Estimation device, estimation model, and model creation device

Abstract

This invention provides an estimation device, estimation model, and model creation device that can quickly and inexpensively determine the visibility conditions for each section of a road. [Solution] The estimation device of the present disclosure includes a storage unit that stores an estimation model for estimating the visibility of each section based on the correlation between each of a plurality of sections constituting a road and the fixed camera among a plurality of fixed cameras installed along the road that has the highest correlation with respect to the visibility of the road, and an estimation unit that estimates the visibility of each section at a predetermined time by inputting a visibility index for each installation point of the fixed camera, calculated using road images generated by the fixed camera at a predetermined time, into the estimation model.

Description

Technical Field

[0001] The present disclosure relates to an estimation device, an estimation model, and a model creation device for estimating the visibility of a road.

Background Art

[0002] Due to meteorological phenomena such as rainfall, fog, snowfall, and blizzard, poor visibility may occur on the road. When poor visibility occurs on the road, smooth traffic is hindered and the occurrence of accidents increases, so various countermeasures have been taken.

[0003] Patent Document 1 discloses a driving support system that accumulates visibility obstruction information of a road based on a visibility index calculated based on a road image captured by a road camera installed along the road, and provides the accumulated visibility obstruction information using a predetermined information transmission means.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] There is a desire to grasp the visibility situation of the road even in sections where fixed cameras are not installed. If fixed cameras could be installed in all sections, it would be fine, but when the number of sections is large, it would cost a great deal.

[0006] An object of the present disclosure is to provide an estimation device, an estimation model, and a model creation device that can grasp the visibility situation for each section included in a road at low cost and in a short time.

Means for Solving the Problems

[0007] An estimation device according to one aspect of the present disclosure includes a storage unit that stores an estimation model for estimating the visibility of each section based on the correlation between each of a plurality of sections constituting a road and the fixed camera among a plurality of fixed cameras installed along the road that has the highest correlation with respect to the visibility of the road, and an estimation unit that estimates the visibility of each section at a predetermined time by inputting a visibility index for each installation point of the fixed camera, calculated using road images generated by the fixed camera at a predetermined time, into the estimation model.

[0008] An estimation model according to one aspect of the present disclosure is an estimation model that outputs visibility indicators for each section in response to inputs of visibility indicators for each installation point of fixed cameras installed along a road including multiple sections, and is created by a computer performing the following steps: reading out visibility indicators for each section generated based on road images generated by an on-board camera of a vehicle traveling on the road at the time the vehicle passes through each section, and visibility indicators for each installation point at the time of passage generated based on road images for each installation point generated by the fixed cameras; performing a correlation analysis using the visibility indicators for each section and the visibility indicators for each installation point at the time the vehicle passes through each section; and creating the estimation model based on the results of the correlation analysis.

[0009] A model creation device relating to one subject of this disclosure includes: a correlation analysis unit that performs correlation analysis using visibility indices for each section generated based on road images generated at the time the vehicle passes through each section included in the road by an on-board camera of a vehicle traveling on the road, and visibility indices for each installation point based on road images of each installation point of a fixed camera installed along the road generated at the time of passage by a fixed camera installed along the road, and determines the fixed camera with the highest correlation to visibility for each section based on the results of the correlation analysis; and a model creation unit that generates a regression equation that calculates the visibility index for a specific section based on the visibility index for each installation point of the fixed camera with the highest correlation to the specific section, and creates an estimation model that outputs the visibility index for each section at a predetermined time when the visibility index for each installation point at a predetermined time is input based on the regression equation for all sections. [Effects of the Invention]

[0010] Visibility conditions for each section of a road can be assessed quickly and at low cost. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows an example of the configuration of an estimation system including an estimation device according to an embodiment of the present disclosure. [Figure 2] A flowchart illustrating an example of the visibility index calculation process using an index calculation device. [Figure 3] Block diagram showing an example of the functional configuration of the estimation device. [Figure 4] A flowchart illustrating an example of estimation processing by an estimation device. [Figure 5] A flowchart illustrating the database update process as an example of the overall operation of the estimation system. [Figure 6] A flowchart illustrating the visibility information provision process, which is an example of the overall operation of the estimation system. [Figure 7] A diagram showing an example of the configuration of a model creation system for generating estimation models. [Figure 8] Block diagram showing an example of the functional configuration of the model creation device [Figure 9] Schematic diagram for explaining the correlation regarding visibility between a section and a fixed camera [Figure 10] Flowchart for explaining the method for creating an estimation model in the model creation device [Figure 11] Graph showing the correlation between the visibility index generated based on the fixed camera image and the visibility index generated based on the in-vehicle camera image [Figure 12] Figure for explaining the correlation regarding visibility between a section and a fixed camera [Figure 13] Figure showing the visibility index for each installation location based on the fixed camera image in the embodiment [Figure 14] Figure showing the visibility index based on the in-vehicle camera image in the embodiment [Figure 15] Figure showing the estimation result of the visibility index for each section, which is output by inputting the visibility index based on the fixed camera image shown in FIG. 13 into the pre-created estimation model in the embodiment [Figure 16] Graph showing the correlation coefficient between the visibility index for each section estimated by the estimation model (regression equation group) and the visibility index based on the in-vehicle camera image in FIG. 14 in the embodiment [Figure 17] Figure illustrating the hardware configuration of a computer [Mode for Carrying Out the Invention]

[0012] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.

