KR102308243B1 - Road Weather Information System using Image Data Checking Process - Google Patents

Abstract

The present invention includes a data receiving unit for receiving external image information, a data processing unit for checking whether the received image information is error or normal step by step, and a data storage unit for storing corrected data after image information inspection, the data receiving unit Receives image data observed in real time from individual observation devices that take images of the road, and the data processing unit performs a missing test on the received image data in a first step, excluding error values, and performing subsequent steps for normal values. of the image data for fog, rain, snow, and road surfaces, among the data filtered through the first step, the data processing unit performs physical limit check in the second step, step check in the third step, and the second step, respectively. A road weather information system is provided that collects normal data by excluding data corresponding to errors in each inspection by sequentially performing a continuity test in step 4 and a bouncing value test in step 5.

Description

Road Weather Information System using Image Data Checking Process

The present invention relates to a road weather information system, and more particularly, it proposes a new system for collecting and analyzing road weather information by processing observed road image data.

Weather information is information that has a great influence on daily life, and as the weather fluctuates due to climate change, interest in weather information is increasing.

If the weather conditions cannot be quickly grasped, human casualties and economic damage are enormous compared to other regions, so interest in weather monitoring technology is increasing. In order to monitor and predict accurate weather information, it is necessary to acquire meteorological data through the installation of meteorological equipment. However, installing meteorological equipment in urban areas is difficult due to problems such as site selection, power supply, safety, and theft. have. Therefore, there is a need for a technology for acquiring rich and accurate weather situation information using the already established infrastructure. According to registered patent No. 10-1461184, by receiving the captured image from the CCTV module, analyzing the received image, and extracting the current weather condition information, the weather using the already established CCTV system infrastructure without the installation of weather equipment, etc. Context information can be acquired in real time.

On the other hand, traffic accidents can occur due to road environmental factors or weather environmental factors. In particular, weather conditions play a decisive role in driver visibility and vehicle friction, so it is important to understand road weather information for stability and mobility when using the road. very important. In addition, it is possible to prevent or reduce traffic accidents in advance by providing accurate road weather information. As in urban areas, weather conditions can be grasped using existing facilities in addition to the additional installation of meteorological equipment on roads. According to Registered Patent No. 10-1858339, there is provided a weather observation apparatus for detecting weather based on an image received through a camera and a method for controlling the same, which is a method for detecting a weather condition based on an image received in real time through a camera. and a controller, wherein the controller designates a partial region of the received image as a region of interest, monitors a real-time image change amount of the region of interest, and detects that the weather condition corresponds to any one of snow, rain, or fog based on this. Observation apparatus and method are provided.

The image information used in the prior art is installed on the road and received from the CCTV for photographing the traffic situation. Except for special cases, the CCTV always shoots an image and transmits the photographed image to a management server. The road where CCTV is installed includes all roads that vehicles can pass, such as high-speed roads, highways, national roads, and local roads. This can fall. In this case, if the weather condition is analyzed based on the received image information, it is difficult to provide accurate weather information, and furthermore, it becomes impossible to provide additional weather service through data processing.

On the other hand, in addition to the fixed CCTV, it is also possible to consider providing weather information using an image capturing device included in a mobile device or a device that collects image information by installing it in a moving vehicle. It is important to extract data suitable for analyzing weather information through

The present invention was created under the above technical background, and an object of the present invention is to provide a new system for receiving image information and examining and analyzing data suitable for extracting weather information.

Another object of the present invention is to provide a road weather information system capable of improving the quality of image data and extracting accurate weather information by examining various types of image information received not only from fixed image equipment but also from mobile image equipment in real time.

In addition, other objects and technical features of the present invention will be presented in more detail in the following detailed description.

In order to achieve the above object, the present invention provides a data receiving unit for receiving external image information, a data processing unit for checking whether the received image information is error or normal step by step, and a data storage for storing corrected data after image information inspection The data receiving unit receives image data observed in real time from an individual observation device that takes an image of the road, and the data processing unit excludes an error value through a missing test on the received image data in a first step. and the normal value is inspected in a subsequent stage, and the data processing unit performs physical limit inspection and a second stage, respectively, on image data for fog, rain, snow, and road surfaces among the data filtered through the first stage. Step test in step 3, continuity check in step 4, and bouncing value check in step 5 are sequentially performed to exclude data corresponding to errors in each check, and normal data is collected as final image data related to road weather. Provided is a road weather information system using image data processing stored in the data storage unit.

