KR102014288B1 - Development pressure prediction method based on artificial intelligence using drone - Google Patents

Abstract

An artificial-intelligence-based development pressure prediction method using a drone is disclosed. According to the present invention, the development pressure prediction method comprises the steps of: obtaining an image of a target region photographed in the air above the target region (S110); acquiring input data from the obtained image (S120); inputting the input data into a trained model (S130); obtaining an output of the trained model (S140); and determining development pressure of the target region based on the obtained output (S150).

Description

Artificial intelligence-based development pressure prediction method using drone {DEVELOPMENT PRESSURE PREDICTION METHOD BASED ON ARTIFICIAL INTELLIGENCE USING DRONE}

The present invention relates to an artificial intelligence based development pressure prediction method using a drone.

The most important field surveys in the selection of candidates for development pressure have been the comparative analysis of relevant documents on the site and aerial photographs taken a year ago (every year) or in person by the administrative staff. However, it was difficult to carry out efficient surveys in real time because there was a shortage of manpower in the field survey and depended on aerial photographs taken a year ago to check the land status.

Therefore, attempts are being made to use drones for land surveys that can capture current status photos and videos in real time regardless of space.

Most of the applications, however, focus only on drones being shot on site in place of existing administrative personnel. While it is possible to compensate for the lack of administrative power and know the local situation in real time, it still invests the time and money of many experts and officials to investigate trends and collect vast data to analyze the images taken by drones directly. In particular, in the selection of candidates for development location to focus on the present invention, the optimal candidates are selected through complex procedures such as data survey, field survey, current status survey, and impact prediction, and often include non-expert evaluators in the selection process. This may lead to fairness due to doubts of professionalism or due to evaluators with interests in the area.

Therefore, it is required to search for and select development pressure candidates by using drones more efficiently.

Artificial Intelligence (AI) system is a computer system that implements human-level intelligence. Unlike conventional rule-based smart systems, the machine learns and judges itself and becomes smart. As the AI system is used, the recognition rate is improved and the user's taste can be understood more accurately. The existing Rule-based smart system is gradually replaced by the deep learning-based AI system.

Artificial intelligence technologies include machine learning (Machine Learning) technology, and deep learning technology, which is widely used for analyzing images, especially machine learning technology.

Deep learning is a set of machine learning algorithms that attempts a high level of abstraction (summarizing key content or functions in large amounts of data or complex data) through a combination of several nonlinear transformations. Is defined. Deep learning can be seen as a field of machine learning that teaches computers how people think in a large framework.

When there is any data, it is represented in a form that can be understood by a computer (e.g., in the case of an image, the pixel information is represented as a column vector), and a lot of research (how to express it better) is applied to learning. How to build techniques and how to build models to learn them). As a result of these efforts, various deep learning techniques have been developed. Deep learning techniques include Deep Neural Networks (DNN), Convolutional Deep Neural Networks (CNN), Reccurent Neural Networks (RNN), and Deep Belief Networks (DBN). have.

Deep Neural Networks (DNNs) are artificial neural networks (ANNs) made up of a plurality of hidden layers between an input layer and an output layer.

The various fields in which artificial intelligence technology is applied are as follows. Linguistic understanding is a technology for recognizing and applying / processing human language / characters and includes natural language processing, machine translation, dialogue system, question and answer, speech recognition / synthesis, and the like. Visual understanding is a technology that recognizes and processes objects as human vision, and includes object recognition, object tracking, image retrieval, person recognition, scene understanding, spatial understanding, and image enhancement. Inference Prediction is a technique for judging, logically inferring, and predicting information. It includes knowledge / probability-based inference, optimization prediction, preference-based planning, and recommendation. Knowledge expression is a technology that automatically processes human experience information into knowledge data, and includes knowledge construction (data generation / classification) and knowledge management (data utilization). Motion control is a technology for controlling autonomous driving of a vehicle and movement of a robot, and includes motion control (navigation, collision, driving), operation control (action control), and the like.

Patent Registration No. 10-1769852, 2017.08.14 Registration

The problem to be solved by the present invention is to provide an artificial intelligence-based development pressure prediction method using a drone.

Problems to be solved by the present invention are not limited to the above-mentioned problems, and other problems not mentioned will be clearly understood by those skilled in the art from the following description.