[0013] [Overview] On roads, the visibility conditions may deteriorate due to meteorological phenomena or the like. Examples of meteorological phenomena that deteriorate the visibility conditions include rainfall, fog, snowfall, blizzard, etc. Blizzard includes ground blizzard that occurs when the snow accumulated on the ground is lifted by the wind even when there is no snowfall.

[0014] Since ground blizzards are primarily caused by wind, their occurrence is easily influenced by topography. For example, in open areas where wind direction changes frequently, the amount of snow whipped up by ground blizzards changes each time the wind direction and speed change. In such terrain, when a ground blizzard occurs, visibility can change rapidly and locally. On the other hand, in areas with windbreaks (such as forests), ground blizzards are less likely to occur even if the wind blows across the entire area.

[0015] When ground blizzards occur on roads passing through various terrains, the presence or absence of ground blizzards changes with changes in terrain, resulting in significantly different visibility conditions depending on the location, even on the same road at the same time. This phenomenon is particularly pronounced with blizzards, including ground blizzards, compared to rain, fog, and snowfall.

[0016] As described above, blizzards, including ground blizzards, can rapidly alter visibility. Furthermore, blizzards can drastically worsen visibility, sometimes resulting in whiteouts. Therefore, when blizzards occur on roads, they have a significant negative impact on vehicles traveling on the roads, disrupting smooth traffic flow.

[0017] Given these circumstances, there is a need to understand the visibility conditions at various points along roads in real time, especially on roads where weather phenomena such as blizzards can cause drastic and localized changes in visibility over short periods of time.

[0018] Traditionally, multiple fixed cameras have been installed along specific roads for purposes such as monitoring traffic conditions and ensuring safety. These fixed cameras along roads are sometimes referred to as fixed-point cameras, road cameras, surveillance cameras, ITV (Industrial Television) cameras, or CCTV (Closed-Circuit Television) cameras.

[0019] Traditionally, road visibility at fixed camera locations has been assessed based on road images obtained from these cameras. Fixed cameras are often installed at intervals of several hundred meters to several kilometers. In the event of rain or snow, visibility conditions tend to be similar over a relatively wide area, making it possible to assess visibility between fixed camera locations based on road images obtained from the cameras. However, as mentioned above, in the event of localized weather phenomena such as blizzards, it is difficult to assess visibility in areas where no fixed cameras are installed. Increasing the number of fixed cameras and shortening the distance between installation locations would increase the number of locations where visibility can be assessed, but the installation, management, and operation of fixed cameras incur significant costs.

[0020] This disclosure is made in light of the circumstances described above, and aims to efficiently and cost-effectively estimate visibility conditions in sections of a road where fixed cameras are not installed, even when localized weather phenomena occur on the road.

[0021] [Configuration of the estimation system, including the estimation device] Figure 1 shows an example of the configuration of an estimation system 100 including an estimation device 110 according to an embodiment of the present disclosure. The estimation system 100 estimates the visibility at a certain time (point in time) for each of several sections included in the road to be estimated for visibility.

[0022] The roads subject to visibility estimation are, for example, roads arbitrarily designated by the administrator of the estimation system 100. The administrator of the estimation system 100 is the person who provides the estimation system 100 to drivers intending to travel on the roads and who operates and manages the estimation system 100. In the following description, the administrator of the estimation system 100 may be simply referred to as the administrator. Furthermore, the roads subject to visibility estimation may be a single continuous road from one starting point to one ending point, or they may include branches along the way. In the following description, the roads subject to visibility estimation may be simply referred to as roads.

[0023] The road is divided into several sections by the administrator, and each section is assigned a unique number. It is desirable that each section is the same length. An example of a section length is 100m. The estimation system 100 estimates the visibility of the road for each section at a predetermined point in time.

[0024] As shown in Figure 1, the estimation system 100 comprises an estimation device 110, a plurality of fixed cameras 120, an index calculation device 130, and a server device 140. The plurality of fixed cameras 120 and the index calculation device 130, the index calculation device 130 and the estimation device 110, and the estimation device 110 and the server device 140 are each connected to each other so as to be able to communicate with each other by wired communication or wireless communication. The plurality of fixed cameras 120 and the index calculation device 130, the index calculation device 130 and the estimation device 110, and the estimation device 110 and the server device 140 may each be connected to each other via a public network such as the Internet.

[0025] Of the components of the estimation system 100, the estimation device 110, the index calculation device 130, and the server device 140 are each composed of computers.

[0026] Multiple fixed cameras 120 are installed along the road and each captures the road to generate road images. The fixed cameras 120 are, for example, cameras that generate moving images of the road. The moving images generated by the fixed cameras 120 may be stored in, for example, an image server device (not shown), associated with the identification information of the fixed camera 120 that generated the moving image, and information regarding the date and time the moving image was generated.

[0027] The multiple fixed cameras 120 may also be cameras that capture still images of the road. For example, the multiple fixed cameras 120 may simultaneously capture still images of the road at predetermined time intervals (for example, every 5 minutes). In this case, the road images (still images) of each installation location will be generated at the same time by the multiple fixed cameras 120.

