Method and system for analyzing green space

Abstract

The present invention relates to a method and system for analyzing green space. More specifically, the green space analysis method according to the present invention may include the steps of: collecting satellite images taken of a specific city; generating a feature map of the satellite images using an image encoder of a green space segmentation model; inputting the feature map into a segmentation module of the green space segmentation model; extracting pixels corresponding to a green space class from the satellite images using the feature map in the segmentation module; and calculating the area of ​​green space included in the satellite images using the extracted pixels.

Description

Technology Field

[0001] The present invention relates to a method and system for analyzing green spaces. More specifically, the present invention relates to a method and system for analyzing green spaces capable of precisely segmenting and quantifying green spaces in satellite imagery. Background Technology

[0002] Recently, rapid urban expansion is taking place due to industrialization and the concentration of the population in cities.

[0003] Such urban expansion offers various benefits, such as improved accessibility to urban infrastructure, the realization of economic agglomeration effects, and increased efficiency in land use.

[0004] However, such urban expansion has limitations, as it entails environmental problems such as climate disturbance, biodiversity loss, ecosystem degradation, and the urban heat island effect. In particular, the urban heat island phenomenon causes localized temperature increases and is pointed out as a serious issue affecting the health of urban residents by increasing the incidence and mortality rates of cardiovascular diseases, respiratory diseases, heat-related illnesses, and mental health disorders.

[0005] As part of Nature-based Solutions (NbS) to address these issues, the management and continuous expansion of urban green spaces are receiving attention, and accordingly, the demand for quantitative monitoring and management technologies for urban green spaces is increasing.

[0006] Urban green space refers to vegetation-based areas located within or adjacent to a city, created naturally or artificially. By performing environmental functions such as air purification, temperature regulation, carbon absorption, and biodiversity conservation, these urban green spaces serve as core infrastructure that strengthens the health and resilience of the city's ecosystem.

[0007] Accordingly, there is a need for monitoring technology that can effectively utilize urban green spaces and identify their location, scope, and area. The problem to be solved

[0008] The present invention relates to a green space analysis method and system for precise urban green space division.

[0009] In addition, the present invention relates to a green space analysis method and system for monitoring the location and area of ​​urban green spaces in real time.

[0010] In addition, the present invention relates to a green space analysis method and system for providing basic information for establishing green space-based urban planning.

[0011] In addition, the present invention relates to a green space analysis method and system for stable green space division even under various conditions and complex urban terrain. means of solving the problem

[0012] To solve the problem described above, the green space analysis method according to the present invention may include the steps of: collecting satellite images taken of a specific city; generating a feature map of the satellite images using an image encoder of a green space segmentation model; inputting the feature map into a segmentation module of the green space segmentation model; extracting pixels corresponding to a green space class from the satellite images using the feature map in the segmentation module; and calculating the area of ​​green space included in the satellite images using the extracted pixels.

[0013] Furthermore, the step of extracting the pixel may include the step of dividing the satellite image captured for the specific city into a plurality of classes based on a pre-set division criterion in the segmentation module, the step of identifying a green space class corresponding to green space among the divided plurality of classes using the green space division model, and the step of extracting the pixel corresponding to the green space class.

[0014] Furthermore, the step of calculating the area of ​​the green space may include the step of generating a binary mask for the satellite image using the pixels corresponding to the green space class, and calculating the area of ​​the green space using the binary mask.

[0015] Furthermore, in the step of calculating the area of ​​the green space, the number of pixels of the binary mask is calculated, and the area of ​​the green space included in the satellite image can be calculated using the calculated number of pixels.

[0016] Furthermore, the above plurality of classes may include at least one of the background, buildings, green spaces, roads, and rivers of the specific city.

[0017] Furthermore, the above green space segmentation model is a model trained using a pre-set loss function, and the pre-set loss function may include a hybrid loss function that combines a cross-entropy loss function and a Dice loss function.

[0018] Furthermore, the green space segmentation model can be trained using a pre-set loss function that combines the cross-entropy loss function defined from the classification error of pixels corresponding to the green space class and the Dice loss function defined from the degree of overlap between the prediction result for the green space class and the green space area.

[0019] Furthermore, the above-mentioned loss function includes weights for each of the cross-entropy loss function and the die loss function, and the weights can be adjusted through the green space partitioning model.

[0020] Meanwhile, in a method and system for analyzing green space, satellite images captured for a specific city are collected, and a feature map for the satellite images is generated using an image encoder of a green space segmentation model. The feature map is input into a segmentation module of the green space segmentation model, and in the segmentation module, pixels corresponding to a green space class are extracted from the satellite images using the feature map. The area of ​​green space included in the satellite images can be calculated using the extracted pixels.