Artificial intelligence-based development pressure prediction method using a drone according to an aspect of the present invention for solving the above problems, the step of acquiring an image of the target region, photographed from above the target region (S110), the obtained Acquiring input data from an image (S120), inputting the input data into a learned model (S130), acquiring an output of the learned model (S140), and the target based on the obtained output Determining the development pressure of the region (S150).

In addition, the step (S110) may further include controlling a drone (S210) and acquiring an image of the target area photographed by the drone (S220).

In addition, determining the development pressure of each of the plurality of target areas photographed by the drone (S310) and generating a development pressure map including the development pressure of each of the plurality of target areas and the plurality of target areas ( S320) may be further included.

In addition, the step (S120) includes the step of generating the at least one tile image (S410) by dividing the obtained image, the step (S130), input the generated tile image to the learned model And a step (S420), wherein the step (S150) includes obtaining a development pressure corresponding to the generated tile image based on the output of the learned model (S430) and one or more of the generated tile images. Determining the development pressure of the target region based on the development pressure of the target region (S440); It may include.

In addition, the step (S430) includes a step (S510) of obtaining a development pressure index corresponding to each of the generated tile image, the step (S440), at a position corresponding to each of the one or more tile images Generating a development pressure map matching the development pressure index corresponding to each of the one or more tile images (S520), and selecting one or more locations having a development pressure index of a predetermined reference value or more from the development pressure map (S530) ), Adjusting the development pressure index corresponding to each of the one or more tile images based on the positional relationship with the one or more positions selected in the step S530 (S540), wherein the adjusted development pressure index is converted into the development pressure. Matching the map (S550) and determining the development pressure of the target region based on the adjusted development pressure index (S560). The.

In addition, acquiring an image of the reference region photographed from above the reference region (S610), acquiring data for calculating a development pressure index for the reference region (S620), based on the obtained data Calculating a development pressure index for the reference region (S630), generating learning data including an image of the reference region and a development pressure index for the reference region (S640), and using the learning data The method may further include training the learned model (S650).

In addition, the step S640 may include generating at least one tile image from the image of the reference area at step S710 and labeling a development pressure index corresponding to each of the at least one tile image at step S720. have.

In operation S720, when the development pressure index of the reference region included in each of the one or more tile images is labeled in each of the one or more tile images, the first tile image includes a plurality of reference regions. Acquiring development pressure indices of each of the plurality of reference regions (S810); calculating a first development pressure index corresponding to the first tile image based on the obtained development pressure indices (S820); and The method may further include labeling the calculated first development pressure index in the first tile image (S830).

Artificial intelligence-based development pressure prediction apparatus using a drone according to an aspect of the present invention for solving the above problems includes a memory for storing one or more instructions and a processor for executing the one or more instructions stored in the memory, The processor executes the one or more instructions to perform an artificial intelligence-based development pressure prediction method using a drone according to the disclosed embodiment.

Artificial intelligence-based development pressure prediction program using a drone according to an aspect of the present invention for solving the above problems, combined with a computer as hardware, stored in a computer-readable recording medium to perform the method of claim 1 do.

Other specific details of the invention are included in the detailed description and drawings.

According to the deep learning-based development pressure index prediction method using the drone and the spatial data according to the disclosed embodiment, the development pressure level of a specific region may be predicted only by an image taken by the drone based on artificial intelligence deep learning technology.

According to this, if a large number of administrative personnel and experts were required, the structural problem that had to go through a long process of selecting candidate sites including field survey and location feasibility review for representative parcels could be solved.

In other words, it is possible to calculate the development pressure index efficiently and to select the candidate candidates for development pressure according to time, space and cost, and also to secure objectivity and reliability in the process. Furthermore, the present invention can be used in various public and industrial fields such as regional gap determination and real estate investment.

Effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 is a diagram illustrating a system according to an exemplary embodiment.
2 is a flowchart illustrating an artificial intelligence-based development pressure prediction method using a drone according to an embodiment.
3 is a flowchart illustrating an image acquisition method using a drone according to an exemplary embodiment.
4 is a flowchart illustrating a method of generating a development pressure map, according to an exemplary embodiment.
5 is a flowchart illustrating a development pressure determination method according to an embodiment.
6 is a flowchart illustrating a development pressure adjusting method according to an embodiment.
7 is a flowchart illustrating a learning method according to an exemplary embodiment.
8 is a flowchart illustrating a method of generating training data according to an exemplary embodiment.
9 is a flowchart illustrating a labeling method according to an exemplary embodiment.
10 is an exemplary diagram illustrating an example of a prediction operation of a CNN.
11 is an exemplary diagram illustrating a method of mapping a drone image.
12 is an exemplary diagram illustrating a detailed tiled method of divided areas based on a development pressure index.
It is a figure explaining the processing method about the development pressure index division area which belongs to each tile.
14 is a diagram illustrating an example of adjusting development pressure based on a development pressure map.
15 is a block diagram of an apparatus according to an embodiment.