In the present invention, the data processing unit sets a minimum physical limit value and a maximum physical limit value for image observation data of fog, rain, snow, and a road surface in the physical limit inspection, respectively, and the observed data is a minimum physical limit value If it is out of the range of and maximum physical limit value, it can be judged as an error, and if it exists within the range, it can be judged as normal.

In addition, in the step inspection, the maximum variation of each data is calculated as a reference value through statistical analysis of data for a certain period of time for image observation data of fog, rain, snow, and road surfaces, and If the difference is out of the range of the maximum fluctuation amount, it can be judged as an error, and if it is within the range, it can be judged as normal.

In addition, in the persistence test, the minimum amount of variation of each data is calculated as a reference value through statistical analysis of the data for a certain period of time for image observation data of fog, rain, snow, and road surfaces, and the cumulative amount of variation during a specific period is the minimum amount of variation. By checking whether it continues within the range, it can be judged as an error if the variation is less than the minimum variation, and normal if it is equal to or greater than the minimum variation.

In addition, the bouncing value test calculates the d value of Equation (1) from the difference of three consecutive observation data, and if the d value satisfies the condition of Equation (2), it is normal, and other values are considered bouncing values. It can be classified as an error.

: 식(1)

: Equation (1)

: 식 (2)

: Equation (2)

where, where MAD is the median absolute deviation of d and Md,

M _d is the median of the deviation, z is 5.5 or 7

In the present invention, with respect to an image observation device showing an outlier in the physical limit test, step test, persistence test, or bouncing value test process, data comparison test between a plurality of observation devices within a certain range, range test, and position test After performing the stage quality inspection, additional spatial inspection can be performed on the data that appears as an error after the stage 3 quality inspection.

In the present invention, the comparison test checks whether the data is the same through simple comparison, and the range test is based on whether the observed data deviates from the standard deviation value using the standard deviation calculated through statistical analysis of the collected data. Determining, the location check checks whether the observation device indicating the error indicates an error in the position of other observation equipment representing the normal value, and the spatial check checks the observation value of the suspicious point within a certain range of data from other observation devices within a certain range. A nonparametric test is performed using the median of , and the upper and lower quantiles.

In addition, in the present invention, it is preferable to transmit only one data when the data received from the observation devices at different locations are the same after the data comparison and inspection between a plurality of observation devices within a certain range.

According to the present invention, it is possible to provide a road weather information system capable of extracting accurate weather information by receiving image information from an existing image observation device installed on a road.

The road meteorological information system of the present invention provides weather information based on normal image information by excluding error values of the collected image data using existing image observation equipment without additionally installing a meteorological observation device. quality can be improved.

In particular, the present invention is suitable for using not only stationary image observation equipment but also mobile image observation equipment, and it is easy to improve the equipment through data verification of equipment with inaccurate image information among a plurality of image observation equipment, Data quality inspection to provide accurate weather information across the country can be processed in real time.

The road weather information system of the present invention may be useful not only for vehicle drivers, but also for road management, road construction, and other specialized institutions that determine road policies.

1 is an overview of a flow chart of processing image information received from individual image observation equipment;
2 is a conceptual diagram of a physical limit test;
3A and 3B are flowcharts showing the physical limit inspection process for fog, respectively, and the fog image observation range step;
4A to 4C are flowcharts showing the physical limit inspection process for ratio, particle cumulative images, and cumulative change amount, respectively.
5 is a flowchart showing the physical limit test process for the eye
6 is a flowchart showing a physical limit inspection process for a road surface;
7 is a conceptual diagram of a step inspection;
8A to 8E are results of statistical analysis of image observation data.
9A to 9D are step inspection flowcharts for fog, rain, snow, and road surface image information.
10 is a conceptual diagram of a persistence check;
11A to 11D are flow charts for checking the persistence of image information of fog, rain, snow, and road surfaces;
12 is a conceptual diagram of a persistence check.
13 is an overall flowchart of image information processing received from individual image observation equipment;
14 is a schematic diagram showing an observation device within the range of influence;
15 is a flowchart showing the step of performing secondary analysis on the observed image information within the influence range;
16 is a flowchart of a comparative test;
17 is a conceptual diagram of a range check.
18 is a conceptual diagram of a position inspection
19 is a flow chart of range check and position check
20 is a flow chart of a spatial inspection;