[0028] Thus, the road images produced by the fixed camera 120 may be moving images or still images. In the following description, the installation location of the fixed camera 120 may be simply referred to as the installation location. Also, in the following description, the road images taken by the fixed camera 120 at the installation location may be referred to as fixed camera images.

[0029] The index calculation device 130 calculates a visibility index related to the visibility of the road captured in the fixed camera image, that is, a visibility index for each fixed camera installation location, based on the fixed camera image generated by the fixed camera 120. The visibility index is an index that indicates how easy the image is to see (visibility) obtained by image processing of the road image.

[0030] The index calculation device 130 calculates the visibility index by, for example, the following method. Figure 2 is a flowchart illustrating an example of the visibility index calculation process by the index calculation device 130.

[0031] In step S1, the index calculation device 130 extracts the area containing the road from the road image as an evaluation image. If the road image is a moving image, in step S1, the index calculation device 130 may extract still images at predetermined time intervals and then extract the evaluation image.

[0032] In step S2, the index calculation device 130 converts the evaluation image extracted in step S1 into grayscale.

[0033] In step S3, the index calculation device 130 decomposes the brightness distribution of each pixel into spatial frequency components for each image converted in step S2 using a two-dimensional Fourier transform, and calculates the power spectrum corresponding to each obtained spatial frequency. The spatial frequency components are components that indicate the density distribution pattern of the image and its intensity.

[0034] In step S4, the index calculation device 130 extracts power spectra within a predetermined range from the power spectrum calculated in step S3, calculates a total value, and outputs this total value as the visibility index value.

[0035] The pre-defined range is the power spectral range in which human contrast sensitivity is high (i.e., image contrast is easily perceived). The spatial frequency bands that are easily perceived by humans are known, and the above-mentioned pre-defined range is set based on these known spatial frequency bands.

[0036] The visibility index calculated through this process will be higher the better the visibility of the road in the evaluation image, and lower the lower the visibility of the road in the evaluation image.

[0037] This visibility index calculation process allows for the accurate quantification of the visibility of roads as they appear in road images.

[0038] The method for calculating the visibility index using the index calculation device 130, as described in Figure 2, is just one example, and the method for calculating the visibility index in this disclosure is not limited to this. For the method of calculating the visibility index using the index calculation device 130 in this disclosure, other known technologies different from the method described above may be appropriately adopted.

[0039] Returning to the explanation of Figure 1, the estimation device 110 estimates the visibility for each section at a predetermined time using a visibility index for each installation location calculated based on fixed camera images generated at a predetermined time by multiple fixed cameras, and a pre-created estimation model.

[0040] Figure 3 is a block diagram showing an example of the functional configuration of the estimation device 110. As shown in Figure 3, the estimation device 110 comprises an acquisition unit 111, a storage unit 112, an estimation unit 113, and an output unit 114.

[0041] The acquisition unit 111 acquires visibility indicators for each installation location of the fixed camera 120 at a predetermined time from the indicator calculation device 130.

[0042] The storage unit 112 stores a pre-created estimation model. The estimation model is a model created in advance by the model creation system 200, which will be described later. The estimation model is constructed to estimate and output the visibility index for each section at a predetermined time when the visibility index for each installation location at a predetermined time is input.

[0043] The estimation unit 113 estimates the visibility for each section at a predetermined time by inputting the visibility index for each fixed camera installation location, calculated using fixed camera images at a predetermined time, into the estimation model. As described above, the estimation model is constructed to output the visibility index for each section at a predetermined time when the visibility index for each installation location at a predetermined time is input. Based on the visibility index for each section at a predetermined time output by the estimation model, the estimation unit 113 generates visibility information regarding the visibility for each section.

[0044] Visibility information is information generated by the estimation unit 113 based on the visibility index for each section. Visibility information is a value obtained by converting the visibility index in a way that is easy for humans to intuitively understand. A specific example of visibility information is the field of view level. The field of view level is a value obtained by ranking the visibility for each section into n levels (n is an integer of 2 or more, for example n=4). This disclosure does not particularly limit the method for calculating the field of view level based on the visibility index, but for example, a method may be adopted in which a subject is shown multiple images for which the visibility index is known in advance, and a correspondence between the visibility index and the field of view level is derived in advance based on which field of view level the subject sets for each image, and the visibility index is converted to the field of view level based on that correspondence.

[0045] The output unit 114 outputs the visibility information for each section generated by the estimation unit 113 to the server device 140. The output unit 114 may also output the visibility index for each section to the server device 140 along with the visibility information for each section.

[0046] Figure 4 is a flowchart illustrating an example of the estimation process performed by the estimation device 110.

[0047] In step S11, the acquisition unit 111 acquires visibility indicators for each installation location of the fixed camera 120 at a predetermined time from the indicator calculation device 130.

[0048] In step S12, the estimation unit 113 inputs the visibility index for each installation location at a predetermined time into the estimation model.

[0049] In step S13, the estimation unit 113 obtains the visibility index for each section at a predetermined time point, which is output from the estimation model.

[0050] In step S14, the estimation unit 113 generates visibility information for each section at a predetermined time based on the visibility index for each section at a predetermined time.

[0051] In step S15, the output unit 114 outputs the visibility information for each section at a predetermined point in time, generated by the estimation unit 113, to the server device 140.