Advantages and features of the present invention and methods for achieving them will be apparent with reference to the embodiments described below in detail with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but can be embodied in various different forms, and the present embodiments only make the disclosure of the present invention complete, and those of ordinary skill in the art to which the present invention belongs. It is provided to fully inform the skilled worker of the scope of the invention, which is defined only by the scope of the claims.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, "comprises" and / or "comprising" does not exclude the presence or addition of one or more other components in addition to the mentioned components. Like reference numerals refer to like elements throughout, and "and / or" includes each and all combinations of one or more of the mentioned components. Although "first", "second", etc. are used to describe various components, these components are of course not limited by these terms. These terms are only used to distinguish one component from another. Therefore, of course, the first component mentioned below may be a second component within the technical spirit of the present invention.

Unless otherwise defined, all terms used in the present specification (including technical and scientific terms) may be used in a sense that can be commonly understood by those skilled in the art. In addition, terms that are defined in a commonly used dictionary are not ideally or excessively interpreted unless they are specifically defined clearly.

As used herein, the term "part" or "module" refers to a hardware component such as software, FPGA, or ASIC, and the "part" or "module" plays certain roles. However, "part" or "module" is not meant to be limited to software or hardware. The “unit” or “module” may be configured to be in an addressable storage medium or may be configured to play one or more processors. Thus, as an example, a "part" or "module" may include components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, Procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. Functions provided within components and "parts" or "modules" may be combined into smaller numbers of components and "parts" or "modules" or into additional components and "parts" or "modules". Can be further separated.

In the present specification, the computer refers to any kind of hardware device including at least one processor, and according to an embodiment, it may be understood as a meaning encompassing a software configuration that operates on the hardware device. For example, a computer may be understood as including, but not limited to, a smartphone, a tablet PC, a desktop, a notebook, and a user client and an application running on each device.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating a system according to an exemplary embodiment.

Referring to FIG. 1, components of a system according to the disclosed embodiment are disclosed. However, at least some of the components shown in FIG. 1 may be omitted, and components not shown in FIG. 1 may be added.

For example, the drone 10 may operate based on the control of the user terminal 20 or the server 100, and the information collected by the drone 10 may be transmitted to the user terminal 20 or the server 100. do. According to an embodiment, the information collected by the drone 10 may be transmitted to the server 100 through the user terminal 20.

In the disclosed embodiment, the user terminal 20 may be one of the above-described computers, but is not limited thereto. For example, the user terminal 20 may be a mobile terminal or a personal computer used by the user, may be a control controller of the drone 10, and means a configuration in which a plurality of devices are coupled or operate by communicating with each other wired or wireless. It may be.

In the disclosed embodiment, the server 100 may be one of the above-described computers, but may also mean a cloud server according to the embodiment.

In the disclosed embodiment, development pressure may be understood as a concept representing a measure of development need for a specific region. In addition, the development pressure index (DPI) may mean that this is quantitatively quantified.

In the present specification, development pressure is understood as a meaning encompassing all means quantitatively or qualitatively capable of expressing development pressure. For example, development pressure is understood as a concept encompassing the development pressure index, and the expression acquiring or calculating the development pressure may be understood to include the concept of acquiring or calculating the development pressure index.

In the disclosed embodiment, the embodiment shown in Figure 1 is to solve the object of the present invention, to provide a deep learning-based development pressure index prediction method using a drone and spatial data for automatically calculating the development pressure index Can be used.

In an embodiment, the system shown in FIG. 1 combines the drone 10 and big data for a deep learning based development pressure index prediction method using the drone 10 and spatial data, and uses deep learning among artificial intelligence technologies. Therefore, the development pressure index will be calculated automatically using only the drone photographed image.

As the name of artificial intelligence, which means technology that realizes human's learning ability, enables deep learning model to predict the development pressure index by looking at the local image as human thinks by 'experience'. It is.