The present invention proposes a system capable of collecting and analyzing road weather information using camera image processing technology, and provides reliable road weather information by excluding error information, while providing additional processing information services such as road risk information. We propose a system that performs quality inspection of road image information.

The road weather information system of the present invention performs image information analysis or inspection in two main aspects. First, sequential quality inspection (QC) is performed on image information received from individual image observation devices through several steps (primary analysis). In addition, the equipment for observing abnormal data is checked by comparing and analyzing image information collected from a plurality of image observation devices (secondary analysis).

In the present invention, the image information subject to quality inspection can be received from various image observation equipment such as a CCTV camera fixedly installed on the road, as well as a mobile camera or an unmanned aerial vehicle equipped with an image observation unit. When video frames received through CCTV, etc. are analyzed, errors in video information occur due to circumstances such as a bird flying into a stationary camera or a mobile camera passing through a tunnel. Inspections such as comparison of front and back values, adjustment of limit values, and filtering should be performed. To this end, the road weather information system of the present invention basically includes a data receiving unit that receives external image normal information, a data processing unit that checks whether the received image information is error or normal step by step, and the corrected data after image information inspection. It includes a data storage unit to store, and may further include a data processing unit for additional processing on data for which quality inspection has been completed, a data transmission unit for transmitting weather information to other systems through a network, and the like.

As the primary analysis, the analysis of image information received from individual image observation equipment is sequentially performed over a total of five steps to improve accuracy by analyzing and removing error values.

I. Primary Analysis - Inspection of image data of individual devices

1 shows an overall flow chart for processing image information received from individual image observation equipment. The image information performs quality inspection in a total of five steps, and through the image information, fog, rain, snow, and road surface (condition) are analyzed as elements related to road weather information, and quality inspection is performed for each image information. . Detailed procedures and methods for each step and individual element are specifically described below.

a) 1st quality inspection - 1st stage missing test

The first is a missing check, which is performed for data integrity. It is determined when observation data is not collected due to camera or communication failure, and when the imaging device is missing, the reason for the missing can be transmitted. , and the missing data is excluded. In the missing inspection, image data without any abnormality is detected or classified as images related to fog, rain, snow, and road surface (state), respectively, and subsequent quality inspection is performed on them.

b) 1st quality inspection - 2nd stage physical limit inspection

The second step is a physical range check, which determines whether or not it is outside the observation range of the observation equipment. It can also be set as a reference value. Specifically, in the physical limit test, it is checked whether there is an error in the maximum and minimum value ranges allowed by the sensor, and if it is within the physical limit range as shown below (see FIG. 2 ), it is normal and the other ranges are determined as errors. In the present invention, unlike general meteorological observation equipment such as a thermometer, a hygrometer, and an anemometer, since the observation is performed using an image, the measurement range is individually determined in consideration of the image screen.

Minimum physical limit (min) ≤ Image data ≤ Maximum physical limit (max)

FIG. 3A is a flowchart showing the physical limit inspection process for fog. For example, the fog image observation range (ROI) is divided into 10 steps (refer to FIG. 3b) and the observed value F of the fog-related image is within the observation range. If there is, it is judged as normal; otherwise, it is judged as an error.

ROI 1 ≤ Fog observations (F) ≤ ROI 10

4A is a flowchart showing the physical limit test process for rain, dividing the observation range of the rain image into the accumulated amount of precipitation (refer to the accumulated particle image of FIG. 4B and the accumulated change of FIG. 4C), and the minimum limit of the range of accumulated change of rain The value is 0.1 and the maximum value is 1, and if the non-observed value (R) is within the following range, it is judged as normal, otherwise it is judged as an error.

