KR-20260062695-A - DEVICE ON PERSON RE-IDENTIFICATION IN THERMAL IMAGES AND METHOD THEREOF

KR20260062695AKR 20260062695 AKR20260062695 AKR 20260062695AKR-20260062695-A

Abstract

The present invention relates to a person re-recognition technology, and more specifically, to a person re-recognition system and method using a thermal imaging camera that re-recognizes a person within an image by performing ultra-high resolution restoration based on deep learning on an image captured by the thermal imaging camera. According to one embodiment of the present invention, a low-resolution thermal image is restored to ultra-high resolution that can be utilized for person re-recognition based on deep learning, thereby improving the accuracy of person re-recognition.

Inventors

박강령
정승용

Assignees

동국대학교 산학협력단

Dates

Publication Date: 20260507
Application Date: 20241029

Claims (10)

In a person re-recognition device using a thermal image, A restoration unit that restores a low-resolution thermal image into an ultra-high-resolution thermal image; A feature extraction unit for extracting features of the restored ultra-high resolution thermal image; An importance calculation unit that calculates importance using the above features; A loss function unit that calculates a loss function applied to the above-mentioned restoration unit; and A person re-recognition device using a thermal image, comprising a similarity calculation unit that calculates distance-based similarity between a restored thermal image and stored images.
In paragraph 1, The above feature extraction unit Extracting a feature map of the above ultra-high resolution thermal image, and Person re-recognition device using a thermal image that generates a feature vector based on the above feature map.
In paragraph 2, The above importance calculation unit A person re-recognition device using a thermal image that extracts an importance map based on the above feature vector and calculates an importance score.
In paragraph 1, The above loss function part A person re-recognition device using thermal images that calculates a loss value based on the average absolute error (L1 Loss) using the absolute value of the difference between the actual (original) value and the restored (predicted) value.
In a method for recognizing a person using a thermal image, performed by a person recognizing device using a thermal image, A step of performing ultra-high resolution restoration of low-resolution thermal images through a deep learning model; A step of extracting feature vectors based on a restored ultra-high resolution thermal image; A step of calculating similarity based on the above feature vector; and A method for recognizing a person using a thermal image, comprising the step of performing person recognizing using the above similarity.
In paragraph 5 The step of extracting feature vectors based on the above-mentioned restored ultra-high resolution thermal image is Extracting feature maps for human re-recognition based on thermal images restored to ultra-high resolution, and A method for recognizing a person using a thermal image that generates a feature vector based on the above feature map.
In paragraph 5 The step of calculating similarity based on the above feature vector A method for recognizing a person using a thermal image, which extracts an importance map based on the above feature vector and calculates an importance score.
In paragraph 5, A method for re-recognizing a person using a thermal image, wherein a loss function is calculated based on the mean absolute error (L1 Loss) using the absolute value of the difference between the actual (original) value and the restored (predicted) value, and the deep learning model applies the above method.
In paragraph 5, The step of performing person re-recognition using the above similarity is to A thermal image-based human re-recognition method that calculates distance-based similarity between the above-mentioned restored ultra-high resolution thermal image and the stored image.
A computer program recorded on a computer-readable recording medium that executes a method for re-identifying a person using a thermal image according to paragraph 5.

Description

Device and method for person re-identification using thermal images {DEVICE ON PERSON RE-IDENTIFICATION IN THERMAL IMAGES AND METHOD THEREOF} The present invention relates to a person re-recognition technology, and more specifically, to a person re-recognition system and method using a thermal imaging camera that re-recognizes a person within an image by performing ultra-high resolution restoration based on deep learning on an image captured by a thermal imaging camera. Recently, Person Re-IDentification (Re-ID) technology is being utilized in the security field and is playing an important role in implementing intelligent surveillance camera systems. However, while visible light cameras improve the accuracy of person re-identification (Re-ID) through high-resolution images in daytime environments, their accuracy is inevitably lower in nighttime or low-light environments, such as cloudy days or at night. To overcome these issues, research on human re-recognition utilizing thermal images from thermal cameras has recently begun, employing a method that combines visible light and thermal images. However, this approach has the disadvantage of being unusable in environments where visible light images cannot be acquired. Unlike visible light cameras, thermal cameras generate infrared-based images rather than standard visible light, allowing them to capture images of subjects even in low-light conditions. However, because thermal cameras designed to produce high-resolution images are expensive compared to visible light cameras, images from standard thermal cameras are often of low resolution. Meanwhile, surveillance systems using cameras often produce low-quality images due to various image degradation factors caused by camera hardware limitations or the influence of the surrounding environment during construction. To address these issues, research is being conducted to improve human recognition performance by utilizing ultra-high resolution reconstruction in human recognition technology. However, while the application of ultra-high resolution reconstruction to human recognition technology has been frequently attempted in technologies utilizing visible light images, it has rarely been attempted in technologies utilizing thermal images. FIGS. 1 to 6 are drawings for explaining a person re-recognition device using a thermal image according to an embodiment of the present invention. FIG. 7 is a diagram illustrating a method for recognizing a person using a thermal image according to an embodiment of the present invention. FIGS. 8 and 9 are drawings for explaining the results of a performance comparison of a person re-recognition device using a thermal image. FIG. 10 is a diagram illustrating a computing device implementing a person re-recognition device using a thermal image according to an embodiment of the present invention. The present invention is susceptible to various modifications and may have various embodiments. Specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. In describing the present invention, detailed descriptions of related prior art are omitted if it is determined that such detailed descriptions would unnecessarily obscure the essence of the invention. Furthermore, singular expressions used in this specification and claims should generally be interpreted as meaning "one or more" unless otherwise stated. In this specification, the term "module" includes a unit composed of hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be a component formed integrally, or a minimum unit or part thereof that performs one or more functions. For example, a module may be composed of an Application-Specific Integrated Circuit (ASIC). Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In describing with reference to the accompanying drawings, identical or corresponding components are given the same reference numerals, and redundant descriptions thereof will be omitted. FIG. 1 is a diagram briefly illustrating a person re-recognition device using a thermal image according to one embodiment of the present invention. Referring to FIG. 1, a person re-recognition device (10) using a thermal image performs super-high resolution restoration of a low-resolution thermal image (LR Image) captured by a thermal camera through a deep learning-based Part Attention and Contrastive Loss-based Super-resolution Reconstruction Network (PCSR-Net), extracts features necessary for person re-recognition from the restored super-high resolution thermal image (SR Image), generates an image nec