KR-20260062136-A - APPARATUS AND METHOD FOR IMAGE PROCESSING AND COMPUTER-READABLE RECORDING MEDIUM INCLUDING THE SAME

KR20260062136AKR 20260062136 AKR20260062136 AKR 20260062136AKR-20260062136-A

Abstract

An image processing device and method, and a computer-readable recording medium having a computer program for executing the method on a computer are provided. An image processing device according to an embodiment of the present invention receives an image of an object, performs preprocessing of the image of the object to generate a Region of Interest (ROI) image, extracts corner information of the object using the ROI image, and performs three-dimensional dimensional measurement of the object using the corner information.

Inventors

김준
김다윗
김보현

Assignees

한국생산기술연구원

Dates

Publication Date: 20260507
Application Date: 20241025

Claims (16)

As an image processing device, One or more processors; and When executed by the above-mentioned one or more processors, the system includes one or more memories in which instructions are stored to cause the above-mentioned one or more processors to perform operations, and The above one or more processors, Receive the target image, Preprocessing of the above target image is performed to generate an ROI (Region of Interest) image, and Using the above ROI image, corner information of the object is extracted, and Performing three-dimensional dimensional measurement of the object using the above corner information, Image processing device.
In Article 1, The above one or more processors, When generating the above ROI image, The image quality of the above-mentioned object image is enhanced to generate a high-quality image, and Generating the ROI image by removing the background from the above high-quality image, Image processing device.
In Article 2, The above one or more processors, When generating the above high-quality image, Adjusting pixel values by calculating the relative illuminance of adjacent pixels located within a Gaussian kernel range for a first pixel in the above-mentioned object image. Image processing device.
In Paragraph 3, The above one or more processors, When generating the above ROI image, Generating the above ROI image using the U2-Net deep learning model, Image processing device.
In Article 1, The above one or more processors, When extracting corner information of the above object, Extract the outline on the above ROI image, and Curvature information is calculated using the perpendicular distance between the first straight line and the points located on the first curve, based on a first triangle formed by a first straight line connecting two endpoints spaced a first distance from the center of the first curve on the above outline. Extracting corner information of the object using the above curvature information, Image processing device.
In Article 5, The above one or more processors, When extracting corner information of the above object, Extracting the point on the first curve where the perpendicular distance between the points located on the first straight line and the first curve is the largest as the first corner of the object. Image processing device.
In Article 6, The above one or more processors, When performing the above three-dimensional dimensional measurement, Calculate the pixel distance between corners using the corner information of the above object, and Performing three-dimensional dimensional measurement of the object using the calculated pixel distance, Image processing device.
One or more processors; and An image processing method by an image processing device comprising one or more memories in which instructions are stored to cause the one or more processors to perform operations when executed by the one or more processors, wherein A step of receiving an object image by one or more processors; A step of generating a Region of Interest (ROI) image by performing preprocessing of the target image by the above one or more processors; A step of extracting corner information of an object using the ROI image by the above one or more processors; and A step comprising performing three-dimensional dimensional measurements of the object using the corner information by the above one or more processors, Image processing method.
In Article 8, The step of generating the above ROI image is, A step of generating a high-quality image by performing image quality enhancement of the target image by the above one or more processors; and The method comprises the step of generating the ROI image by removing the background from the high-quality image by the above one or more processors. Image processing method.
In Article 9, The step of generating the above high-quality image is, A step comprising adjusting pixel values by calculating the relative illuminance of adjacent pixels located within a Gaussian kernel range for a first pixel in the object image by the above-mentioned one or more processors, Image processing method.
In Article 10, The step of generating the above ROI image is, A method comprising the step of generating the ROI image using a U2-Net deep learning model by one or more of the above processors, Image processing method.
In Article 8, The step of extracting corner information of the above object is, A step of extracting an outline on the ROI image by one or more processors; A step of calculating curvature information using the perpendicular distance between points located on the first straight line and the first curve, based on a first triangle formed by a first straight line connecting two endpoints spaced a first distance from the center of the first curve on the outline, by the above-mentioned one or more processors; and A step comprising extracting corner information of the object using the curvature information by the above one or more processors, Image processing method.
In Article 12, The step of extracting corner information of the above object is, The method comprises the step of extracting, by the above-mentioned one or more processors, the point on the first curve where the perpendicular distance between the first straight line and the point on the first curve is the largest as the first corner of the object. Image processing method.
In Article 13, The step of performing the above three-dimensional dimensional measurement is, A step of calculating pixel distances between corners using corner information of the object by the above-mentioned one or more processors; and A step comprising performing a three-dimensional dimensional measurement of the object using pixel distances calculated by the one or more processors above, Image processing method.
One or more processors; and A computer-readable recording medium having a computer program for execution on a computer comprising one or more memories in which instructions are stored for the one or more processors to perform operations when executed by the one or more processors, wherein A step of receiving an object image by one or more processors; A step of generating a Region of Interest (ROI) image by performing preprocessing of the target image by the above one or more processors; A step of extracting corner information of an object using the ROI image by the above one or more processors; and A computer-readable recording medium having a computer program recorded thereon for executing, by one or more processors, the step of performing three-dimensional dimensional measurements of the object using the corner information.
In Article 15, The step of extracting corner information of the above object is, A step of extracting an outline on the ROI image by one or more processors; A step of calculating curvature information using the perpendicular distance between points located on the first straight line and the first curve, based on a first triangle formed by a first straight line connecting two endpoints spaced a first distance from the center of the first curve on the outline, by the above-mentioned one or more processors; and A computer-readable recording medium having a computer program recorded thereon for executing a step of extracting corner information of an object using the curvature information by the above one or more processors.