[0021] Meanwhile, the program according to the present invention is a program that is executed by one or more processes in an electronic device and can be stored on a computer-readable recording medium, and may include instructions for performing the steps of: collecting satellite images taken for a specific city; generating a feature map for the satellite images using an image encoder of a green space segmentation model; inputting the feature map into a segmentation module of the green space segmentation model; extracting pixels corresponding to a green space class from the satellite images using the feature map in the segmentation module; and calculating the area of ​​green space included in the satellite images using the extracted pixels. Effects of the invention

[0022] As described above, the green space analysis method and system according to the present invention can precisely segment narrow green zones and areas with unclear boundaries even in complex satellite imagery. Through this, the actual spatial distribution of green spaces within a city can be accurately identified.

[0023] Furthermore, the green space analysis method and system according to the present invention can secure high boundary accuracy without loss of detailed structure. Through this, narrow and fragmented green spaces can be accurately identified even in complex terrain structures such as roads.

[0024] Furthermore, the green space analysis method and system according to the present invention enables accurate area calculation. Accordingly, it can provide highly reliable information that can be utilized for various policy formulation and environmental analysis tasks.

[0025] Furthermore, the green space analysis method and system according to the present invention can analyze green spaces in real time. Accordingly, the green space monitoring system can be automated. Brief explanation of the drawing

[0026] FIGS. 1, FIGS. 2, FIGS. 3 and FIGS. 4 are conceptual diagrams for explaining a green space analysis method and system according to the present invention. FIG. 5 is a flowchart illustrating a green space analysis method and system according to the present invention. FIGS. 6, FIGS. 7, and FIGS. 8 are conceptual diagrams for explaining a green space analysis method and system according to the present invention. FIG. 9 is a formula for explaining the green space analysis method and system according to the present invention. Specific details for implementing the invention

[0027] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components are assigned the same reference number regardless of the drawing symbols, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not have distinct meanings or roles in themselves. Furthermore, in describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0028] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0029] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0030] Singular expressions include plural expressions unless the context clearly indicates otherwise.

[0031] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0032] The present invention relates to a method and system for analyzing green space. The method and system for analyzing green space according to the present invention may be a system that performs precise segmentation and area quantification of green space in satellite imagery using an image encoder and a segmentation module.

[0033] Furthermore, the green space analysis method and system according to the present invention may include a green space segmentation model. In particular, the present invention aims to provide a green space segmentation model capable of generating accurate green space segmentation results by combining an image encoder and a segmentation module to accurately segment green space in satellite imagery.

[0034] In particular, the green space segmentation model is trained based on a hybrid loss function, aiming to mitigate class imbalance issues and accurately segment even fine green spaces.

[0035] Hereinafter, the present invention will be examined in more detail with reference to the attached drawings. FIGS. 1, 2, 3, and 4 are conceptual diagrams for explaining a green space analysis method and system according to the present invention. FIGS. 5 is a flowchart for explaining a green space analysis method and system according to the present invention. FIGS. 6, 7, and 8 are conceptual diagrams for explaining a green space analysis method and system according to the present invention. FIGS. 9 is a formula for explaining a green space analysis method and system according to the present invention.

[0036] As illustrated in FIG. 1, the green space analysis method and system according to the present invention (hereinafter referred to as the “green space division system”) may include at least one of a communication unit (110), a storage unit (120), a control unit (130), an input unit (140), an image encoder (150), a segmentation module (160), and an area calculation module (170).

[0037] Although not described, the green space division system according to the present invention may include one or more processors, and such processors may include one or more general-purpose processors and / or one or more special-purpose processors (e.g., digital signal processors, tensor processing units (TPUs), graphics processing units (GPUs), neural network processing units (NPUs), application integrated circuits, application semiconductors (ASICs), etc.). One or more processors may be configured to execute instructions, computer-readable instructions, and / or other instructions described herein that are stored (or included) in the storage unit (120). The deep learning model-based concrete bridge component crack classification method and system according to the present invention may perform the data processing described below in cooperation with memory and at least one processor. The processor may perform a series of operations and data processing using data and information stored in memory. In this case, memory may be a component of the storage unit (120).

[0038] Meanwhile, the communication unit (110) may be connected to a user terminal, device (or device), server (e.g., external server, central server, etc.) and at least one network via a wireless or wired network to receive or transmit overall data and information necessary for the operation of the green space division system.

[0039] For example, the communication unit (110) can receive at least one of a satellite image and metadata related to the satellite image (e.g., location (or coordinate) information where the satellite image was taken, shooting time information, etc.) from at least one user terminal (mobile, tablet, laptop, etc.) and an external device (satellite, drone, CCTV, etc.).

[0040] The communication unit (110) may include one or more communication modules that enable wireless and / or wired communication between the green space division system and the communication target. Additionally, the communication unit (110) may include one or more communication modules that connect the green space classification system to one or more networks.

[0041] Furthermore, the communication unit (110) can support various communication methods according to the communication standards of the communicating device.

[0042] For example, the communication unit (110) may be configured to communicate with a communication target using at least one of the following technologies: WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G (5th Generation Mobile Telecommunication), Bluetooth (Bluetooth™ RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra-Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus).

[0043] Next, the storage unit (120) serves to store various data related to the present invention and may include one or more non-transient computer-readable storage media that can be read and / or accessed by at least one of one or more processors.

