JP4774807B2 - Image correction device, feature recognition device, and feature recognition method - Google Patents

Description

  The present invention relates to an image correction device, a feature recognition device, and a feature recognition method.

  Conventionally, in a navigation device, for example, the current position of a vehicle, that is, the current position is detected by GPS (Global Positioning System), and the direction of the vehicle, that is, based on the turning angle of the vehicle detected by a gyro sensor, that is, The vehicle direction is detected, map data is read from the data recording unit, a map screen is formed on the display unit, the vehicle position indicating the current location, a map around the vehicle position, and the vehicle The direction is displayed. Therefore, the driver who is an operator can drive the vehicle according to the vehicle position displayed on the map screen, the map around the vehicle position, and the vehicle direction.

  When the driver inputs a destination and sets search conditions, a route search process is performed based on the search conditions, and a route from the starting point represented by the current location to the destination is searched according to the map data. Is done. Then, the searched route, that is, the searched route is displayed on the map screen together with the vehicle position, and guidance for the searched route, that is, route guidance is performed. Therefore, the driver can drive the vehicle along the displayed search route.

  By the way, for example, a stop line as a feature may be formed on the road before the intersection by painting or the like. In that case, the stop line is photographed by a camera mounted on the vehicle. The stop line can be recognized based on the image of the stop line (see, for example, Patent Document 1).

For this purpose, the navigation device reads road data from the mounted data recording unit to acquire the width of the road, the width of the stop line, etc., and recognizes the image in the horizontal direction and the front-rear width direction of the stop line with respect to the image. I do.
JP 2004-295597 A

  However, in the conventional navigation device, the accuracy of recognizing the stop line may be lowered. For example, the stop line may be rubbed (rubbed) when the vehicle repeatedly travels on the stop line, and a part of the stop line may be lost. In that case, even if image recognition in the horizontal width direction and the front-rear width direction of the stop line is performed based on the width of the road, the width of the stop line, and the like, the accuracy of the image recognition is reduced by the amount of missing. Therefore, the stop line recognition accuracy is lowered.

  As a result, the accuracy of detecting the intersection and calculating the distance to the intersection is lowered.

  An object of the present invention is to provide an image correction device, a feature recognition device, and a feature recognition method capable of solving the problems of the conventional navigation device and increasing the recognition accuracy of the feature.

  For this purpose, in the image correction device of the present invention, a photographing device for photographing a feature and a surrounding of the target feature when the vehicle travels on the target feature for which the photographed image is to be corrected. A traveling locus estimation processing means for estimating a steady traveling locus based on road shape information of a road, and a corrected portion calculation processing means for calculating a corrected portion of a photographed image on the target feature based on the estimated traveling locus And a rubbing estimation parameter calculation processing means for calculating a rubbing estimation parameter representing the degree of image loss at the calculated corrected portion based on the vehicle running index on the target feature, and based on the calculated rubbing estimation parameter Interpolating processing means for correcting a defect in the image of the corrected portion.

  According to the present invention, a steady traveling locus when the vehicle travels on the target feature is estimated, and based on the traveling locus, the corrected portion of the image of the target feature is calculated, and the image of the corrected portion of the image is corrected. Since the defect is corrected, the feature recognition accuracy can be increased.

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

  FIG. 1 is a diagram showing a navigation system according to an embodiment of the present invention.

  In the figure, reference numeral 10 denotes an automatic transmission control unit as a power train control unit, and 11 denotes a power train that is used together with a driving power source such as an engine and an electric motor (not shown) and performs shifting at a predetermined gear ratio. In the present embodiment, a belt-type, toroidal-type continuously variable transmission (CVT) such as an automatic transmission is used as the power train 11, but a stepped transmission (automatic transmission) as an automatic transmission is used. Can be used. Further, as the power train 11, an electric drive device that is used in a hybrid vehicle or an electric vehicle including an engine and an electric motor and performs a continuously variable transmission by driving the electric motor can be used.

  Reference numeral 14 is an information terminal, for example, a navigation device as an in-vehicle device mounted on a vehicle, 63 is a network, 51 is an information center as an information provider, and the automatic transmission control unit 10, the navigation device 14, the network 63, the information center 51 and the like constitute a navigation system.

  The navigation device 14 includes a GPS sensor 15 as a current location detection unit for detecting a current location, a data recording unit 16 as an information recording unit in which various information is recorded in addition to map data, and navigation based on the input information. A navigation processing unit 17 that performs various arithmetic processing such as processing, a direction sensor 18 that serves as a direction detection unit that detects the vehicle direction, and a first input for performing predetermined input when operated by a driver as an operator. An operation unit 34 as an input unit, various displays by an image displayed on a screen (not shown), a display unit 35 as a first output unit for notifying the driver, and a predetermined input by voice A voice input unit 36 as a second input unit, a voice output unit 3 as a second output unit for performing various displays by voice and notifying the driver , And a communication unit 38 as a transmission / reception unit functioning as a communication terminal. The navigation processing unit 17 includes a GPS sensor 15, a data recording unit 16, an orientation sensor 18, an operation unit 34, a display unit 35, a voice input unit 36, An audio output unit 37 and a communication unit 38 are connected.