Description

Apparatus and method for image processing, and a computer-readable recording medium having a computer program for executing the method on a computer. The present invention relates to an image processing device and method, and a computer-readable recording medium having a computer program for executing the method on a computer. More specifically, it relates to an image processing device and method capable of measuring three-dimensional dimensions by detecting corners from the outlines of low-quality images, and a computer-readable recording medium having a computer program for executing the method on a computer. With the recent growth of e-commerce and the expansion of the online market, the demand for delivery of various consumer products is increasing, leading to a corresponding rise in the importance and scale of the courier industry. Courier companies are employing strategies such as optimizing delivery routes, introducing electric vehicles, and diversifying delivery methods based on cargo characteristics to maximize profits through the provision of efficient delivery services. In particular, there is a recent trend to further subdivide freight rates by incorporating information on cargo size and volume, in addition to the previously considered weight, when establishing pricing criteria. In order to utilize the size and volume information of parcel shipments as a basis for determining freight rates, the automation of cargo volume measurement is necessary. To this end, high-performance sensor-based equipment is being developed to accurately measure the three-dimensional dimensions of cargo; however, commercialization and field application are difficult due to the high cost of introducing such equipment. As an alternative, methods to measure dimensions from images of parcel shipments using computer vision technology are being researched, and a representative related research field is '3D reconstruction'. 3D reconstruction is a technology that estimates the 3D shape of an object from a set of 2D images captured from multiple angles, and it has been primarily studied in the field of surveying engineering. Early research focused mainly on generating 3D models of objects using deep learning techniques; however, as the accuracy of 3D model generation has improved, research has recently expanded to measuring dimensions from these 3D models. While 3D reconstruction technology offers the advantage of generating 3D models with high accuracy, it also has limitations, such as the need for complex labeling tasks for training, the necessity of constructing image sets capturing objects from multiple angles, and relatively long computation times. For these reasons, there are difficulties in applying and commercializing automated volume measurement of parcels using 3D reconstruction technology in logistics environments where the daily volume of parcels reaches millions and the installation of complex hardware is challenging. One method that utilizes relatively fewer computing resources, although less accurate than 3D reconstruction technology, involves detecting corner points of an object within an image and calculating pixel distances between corners to estimate the size of the object. This method can be utilized if information such as the distance between the lens and the object, the camera's installation angle and position, and the size of a reference object is known in advance; however, it has limitations, such as the potential for significant errors in accuracy if camera distortion is present, and it is more suitable for measuring 2D dimensions than 3D dimensions. FIG. 1 is an exemplary diagram of an image of an object according to an embodiment of the present invention. FIG. 2 is a conceptual diagram of an image processing device according to an embodiment of the present invention. FIG. 3 is an example diagram comparing an object image and a high-quality image according to an embodiment of the present invention. Figure 4 is an example of an ROI image finally generated in an SOD module by applying a U2-Net algorithm to an image with improved quality by applying an SSR algorithm according to an embodiment of the present invention. FIG. 5 is an exemplary diagram showing the relative position of a pixel corresponding to a major corner of an object according to an embodiment of the present invention. FIG. 6 is a conceptual diagram illustrating the curvature estimation scale defined in the LPD-CD algorithm according to an embodiment of the present invention. FIG. 7 is a conceptual diagram illustrating the operation of the LPD-CD algorithm according to an embodiment of the present invention. FIG. 8 is an example diagram showing the pseudocode of the LPD-CD algorithm according to an embodiment of the present invention. FIG. 9 is an example diagram comparing the original image of an object according to an embodiment of the present invention with a corner image detected after applying a CD module. FIG. 10 is an example diagra