[0044] One or more computer-readable storage media may include volatile and / or non-volatile storage components, such as optical, magnetic, organic, or other memory or disk storage devices. In some examples, the storage unit (120) may be implemented using a single physical device (e.g., one optical, magnetic, organic, or other memory or disk storage device), whereas in other examples, the storage unit (120) may be implemented using two or more physical devices.

[0045] The storage unit (120) may include computer-readable instructions and additional data. The storage unit (120) may include a storage necessary to perform at least some of the methods, scenarios, and techniques described herein and / or at least some of the functions of the device and network.

[0046] Furthermore, at least a portion of the storage unit (120) may be a cloud storage or a cloud server. At least a portion of the data corresponding to user input and training data received from the input unit (140) may be stored in the storage unit (120).

[0047] That is, the storage unit (120) is sufficient as a space where information necessary for the operation of the green space division system (100) according to the present invention is stored, and it can be understood that there are no restrictions on the physical space.

[0048] Meanwhile, the input unit (140) can be configured in various ways as a means of data input. For example, the input unit (140) can be configured to receive user input. The input unit (140) can be configured to receive user input from a user terminal. Here, “receiving input” may mean receiving an input signal (or selection signal) corresponding to the user’s input based on the input being made by the user through the configuration of the input unit (140) provided in the user terminal. In the present invention, the input unit (140) does not necessarily mean a hardware means, but can be understood as a channel for receiving input from a user.

[0049] The input unit (140) may also be referred to as a user interface module. The input unit (140) may include a touch screen, a computer mouse, a keyboard, a keypad, a touchpad, a trackball, a joystick, a voice recognition module, or other similar devices. However, the present invention does not limit the type of input unit (140).

[0050] Here, user input may include documents, text, images (or videos), voice, etc. In this case, the green space partitioning system may further include a module that converts voice into text.

[0051] Next, the image encoder (150) can generate a feature map for the satellite image that includes visual features such as color, texture, and boundaries, as well as the relationship with surrounding areas and the spatial arrangement structure.

[0052] Next, the segmentation module (160) restores the resolution of the feature map to the same level as the existing satellite image based on the feature map generated from the image encoder (150), and classifies each pixel into one of a plurality of predefined classes to identify objects (e.g., greenery, buildings, roads, bodies of water, etc.) within the satellite image for each restored pixel location.

[0053] Next, the area calculation module (170) can selectively extract pixels corresponding to green space among the plurality of classes and generate a binary mask based on the pixels. Subsequently, the actual area occupied by green space within the entire satellite image can be calculated using the binary mask corresponding to the green space.

[0054] Next, the control unit (130) can perform the role of controlling the overall operation of the green space division system related to the present invention. The control unit (130) can process signals, data, information, etc. that are input or output through the components of the green space division system (100) described above, or perform a series of data processing to provide or process appropriate information and functions to the user. The control unit (130) can be physically implemented by the processor described above.

[0055] The control unit (130) can detect and classify green spaces in satellite images by using (or utilizing) a deep learning model-based green space segmentation system for precise detection of satellite images (10).

[0056] Furthermore, the green space division system (100) according to the present invention can be implemented through a computing device described below and can perform data processing related to the green space division method described above.

[0057] Meanwhile, FIG. 2 illustrates an example of a block diagram of a computing system in which the present invention can be implemented.

[0058] Referring to FIG. 2, a computing system (10000) for performing a green space division and quantification method according to one embodiment of the present invention may include at least one computing device. At this time, the at least one computing device may be a single processor or a multi-processor computing device.

[0059] The components of at least one computing device of the present invention may include various hardware components such as one or more processors, memory, other hardware, and a system bus (not shown) that connects various system components so that they can transmit and receive data to and from each other (e.g., telecommutatively connected, physically connected, electrically connected), and the components of at least one computing device are not limited thereto and may be very diverse.

[0060] Meanwhile, at least one computing device included in the computing system (10000) that performs the green space division and quantification method may be connected to communicate via a network (1070). For example, at least one computing device included in the computing system (10000) may be clustered or part of a local area network (LAN). Additionally, at least one computing device may be part of a wide area network (WAN) or connected to at least one of a client-server network and a peer-to-peer network within the cloud.

[0061] Meanwhile, when at least one computing device is used in at least one of a network environment and a cloud computing environment, the at least one computing device may be connected to at least one of a public and private network through a network interface or adapter. In one embodiment, other communication connection devices, such as a modem, may be used to establish communication through the network. The modem may be at least one of an internal modem and an external modem, and may be connected to a system bus through a network interface or a specific mechanism, etc. A wireless network component consisting of an interface and an antenna may be coupled to the network through a device such as an access point, a peer computer, etc. In the present invention, the method of connecting at least one computing device to communicate through the network (1070) is not limited, and it may be connected to communicate in a manner different from the described example.

[0062] Furthermore, other computer-type devices and / or systems not shown in FIG. 2 may also interact technically with at least one computing device or other system through one or more connections to the network (1070) via a network interface. Here, the network interface may include network interface equipment such as a physical network interface controller (NIC) or a virtual network interface (VIF).