  The navigation processing unit 17 includes the automatic transmission control unit 10, a front monitoring device 48 that monitors the front of the vehicle, and an imaging device that captures the front, side, or rear of the vehicle, that is, the periphery of the vehicle. A camera 49, an accelerator sensor 42 as a load detection unit that detects the operation of the accelerator pedal by the driver by the accelerator opening, a brake sensor 43 as a brake detection unit that detects the operation of the brake pedal by the driver by the brake depression amount, vehicle speed A vehicle speed sensor 44 or the like is detected. The accelerator sensor 42, the brake sensor 43 and the like constitute a vehicle information detection unit for detecting vehicle operation information by the driver.

  The GPS sensor 15 detects the current location on the earth by receiving radio waves generated by an artificial satellite, and also detects the time. In the present embodiment, the GPS sensor 15 is used as the current position detection unit, but a distance sensor, a steering sensor, an altimeter, etc. (not shown) are used alone or in combination instead of the GPS sensor 15. You can also Further, a gyro sensor, a geomagnetic sensor, or the like can be used as the direction sensor 18.

  The data recording unit 16 includes a map database including map data files, and map data is recorded in the map database. The map data includes intersection data relating to intersections, node data relating to nodes, road data relating to road links and lanes, search data processed for searching, facility data relating to facilities, and other feature data relating to features. included. The features include a stop line formed on the road, a pedestrian crossing, a traffic band display line that divides the lane (lane), a traffic division that indicates the traveling direction in the lane with an arrow, a road sign, a manhole on the road, a traffic light The feature data includes position information representing the position of each feature, image information representing an image of each feature, and the like. The stop line is formed before an intersection, a crossing, or the like. The road data related to the lane includes lane data including a lane number assigned to each lane on the road, lane position information, and the like.

  Further, the data recording unit 16 includes a statistical database composed of statistical data files, a travel history database composed of travel history data files, and the like. Statistical data is stored in the statistical data file, and travel history data is stored in the travel history data file. However, both are recorded as performance data.

  The statistical data is a history of traffic information provided in the past, that is, history information representing a history, and a road traffic information center (not shown) such as a VICS (registered trademark: Vehicle Information and Communication System) center as an information provider. Traffic information provided in the past, etc., data indicating traffic volume by road traffic census provided by the Ministry of Land, Infrastructure, Transport and Tourism (hereinafter referred to as “road traffic census information”), road timetable information provided by the Ministry of Land, Infrastructure, Transport and Tourism, etc. Are used alone or in combination, and if necessary, they are processed and subjected to statistical processing. It is possible to add traffic jam prediction information for predicting traffic jams to the statistical data. In that case, in creating the statistical data, detailed conditions such as date and time, day of the week, weather, various events, seasons, facility information (presence / absence of large facilities such as department stores and supermarkets) are added to the history information.

  The data items of the statistical data include the road on which the vehicle has traveled in the past, that is, the link number for each road link constituting each traveling road, the direction flag representing the traveling direction, the information type representing the type of information, a predetermined type The degree of congestion at each timing, the link required time indicating the required time for each predetermined timing when traveling on each road link, average data for each day of the link required time (for example, average data for day of week), etc. Consists of.

  The travel history data is the track record of the travel of the vehicle on the travel road collected by the information center 51 from a plurality of vehicles (own vehicle or other vehicle), that is, the track record information indicating the track record. Calculated and accumulated as data.

  The data items of the travel history data include the degree of congestion at each predetermined timing, the required link time at each predetermined timing when traveling on each road link, and the average at each predetermined timing when traveling on each road link. Consists of vehicle speed and traffic volume for each type of vehicle. Note that travel history data can be added to the statistical data. Further, in the present embodiment, the congestion level is used as a congestion index that represents the degree of congestion, and is represented separately for congestion, congestion, and non-congestion.

  Further, data for outputting predetermined information by the audio output unit 37 is also recorded in the data recording unit 16.

  The data recording unit 16 includes a disk (not shown) such as a hard disk, a CD, a DVD, and an optical disk in order to record the various data, and also reads / writes for reading and writing various data. A head (not shown) such as a head is provided. A memory card or the like can be used for the data recording unit 16.

  In the present embodiment, the data recording unit 16 is provided with the map database, statistical database, travel history database, and the like. However, in the information center 51, the map database, statistical database, travel history, etc. A history database or the like can also be provided.

  The navigation processing unit 17 is a control device that performs overall control of the navigation device 14 and a CPU 31 as an arithmetic device, a RAM 32 that is used as a working memory when the CPU 31 performs various arithmetic processes, In addition to the above program, a ROM 33 in which various programs for searching for a route to the destination, route guidance and the like are recorded, and a flash memory (not shown) used for recording various data, programs and the like are provided. . In the present embodiment, the CPU 31 constitutes an image correction device and a feature recognition device.