0.1 ≤ non-observed (R) ≤ 1

5 is a flowchart showing the physical limit test process for snow. The observation range of the snow image is divided by the accumulated amount of snow similarly to the rain image, so that the weather stage can be determined and used for the physical limit test.

By setting the maximum value of the range of accumulated change in snow to 1, the weather stage according to the accumulated change can be set as follows (Table 1). case is considered as an error.

0.1 < Observed snow cover (S) < 1

6 is a flowchart showing a physical limit inspection process for a road surface. The observation range of a road surface image can be determined based on, for example, the accumulated amount of precipitation (see Table 2), and the accumulated amount of precipitation is converted into a ratio for the unit area and accumulated The amount of change is derived, and if the observed value (RS) is in the range between the minimum and maximum limits of the cumulative change, it is judged as normal, otherwise it is judged as an error. Here, the minimum and maximum thresholds may be arbitrarily determined in consideration of weather conditions and road conditions.

0.06% (minimum limit value) < road surface observation (RS) < 50% (maximum limit value)

Data determined as an error in the physical limit inspection are excluded, and for normal image data, the next stage of quality inspection is performed for fog, rain, snow, and road surfaces, respectively.

c) 1st Quality Inspection - Step 3 Step Inspection

The third step is a step check, which calculates the limit value of the variation in unit time, and determines whether the observed data is normal by checking whether there is an abnormal variation in continuous time data (see FIG. 7 ). Although the WMO and the Korea Meteorological Administration recommend and suggest maximum fluctuation values for temperature, humidity, wind speed, etc., the present invention calculates the maximum fluctuation value through statistical analysis of data for a certain period of time. Here, the maximum variation is a reference value required for the step inspection, and the maximum variation is calculated by calculating the variation between the data value and the previous value to be inspected as follows.

After checking the amount of time variation for time series data, if the variation between the immediately previous value and the current value is outside the range of the maximum variation, it is judged as an error.

Difference between data values (│current data value - previous data value│) ≤ maximum variation

8a to 8e show the statistical analysis results of 3-month image observation data of Jangpyeong, Woljeong-ri, Hoenggye-ri, Seongsan 2-gyo, and Daegwallyeong. It is determined as a reference value, and the minimum variation for fog, rain, and snow records 0 for all 5 observation images, and 0 is determined as the reference value. 0.7 and 0.7 were recorded and these values were determined as the maximum variation. However, in the case of the road surface, there was no observation data, so an arbitrary value was set. The minimum fluctuation duration was 30 days, the minimum fluctuation amount was 0, the maximum fluctuation amount was 20%, and the physical limit value was 50%.

9A shows a step check for fog image data. If _{the difference between the current observation value (F t} ) and the previous observation value (F _t-1 ) is smaller than the maximum variation reference value (ROI 7), it is determined as normal, and the If it is any other value, it is judged as an error. 9B shows a step check for non-image data. If _{the difference between the current observation value (R t} ) and the previous observation value (R _t-1 ) is smaller than the maximum variation reference value (0.7), it is determined as normal, and otherwise If the value is , it is judged as an error. 9C shows a step test for eye image data. If _{the difference between the current observation value (S t} ) and the previous observation value (S _t-1 ) is smaller than the maximum variation reference value (0.7), it is determined as normal, and otherwise If the value is , it is judged as an error. Figure 9d shows a step inspection for road image data. If _{the difference between the current observation value (RS t} ) and the previous observation value (RS _t-1 ) is less than the maximum variation reference value (20%), it is determined as normal, and the If it is any other value, it is judged as an error.

Data determined as an error in the step inspection are excluded, and for normal image data, the next stage of quality inspection is performed for fog, rain, snow, and road surfaces, respectively.

d) 1st quality inspection - 4th stage continuity inspection

In the fourth step, the persistence check, if the observed value does not change at all due to a failure in the observation equipment, it is determined whether there is a change in the value for a certain period of time using the unit time variation. If a value that does not change at all for a long time is measured in time series observation data, it is determined as an error value (see FIG. 10).

Specifically, it is checked whether the accumulated fluctuation amount is maintained within the minimum fluctuation amount for a specific period, and if the fluctuation amount is less than the minimum fluctuation amount, it is determined as an error, and if it is equal to or greater than the minimum fluctuation amount, it is determined as normal.

