KR-20260062769-A - SERVER AND METHOD FOR CLUSTERING OBJECT AND SYSTEM INCLUDING THE SAME

KR20260062769AKR 20260062769 AKR20260062769 AKR 20260062769AKR-20260062769-A

Abstract

The present invention relates to an object clustering server and method capable of performing object segmentation as well as object clustering by utilizing not only electro-optical images but also multi-spectral images, and a system including the same. The object clustering server may include a data collection module that receives satellite images, a segmentation module that segments objects included in the satellite images to generate segmentation results, an extraction module that extracts features of the segmented objects included in the segmentation results, and a clustering module that performs clustering on the objects included in the satellite images based on the extracted features to generate clustering results.

Inventors

구아롬
박현선
김동철
강명훈
사공용협
이여울
이용주

Assignees

주식회사 메이사

Dates

Publication Date: 20260507
Application Date: 20241230

Claims (16)

Memory for storing at least one instruction; and It includes at least one processor that executes the above at least one instruction, and The above processor is, Receive satellite imagery, The object included in the above satellite image is segmented to generate a segmentation result, and Extract the features of the divided object included in the above division result, and Based on the extracted features above, clustering is performed on the objects included in the satellite image to generate a clustering result. Object Clustering Server.
In Article 1, The processor generates the partitioning result by partitioning the object using a predefined partitioning model. Object Clustering Server.
In Article 2, The above-mentioned segmentation model includes at least one of the SAM model (Segment Anything Model) and the DINO model (Distillation with NO labels Model). Object Clustering Server.
In Article 1, The above satellite image includes electro-optical imagery and multi-spectral imagery. Object Clustering Server.
In Paragraph 4, The above electro-optical image includes an RGB (Red, Green, Blue) image. Object Clustering Server.
In Paragraph 4, The above multispectral image includes at least one of a near-infrared (NIR) image and a short-wave infrared (SWIR) image. Object Clustering Server.
In Paragraph 4, The above processor is, Extracting the feature from at least one of the segmentation result for the electro-optical image and the segmentation result for the multispectral image. Object Clustering Server.
In Article 7, The above processor is, Extracting at least one of the color, texture, and pattern of the segmented object as a feature from the electro-optical image Object Clustering Server.
In Article 7, The above processor is, Extracting at least one of the reflectance and spectrum of the segmented object as a feature from the multispectral image Object Clustering Server.
In Article 7, The above processor is, Preprocessing is performed on the above satellite image to generate a preprocessing result, and Generating the clustering result using the above preprocessing result, the above features, and a pre-trained clustering model Object Clustering Server.
In Article 10, The processor generates the preprocessing result by performing preprocessing related to the number of channels of the satellite image. Object Clustering Server.
In Article 11, The above processor is, A first preprocessing result having a predefined reference number of channels is generated by combining the electro-optical image and the multispectral image, or Copying the above multispectral image to generate a second preprocessing result having the above reference number of channels Object Clustering Server.
In Article 10, The above processor is, Generating the clustering result by inputting the above preprocessing result and the above feature into the clustering model Object Clustering Server.
In Article 13, The above clustering model generates the clustering result by performing clustering on the object based on the similarity between a plurality of the above features. Object Clustering Server.
In Article 14, The above clustering model is, If a predefined prior classification exists regarding the type of the above object, the object is classified into one of the existing classes included in the prior classification or a new class not included in the prior classification based on the above features, and the classified result is determined as the clustering result. If there is no predefined prior classification regarding the types of the above objects, a group of similar objects among the above objects that have similar characteristics is determined, and the determined group is determined as the clustering result. Object Clustering Server.
In Article 15, The above clustering model is, A first model based on K-means clustering and a second model based on a neural network, comprising at least one of the two Object Clustering Server.

Description

SERVER AND METHOD FOR CLUSTERING OBJECT AND SYSTEM INCLUDING THE SAME The present invention relates to an object clustering server, a method, and a system including the same. Specifically, the present invention relates to an object clustering server and method capable of performing object segmentation as well as object clustering by utilizing not only electro-optical images but also multi-spectral images, and a system including the same. The content described in this section merely provides background information regarding the present embodiment and does not constitute prior art. Recently, satellite imagery is being utilized to acquire information about buildings, roads, and nature. Such satellite imagery is being used in various fields, including detecting the risk of wildfires and assessing vegetation indices. In the process of utilizing such satellite imagery, it is necessary to identify the objects included in the images. Generally, while technological development is progressing regarding the segmentation of objects contained in these satellite images, technological development related to clustering and classifying the segmented objects has not progressed sufficiently. In addition, conventional technology performed object segmentation using only electro-optical images, such as RGB (Red, Green, Blue) images, which resulted in a problem of reduced accuracy during the object segmentation process. FIG. 1 illustrates an object clustering system according to some embodiments of the present invention. FIG. 2 is a diagram illustrating the neural network structure of a neural network model used by an object clustering server according to some embodiments of the present invention. FIG. 3 is a block diagram of an object clustering server according to some embodiments of the present invention. FIG. 4 is a conceptual diagram illustrating the operation of an object clustering server according to some embodiments of the present invention. FIG. 5 is a detailed block diagram of a split module according to some embodiments of the present invention. FIGS. 6 and 7 are drawings for explaining the learning and execution steps of a partitioning model used by a partitioning module according to some embodiments of the present invention. FIG. 8 is a detailed block diagram of an extraction module according to some embodiments of the present invention. FIG. 9 is a diagram illustrating the operation of an extraction module according to some embodiments of the present invention. FIGS. 10 and FIGS. 11 are drawings for explaining features extracted by an extraction module according to some embodiments of the present invention. FIG. 12 is a detailed block diagram of a clustering module according to some embodiments of the present invention. FIG. 13 is a detailed block diagram of a preprocessing unit included in a clustering module according to some embodiments of the present invention. FIGS. 14 to 16 are drawings for explaining the operation of a preprocessing unit according to some embodiments of the present invention. FIG. 17 is a diagram illustrating the types of clustering models according to some embodiments of the present invention. FIGS. 18 and 19 are drawings for illustrating the learning and execution steps of a second model included in a clustering model according to some embodiments of the present invention. FIG. 20 is a block diagram of an object clustering server according to some other embodiments of the present invention. FIG. 21 is a flowchart of an object clustering method according to some embodiments of the present invention. FIG. 22 is a detailed flowchart of the step of generating a segmented result according to some embodiments of the present invention. FIG. 23 is a detailed flowchart of a feature extraction step according to some embodiments of the present invention. FIG. 24 is a detailed flowchart of the step of extracting features for an electro-optical image included in the feature extraction step according to some embodiments of the present invention. FIG. 25 is a detailed flowchart of the step of extracting features for a multispectral image included in the feature extraction step according to some embodiments of the present invention. FIG. 26 is a detailed flowchart of a clustering step according to some embodiments of the present invention. FIG. 27 is a detailed flowchart of the preprocessing step included in the clustering step according to some embodiments of the present invention. FIG. 28 is a detailed flowchart of the preprocessing step included in the clustering step according to some other embodiments of the present invention. FIG. 29 is a detailed flowchart of the execution steps included in the clustering step according to some embodiments of the present invention. FIG. 30 is a diagram illustrating a hardware implementation of an object clustering server according to some embodiments of the present invention. Terms and words used in this specification and claims shall not be interpreted as being li