For example, the system shown in Fig. 1 is a method of automatically calculating the objectivity and reliable development pressure index by learning CNN, the deep learning model most often used for image recognition, by using local image of the current point of time photographed by a drone. It may be performed, but is not limited thereto.

To sum up the process easily, we use a lot of map image tiles as training data to learn how the deep learning model is characterized by tiles with different development pressure indices. Entering new data into the trained deep learning model in turn determines whether the image has its features and calculates the development pressure index accordingly. Referring to FIG. 10, an example of visualization of the above-described embodiment is illustrated.

Hereinafter, each step of the method according to the disclosed embodiment will be described in detail with reference to the accompanying drawings.

Each step described herein is described as being performed by a computer, but the subject of each step is not limited thereto, and at least some of the steps may be performed by different devices according to embodiments.

For example, the computer may mean one of the server 100 and the user terminal 20 shown in FIG. 1, but may mean a system including both.

In addition, each step described below may be performed in one of the server 100 and the user terminal 20, but at least some or all of them may be performed in another device.

2 is a flowchart illustrating an artificial intelligence-based development pressure prediction method using a drone according to an embodiment.

In step S110, the computer acquires an image of the target area photographed over the target area.

In the disclosed embodiment, the target area refers to a target area for determining the development pressure, and is not limited to a specific type or location. In addition, the area of the target area is also not limited.

In one embodiment, the method of acquiring the captured image of the target area is not limited. According to an embodiment, the method of acquiring a photographed image of the target area may be photographing using a drone, but is not limited thereto.

In addition, the acquisition of the captured image of the target area may be obtained by capturing an image of a specific moment, but may also be obtained by capturing an image through image capturing and capturing an image of a specific moment.

In step S120, the computer obtains input data from the obtained image.

In one embodiment, the type or type of the input data is not limited. For example, the input data may mean the acquired image itself, or may mean the result of performing various preprocessing processes such as cropping, splitting, enlarging, reducing, resizing, rotating, filtering, and compressing the acquired image. It is not limited to this.

In an embodiment, the input data may be generated based on an image generated by concatenating the obtained plurality of images through a predetermined preprocessing process or a corresponding image.

For example, the input data may refer to the image itself including the target region, may indicate a portion including the target region in the image, or may mean a cropped image, or may divide the image into a plurality of portions. Each image may be referred to, and may be understood to include images including a target area of each divided image, but is not limited thereto.

In step S130, the computer inputs the input data into the trained model.

In the present specification, the learned model may mean a model trained based on artificial intelligence technology, for example, machine learning, deep learning, and the like, but is not limited thereto.

Specifically, the trained model may mean a trained model based on CNN, but is not limited thereto.

In step S140, the computer obtains the output of the learned model.

The type of output is not limited, for example, the output may be obtained as a specific value or a result, may be obtained as a probability, may be obtained as a probability for each value, but is not limited thereto.

In step S150, the computer determines the development pressure of the target region based on the obtained output.

For example, the computer may determine a predetermined value based on the output of the learned model. In addition, the computer may calculate the development pressure index of the target region by performing an operation based on the obtained value.

In an embodiment, the computer may obtain a development pressure index of each of the plurality of divided zones included in the target region, and calculate the development pressure index of the target region based on the development pressure index. For example, the development pressure index of the target region may be calculated based on the sum or average of the development pressure indexes of each of the divided zones included in the target region, but is not limited thereto.

3 is a flowchart illustrating an image acquisition method using a drone according to an exemplary embodiment.

In operation S110, the computer may perform operation S210 of controlling the drone.

For example, the drone may operate based on a preset automatic control program.

For example, the drone may be controlled by receiving a wireless RF signal directly from a computer.

For example, the drone may be controlled by receiving a control signal transmitted from a computer through a network communication network.

For example, the drone may be controlled by receiving an RF signal from a controller that receives a control signal transmitted from a computer.

In addition, the computer can control the drone in various ways, the method is not limited.

In addition, the computer may perform an operation (S220) of acquiring an image of the target area photographed by the drone.

The shape of the acquired image is not limited. For example, the entire captured image may be acquired, the captured image may be captured, or the drone may acquire the captured image.

In an embodiment, the computer may control the drone based on the development pressure calculated based on the image photographed by the drone. For example, the computer calculates the development pressure for the target area underneath the drone, which is taken in real time from the drone, and controls the drone in a direction in which the development pressure calculated according to the movement of the drone increases, thereby having the highest development pressure. It can be controlled to find a position or to search and move to a position having a development pressure of a predetermined reference value or more.