[0063] The network (1070) of the present invention may include various forms such as the Internet, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G (5th Generation Mobile Telecommunication), Bluetooth (Bluetooth™ RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra-Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi Direct, Wireless USB (Wireless Universal Serial Bus), etc., and in the present invention, data transmission may be performed based on standard communication protocols such as TCP / IP, HTTP, SSL, etc.

[0064] A computing system (10000) for performing a green space division and quantification method according to the present invention may include at least one of a user computing device (1010), a training computing system (1050), and a server computing system (1030).

[0065] A user computing device (1010) according to the present invention may be understood as a computing device comprising at least one processor (1011) and a memory (1012) for performing a green space division and quantification method. For example, the user computing device (1010) may include at least one computing device among a smartphone, a smart TV, a laptop computer, a desktop computer, a digital broadcasting terminal, a PDA (personal digital assistants), a PMP (portable multimedia player), a navigation device, a slate PC, a tablet PC, an ultrabook, a wearable device (e.g., a smartwatch, a smart glass, and a head-mounted display).

[0066] At least one processor (1011) constituting the user computing device (1010) may include one or more general-purpose processors and / or one or more special-purpose processors. For example, at least one processor (1011) constituting the user computing device (1010) may be composed of at least one or a plurality of electrically connected processors among a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU), a neural network processing unit (NPU), an arithmetic logic unit (ALU), a floating-point arithmetic unit (FPU), an application integrated circuit, an application semiconductor (ASIC), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, and / or other electrical units for performing functions.

[0067] Furthermore, at least one processor (1011) may be configured to execute computer-readable instructions contained in memory (1012) and / or other instructions described herein.

[0068] The memory (1012) constituting the user computing system (1010) according to the present invention may include volatile memory, non-volatile memory, fixed media, removable media, magnetic media, optical media, semiconductor media and / or other types of physically durable storage media.

[0069] For example, the memory (1012) may include one or more non-transient / transient computer-readable storage media such as RAM, ROM, HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), EEPROM, EPROM, flash memory device, magnetic disk, and combinations thereof, and may include web storage of a server that performs the storage function of memory on the internet. This memory (1012) may store data and instructions necessary for the at least one processor (1011) to perform the operation of an application for green space division and quantification methods.

[0070] A user computing device (1010) may include one or more user input components (1021) that detect user input. For example, the user input component (1021) may also be referred to as a user interface module. The user input component (1021) may include a touch screen, a computer mouse, a keyboard, a keypad, a touchpad, a trackball, a joystick, a voice recognition module, or other similar devices. However, the present invention does not limit the type of user input component (1021). In this case, the user input component (1021) in the present invention does not necessarily mean a hardware means, but can be understood as a channel for receiving input from a user. Meanwhile, the user of the present invention may refer to an automated agent, script, playback software, etc., that operates on behalf of one or more people.

[0071] A user can interact with a computing system (10000) including at least one computing device through input text, touch, voice, movement, computer vision, gestures and / or other forms of input / output using a user input component (1021). For example, the user input component (1021) may include one or more of a command line interface (CLI), a graphical user interface (GUI), a natural user interface (NUI), a voice command interface and / or other user interface (UI) representations.

[0072] Between the user input component (1021) and the user computing device (1010), one or more application programming interface (API) calls may be made based on user input received from the user interface and / or network.

[0073] Here, the expression "based on" may be interpreted to include cases where it is based on the use of a specific configuration, modified from, derived from, influenced by, dependent on, or otherwise derived from a specific configuration. In some embodiments, an API call may be configured for a specific API, which may be interpreted or converted into an API call configured for another API. Here, an API may refer to a defined interface or connection between computers or between computer programs.

[0074] In one embodiment, the user computing device (1010) may store at least one machine learning model (1020). For example, the user computing device (1010) may be various machine learning models, such as a plurality of neural networks (e.g., deep neural networks) that perform green space segmentation and quantification using satellite imagery, or other types of machine learning models including non-linear models and / or linear models, and may be composed of a combination thereof.

[0075] According to an embodiment of the present invention, a user computing device (1010) may perform a green space division and quantification method using a local or / and external machine learning model (1020). Alternatively, the user computing device (1010) may perform a green space division and quantification method using a machine learning model (1040) provided by a server.

[0076] In addition, according to another embodiment of the present invention, a server computing system (1030) communicating with a user computing device (1010) may provide a calculation result for a green area to the user computing device (1010) on an application or / and the web in accordance with a request from a user received through the user computing device (1010).

[0077] In addition, according to another embodiment of the present invention, at least a part of the user computing device (1010) and the server computing system (1030) are linked together to perform a green space division and quantification method, thereby providing the user with a calculation result for the green space area.

[0078] Additionally, according to various embodiments of the present invention, a user computing device (1010) and / or a server computing system (1030) can learn machine learning models (1020, 1040) performed in a green space segmentation and quantification method through interaction with a training computing system (1050) that is communicatedly connected via a network (1070). In this case, the training computing system (1050) may be a computing system separate from the server computing system (1030). Alternatively, in some embodiments, the training computing system (1050) may be part of the server computing system (1030) or part of the user computing device (1010).