Cumulative amount of change over a specific period ≥ Minimum amount of change (O _min )

The minimum amount of variance, which is the reference value required for the persistence test, calculates the minimum value of 0 or more by calculating the amount of variance between the observed value and the previous value to be tested.

According to the statistical analysis of data for the past three months presented in relation to the step test, the minimum amount of change in fog, rain, and snow was recorded as 0 in all 5 observation images, and 0 was determined as the minimum amount of change in this example, and in the case of the road surface, any 0 was determined as the value.

11A shows a continuity test for fog image data, _{and it is determined as an error that the accumulated variation F a} is smaller than the minimum variation, and a value other than that is determined as normal. 11B shows the persistence test for non-image data, and it is determined that the accumulated variation (R _a ) is smaller than the minimum variation as an error, and other values are determined as normal. 11C shows the persistence test for non-image data, and it is determined that the accumulated variation S _a is smaller than the minimum variation and thus it is determined as an error, and other values are determined as normal. 11D shows a continuity test for road image data, and it is determined that the accumulated variation (RS _a ) is smaller than the minimum variation as an error, and other values are determined as normal.

Data determined as an error in the continuity test are excluded, and for normal image data, the next stage of quality inspection is performed for fog, rain, snow, and road surfaces, respectively.

e) 1st Quality Inspection - 5th Step Inspection of Bouncing Values

The next step, the bouncing value test, is a test to determine the observation error value that occurs due to various reasons. The result of Eq. The d value is calculated, and if the d value satisfies the condition of Equation (2), it is normal, and any other value is classified as a bouncing value and determined as an error (refer to FIG. 12).

: 식(1)

: Equation (1)

: 식 (2)

: Equation (2)

), where MAD is the median absolute deviation of d and Md (

),

M _d is the median of the differences, z is 5.5 or 7

In each of the above inspection steps, data corresponding to errors is excluded, and normal data is collected and stored as final image data related to road weather, which can be used for road weather information services, etc. 13 shows the entire process of processing image data related to fog, rain, snow, and road surfaces observed in real time from individual observation devices that capture images of roads by the above-described primary inspection method.

II. Secondary analysis - comparative analysis of image information of multiple image observation devices

In extracting meteorological information from road image data, in order to provide high-quality weather information services over a wide area, it is necessary to have image observation devices installed in different areas or a plurality of mobile observation devices that acquire image information while moving between different areas. It is very important to receive accurate data through data transmission/reception with these devices.

In particular, among the plurality of image observation apparatuses, if data quality is deteriorated or an abnormality occurs in the operation of the apparatus, the management of the apparatus, such as repair or replacement, is required, and it is necessary to analyze and determine such a situation in advance and deal with it. For example, as shown in FIG. 14 , when a plurality of (three or more) fixed or mobile image observation apparatuses exist in a certain area, data of observation equipment showing different values among them is used as image information between observation apparatuses. Thus, it can be judged whether it is normal or not. In the case of fog, rain, and snow, spatial heterogeneity due to topographical factors is very large, and in order to determine whether there is an error in image information for observation devices at multiple points within a certain area, the range of influence can be determined by considering topographic factors. In the present invention, an error value is detected by comparing the image information of the observation devices within this range by setting the influence range of a fixed observation device or a mobile observation device to 1 km to 5 km, and performing a spatial inspection.

The image information comparison and analysis process (secondary analysis) of the image observation apparatus performed in the road weather information system of the present invention can be performed on the image observation apparatus showing an abnormal value in the inspection process of image information through the above-described primary analysis, , perform three-step quality inspection: data comparison between observation devices, range inspection, and position inspection. After the 3rd step inspection, an additional spatial inspection is performed on the result expressed as an error value. In this case, for the secondary analysis, a separate data processing unit (second data processing unit) for comparing and examining image information received from a plurality of image observation apparatuses may be included.

In addition, this secondary analysis may be utilized to extract normal data excluding error values by performing a mutual comparison check step without performing the above-described primary analysis for a plurality of observation devices in a predetermined space.