4 is a flowchart illustrating a method of generating a development pressure map, according to an exemplary embodiment.

In operation S310, the computer may determine development pressure of each of the plurality of target areas photographed by the drone.

In operation S320, the computer may generate a development pressure map including development pressures of the plurality of target areas and each of the plurality of target areas.

The generated development pressure map may be used for selecting a target area for development or valuation of each area, but is not limited thereto.

In one embodiment, the computer may respond to the query based on the generated development pressure map. For example, the computer may respond to a query for selecting a region requiring development in a predetermined region, a query for a region to be invested in, or a query for a value for each region, but is not limited thereto.

5 is a flowchart illustrating a development pressure determination method according to an embodiment.

In operation S120, the computer may perform operation S410 of generating the at least one tile image by dividing the obtained image.

For example, as shown in the image 600 of FIG. 12, the image may be divided into a plurality of tiles. The size or shape that is the basis of the division is not limited, but for example, the tile may be divided into a predetermined size and shape so that the input to the trained model is possible.

In operation S130, the computer may perform an operation S420 of inputting the generated tile image to the trained model.

In operation S150, the computer may perform operation S430 of obtaining a development pressure corresponding to the generated tile image based on the output of the learned model.

That is, the computer may acquire development pressure for each of the divided tiles, and match and display the obtained development pressure on the image.

In operation S440, the computer may determine the development pressure of the target region based on the development pressure of one or more of the generated tile images.

For example, as described above, the computer may calculate the development pressure of the target region based on the development pressure of the tiles included in the target region, and the specific method is not limited.

6 is a flowchart illustrating a development pressure adjusting method according to an embodiment.

In operation S430, the computer may perform operation S510 of obtaining a development pressure index corresponding to each of the generated tile images.

In operation S440, the computer may generate the development pressure map in which the development pressure index corresponding to each of the one or more tile images is matched to a position corresponding to each of the one or more tile images, in operation S520. have.

That is, the computer may match and display a development pressure index of each tile image at a position corresponding to each tile image in the image including the target region, and generate a development pressure map based on the matching pressure index.

In addition, the computer may perform the step S530 of selecting one or more locations having a development pressure index of a predetermined reference value or more from the development pressure map.

In an embodiment, the computer may generate a contour map by connecting positions having development pressure within a predetermined range with respect to the development pressure displayed at each position of the image or the development pressure map. In addition, the computer can display zones divided into contour lines in different colors, patterns, and patterns based on the development pressure index.

In an embodiment, the computer may select one or more locations having a development pressure index greater than or equal to a preset reference value, and connect the selected locations to generate a figure or a closed curve.

Alternatively, the computer may select and connect one or more locations having a development pressure index of a predetermined reference value or more, and display an area having a predetermined range from the connected line.

In an embodiment, the computer may display tiles having a development pressure index that is greater than or equal to a preset reference value.

In addition, the computer may perform the step S540 of adjusting the development pressure index corresponding to each of the one or more tile images based on the positional relationship with the one or more positions selected in the step S530.

In addition, the computer may perform a step S550 of matching the adjusted development pressure index to the development pressure map.

In addition, the computer may determine the development pressure of the target region based on the adjusted development pressure index (S560).

Referring to FIG. 14, an example of adjusting the development pressure based on the development pressure map is illustrated.

The image 700 illustrated in FIG. 14 divides an image including a target area into tile units, matches a development pressure index corresponding to each tile to the image, and displays an area having a development pressure index of a predetermined reference value or more. will be.

In the embodiment illustrated in FIG. 14, regions 710, 720, and 750 are displayed in the image 700 having a development pressure index higher than or equal to a preset reference value.

In FIG. 14, the remaining regions except for the regions 710, 720, and 750 may be understood to have a development pressure index of less than or equal to a preset reference value. However, this is assumed to be the result of separately determining the development pressure index for each image divided by tiles according to the disclosed embodiment.

At this time, there may be an area 730 surrounded by the area 720. In FIG. 14, the development pressure of the region 730 may be equal to or less than a preset reference value. However, the development pressure of the region 730 may be adjusted according to the terrain and the positional relationship between the region 720 and the region 730 and the terrain of the region 730.

For example, if there is a lake or a mountain in the region 730, the development pressure of the region 730 itself may be low. On the other hand, when the area 730 is an empty lot connected to the area 720 or a terrain that can be easily cleaned up, the development value of the area 730 may also increase together when the area 720 is developed.