[0079] Meanwhile, the server computing system (1030) may include at least one processor (1031) and memory (1032). Here, the processor (1031) may be composed of at least one or a plurality of electrically connected processors among a central processing unit (CPU), a graphics processing unit (GPU), a tensor processing unit (TPU), a neural network processing unit (NPU), an application integrated circuit, an application semiconductor (ASIC), an arithmetic logic unit (ALU), a floating-point arithmetic unit (FPU), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, and / or other electrical units for performing functions. For example, at least one processor (1031) may include a circuit and a transistor configured to execute instructions from memory (1032).

[0080] The memory (1032) constituting the server computing system (1030) according to the present invention may include volatile memory, non-volatile memory, fixed media, removable media, magnetic media, optical media, semiconductor media, and / or other types of physically durable storage media. For example, the memory (1032) may include one or more non-transient / transient computer-readable storage media such as RAM, ROM, HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), EEPROM, EPROM, flash memory device, magnetic disk, etc., and combinations thereof, and may include web storage of a server that performs the storage function of memory over the internet. Additionally, the server computing system (1030) may further include a data storage (data store). For example, the data storage may be composed of at least one of a relational database, a NoSQL database, a data warehouse, and a local file system.

[0081] In the memory (1032) constituting the server computing system (1030) according to the present invention, data and instructions necessary for the at least one processor (1031) to perform the operation of an application for solving the green space division and quantification method may be stored.

[0082] In one embodiment, the server computing system (1030) may be composed of a single device or a plurality of computing devices, and may be configured to operate according to a sequential or parallel computing architecture. Additionally, a distributed processing system may be configured with a plurality of networked devices.

[0083] Meanwhile, the training computing system (1050) may include at least one processor (1051) and memory (1052). The model trainer (1060) is a logical component that executes the training of at least one machine learning model (1020, 1040) and may be implemented in the form of hardware, firmware, or software. For example, the model trainer (1060) may be executed by the processor (1051) after loading training data (1061) stored in a storage device into memory (1052). For example, the model trainer (1060) may be configured to execute one or more operations (e.g., model training, model reconstruction, model validation, model testing) on ​​at least one machine learning model.

[0084] The machine learning model of the present invention may include at least one of a statistical model, an algorithm, a neural network (NN), a convolutional neural network (CNN), a generative neural network (GNN), a Word2Vec model, a Bag of Words model, a TF-IDF (document frequency-inverse document frequency) model, a GPT (Generative Pre-trained Transformer) model (or other autoregressive models), a PPO (Proximal Policy Optimization) model, a nearest neighbor model (e.g., a k-nearest neighbor model), a linear regression model, a K-means clustering model, a Q-learning model, a TD (Temporal Difference) model, a Deep Adversarial Network model, and all other types of models further described herein.

[0085] Specifically, the model trainer (1060) may execute operations to train a machine learning model, and said operations may include at least one of adding, removing, and modifying model parameters. At this time, the training of the machine learning model may be at least one of supervised learning, semi-supervised learning, and unsupervised learning. In one embodiment, the training of the machine learning model may include the step of repeatedly inputting training data (1061) based on epochs and repeatedly performing the machine learning model training process configured in this way. Here, an epoch may refer to a unit in which the entire set of training data (1061) undergoes forward and backpropagation processing once. In some implementations, different levels of training methods (e.g., supervised learning, semi-supervised learning, unsupervised learning) may be used for different epochs.

[0086] The training data (1061) of the present invention may include input data and / or data previously output from at least one machine learning model (e.g., recursive learning feedback).

[0087] At least one parameter of a machine learning model may include at least one of a seed value, a model node, a model layer, an algorithm, a function, connections between different machine learning models, connections between parameters, machine learning model constraints, and other digital components that influence the output of the machine learning model. In this case, model connections between different machine learning models may include or represent relationships between model parameters and / or models, which may be dependent or interdependent, hierarchical, and / or static or dynamic. The combinations and configurations of model parameters described herein may be too complex to be maintained or utilized by human cognitive abilities.

[0088] In the present invention, the machine learning parameters described according to the embodiments are not limited, and a single machine learning model may further include a plurality of model parameters.

[0089] Meanwhile, FIG. 3 illustrates an example of a block diagram of a computing device (1100) that may be included in a user computing device (1010), a server computing system (1030), and a training computing system (1050), as an embodiment of a computing system (10000) in which the present invention can be implemented.

[0090] As illustrated in FIG. 3, the computing device (1100) may include at least one application (e.g., Application 1 to Application N), and each of the at least one application may include a machine learning library and a model execution environment for performing a machine learning-based green space segmentation and quantification method. The at least one application included in the computing device (1100) may communicate with the sensor, context manager, device state manager, or additional component(s) within the computing device (1100) via an Application Programming Interface (API). In one embodiment, the at least one application may interface with device components, such as receiving sensor data or state data or transmitting prediction results to an output device via a public or private API.