[0052] This estimation process allows for the generation of visibility information for each section based on the visibility index for each installation point of the fixed cameras 120. This makes it possible to estimate visibility for each section collectively, even if the number of fixed cameras 120 is less than the number of sections included in the road. How the estimation model estimates the visibility index for each section based on the visibility index for each installation point will be described later.

[0053] Returning to the explanation of Figure 1, the server device 140 has a visibility database containing visibility information for each time point and each interval. When the server device 140 receives visibility information for each interval at a new time point from the estimation device 110, it adds the new visibility information to the visibility database. As a result, the visibility database of the server device 140 accumulates estimation results for each interval at multiple past time points.

[0054] The server device 140 is connected to external information terminals via a network such as the Internet. These external information terminals are terminals outside the estimation system 100, such as terminals used by drivers who are about to drive on a road. Examples of information terminals include PCs (Personal Computers), smartphones, tablet terminals, and car navigation systems. Based on requests received from various external information terminals, the server device 140 refers to the visibility database and transmits visibility information for the requested time and section to the information terminals.

[0055] The visibility database may further include corresponding visibility indices associated with visibility information for each section at a predetermined time. The visibility database may further include corresponding fixed camera images associated with visibility information for each section at a predetermined time. Furthermore, the visibility database may further include, for example, weather information and traffic information for that time and section, associated with visibility information for each section at a predetermined time. In this case, the server device 140 can transmit at least one of the visibility information, road images, weather information, and traffic information for the requested time and section to the information terminal.

[0056] [Example of operation of estimated system 100: Database update process] The following describes an example of the operation of the estimation system 100. Figure 5 is a flowchart illustrating the database update process, which is one of the examples of the operation of the estimation system 100.

[0057] In step S21, the indicator calculation device 130 acquires fixed camera images for each installation location at a predetermined time.

[0058] In step S22, the index calculation device 130 performs the visibility index calculation process described in Figure 2.

[0059] In step S23, the estimation device 110 performs the estimation process described in Figure 3.

[0060] In step S24, the server device 140 adds visibility information for each section at a predetermined time to the visibility database.

[0061] Through the database update process described above, the server device 140 accumulates visibility information for each section at multiple points in the past. This enables the estimation system 100 to respond to requests for visibility information from any point in the past when an external information terminal requests it.

[0062] Furthermore, it is desirable that the database update process be executed each time the fixed camera 120 generates a new fixed camera image. That is, if the fixed camera 120 generates a new fixed camera image every 5 minutes, the database update process should also be performed every 5 minutes. Alternatively, for example, the database update process may wait until a sufficient number of fixed camera images have been accumulated from the fixed camera 120, and then begin.

[0063] [Example of operation of estimation system 100: Visibility information provision processing] Figure 6 is a flowchart illustrating the visibility information provision process, which is one example of the operation of the estimation system 100.

[0064] In step S31, the server device 140 receives a request for visibility information from an external information terminal.

[0065] In step S32, the server device 140 reads the visibility information for the requested time and section from the visibility database.

[0066] In step S33, the server device 140 transmits the visibility information read in step S32 to the requesting information terminal.

[0067] Through the visibility information provision process described above, the estimation system 100 can suitably provide visibility information for a given time and section requested by a driver or other user holding an information terminal. For example, a driver can request the latest visibility information for the section they plan to travel, as well as corresponding road images, weather information, and traffic information, from the estimation system 100, and by confirming the information, they can safely travel that section. Furthermore, on roads where multiple routes can be set, including branching paths, a driver can obtain visibility information for the sections included in each route, thereby appropriately selecting a route with good visibility.

[0068] Furthermore, the visibility information provision process only needs to be executed as needed whenever the server device 140 receives a request from an external information terminal.

[0069] [How to create an estimation model] The following describes how to create an estimation model used by the estimation device 110 of the estimation system 100 to estimate visibility information. Figure 7 shows an example of the configuration of the model creation system 200 for creating the estimation model.

[0070] As shown in Figure 7, the model creation system 200 comprises a plurality of fixed cameras 120, a vehicle 230 equipped with an on-board camera 210 and a position detection unit 220, an index calculation device 240, and a model creation device 250. The plurality of fixed cameras 120 and the index calculation device 240, the on-board camera 210 and the position detection unit 220 and the index calculation device 240, and the index calculation device 240 and the model creation device 250 are each connected to each other so as to be able to communicate with each other by wired communication or wireless communication. The plurality of fixed cameras 120 and the index calculation device 240, the on-board camera 210 and the position detection unit 220 and the index calculation device 240, and the index calculation device 240 and the model creation device 250 may each be connected to each other via a public network such as the Internet.

[0071] The fixed camera 120 included in the model creation system 200 is the same fixed camera 120 included in the estimation system 100 shown in Figure 1. Furthermore, the index calculation device 240 included in the model creation system 200 may be the same device as the index calculation device 130 included in the estimation system 100 shown in Figure 1, or it may be a different device. If the index calculation device 240 and the index calculation device 130 are different devices, then the index calculation device 240 and the index calculation device 130 should calculate the visibility index using the same visibility index calculation process (see Figure 3).

[0072] Of the components of the model creation system 200, the index calculation device 240 and the model creation device 250 are each composed of computers.