FIG. 15 schematically shows a process of performing secondary analysis on image data observed at three locations within an impact range of 1 km. If the same information is confirmed by performing a comparison check in step 1, only one data is transmitted to shorten data processing, which will be described later. If an outlier is found through comparative inspection, steps 2 (range inspection) and 3 (location inspection) are performed. If an error is confirmed in either step 2 or 3, the data corresponding to the error is excluded. exclude

a) 2nd quality inspection - 1st stage comparative inspection

As a step of detecting outliers through simple data comparison between observation devices within a certain range, the next stage quality check is performed on the data of the observation devices showing the outliers through data comparison inspection, and subsequent inspection is performed only for outliers among three or more data. proceed, and if all data are not identical to each other, proceed to the next step for all data (see FIG. 16 ). When all the observed image information is confirmed to be the same data, only one data can be transmitted to the central processing unit, etc. to end the inspection and reduce the data processing speed.

b) 2nd Quality Inspection - 2nd Stage Scope Inspection

The range check detects a case where the data of the observation device within a certain range shows a large difference compared to the existing data, and uses the standard deviation (σ) calculated through statistical analysis of the existing collected data, and the observed data is ± It is determined whether or not 3σ is out of range (see FIG. 17).

The criterion for the range inspection (Ra) is normal when the observed image information for fog (F), rain (R), and snow (S) is within 3σ, otherwise it is judged as an error.

c) 2nd quality inspection - 3rd stage location inspection

The position test is to check whether an observation device showing an error in the data comparison test shows an error in the position of another observation device showing a normal value. The location inspection can be performed simultaneously or sequentially with the aforementioned range inspection, and spatial heterogeneity of weather changes caused by topographical factors is resolved. Referring to FIG. 18 , when the observation device indicating an error in the data comparison test is C, when the position (GPS) of C reaches the position (GPS) of the normal observation device A or B, a re-observation command is issued and A or B Determine whether the observed data at the corresponding location is normal by checking whether the observed result indicated by is displayed.

The criterion for determining the position inspection (Ro) is that when the device X shows an error value, when X reaches the Y or Z position, the (fog (Fx), rain (Rx), snow (Sx)) image information of X is If it is the same as the observed value of Y (fog (Fy), rain (Ry), snow (Sy)) or the observed value of Z (fog (Fz), rain (Rz), snow (Sz)), it is judged as normal, and the Otherwise, it is judged as an error.

19 shows a determination method and an inspection sequence for the range inspection and the position inspection. If both inspection results are determined to be normal, the inspection is terminated, and if any one result is determined to be an error, a subsequent inspection is performed.

d) Secondary Quality Inspection - Stage 4 Spatial Inspection

Spatial inspection detects erroneous observations by comparing observed values at a suspected point with other observation values within a certain range. In the present invention, the Madsen-Allerup method is used. A nonparametric test is performed using the values and the upper and lower quantiles.

Specifically, as a spatiality test using the median (median value), it is determined whether the test is normal or not by obtaining an intermediate test statistical value according to the following equation.

here,

T _i,t is the intermediate test statistic,

x _i,t : the observed value at the observation device i at time t

M _t : the median value obtained from N observations at time t

q _t,25 : 25% (quarter quartiles) of observations at N points

Q _t,75 : 75% of N point observations (3/4 quartiles)

Range check (Ra) and location check (Po) If only one of the two result values is an error, Madsen-Allerup spatial check is performed to determine the final error (see Fig. 20), and range check (Ra) and location check ( Po) If both result values are errors, it is immediately judged as an error and the corresponding data is excluded.

The criterion for spatial inspection is that if T _i,t > 2 or q _t,75 -q _t,25 = 0, it is determined as an error and the corresponding data is excluded.

Speed up data processing

When a number of mobile observation devices exist within a certain range of the same area, processing all the data of these devices takes a long time and results in inefficient observation. Meanwhile, in order to improve the data processing speed, it is most effective to reduce the amount of processed data.

The road weather information system of the present invention transmits information from only one image observation device when data from a plurality of observation devices within a certain range of the same area indicates the same image information values such as fog, rain, snow, etc. through comparative inspection. It can reduce the data processing speed and perform efficient observation. Such data transmission is performed not only when the above-mentioned primary and secondary analysis are performed together in a single road weather information system, but in particular, individual data quality management (primary analysis) for each image information collection device is performed in the road weather information system. and data quality management (secondary analysis) through comparison analysis of data collected from a plurality of image observation devices is more effective when it is performed in a remote server or central processing system.