[0091] Meanwhile, FIG. 4 illustrates an example of a block diagram in another aspect of a computing device (1200), which is one of the components of a computing system (10000) that performs a green space division and quantification method according to an embodiment of the present invention.

[0092] A computing device (1200) according to the present invention may include at least one application (e.g., Application 1 to Application N), and at least one application may communicate with a central intelligence layer (1210). Each application may interact with a shared model within the central intelligence layer (1210) through an API (e.g., a common API).

[0093] The central intelligence layer (1210) includes one or more machine learning models and may share them among multiple applications or provide them independently to each. In one embodiment, the central intelligence layer (1210) may be integrated as part of an operating system or implemented as a separate logical layer.

[0094] Additionally, the central intelligence layer (1210) can communicate with the central device data layer (1220). The central device data layer (1220) can integrate and store satellite images and information about greenery required for the greenery segmentation and quantification method stored within the computing device (1200), and provide this as input data required for solving the greenery segmentation and quantification method. Each device component (e.g., sensor, state manager, etc.) can communicate with the central device data layer (1220) through a private API, etc.

[0095] The technology described in this specification may be composed of a single or multiple computing devices, and the machine learning model performing the green space segmentation and quantification method may be executed sequentially or in parallel on one component or multiple distributed components. The data storage, machine learning model, and application may be distributed and operated locally or over a network, and these configurations can be flexibly applied to various system architectures.

[0096] The green space partitioning system according to the present invention can be trained based on a hybrid loss function.

[0097] Here, a loss function refers to a function that calculates the difference between the model's output (or predicted value) and the actual ground truth and numerically expresses how large that difference is.

[0098] The larger the value of the loss function, the greater the difference between the model's prediction and the actual correct answer, whereas the smaller the value of the loss function, the closer the prediction is to the correct answer.

[0099] For example, in the present invention, a hybrid loss function combining a Cross-Entropy Loss function and a Dice Loss function may be used, and such a loss function can solve the class imbalance problem occurring in image segmentation and improve model performance.

[0100] The above hybrid loss function can be expressed as shown in (a) of Fig. 9 based on a weighted combination method.

[0101] Here, α is the weight of the cross-entropy loss function, and β is the weight of the Dice loss function.

[0102] At this time, it can be set to satisfy the condition α+β=1.

[0103] According to one embodiment, when α=β=0.5 is set, an optimal balance between classification accuracy and segmentation precision can be achieved.

[0104] The hybrid loss function may include weights for the cross-entropy loss function and the die loss function, respectively, and the weights can be adjusted through the green space partitioning model.

[0105] Here, cross-entropy loss can be understood as a loss function used to numerically express how close the prediction result is to the correct answer by calculating the difference between the probability distribution predicted by the model and the actual correct answer through log probability. Cross-entropy loss can be expressed as shown in (b) of Fig. 9.

[0106] Here, N is the number of samples, C is the number of classes, y_(i,c) is the actual distribution, and is the predicted probability for class C.

[0107] Next, Dice loss can be understood as a loss function for maintaining boundary accuracy even in semantic partitioning problems where class imbalance exists, by calculating the loss based on the overlap ratio between the model's predicted mask and the actual mask. Dice loss can be expressed as shown in (c) of Fig. 9.

[0108] Here, is the predicted probability, is the One-Hot encoded actual answer, and is a smoothing factor to prevent division by zero.

[0109] The above hybrid loss function can balance pixel-level accuracy and region consistency.

[0110] In other words, cross-entropy loss provides a fine-grained supervised signal to improve pixel-level classification accuracy, and dice loss emphasizes overlap between regions, thereby resolving class imbalance issues in remote sensing scenarios where there are large size differences between green areas. Additionally, a flexible weighting mechanism allows for adjusting the relative importance of the two loss items, further enhancing adaptability to meet the requirements of various tasks.

[0111] Meanwhile, the green space segmentation system according to the present invention may collect satellite images taken of a specific city (S510, see FIG. 5) and generate a feature map of the satellite images using an image encoder of the green space segmentation model (S520, see FIG. 5).

[0112] The green space segmentation system may be composed of an image (Vision Transformer) encoder including a Multi-Head Self-Attention (MHSA) structure to effectively extract a feature map from a satellite image (10).

[0113] The image encoder (150) may be composed of a total of 24 layers, and each layer may be composed of a transformer block (330) including 16 multi-head self-attention heads.

[0114] Here, self-attention can be understood as a means of identifying the interrelationships between each patch, and multi-head self-attention can be understood as a means of analyzing satellite images (10) from different perspectives (e.g., location, color, texture, etc.) using multiple attention heads.

[0115] For example, the high-rise building dense area and the nearby green space in the satellite image (10) are spatially adjacent but have different color and texture characteristics, so the green space can be effectively distinguished even in the complex satellite image (10) by utilizing multi-head self-attention.

[0116] The green space segmentation system can simultaneously capture various semantic and spatial relationships of the satellite image (10) using an image encoder (150).

[0117] For example, by capturing adjacent structures between building roofs and green spaces and green zones distributed along rivers in satellite imagery (10), precise identification of green spaces within complex urban structures can be performed.