[0073] Vehicle 230 travels on a specific road. In this disclosure, vehicle 230 may be any type of vehicle, but preferably vehicle 230 is a vehicle that travels regularly on a road from a predetermined starting point to an end point in a fixed direction. Vehicle 230 may travel back and forth on a road from a predetermined starting point to an end point, or it may travel in one direction. Preferred examples of vehicle 230 include route buses, road patrol vehicles of road administrators, or scheduled trucks.

[0074] The on-board camera 210 is a camera mounted on the vehicle 230. The on-board camera 210 captures images of the road around the vehicle 230 at predetermined time intervals while the vehicle 230 is traveling on a specific road, and generates road images. For example, the on-board camera 210 may capture images of the road in front of the vehicle 230 to generate road images. The road images generated by the on-board camera 210 may be moving images or still images. In the following description, the road images generated by the on-board camera 210 may be referred to as on-board camera images.

[0075] The position detection unit 220 detects the position of the vehicle 230 at the time the in-vehicle camera 210 photographs the road and outputs it as position information. The position detection unit 220 is, for example, a GPS sensor and detects the position of the vehicle 230 by communicating wirelessly with GPS (Global Positioning System) satellites.

[0076] Each in-vehicle camera image is associated with location information detected by the location detection unit 220, which indicates the location where the in-vehicle camera 210 generated the image. The in-vehicle camera 210 and the location detection unit 220 transmit the in-vehicle camera image and the location information associated with the in-vehicle camera image to the index calculation device 240.

[0077] The index calculation device 240 calculates a visibility index for each installation location of the fixed camera 120 based on the fixed camera images received from the fixed camera 120. The index calculation device 240 also calculates a visibility index for each section based on the in-vehicle camera images and location information received from the in-vehicle camera 210. Specifically, the index calculation device 240 determines, based on the location information, that an in-vehicle camera image was taken by the in-vehicle camera 210 while the vehicle 230 was traveling through a certain section, and uses the visibility index calculated based on that in-vehicle camera image as the visibility index for that section. If there are multiple in-vehicle camera images taken within a single section, the index calculation device 240 may, for example, use the average or minimum value of the visibility indices calculated based on multiple in-vehicle camera images taken within that section as the visibility index for that section.

[0078] The indicator calculation device 240 calculates visibility indicators for each installation location based on fixed camera images and visibility indicators for each section based on vehicle-mounted camera images by performing the visibility indicator calculation process described in Figure 2.

[0079] The model creation device 250 accumulates visibility indices for each installation location and each section at multiple points in time, and creates an estimation model based on the large amount of accumulated visibility indices for each installation location and each section. The estimation model is a model that, when the visibility indices for each installation location at a given time are input, estimates and outputs the visibility indices for each section at that time.

[0080] Figure 8 is a block diagram showing an example of the functional configuration of the model creation device 250. As shown in Figure 8, the model creation device 250 comprises an accumulation unit 251, a correlation analysis unit 252, and a model creation unit 253.

[0081] The storage unit 251 stores visibility indicators for each installation location and for each section at multiple points in the past.

[0082] The correlation analysis unit 252 performs a correlation analysis between the visibility of a section and a fixed camera, based on the visibility index for each section generated from the on-board camera image generated by the on-board camera 210 at the time the vehicle 230 passes through a certain section, and the visibility index for each installation point at that time. The correlation analysis unit 252 then determines the fixed camera with the highest correlation to visibility for each section. This makes it possible to associate the visibility of each section included in the road with the fixed camera that has the highest correlation to that section.

[0083] Past experience has shown that when visibility conditions change locally in a short period of time due to localized weather phenomena such as blizzards, there can be a high correlation between the visibility in a certain section and the visibility at an installation site located at a distance from that section at a given point in time. This phenomenon can occur when the environmental conditions of a certain section and an installation site located at a distance from that section are similar, when wind blows across the region, the main wind direction on both the section and the installation site is wide open, and when a blizzard occurs simultaneously in both the section and the installation site.

[0084] Figure 9 is a schematic diagram illustrating the correlation between the visibility of sections and fixed cameras. In the example shown in Figure 9, the road contains 10 sections s1 to s10, and fixed cameras c1 to c3 are installed along the road. Fixed camera c1 is installed within section s1, fixed camera c2 is installed within section s5, and fixed camera c3 is installed within section s9.

[0085] In Figure 9, the fixed camera with the highest correlation to each section is shown connected by a straight line. In Figure 9, the section with the highest correlation to fixed camera c1 is connected by a solid line, the section with the highest correlation to fixed camera c2 is connected by a relatively coarse dashed line, and the section with the highest correlation to fixed camera c1 is connected by a relatively fine dashed line.

[0086] In the example shown in Figure 9, the correlation between the section and the visibility of the fixed camera is as follows: • Section s1 has the highest correlation with the fixed camera c1. • Section s2 has the highest correlation with the fixed camera c2. • Section s3 has the highest correlation with the fixed camera c2. • Section s4 has the highest correlation with the fixed camera c1. • Section s5 has the highest correlation with the fixed camera c2. Section s6 has the highest correlation with the fixed camera c3. Section s7 has the highest correlation with the fixed camera c3. • Section s8 has the highest correlation with the fixed camera c1. • Section s9 has the highest correlation with the fixed camera c3. • Section s10 has the highest correlation with the fixed camera c2.