Therefore, in this case, the development pressure of the region 730 may be adjusted based on a level corresponding to the development pressure of the region 720 or a predetermined ratio of the development pressure of the region 720.

Similarly, in the case of region 740 located in the space between regions 710 and 720, the development pressure index itself may be low. However, in some cases, there may be a development value as a space connecting the region 710 and the region 720, and thus the development pressure of the region 740 depends on the positional relationship with the region 710 and the region 720. Can be adjusted.

For example, area 740 may have development value depending on the roads connecting area 710 and area 720 and the associated infrastructure, or accessibility to area 710 and area 720. Depending on the development value may be.

In addition, in the case of the region 750, the development pressure of each of the regions 760 and 770 may be adjusted based on the distance from the region 750 with respect to the regions 760 and 770 located at a predetermined distance from the region 750. .

That is, according to the embodiment illustrated in FIG. 14, a development pressure adjusting method based on a positional relationship between different regions, which may complement development pressure analysis for each image divided into tiles, may be provided.

In one embodiment, development pressure adjustment based on the positional relationship between different regions may also be performed based on the learned model.

For example, a position at which a development pressure of a predetermined reference value or more is calculated based on a trained model according to an embodiment disclosed in the image is displayed, and in addition, a difference between the development pressure at which the actual development pressure is calculated and a predetermined reference value or more occurs. Locations may be indicated.

When the model is trained based on the training data, an image indicating a position at which a development pressure of a predetermined reference value or more is calculated based on the trained model according to the disclosed embodiment is input, and the development pressure of other areas is based on the input. Models for adjusting or calculating or outputting information for determining one or more positions that are expected to have development pressure above a predetermined reference value can be learned.

7 is a flowchart illustrating a learning method according to an exemplary embodiment.

In operation S610, the computer may acquire an image of the reference region, which is photographed over the reference region.

In the disclosed embodiment, the reference area may mean, but is not limited to, an area for collecting information for learning a model according to the disclosed embodiment. For example, the reference area may be set at various locations, and may also be set at different times at the same location.

According to the disclosed embodiment, the target area for predicting the development pressure may also be a reference area used for learning a later learned model.

In operation S620, the computer may acquire data for calculating a development pressure index for the reference region.

For example, a computer may collect data about the region's population, business, and public announcement data for the region for the five years from the previous year in the year.

In operation S630, the computer may calculate a development pressure index for the reference region based on the obtained data.

For example, the computer may calculate the development pressure index using the data collected in step S620. For example, the development pressure index can be obtained by calculating the standardized value of the average increase / decrease rate for each of the collected data. Value)] to calculate the raw DPI and finally have a value between 0 and 100 using the formula [(raw DPI / max DPI) * 100], but is not limited thereto. .

In operation S640, the computer may generate learning data including an image of the reference region and a development pressure index for the reference region. For example, the computer may generate, but is not limited to, learning data labeled with the development pressure index calculated on the image of the reference region.

In operation S650, the computer may train the trained model using the training data.

For example, a computer can use the CNN (Convolutional Neural Network) learning module. Specifically, a computer can train a CNN model that calculates the development pressure index using the DenseNet structure, which has recently shown excellent performance in image recognition. Can be. DenseNet has features that make learning well in deeper structures by connecting all feature maps of the same size at the input of the composite product layer of the existing CNN structure.

In this case, the model generator may further include a process of visualizing the learning of the learned deep learning model. For example, CNN can visualize many elements because of its ability to learn visual concepts. Therefore, unlike other deep learning techniques, CNN can be interpreted according to visualization. More specifically, visualize and analyze the learning of this deep learning model by visualizing the activation visualization in the CNN's middle layer, the CNN's filter visualization, and the heatmap for class activation. Easy multi-angle analysis

The CNN learning module may include a process of analyzing what features or elements the model has been trained using the visualization methods. This means, for example, that the model can be trained under the influence of any image element or feature (color, terrain, building, etc.), that is, which element contributes to the development pressure index.

In one embodiment, the computer may predict the development pressure index by inputting input data into the CNN model of the DenseNet structure learned based on the above-described example.

In one embodiment, the preprocessing module used to process the input data may be the same as the preprocessing module used to process the training data, but is not limited thereto.