In the road weather information system of the present invention, a method of performing various inspection steps for road image data can be operated by being embedded in a PCB circuit or a control process chip, and the overall system hardware form is a personal terminal such as a personal computer or a weather It may be constructed as a server-type system installed in an information providing institution, or may be implemented as a portable small device so that image data can be analyzed in real time while being mounted on a moving vehicle.

In addition, after establishing unit systems for analyzing image information for a specific area, each unit system may be integrated into a wide area unit to provide comprehensive weather information.

Although the present invention has been exemplarily described through preferred embodiments above, the present invention is not limited to such specific embodiments, and various forms within the scope of the technical idea presented in the present invention, specifically, the claims may be modified, changed, or improved.

Claims (5)

A data receiving unit for receiving external image information, a data processing unit for checking whether the received image information is error or normal step by step, and a data storage unit for storing corrected data after image information inspection,
The data receiving unit receives the image data observed in real time from individual observation devices that take images of the road,
The data processing unit performs a follow-up test for normal values except for error values through a missing test as a first step for the received image data,
The data processing unit includes a physical limit check in a second step, a step check in a third step, a continuity check in a fourth step, and image data on fog, rain, snow, and road surfaces among the data filtered through the first step, respectively. In the fifth step, the bouncing value test is sequentially performed to exclude data corresponding to errors in each test, and normal data is collected and stored in the data storage unit as final image data related to road weather,
The data processing unit
In the physical limit test, a minimum physical limit value and a maximum physical limit value for image observation data of fog, rain, snow, and a road surface are respectively set, and the observed data is out of the range of the minimum physical limit value and the maximum physical limit value If it is an error, if it exists within the range, it is judged as normal,
In the step inspection, the maximum variation of each data is calculated as a reference value through statistical analysis of the data for a certain period of time for image observation data of fog, rain, snow, and road surfaces, and the difference between the current data and previous data to be tested is If it is out of the range of the maximum fluctuation amount, it is judged as an error, if it is within the range, it is judged as normal.
In the persistence test, the minimum variation of each data is calculated as a reference value through statistical analysis of the data for a certain period of time for image observation data of fog, rain, snow, and road surfaces, and the cumulative variation is maintained within the minimum variation for a specific period. If the variation is less than the minimum variation, it is judged as an error, and if it is equal to or larger than the
The bouncing value test calculates the d value of equation (1) from the difference of three consecutive observation data, and if the d value satisfies the condition of equation (2), it is normal, and other values are classified as bouncing values. characterized in that it is judged as an error

: Equation (1)

: Equation (2)
where MAD is the median absolute deviation of d and Md,
M _d is the median of the deviation, z is 5.5 or 7
Road weather information system using image data processing.
According to claim 1,
The image information is a road weather information system using image data processing, characterized in that it is received from a CCTV camera fixedly installed on the road, a mobile camera running on the road, or an unmanned aerial vehicle equipped with an image observation unit.
According to claim 1,
For the image observation device showing an outlier in the physical limit inspection, step inspection, continuity inspection, or splash value inspection process, three-stage quality inspection is performed: data comparison inspection, range inspection, and position inspection between a plurality of observation devices within a certain range and, after the three-step quality inspection, additionally a spatial inspection is performed on data that appears as an error.
4. The method of claim 3,
The comparison check confirms whether the data are the same through a simple comparison,
The range check determines whether the observed data deviates from the standard deviation value using the standard deviation calculated through statistical analysis of the collected data,
The position check checks whether the observation device indicating an error indicates an error in the position of another observation device indicating a normal value,
The spatial test detects erroneous observations by comparing the observation values of the suspected point with other observation values within a certain range, and performs a non-parametric test using the median and upper and lower quantiles of the observation device data within a certain area. Road weather information system using image data processing, characterized in that.
4. The method of claim 3,
A road weather information system using image data processing, characterized in that after data comparison and inspection between a plurality of observation devices within a certain range, only one data is transmitted when the data received from the observation devices at different locations are the same.

Images