KR-20260065341-A - SUPER RESOLUTION-BASED OBJECT FEATURE IDENTIFICATION SYSTEM

KR20260065341AKR 20260065341 AKR20260065341 AKR 20260065341AKR-20260065341-A

Abstract

The present invention relates to an object feature identification system utilizing super-resolution, and in an object identification system utilizing super-resolution introduced in a component alignment machine, the system comprises: an original data acquisition unit that acquires an original image containing one or more target objects in the component alignment machine; a super-resolution unit that generates a super-resolution image through a resolution enhancement process for the original image; an object region acquisition unit that detects an object region from the super-resolution image; and an object feature extraction unit that extracts object feature data including the features and arrangement of an object from the object region.

Inventors

김웅섭
한요한

Assignees

주식회사 디밀리언

Dates

Publication Date: 20260508
Application Date: 20241101

Claims (10)

In an object identification system utilizing super-resolution introduced in a parts sorter, Original data acquisition unit for acquiring an original image containing one or more target objects in the above-mentioned part aligner; A super-resolution unit that generates a super-resolution image through a resolution enhancement process for the original image above; An object region acquisition unit that detects an object region from the above super-resolution image; and An object feature extraction unit that extracts object feature data including object features and arrays from the object region; comprising Object feature identification system utilizing super-resolution.
In paragraph 1, The above-mentioned component alignment device is, The components are aligned by applying vibrations to the components placed on the diaphragm to align them, and The above original data acquisition unit includes acquiring original images regarding the vibration process of a component on a diaphragm, Object feature identification system utilizing super-resolution.
In paragraph 1, The object feature extraction unit described above includes verifying the vibration response patterns of each object to recognize whether they are different objects, and extracting relative weight information between them. Object feature identification system utilizing super-resolution.
In paragraph 1, The above original data acquisition unit includes a plurality of image acquisition devices, and The plurality of image acquisition devices are arranged to have different directions and heights based on the target objects. Object feature identification system utilizing super-resolution.
In paragraph 1, The above original data acquisition unit includes irradiating a plurality of light sources onto the target object and acquiring a reflected image, and The object feature extraction unit described above includes extracting information about materials included in the objects. Object feature identification system utilizing super-resolution.
In paragraph 1, Further comprising an image preprocessing unit that removes noise from the original image or adjusts the contrast ratio of the original image; The above super-resolution unit generates the super-resolution image from the image preprocessed by the above image preprocessing unit, Object feature identification system utilizing super-resolution.
In paragraph 1, An object position recognition unit that recognizes the position coordinates and rotation angle of the object from the super-resolution image and extracts the position and orientation of the object; A data learning unit that trains a deep learning model by expanding training data through rotation, translation, and scaling of recognized objects; and A real-time recognition unit that re-recognizes an object in real time using the above-mentioned deep learning model; further comprising Object feature identification system utilizing super-resolution.
In paragraph 1, A result providing unit that visually provides information of the recognized object through a bounding box or label; and A feedback unit that analyzes falsely detected or undetected objects to retrain the deep learning model; further comprising Object feature identification system utilizing super-resolution.
In paragraph 1, The above super-resolution unit is, A method comprising obtaining a sampling image, which is a low-resolution image, through a quality degradation process on an original high-resolution image, and generating the super-resolution image by a super-resolution model trained with image data paired with the high-resolution image and the low-resolution image. Object feature identification system utilizing super-resolution.
In Paragraph 9, The low-resolution sampled image further includes the original low-resolution image actually captured, and The above super-resolution model is trained on a pair of actual original low-resolution and high-resolution images obtained from a single object, Object feature identification system utilizing super-resolution.

Description

Object Feature Identification Using Super Resolution {SUPER RESOLUTION-BASED OBJECT FEATURE IDENTIFICATION SYSTEM} The present invention relates to an object identification system that can be introduced into a part sorter, and more specifically, to a system that helps accurately recognize parts of a part sorter by utilizing super-resolution images. To enhance the production efficiency of part sorters used in existing manufacturing and assembly lines, technologies that recognize and classify parts using various types of sensors, particularly cameras, are being applied. Camera-based part sorters analyze the precise size, shape, and position of parts, enabling automated part sorting based on this analysis. In this case, the resolution of the video captured by the camera significantly impacts the system's performance. Particularly when even small parts need to be recognized in detail, low-resolution images fail to clearly capture the fine features of the parts, which can lead to recognition errors or reduce the accuracy of part classification. This negatively affects the speed and accuracy of part classification and can lower the overall efficiency of the production process. FIG. 1 is a diagram showing the configuration of an object feature identification system according to one embodiment of the present invention. FIG. 2 is a flowchart of an object feature identification method according to one embodiment of the present invention. FIG. 3 is a perspective view of a part aligner to which an object feature identification system according to one embodiment of the present invention is applied. FIG. 4 is an example diagram of an original data acquisition unit of an object feature identification system according to one embodiment of the present invention. The embodiments of the present invention are illustrative for the purpose of explaining the technical concept of the present invention. The scope of rights according to the present invention is not limited to the embodiments presented below or the specific description thereof. All technical and scientific terms used in this invention, unless otherwise defined, have the meaning generally understood by those skilled in the art to which this invention pertains. All terms used in this invention are selected for the purpose of further explaining this invention and are not selected to limit the scope of rights according to this invention. Expressions such as "comprising," "having," "having," etc. used in the present invention should be understood as open-ended terms implying the possibility of including other embodiments, unless otherwise stated in the phrase or sentence containing such expressions. Unless otherwise stated, singular expressions described in the present invention may include the meaning of the plural form, and this applies likewise to singular expressions described in the claims. Embodiments of the present invention will be described in more detail below based on the drawings illustrated in FIGS. 1 to 4. In the following, each component is described in detail with reference to FIGS. 1 to 4. However, the embodiments illustrated in FIGS. 1 to 4 are merely examples, and can be replaced with other structures within the scope where the technical concept of the present invention can be extended. FIG. 1 is a diagram showing the configuration of an object feature identification system utilizing super-resolution according to an embodiment of the present invention. An object identification system utilizing super-resolution introduced into a component alignment device according to one embodiment of the present invention may include: an original data acquisition unit; a super-resolution unit; an object region acquisition unit; and an object feature extraction unit. The above-mentioned part aligner may be a device that aligns multiple types of parts simultaneously. The original data acquisition unit acquires an original image containing one or more target objects from the component aligner, the super-resolution unit generates a super-resolution image through a resolution enhancement process for the original image, the object region acquisition unit detects an object region from the super-resolution image, and the object feature extraction unit extracts object feature data including the features and arrangement of the object from the object region. Specifically, the original data acquisition unit may use one or more of the following devices. For example, the original data acquisition unit may use one or more means of image acquisition devices such as a digital camera, a line scan camera, a multispectral or hyperspectral camera, or a proximity sensor camera. The above digital cameras include CCD cameras, CMOS cameras, 3D cameras, high-speed cameras, smartphone cameras, etc. The above super-resolution unit may include a super-resolution model. The above super-resolution model may include SRCNN (Super-Resolution Convolutional Neural Network), VDSR (Very Deep Super-Resoluti