[0118] More specifically, the control unit (130) can process the satellite image (10) as an input to the image encoder (150). At this time, the size of the satellite image (10) may be 244x244 pixels, and the satellite image (10) may be divided into non-overlapping 14x14 patches.

[0119] Here, a patch may refer to a piece of the entire image cut into a certain size. For example, if a satellite image (10) of size 244x244 is divided into 14x14 patches, it may consist of 16 horizontal patches x 16 vertical patches = a total of 256 patches.

[0120] Each patch may undergo a preprocessing process to be processed as an input to the transformer block (330) during the encoding process.

[0121] More specifically, each patch can be converted into a fixed-dimensional vector value through a patch embedding (310) process, and this conversion can be performed, for example, through a Conv2D operation with a kernel size of 14 and a stride of 14.

[0122] Each converted patch vector can be position encoded (320) to preserve the spatial order on the original image and then processed as an input to a transformer block (330).

[0123] The transformer block (330) can integrate image information by modeling the relationship between each patch through multiple head self-attention.

[0124] The multi-head self-attention of the transformer block (330) can generate features for the satellite image (10) by converting the satellite image (10) into a query, key, and value matrix, respectively, calculating an attention weight through the dot product between the query and the key, and applying the weight to the value.

[0125] In particular, the transformer block (330) can divide the satellite image (10) into multiple attention heads and perform independent attention operations on each, and each head focuses on a different aspect of the input (e.g., spatial location, texture, boundary, etc.), thereby enabling the green space segmentation system to simultaneously capture various semantic and spatial relationships (e.g., texture, boundary, color contrast, relative positional relationship between objects, etc.) of the satellite image (10).

[0126] The result of the multi-head attention operation is passed to the next block after merging the outputs per head, and this structure can enable richer and more stable representation learning compared to single-head attention.

[0127] That is, the above multi-head self-attention-based encoding method can improve the accuracy and generalization performance of green area segmentation by allowing the green area segmentation system to precisely recognize complex structures (e.g., green areas, roads, boundaries between buildings, etc.) within the satellite image (10) and learn information of various scales and visual patterns.

[0128] Meanwhile, the green space classification system according to the present invention processes the feature map as input to a segmentation module (S530, see FIG. 5), and a process of extracting pixels corresponding to a green space class from a satellite image may be performed (S540, see FIG. 5).

[0129] The control unit (130) can process the generated feature map as an input to the segmentation module (160). The segmentation module (160) may be a model that performs segmentation of green space within the satellite image (10) and may be composed of a plurality (e.g., 4) of upsampling pyramids.

[0130] More specifically, the segmentation module (160) can perform resolution restoration and channel reduction on the feature map, and can generate an intermediate feature map with multiple different resolution levels during the process of upsampling.

[0131] An intermediate feature map may refer to a feature map with different resolutions and number of channels generated during the process in which the input feature map undergoes an upsampling process through the segmentation module (160).

[0132] For example, a low-resolution feature map can identify the overall structure of the satellite image (10), and as the high-resolution feature map is restored, detailed structures such as narrow green spaces or roads can be identified. Intermediate feature maps are linked through residual connections, preserving the flow of information from low resolution to high resolution, preventing gradient vanishing, and maintaining structural consistency.

[0133] More specifically, the segmentation module (160) can perform a 1×1 convolution on the feature map generated from the encoder to compress the channels, and then perform BatchNorm and ReLU operations to refine the feature map.

[0134] Furthermore, the segmentation module (160) can restore the resolution of the feature map by performing progressive upsampling (e.g., bilinear interpolation upsampling).

[0135] For example, the segmentation module (160) can gradually restore the resolution of the feature map from 14×14→28×28→56×56→112×112→224×224 by applying an upsampling scale factor sequentially from 2→2→2→1.75.

[0136] In addition, gradient flow can be maintained by applying residual connections between each upsampling step.

[0137] The segmentation module (160) can perform division into 5 classes for the finally restored 224×224 feature map.

[0138] More specifically, the segmentation module (160) can perform a 3×3 convolution operation on the final feature map to adjust the output channel to 5.

[0139] For example, each channel may correspond to at least one of five classes (e.g., background, greenery, buildings, roads, water bodies, etc.) within the satellite image (10).

[0140] Meanwhile, the green space analysis method and system according to the present invention may proceed with a process of calculating the area of ​​green space included in a satellite image using extracted pixels (S550, see FIG. 5).

[0141] The control unit (130) can process an area corresponding to at least one specific class among a plurality of classes included in the segmentation result output from the decoder as an input to the area calculation module (170). The area calculation module (170) can selectively extract pixels corresponding to the input class and generate a binary mask (20) for the corresponding class.

[0142] For example, the area calculation module (170) can selectively extract only the pixels corresponding to the green space classified into the first class and generate a binary mask (20) for the pixels.

[0143] The binary mask (20) can be configured to display pixels (510) determined to be green as '1' and the remaining classes as '0'.