[0087] When the correlation between the interval sx (x=1~10) and the fixed camera cy (y=1~3) is highest, if a blizzard occurs in interval sx and the visibility index decreases, the visibility index will also decrease at the fixed camera cy installation site. Conversely, if the wind subsides and the visibility index increases in interval sx, the visibility index is likely to increase at the fixed camera cy installation site as well.

[0088] The correlation analysis unit 252 can, for example, analyze the correlation between the section and the visibility of the fixed camera using a known correlation analysis method.

[0089] The model creation unit 253 creates a regression equation using a set of visibility indicators for a certain section as the dependent variable, and a set of visibility indicators for each fixed camera installation location that has the highest correlation with that section as the independent variables. An example of a regression equation created by the model creation unit 253 is a linear regression equation using the least squares method.

[0090] The regression equation can be as follows, for example, equation (1) below.

number

[0091] In equation (1), WIPS S This is a visibility index for each section, and WIPS C The 'b' is a visibility index for each installation location, the subscript 'k' is the identification number for the section, the subscript 'i' is the identification number indicating the number of times a vehicle has passed, and the subscript 'j' is the identification number for the installation location. The 'b' is the slope. 1kj and intercept b 0kj This is calculated based on the visibility index for each installation location and the visibility index for each section, which are stored in the storage unit 251.

[0092] The model creation unit 253 creates the above regression equation for all sections included in the road. Based on the visibility index for each installation point at a given time, the visibility index for each section at that time can be calculated using this regression equation.

[0093] The model creation unit 253 outputs a set of regression equations for the entire section as an estimation model. In this way, since the estimation model includes regression equations that estimate the visibility index for the entire section, it can output the visibility index for each section at a given time based on the input of the visibility index for each installation location at that time.

[0094] In the above explanation, the model creation unit 253 assumed that the regression equation for all intervals was used as the estimation model, but this disclosure is not limited to this. For example, the model creation unit 253 may use known machine learning methods to create an estimation model that estimates the visibility index for each interval.

[0095] The estimated model created by the model creation device 250 is stored in the estimation device 110 of the estimation system 100 shown in Figure 1. This allows the estimation device 110 to perform the estimation process shown in Figure 2.

[0096] [How to create an estimation model] Figure 10 is a flowchart illustrating the method for creating an estimated model in the model creation device 250.

[0097] In step S41, the model creation device 250 reads out the visibility index for each installation location and the visibility index for each section at multiple points in the past.

[0098] In step S42, the model creation device 250 performs a correlation analysis on the visibility between a section and a fixed camera, based on the visibility index for each installation point and the visibility index for each section at the same time point.

[0099] In step S43, the model creation device 250 creates a regression equation that calculates the visibility index for the entire section based on the results of the correlation analysis, using the visibility index generated based on the fixed camera image of the fixed camera that has the highest correlation with visibility for that section.

[0100] In step S44, the model creation device 250 creates an estimation model based on the set of regression equations for the entire interval.

[0101] This process allows for the creation of an estimation model that can accurately estimate the visibility index for each section based on the visibility index for each installation location. According to this estimation model, by inputting the visibility index for each installation location at a predetermined time, the visibility index for each section at that time can be obtained all at once, thus providing information on the visibility of the entire road at low cost and in a short time.

[0102] [Examples] The following describes an example of creating an estimation model in the above-described embodiment and estimating visibility indices for each section using the estimation model.

[0103] In this example, the specific road chosen was National Route 232 (Hokkaido), specifically the section from Teshio Town to Rumoi City, excluding the section that overlaps with National Route 40. National Route 232 is a national highway connecting Wakkanai City and Rumoi City. In the following description, the intersection of National Route 232 and National Route 40 may be referred to as the starting point (0 km point) of National Route 232, and Rumoi City as the ending point (128 km point).

[0104] Nineteen fixed cameras are installed along National Route 232 from Teshio Town to Rumoi City. With the starting point set at 0 km, the 19 fixed cameras are installed at the following distances from the starting point: c1:0km c2: 5.272km c3: 9,000km c4: 18.583km c5:31.086km c6:35.805km c7: 39.854km c8:44.033km c9: 45.300km c10: 49.800km c11: 61.129km c12: 67.775km c13: 71.100km c14: 75.675km c15: 80.550km c16: 91.093km c17: 96.324km c18: 104.622km c19: 126.400km

[0105] These fixed cameras simultaneously generated fixed camera images every 5 minutes.

[0106] National Route 232 is used by intercity buses that run six round trips a day between Wakkanai City and Sapporo City. Onboard cameras were installed on the intercity buses to capture images of the road ahead at 8-second intervals.

[0107] National Route 232 was divided into sections of 0.1 km each. This resulted in a total of 1281 sections being designated.

[0108] From December 1, 2021 to February 28, 2022 (a 90-day period), fixed camera images and in-vehicle camera images generated between 7:00 AM and 4:00 PM were collected, and a visibility index was calculated. An estimation model was then created using the calculated visibility index.

