", "outer", "upper", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention.

[0019] like Figure 1 As shown, the online defect detection system for fabric label washing marks includes an image acquisition module for acquiring images of fabric label washing marks to be inspected, and a sample management module for pre-acquiring standard product label samples and setting exposure parameters. The images to be inspected acquired by the image acquisition module are transmitted to an image comparison module, which compares the images to be inspected with the standard product label samples and sends the comparison results to a defect judgment module. The defect judgment module identifies the defect information in the images to be inspected. The defect information includes at least one or more of the following: typos, numerical printing errors, graphic defects, or font size deviations.

[0020] The image acquisition module consists of an industrial camera, lens, and light source. It adopts a trigger-based acquisition method to ensure that each wash label is stably imaged at a fixed position.

[0021] The sample management module includes a sample sampling unit and a parameter setting unit. The sample sampling unit is used to sample the standard products to be tested, and the parameter setting unit is used to set the exposure parameters of the samples to serve as a benchmark sample for comparison during the testing process. More specifically, the pre-setting of standard samples and optimal parameters provides a stable and reliable benchmark for subsequent automated comparisons. At the same time, by solidifying the experience confirmed by human eyes into the system's digital benchmark, the testing process becomes more objective and standardized, effectively eliminating the quality risks caused by fluctuations in testing standards due to different operators or different time periods.

[0022] The sample management module selects qualified products as standard samples, acquires their images, and extracts their feature templates. It is characterized by the following steps: Step 1: Select qualified products as standard samples, collect their images, and extract their feature templates.

[0023] Step 2: Determine the optimal exposure parameters using an automatic exposure algorithm. Here, grayscale histogram analysis is used to concentrate the image's grayscale distribution within the target range. in: (I) represents the average gray level of the image; target The target grayscale mean.

[0024] The image comparison module performs pixel-level or feature-level comparison between the image to be detected and a standard product label sample to identify the regions of difference between the image to be detected and the standard sample. It is characterized by including the following steps: Step 1, Region Extraction: Crop roi_graphics from the difference regions and extract the standard graphic template (pre-stored). Step 2, Image Preprocessing: Convert to grayscale, apply Gaussian filter for noise reduction, and use histogram equalization to enhance contrast.

[0025] Step 3, Template Matching: Calculate the matching degree using Normalized Cross-Correlation (NCC): If the NCC is below the threshold, it is determined to be a graphic defect.

[0026] Step 4, Shape Analysis: Extract the contour and calculate Hu moments. Compare the Hu moment differences between the tested image and the standard image: If the difference exceeds the threshold, it is determined to be a graphic distortion.

[0027] Step 5, Defect Judgment: Based on the NCC and shape difference, output the defect type (damaged / deformed / normal).

[0028] The defect judgment module further includes a character recognition unit, an image recognition unit, and a size detection unit. The character recognition unit is responsible for extracting text information from the image and comparing it with the content read from the barcode to verify the correctness of the barcode information. This unit adopts an end-to-end text recognition method based on deep learning, which has good adaptability to text of variable length, complex backgrounds, and printed fonts. Its key feature is that it includes the following steps: Step 1: First, convert the color ROI to grayscale and normalize the pixel values ​​to [0,1]. Then, scale the image to a fixed height H. (1) I gray = 0.299R + 0.587G + 0.114B (2)I gray = I gray / 255 (3) cv2.resize(roi, (new_width,H)) Step 2: Then extract high-dimensional feature sequences from the image, with each feature vector corresponding to a local region of the image (usually a vertical strip).

[0029] Network structure: Multiple convolutional layers and pooling layers are used, and finally, a time-dimensional compression is applied to convert the feature map into a sequence of time-step feature dimensions.

[0030] Assuming the input image size is HW, after passing through a CNN, the height is compressed to 1 (or very small), the width becomes T, and the feature vector length at each time step is D.