[0144] Subsequently, the area calculation module (170) can calculate the actual area occupied by the corresponding class within the entire satellite image (10) based on the number of 1 pixels included in the binary mask (20). The actual area can be calculated based on the geographical area of ​​the satellite image (10) and can be expressed as shown in (d) of FIG. 9 as follows.

[0145] Here, is the number of pixels of the corresponding class, is the geographical area of ​​the satellite image (10), and may be the area per pixel.

[0146] For example, assuming the geographical area of ​​the satellite image (10) is 3.5 ≈ km ≈ 2 and the number of pixels in that class is 50,176, 1 pixel can be approximately 0.00487 square meters.

[0147] The above area per pixel can be used to evaluate the accuracy of the model later.

[0148] Meanwhile, the present invention described above can be implemented as a program that is executed by one or more processes on a computer and can be stored on a computer-readable medium (or recording medium).

[0149] Furthermore, the present invention described above can be implemented as computer-readable code or instructions on a medium on which a program is recorded. That is, the present invention can be provided in the form of a program.

[0150] Meanwhile, computer-readable media include all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0151] Furthermore, the computer-readable medium may be a server or cloud storage that includes a storage and is accessible to an electronic device via communication. In this case, the computer may download the program according to the present invention from the server or cloud storage via wired or wireless communication.

[0152] Furthermore, in the present invention, the computer described above is an electronic device equipped with a processor, namely a CPU (Central Processing Unit), and no special limitations are placed on its type.

[0153] Meanwhile, the above detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

Claims

Claim 1 A method for analyzing green space, characterized by comprising: a step of collecting satellite images taken of a specific city; a step of generating a feature map for the satellite images using an image encoder of a green space segmentation model; a step of inputting the feature map into a segmentation module of the green space segmentation model; a step of generating a plurality of intermediate feature maps having different resolution levels during the process of performing resolution restoration and channel reduction on the feature map in the segmentation module and performing upsampling, and a step of restoring a high-resolution feature map by linking the intermediate feature maps through residual connections; a step of extracting pixels corresponding to a green space class from the satellite images using the restored high-resolution feature map; and a step of calculating the area of ​​green space included in the satellite images using the extracted pixels. Claim 2 A method for analyzing green space according to claim 1, wherein the step of extracting the pixel comprises: a step of dividing the satellite image taken of the specific city into a plurality of classes based on a division criterion set in the segmentation module; a step of identifying a green space class corresponding to green space among the divided plurality of classes using the green space division model; and a step of extracting the pixel corresponding to the green space class. Claim 3 A method for analyzing green space according to claim 2, wherein the step of calculating the area of ​​the green space comprises: a step of generating a binary mask for the satellite image using the pixel corresponding to the green space class; and a step of calculating the area of ​​the green space using the binary mask. Claim 4 A method for analyzing green space according to claim 3, characterized in that, in the step of calculating the area of ​​the green space, the number of pixels of the binary mask is calculated, and the area of ​​the green space included in the satellite image is calculated using the calculated number of pixels. Claim 5 A method for analyzing green space according to paragraph 2, wherein the plurality of classes include at least one of the background, buildings, green space, roads, and rivers of the specific city. Claim 6 A method for analyzing green space according to claim 1, wherein the green space segmentation model is a model learned using a preset loss function, and the preset loss function includes a hybrid loss function that combines a cross-entropy loss function and a Dice loss function. Claim 7 A method for analyzing green space according to claim 6, wherein the green space segmentation model is trained using a pre-set loss function that combines the cross-entropy loss function defined from the classification error of pixels corresponding to the green space class and the dice loss function defined from the degree of overlap between the prediction result for the green space class and the green space area. Claim 8 A green space analysis method according to claim 7, wherein the above-mentioned loss function includes weights for each of the above-mentioned cross-entropy loss function and the above-mentioned die loss function, and the weights are adjusted through the above-mentioned green space segmentation model. Claim 9 A green space analysis system, wherein the system collects satellite images taken of a specific city, generates a feature map for the satellite images using an image encoder of a green space segmentation model, inputs the feature map into a segmentation module, performs resolution restoration and channel reduction on the feature map in the segmentation module, generates a plurality of intermediate feature maps with different resolution levels during the process of performing upsampling in the segmentation module, restores a high-resolution feature map by linking the intermediate feature maps through residual connections in the segmentation module, extracts pixels corresponding to a green space class using the restored high-resolution feature map, and calculates the area of ​​the green space using the extracted pixels. Claim 10 A program stored on a computer-readable recording medium that is executed by one or more processes in an electronic device, the program comprising instructions for performing steps such as: collecting satellite images taken of a specific city; generating a feature map for the satellite images using an image encoder of a green space segmentation model; inputting the feature map into a segmentation module; generating intermediate feature maps having multiple different resolution levels in the process of performing resolution restoration and channel reduction on the feature map and upsampling in the segmentation module, and restoring a high-resolution feature map by linking the intermediate feature maps through residual connections; extracting pixels corresponding to a green space class using the restored high-resolution feature map; and calculating the area of ​​the green space using the extracted pixels.

Images