[0109] Figure 11 is a graph showing the correlation between visibility indices generated from fixed camera images and visibility indices generated from vehicle-mounted camera images. The vertical axis of Figure 11 is the correlation coefficient, and the horizontal axis is the distance from the starting point of National Route 232. Figure 11 shows the correlation between the visibility index for each installation location, generated from fixed camera images, and the visibility index for each installation location, generated from vehicle-mounted camera images generated at the time a vehicle (bus) passes through the installation location. According to Figure 11, it can be seen that, depending on the location, the correlation between the visibility index generated from fixed camera images and the visibility index generated from vehicle-mounted camera images changes even if the images were generated at the same time and location. Furthermore, the change in the correlation between the visibility index generated based on fixed camera images and the visibility index generated based on in-vehicle camera images is thought to be due to factors such as the difference in height between the fixed camera and the in-vehicle camera (fixed camera height: approximately 10m, in-vehicle camera height: approximately 1.5m), and the difference in field of view resulting from the difference in height (the fixed camera captures a wider area).

[0110] Correlation analysis was performed using visibility indices for each installation location generated based on fixed camera images, and visibility indices for each section generated based on vehicle-mounted camera images, to determine the fixed camera with the highest correlation to visibility for each section. Figure 12 is a diagram illustrating the correlation between visibility between sections and fixed cameras, as created in the example.

[0111] Figure 12 shows, with circles and arrows, examples where a particular section has the highest correlation with a fixed camera installed at a distance from that section. Figure 12 shows an example where a section around 45km has the highest correlation with fixed camera c1 installed at 0km. Furthermore, Figure 12 shows an example where multiple sections around 25km have the highest correlation with fixed camera c15 installed at approximately 80km. These examples demonstrate that even geographically distant locations can have similar visibility conditions at the same time.

[0112] In this way, the fixed camera with the highest correlation in visibility was determined for each of the 1281 sections. Then, for each of the 1281 sections, a regression equation (see equation (1) above) was created to calculate the visibility index of the section based on the visibility index of the installation location of the fixed camera with the highest correlation. The 1281 regression equations were combined to form an estimation model that estimates the visibility index for each section. This estimation model can output the visibility index for each section (in this example, the visibility index for 1281 sections) at a predetermined time when the visibility index for each installation location at a predetermined time (in this example, the visibility index for 19 locations) is input.

[0113] The following sections provide specific examples of visibility indicators for each of the 19 fixed camera locations along National Route 232, visibility indicators for each section obtained from on-board cameras of vehicles traveling on National Route 232, and visibility indicators for each section obtained from an estimation model.

[0114] Figure 13 shows the visibility index for each installation location based on fixed camera images from 7:00 to 11:00 on January 4, 2022. On January 4, 2022, visibility impairment due to ground blizzards actually occurred in at least some sections of National Route 232. The horizontal axis of Figure 13 shows the passage of time, and the vertical axis shows the distance from the starting point of National Route 232.

[0115] The circles and triangles in Figure 13 represent four levels of visibility based on the visibility index. White circles indicate visibility level 1, where the visibility index is highest and visibility is best. Shaded circles indicate visibility level 2, where the visibility index is lower than level 1. Shaded triangles indicate visibility level 3, where the visibility index is lower than level 2. Black triangles indicate visibility level 4, where the visibility index is lowest and visibility is worst. Note that in the example shown in Figure 13, there were no installation locations exhibiting visibility level 4.

[0116] In the example shown in Figure 13, the visibility index for all installation locations is recorded every 5 minutes. Note that blank spaces are provided for locations where fixed camera images unsuitable for calculating the visibility index were acquired due to snow accumulation in front of the camera, or where images could not be acquired due to communication failures, etc.

[0117] Figure 13 shows that data indicating visibility indicators was suitably acquired at predetermined time intervals for each installation location.

[0118] Figure 14 shows the visibility index for each section, based on onboard camera images generated by an onboard camera installed on an intercity bus that passed the 0km point on National Route 232 at approximately 7:50 AM and the 128km point at approximately 10:25 AM on January 4, 2022.

[0119] Figure 14 shows how visibility indicators are obtained for each section as an intercity bus travels along National Highway 232. Note that data is missing between 9:10 and 9:30 because the bus was stopped for a rest.

[0120] Figure 14 shows that data indicating visibility indices in the section through which the vehicle passes was successfully acquired.

[0121] Figure 15 shows the estimated visibility index for each section, which was output by inputting the visibility index based on the fixed camera image shown in Figure 13 into a pre-created estimation model. In addition to the estimated results, Figure 15 also shows the visibility index based on the fixed camera image shown in Figure 13, and the visibility index based on the in-vehicle camera image shown in Figure 14.

[0122] In Figure 15, the vertical lines at each 5-minute interval represent the visibility level corresponding to the visibility index for each section, which is the estimated result. Figure 15 shows that the visibility index was estimated at each time point in almost all sections where no fixed cameras were installed. Note that in Figure 15, data is missing for sections and time points where estimation was not possible for some reason.

[0123] Figures 13 to 15 reveal the following: First, the visibility index based on fixed camera images, shown in Figure 13, and the visibility index based on in-vehicle camera images, shown in Figure 14, are generally at similar levels at the same location and time. Even when the visibility index based on fixed camera images and the visibility index based on in-vehicle camera images do not coincide at the same time, they often coincide at a nearby time.

[0124] The visibility index for each section shown by the estimation model in Figure 15 is roughly similar to the visibility index based on fixed camera images shown in Figure 13, and the visibility index based on vehicle-mounted camera images shown in Figure 14, for the same section and location. This indicates that an estimation model has been obtained that can estimate the visibility index with a certain degree of accuracy even in sections where fixed cameras are not installed.

