KR-102963936-B1 - System and method for inspecting of laminating floor using AI

Abstract

The present invention relates to a floor inspection system and method using artificial intelligence, comprising: a high-speed marker sticker device for processing markers or stickers to distinguish samples of a floor to be inspected; a lighting means for irradiating light onto the floor to be inspected; a first group column of shooting means arranged at regular intervals and a second group column of shooting means arranged at regular intervals and forming a regular angle with the first group column of shooting means, which capture an image of the floor by the reflected light of the light irradiated from the lighting means; an artificial intelligence modeling processing judgment unit that extracts an image from the image transmitted by the shooting means to collect data, refines the data, labels the data, constructs a training data set, and builds an AI model by repeating a deep learning process, thereby immediately displaying an inspection result based on a defect judgment to a user in real time when an inspection image of the floor to be inspected is input; and a real-time inspection monitoring unit linked to the artificial intelligence modeling processing inspection judgment unit to monitor floor defects produced in real time.

Inventors

• 박상열

Dates

Publication Date

20260513

Application Date

20250314

Claims (6)

1. In a floor inspection system for inspecting defects in multiple floors loaded and transported on a conveyor belt, A high-speed sticker marker device for processing markers or stickers to distinguish and recognize samples on floors under inspection; Lighting means for irradiating light onto the floor subject to inspection; An image of the floor is captured by the reflected light of the light irradiated from the above lighting means, and a first group column of shooting means arranged at regular intervals and a second group column of shooting means arranged at regular intervals, forming a regular angle with the first group column of shooting means; An artificial intelligence modeling processing judgment unit that extracts an image from a picture transmitted by the above-mentioned shooting means to collect data, refines the data, labels the data, constructs a training data set, and constructs an AI model by repeating a deep learning process, thereby immediately transmitting an inspection result based on a defect judgment to a user in real time when an inspection image of a floor subject to inspection is input; and It includes a real-time inspection monitoring unit in which products that have entered the classification process—which classifies products into good and defective items based on inspection results by an artificial intelligence modeling processing judgment unit—are discharged, with good products going to the packaging process and defective products going to the defective processing process. The artificial intelligence modeling processing judgment unit, A training data construction unit that extracts images to collect data, refines the data, labels the data, and constructs a training data set; A data preprocessing unit that defines business processes based on image extraction, collects data by scenario, deletes damaged images from extracted captured images, reduces dimensions for data refinement, and processes refined images; A model selection learning unit that extracts image feature points by self-analyzing a training data set through a training data neural network and establishes image judgment criteria in an artificial intelligence model based on the extracted image feature points; An AI modeling process unit that constructs an AI modeling training model through data stratification and clustering from a constructed training dataset, goes through a model performance evaluation process using an evaluation dataset with training data and validation data set in a certain ratio to set hyperparameters for model optimization, and provides an inspection monitoring service by improving the AI modeling training model through feedback based on model performance verification; and A floor inspection system using artificial intelligence, characterized by including an artificial intelligence (AI) modeling evaluation verification unit that verifies model performance corresponding to a target floor defect prediction rate, analyzes performance improvement, and feeds back to an artificial intelligence (AI) modeling process unit.
2. In Article 1, A floor inspection system using artificial intelligence characterized by AI modeling being performed through a local convolutional neural network (RCNN) for model selection or training.
3. In Article 1, A floor inspection system using artificial intelligence characterized by using SSD (Single Shot MultiBox Detector) as a deep learning-based algorithm for object detection in model selection or training.
4. delete
5. delete
6. delete

Description

System and method for inspecting laminating floor using AI The present invention relates to a floor inspection system and method using artificial intelligence, and more specifically, to a floor inspection system and method using artificial intelligence that can provide inspection results to a user when an image of a floor to be inspected is input, by extracting a large number of floor images through multiple camera shots to determine defects in floor products and building an AI model through data labeling and deep learning. In general, in production lines for manufacturing or processing flooring, an inspection device is used to detect abnormalities in the flooring (such as foreign matter contamination, contamination, cracks, etc., hereinafter referred to as defects) by using an image obtained by irradiating visible light or ultraviolet light onto the flooring and photographing the surface and appearance of the flooring with a camera using the transmitted or reflected light. In conventional inspection devices, to inspect abnormal parts of the floor using feature quantities obtained by image processing of captured images, inspection criteria (hereinafter referred to as threshold values) corresponding to each type of abnormality are set based on a limited number of defective samples, and the determination of good and bad is performed accordingly. According to the above method, in order to set the threshold value for anomaly detection regarding the aforementioned characteristic quantity, samples are required to distinguish between good and defective products for each type of anomaly. Typically, since it is not easy to obtain limit samples capable of distinguishing between good and defective products, threshold values are set by estimating non-uniformity based on available samples; however, there were cases where the optimal threshold value could not be set when the actual non-uniformity differed from the assumption. As described above, in situations where limit samples are not sufficiently available, such as when installing a new device or adding new items to the inspection target, there was a problem in that inspections could not be performed with appropriate precision and, in some cases, the inspection device could not be operated easily. According to published patent 10-2023-0139167 regarding a technology to solve this problem, a laminate floor inspection device is provided that can automatically perform defect inspection with high precision using an external image of the laminate floor. However, according to the above technology, the process of acquiring the feature quantity distribution of each pixel in the image of the laminate flooring and correcting the threshold value according to the pattern map must be repeated, and the inspection criteria are modified by repeating the process of correcting the pattern map again, so the problem of the inspection process being complex and the difficulty of consistently maintaining precision according to the criteria still persists. FIG. 1 is an overall block diagram of a floor inspection system using artificial intelligence according to the present invention. FIG. 2 is a drawing of an embodiment implementing a floor inspection system using artificial intelligence according to the present invention. FIG. 3 is an example showing the location and installation method of an inspection device for applying a floor inspection system using artificial intelligence according to one embodiment of the present invention shown in FIG. 2 to a floor production line. Figure 4 is a photograph showing existing visual inspection sample data in reinforced floor inspection. Figure 5 shows inspection type items in a floor inspection system using artificial intelligence according to the present invention. FIG. 6 is a detailed block diagram of the artificial intelligence modeling processing judgment unit in a floor inspection system using artificial intelligence according to the present invention. FIG. 7 is a process of constructing training data in the training data construction unit of a floor inspection system using artificial intelligence according to the present invention. FIG. 8 is a diagram of the modification preprocessing process of the data preprocessing unit in a floor inspection system using artificial intelligence according to the present invention. FIG. 9 is a diagram of the model selection and learning process in a floor inspection system using artificial intelligence according to the present invention. FIG. 10 illustrates the artificial intelligence model processing and service operation process in a floor inspection system using artificial intelligence according to the present invention. FIG. 11 is an example of a real-time monitoring system linked to a floor inspection system using artificial intelligence according to the present invention. FIG. 12 is a flowchart illustrating a floor inspection method using artificial intelligence according to the present invention. FIG. 13 s