The computer may preprocess the image taken from the drone and then input the preprocessed image into the learned model to finally predict the development pressure index of the region.

For example, the index prediction module inputs the acquired map image of the analysis target spot into the learned CNN model for each tile, and finally outputs the development pressure index of the tile. In this case, the exponent output value is a continuous value having a value between 0 and 100. Therefore, the average loss function (Mean Square Error (MSE)) is used as a proper cost function, and an output layer generally used for classification in neural networks. The development pressure index can be output by removing the Softmax function, which is the activation function of.

Hereinafter, a method of preprocessing an image in order to generate and process the training data and a method of generating the training data based thereon will be described in detail.

8 is a flowchart illustrating a method of generating training data according to an exemplary embodiment.

In operation S640, the computer may perform operation S710 of generating one or more tile images from the image of the reference region.

First, as illustrated in FIG. 11, the computer may correct the images 400 photographed from the drone 10 and may preprocess the same as the image 500.

For example, a computer may correct images such as lens distortion or noise, and then combine the images by allowing a certain amount of overlapping between the images to form a single whole map image ( 500).

In addition, as described above, the computer may calculate the development pressure index of each region by using the population number, the published land price, the number of businesses of each region, and the like, for each of the target regions included in the image 500. The calculated development pressure index may be matched with the image 500 and stored.

In an embodiment, in the process of matching the calculated development pressure index and the map image 500, the computer may tile each of the regions in which the exponents are divided as shown in FIG. 12. ) You can perform the process of dividing finely into sizes. That is, the computer may generate the image 600 based on the image 500 of FIG. 12.

The process is to learn the model for more detailed regions by dividing the units of the development pressure index calculated earlier into narrower units, and to derive the development pressure index for specific regions. The mapping module provides training data for the deep learning model by matching development pressure indices for each tile in the map.

In one embodiment, the computer may perform preprocessing so that the preprocessed image may be used as an input value for deep learning learning, for example, image splitting, data augmentation, An image processing method such as adjusting an image size (pixel value) may be performed. Data propagation is a technique of artificially increasing the amount of training data by generating new data from existing data. As a method, scaling, rotation, and flipping may be used. However, the present invention is not limited thereto.

In addition, the computer may perform a step S720 of labeling the development pressure index corresponding to each of the one or more tile images.

9 is a flowchart illustrating a labeling method according to an exemplary embodiment.

In operation S720, when the computer labels each of the one or more tile images, the development pressure index of the reference region included in each of the one or more tile images, but includes a plurality of reference regions in the first tile image. In operation S810, the development pressure indices of each of the plurality of reference regions may be obtained.

In operation S820, the computer may calculate a first development pressure index corresponding to the first tile image based on the obtained development pressure indices.

In addition, the computer may perform the step S830 of labeling the calculated first development pressure index on the first tile image.

Referring to FIG. 13, an example of a method of calculating a development pressure index corresponding to each tile is illustrated.

That is, when only one region is included in each of the divided tiles, or regions in which development pressure indices coincide with each other, the development pressure index of the corresponding tile may be determined accordingly.

However, if a tile includes a plurality of regions having different development pressure indices, the computer calculates the development pressure index of the region.

For example, the computer may calculate the development pressure index of the tile based on the average of the development pressure indices of each region included in one tile, and the weight may be given based on the region of each region included in the tile. After that, the average of the development pressure index of the tile may be calculated. That is, the development pressure index of the tile may be calculated by calculating an average based on the area ratio of each region included in the tile.

15 is a block diagram of an apparatus according to an embodiment.

The processor 102 may include a connection passage (eg, a bus, etc.) that transmits and receives signals with one or more cores (not shown) and graphics processor (not shown) and / or other components. .

The processor 102 according to one embodiment executes one or more instructions stored in the memory 104 to perform the method described in connection with FIGS. 1-14.

Meanwhile, the processor 102 may include random access memory (RAM) and read-only memory (ROM) for temporarily and / or permanently storing a signal (or data) processed in the processor 102. , Not shown) may be further included. In addition, the processor 102 may be implemented in the form of a system on chip (SoC) including at least one of a graphic processor, a RAM, and a ROM.

The memory 104 may store programs (one or more instructions) for processing and controlling the processor 102. Programs stored in the memory 104 may be divided into a plurality of modules according to functions.