[0125] Figure 16 is a graph showing the correlation coefficient between the visibility index for each section estimated by the estimation model (regression equation group) and the visibility index based on onboard camera images generated when a vehicle actually traveled through each section (i.e., the visibility index shown in Figure 14). As shown in Figure 16, the visibility index for each section estimated by the estimation model can be seen to agree to a certain extent with the visibility index based on onboard camera images generated when a vehicle actually traveled through each section, across almost the entire length of National Route 232.

[0126] Thus, as shown in the embodiment, when the visibility index for each section is estimated using the estimation model of this disclosure, it can be seen that the visibility for each section can be estimated all at once, quickly and accurately by inputting the visibility index for each installation location at a predetermined time into the estimation model.

[0127] [Example of computer hardware configuration] In the embodiments described above, the estimation device 110, the index calculation device 130, the server device 140, the index calculation device 240, and the model creation device 250 are all computers, and their functional configurations are realized by the computer executing a predetermined program. Below, we will describe examples of the computer hardware configurations that realize the functions of the estimation device 110, the index calculation device 130, the server device 140, the index calculation device 240, and the model creation device 250.

[0128] Figure 17 illustrates the hardware configuration of computer 2100. As shown in Figure 17, computer 2100 includes input devices 2101 such as input buttons and a touchpad, output devices 2102 such as a display and speakers, a CPU (Central Processing Unit) 2103, ROM (Read Only Memory) 2104, and RAM (Random Access Memory) 2105. Computer 2100 also includes storage devices 2106 such as a hard disk drive and SSD (Solid State Drive), a reader 2107 that reads information from recording media such as DVD-ROM (Digital Versatile Disk Read Only Memory) and USB (Universal Serial Bus) memory, and a transceiver 2108 that communicates via a network. The above components are connected by a bus 2109.

[0129] The reading device 2107 reads the program for realizing the functions of each of the above-mentioned parts from the recording medium on which the program is recorded and stores it in the storage device 2106. Alternatively, the transmitting / receiving device 2108 communicates with a server device connected to the network and stores the program for realizing the functions of each of the above-mentioned parts downloaded from the server device in the storage device 2106.

[0130] The CPU 2103 copies the program stored in the memory device 2106 to the RAM 2105, and then sequentially reads and executes the instructions contained in that program from the RAM 2105, thereby realizing the functions of each of the above-mentioned parts. Furthermore, when the program is executed, information obtained from the various processes described in each embodiment is stored in the RAM 2105 or the memory device 2106 and used as appropriate. [Industrial applicability]

[0131] The estimation device described herein is useful for estimating visibility in sections where fixed cameras are not installed. [Explanation of symbols]

[0132] 100 Estimation Systems 110 Estimation device 111 Acquisition Department 112 Storage Unit 113 Estimation Department Output section of 114 120 Fixed Cameras 130 Indicator calculation device 140 Server Devices 200 Model Creation Systems 210 In-car cameras 220 Position detection unit 230 vehicles 230 Index calculation device 230 vehicles National Highway 232 240 Index calculation device 250 Model Creation Devices 251 Storage Unit 252 Correlation Analysis Department 253 Model Creation Department

Claims

1. A storage unit for storing an estimation model that estimates the visibility for each section based on the correlation between each of the multiple sections that make up the road and the fixed camera among the multiple fixed cameras installed along the road that has the highest correlation with the visibility of the road, An estimation unit estimates the visibility for each section at a predetermined time by inputting the visibility index for each fixed camera installation point, calculated using road images generated by the fixed camera at a predetermined time, into the estimation model. An estimation device equipped with the following features.

2. The storage unit stores the estimation model constructed by performing a correlation analysis using the visibility index for each section, calculated based on the road image generated by the vehicle's onboard camera at the time a vehicle traveling on the road passes through each of the sections, and the visibility index for each installation point at the time of passage. The estimation device according to claim 1.

3. An estimation model that takes as input the visibility index for each installation point of a fixed camera installed along a road that includes multiple sections, and outputs the visibility index for each section, The process involves reading out visibility indicators for each section, generated based on road images generated by an onboard camera of a vehicle traveling on the road at the time the vehicle passes through each section, and visibility indicators for each installation point at the time of passage, generated based on road images for each installation point generated by a fixed camera. A process of performing correlation analysis using the visibility index for each section and the visibility index for each installation point at the time the vehicle passes through each section, A process for creating the estimation model based on the results of the correlation analysis, An estimation model created by a computer executing the following.

4. A correlation analysis unit performs a correlation analysis using road images generated by an onboard camera of a vehicle traveling on a road at the time the vehicle passes through each section of the road, and visibility indices for each installation location generated by road images of fixed cameras installed along the road at the time of passage, and determines the fixed camera with the highest correlation to visibility for each section based on the results of the correlation analysis. A model creation unit generates a regression equation that calculates the visibility index for a specific section based on the visibility index for each fixed camera installation location that has the highest correlation with the specific section, and creates an estimation model that outputs the visibility index for each section at a predetermined time when the visibility index for each installation location at a predetermined time is input based on the regression equation for the entire section. A model creation device equipped with the following features.

Images