[0031] Step 3: Capture the contextual dependencies between characters. Use a bidirectional LSTM, where the output of each time step is a concatenation of the forward and backward LSTMs.

[0032] Input: sequence X =( x 1, x 2,, x T ), each x t R D .

[0033] Forward LSTM: Backward LSTM: splicing output: Where H is the number of hidden units in the LSTM. Finally, a linear (fully connected) layer is used to connect h. t Mapped to character category number C (including whitespace tags): y t =softmax(Wh t +b)R C Step 4: Finally, perform decoding. (1) At each time step t, retain the k candidate paths with the highest current probability.

[0034] (2) For each path, expand to all possible characters (including whitespace) in the next time step to generate new candidate paths.

[0035] (3) Only keep the k paths with the highest total probability after expansion, and discard the rest of the paths.

[0036] (4) Finally, the path with the highest probability is selected, and the output sequence is obtained after B mapping (removing duplicates and empty strings).

[0037] The graphic recognition unit is used to identify graphic elements in the image to be detected and compare them with standard graphic data to determine whether there are graphic defects.

[0038] The size detection unit is used to detect the size of characters or graphics in the image to be detected and compare it with standard size data to determine whether there is a font size deviation. It is characterized by including the following steps: Step 1, Calibration: During system initialization, use a calibration board to obtain the conversion coefficients between pixels and actual sizes. For example, if a pattern on a calibration board is known to be 10mm wide and 200px wide, then... k =0.05mm / px Step 2, Target Extraction: If detecting character size, use OCR to segment and obtain the bounding box of a single character. If detecting graphic size, directly use the region contour provided by the image comparison module.

[0039] Step 3, Size Calculation: For each target, calculate its bounding rectangle: Convert to actual size: Step 4, Size Comparison: Standard size is W std , H std Tolerances are allowed.

[0040] Judgment conditions: It should be noted that the character recognition unit (such as an OCR engine) identifies the characters in the discrepancy area and compares them with preset standard character content to determine if there are any typos or numerical errors. The image recognition unit analyzes the graphic features of the discrepancy area and matches them with a standard graphic template to determine if there are any graphic defects or deformations. The size detection unit measures the pixel size of the characters or graphics in the discrepancy area, converts it to the actual physical size, and compares it with the standard tolerance range to determine if there are any deviations in font size or graphic size. Through this two-level processing mode of comparing differences and unit judgment, the detection speed is guaranteed to meet online requirements, while the specialized sub-units improve the recognition accuracy of complex defects (such as similar-looking character errors and minor size deviations).

[0041] It also includes a data verification module connected to the Manufacturing Execution System (MES). This module verifies the correctness of label data read from the image to be inspected against the source data in the MES, identifying data printing errors, duplications, or inconsistencies caused by human error, software malfunctions, data conversion errors, network failures, or printer malfunctions. It's important to note that this module establishes a communication connection with the factory's MES (e.g., via OPC UA, TCP / IP, etc.). The data verification module receives successfully identified label data (such as product batch number, specifications, ingredient codes, etc.) from the image to be inspected and sends a query request to the MES to retrieve the source data corresponding to the production order. The module then compares the two data item by item. If inconsistencies are found, such as an incorrect batch number printed on the label, a missing ingredient information, or redundant data not present in the MES, a data printing error is identified, and an alarm is triggered. The data verification module 5 elevates the system's detection capabilities from the physical image level to the information logic level, linking with the production management center (MES). Through real-time comparison, it can effectively intercept the most hidden quality problem—correct printing but incorrect content—caused by upstream software configuration errors, network transmission packet loss, printer memory corruption, or manual input errors, thus better meeting the core data traceability requirements of modern intelligent manufacturing.