The steps of a method or algorithm described in connection with an embodiment of the present invention may be implemented directly in hardware, in a software module executed by hardware, or by a combination thereof. Software modules may include random access memory (RAM), read only memory (ROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, hard disk, removable disk, CD-ROM, or It may reside in any form of computer readable recording medium well known in the art.

The components of the present invention may be embodied as a program (or an application) and stored in a medium for execution in combination with a computer which is hardware. The components of the present invention may be implemented as software programming or software elements, and likewise, embodiments may include various algorithms implemented in combinations of data structures, processes, routines or other programming constructs, such as C, C ++. It may be implemented in a programming or scripting language such as Java, an assembler, or the like. The functional aspects may be implemented with an algorithm running on one or more processors.

In the above, embodiments of the present invention have been described with reference to the accompanying drawings, but those skilled in the art to which the present invention pertains may realize the present invention in other specific forms without changing the technical spirit or essential features thereof. I can understand that. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive.

10: drone
20: user terminal
100: server

Claims (10)

In a method performed by a computer,
Acquiring an image of the target area, which is photographed from a target area (S110);
Acquiring input data from the obtained image (S120);
Inputting the input data into a trained model (S130);
Acquiring an output of the learned model (S140); And
Determining development pressure of the target area based on the obtained output (S150); Including,
The step (S120),
Generating at least one tile image by dividing the obtained image (S410); Including,
The step (S130),
Inputting the generated tile image to the learned model (S420); Including,
The step (S150),
Acquiring development pressure corresponding to the generated tile image based on the output of the learned model (S430); And
Determining development pressure of the target region based on development pressure of at least one of the generated tile images (S440); Including,
The step (S430),
Obtaining a development pressure index corresponding to each of the generated tile images (S510); Including,
The step (S440),
Generating a development pressure map in which a development pressure index corresponding to each of the one or more tile images is matched to a position corresponding to each of the one or more tile images (S520);
Selecting one or more locations having a development pressure index of a predetermined reference value or more from the development pressure map (S530);
Adjusting the development pressure index corresponding to each of the one or more tile images based on the positional relationship with the one or more positions selected in the step S530 (S540);
Matching the adjusted development pressure index to the development pressure map (S550); And
Determining development pressure of the target region based on the adjusted development pressure index (S560); Including,
Artificial intelligence based development pressure prediction method using drones. According to claim 1,
The step (S110),
Controlling the drone (S210); And
Acquiring an image of the target area photographed by the drone (S220); Further comprising,
Artificial intelligence based development pressure prediction method using drones. The method of claim 2,
Determining the development pressure of each of the plurality of target areas photographed by the drone (S310); And
Generating a development pressure map including development pressures of the plurality of target areas and each of the plurality of target areas (S320); Further comprising,
Artificial intelligence based development pressure prediction method using drones. delete delete In a method performed by a computer,
Acquiring an image of the target area, which is photographed from a target area (S110);
Acquiring input data from the obtained image (S120);
Inputting the input data into a trained model (S130);
Acquiring an output of the learned model (S140); And
Determining development pressure of the target area based on the obtained output (S150); Including,
The method,
Acquiring an image of the reference region, which is photographed over the reference region (S610);
Acquiring data for calculating a development pressure index for the reference region (S620);
Calculating a development pressure index for the reference region based on the obtained data (S630);
Generating learning data including an image of the reference region and a development pressure index for the reference region (S640); And
Training the trained model using the training data (S650); Further comprising,
Artificial intelligence based development pressure prediction method using drones. The method of claim 6,
The step (S640),
Generating at least one tile image from the image of the reference region (S710); And
Labeling a development pressure index corresponding to each of the one or more tile images (S720); Including,
Artificial intelligence based development pressure prediction method using drones. The method of claim 7, wherein
The step (S720),
Labeling each of the one or more tile images a development pressure index of a reference region included in each of the one or more tile images,
If the first tile image contains a plurality of reference regions,
Obtaining development pressure indices of each of the plurality of reference regions (S810);
Calculating a first development pressure index corresponding to the first tile image based on the obtained development pressure indices (S820); And
Labeling the calculated first development pressure index on the first tile image (S830); Further comprising,
Artificial intelligence based development pressure prediction method using drones. Memory for storing one or more instructions; And
A processor for executing the one or more instructions stored in the memory;
The processor executes the one or more instructions,
An apparatus for carrying out the method of claim 1 or 6. A computer program coupled to a computer, which is hardware, and stored on a computer-readable recording medium for carrying out the method of claim 1.

Images