[0042] The image acquisition module includes an industrial camera and a light source device that provides uniform illumination for the to-be-detected cloth label washing instructions. The industrial camera acquires the to-be-detected image under preset exposure parameters. More specifically, the industrial camera preferably uses a high-resolution area array CCD or CMOS camera and is fixed above the production line conveyor belt. The light source device provides uniform and shadowless illumination for the to-be-detected cloth label washing instructions. For example, a ring LED light source or a coaxial light source is used to eliminate the reflection and shadow interference caused by the texture or curved surface of the cloth label. The industrial camera precisely triggers and acquires a clear to-be-detected image under the exposure parameters set by the sample management module. The high-quality industrial camera and the professional light source device work together to ensure that the input to the subsequent processing module is a high-quality and feature-distinct original image. At the same time, by presetting the exposure parameters, the system can adapt to the imaging environment of different products, ensuring the stability of the image quality during continuous production.

[0043] To sum up: When the system starts, first, the sample management module establishes the standard image samples and imaging parameters of the current product. On the production line, every flowing cloth label washing instruction is triggered to acquire a high-definition to-be-detected image when passing under the image acquisition module. This image is immediately sent to the image comparison module for rapid comparison with the standard sample to locate any potentially abnormal areas. These areas are further submitted to the defect judgment module, where professional units such as characters, graphics, and dimensions inside it conduct precise analysis and classification to finally determine whether there are appearance defects and their types. At the same time, the key text or barcode data identified from the image is extracted and sent by the data verification module to the MES system for authenticity and correctness verification. Only the labels that have passed both the appearance defect detection and the data source verification can be determined as qualified by the system. Therefore, by constructing a comprehensive system integrating automated image processing, intelligent defect judgment, and real-time production data verification, the present invention realizes the all-round and online intelligent detection of the quality of cloth label washing instructions, replacing traditional manual detection with machine vision and data fusion technology to ensure the reliability of product quality and data traceability.

[0044] The above shows and describes 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 by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for detecting defects in an online label washing mark, characterized by, The system includes an image acquisition module for acquiring images of fabric labels and washing labels to be inspected, and a sample management module for pre-acquiring standard product label samples and setting exposure parameters. The images to be inspected acquired by the image acquisition module are transmitted to an image comparison module, which compares the images to be inspected with the standard product label samples and sends the comparison results to a defect judgment module. The defect judgment module identifies defect information in the images to be inspected, and the defect information includes at least one or more of the following: typos, numerical printing errors, graphic defects, or font size deviations.

2. The fabric marking water mark on-line defect detection system according to claim 1, wherein, The sample management module includes a sample sampling unit and a parameter setting unit. The sample sampling unit is used to sample the standard products to be tested, and the parameter setting unit is used to set the exposure parameters of the sampled products as a reference sample for comparison during the testing process.

3. The fabric marking online defect detection system according to claim 1, wherein, The image comparison module performs pixel-level or feature-level comparisons between the image to be detected and standard product label samples to identify the difference regions between the image to be detected and the standard samples.

4. The online defect detection system for fabric label washing tags according to claim 1, characterized in that, The defect judgment module further includes a character recognition unit, a graphic recognition unit, and a size detection unit. The character recognition unit is used to recognize the character content in the image to be detected and compare it with standard character data to determine whether there are typos or numerical errors. The graphic recognition unit is used to recognize the graphic elements in the image to be detected and compare them with standard graphic data to determine whether there are graphic defects. The size detection unit is used to detect the size of the characters or graphics in the image to be detected and compare it with standard size data to determine whether there are font size deviations.

5. The online defect detection system for fabric label washing tags according to claim 1, characterized in that, It also includes a data verification module that connects to the Manufacturing Execution System (MES) data. This data verification module is used to verify the correctness of the label data read from the image to be inspected against the source data of the MES, in order to identify data printing errors, duplicates or confusion caused by human error, software failure, data conversion errors, network failure or printer failure.

6. The online defect detection system for fabric label washing marks according to claim 1, characterized in that, The image acquisition module includes an industrial camera and a light source device that provides uniform illumination for the fabric label / wash label to be inspected. The industrial camera acquires the image to be inspected under preset exposure parameters.

Images