  In the present embodiment, various programs can be recorded in the ROM 33 and various data can be recorded in the data recording unit 16, but the programs, data, and the like can also be recorded on a disk or the like. In this case, the program, data, etc. can be read from a disk or the like and written to the flash memory. Therefore, the program, data, etc. can be updated by exchanging the disk or the like. Further, the control program, data, and the like of the automatic transmission control unit 10 can be recorded on the disk or the like. Further, the program, data, and the like can be received via the communication unit 38 and written into the flash memory of the navigation processing unit 17.

  The operation unit 34 is operated by the driver to correct the current location at the start of traveling, to input a departure point and a destination, to input a passing point, and to activate the communication unit 38. As the operation unit 34, a keyboard, a mouse, or the like disposed independently of the display unit 35 can be used. Further, as the operation unit 34, a predetermined input operation is performed by touching or clicking an image operation unit such as various keys, switches, and buttons displayed as images on the screen formed in the display unit 35. It is possible to use a touch panel that can be used.

  A display is used as the display unit 35. Then, on various screens formed on the display unit 35, the vehicle position indicating the current location, a map, a searched route, guidance information along the searched route, traffic information, etc. are displayed, or until the next intersection in the searched route. In addition to displaying the distance and the direction of travel at the next intersection, the operation guidance, operation menu, and key guidance of the image operation unit, operation unit 34, voice input unit 36, etc. can be displayed. A multiplex broadcast program or the like can be displayed.

  The voice input unit 36 includes a microphone (not shown) and the like, and can input necessary information by voice. Further, the voice output unit 37 includes a voice synthesizer and a speaker (not shown), and the search route, guidance information, traffic information, and the like are output from the voice output unit 37, for example, as voice synthesized by the voice synthesizer. .

  The communication unit 38 transmits various information such as current traffic information and general information transmitted from the road traffic information center to radio wave beacons via radio wave beacon devices, optical beacon devices and the like arranged along the road. A beacon receiver for receiving as an optical beacon, an FM receiver for receiving as an FM multiplex broadcast via an FM broadcast station, and the like. The traffic information includes traffic jam information, regulation information, parking lot information, traffic accident information, service area congestion status information, and the like, and general information includes news, weather forecasts, and the like. Further, the beacon receiver and the FM receiver are unitized and arranged as a VICS receiver, but can be arranged separately.

  The traffic information includes an information type that indicates the type of information, a mesh number for specifying a mesh, a road link that connects two points (for example, intersections), and a link that indicates whether it is up / down. Number, including link information representing the content of information provided corresponding to the link number. For example, when the traffic information is traffic jam information, the link information is a distance from the start point of the road link to the head of the traffic jam Congestion head data representing traffic congestion degree, congestion section, traffic jam length representing the distance from the head of the traffic jam to the end of the traffic jam, link required time, and the like.

  The communication unit 38 can receive various information such as traffic information and general information in addition to the map data, statistical data, and travel history data from the information center 51.

  For this purpose, the information center 51 includes a server 53, a communication unit 57 connected to the server 53, a database (DB) 58 as an information recording unit, and the like. A CPU 54, a RAM 55, a ROM 56, etc. are provided as devices. The database 58 records data similar to the various data recorded in the data recording unit 16, for example, the map data, statistical data, travel history data, and the like. Further, the information center 51 provides various information such as current traffic information and general information transmitted from the road traffic information center, and travel history data collected from a plurality of vehicles (own vehicle or other vehicles) in real time. can do.

  The forward monitoring device 48 includes a radar such as a laser radar or a millimeter wave radar, an ultrasonic sensor, or a combination thereof. As the vehicle surrounding information, an inter-vehicle distance, an inter-vehicle time, an approach speed with respect to a preceding vehicle, a temporary stop point ( The approach speed to the priority road from the non-priority road, the crossing, the intersection where the red light blinks, etc.), the approach speed to the obstacle, etc. are calculated, and the preceding vehicle, the temporary stop, the obstacle, etc. are monitored. The camera 49 is composed of a CCD, a C-MOS, and the like, and as a subject vehicle surrounding information, in addition to the above-mentioned various features, the surrounding vehicles, buildings, intersections, information signs, etc. are photographed as objects to be photographed. The image data of the captured object image is sent to the CPU 31.

  When the CPU 31 receives the image data, the CPU 31 processes the image data to recognize an object to be photographed and monitors the periphery of the vehicle.

  The navigation system, the navigation processing unit 17, the CPUs 31, 54, the server 53, and the like function as a computer based on various programs, data, and the like. The data recording unit 16, RAMs 32 and 55, ROMs 33 and 56, database 58, flash memory, and the like constitute a recording medium. An MPU or the like can be used instead of the CPUs 31 and 54 as the arithmetic unit.

  Next, a basic operation when the navigation system having the above configuration is used as a route guidance system will be described.

  First, when the operation unit 34 is operated by the driver and the navigation device 14 is activated, a navigation initialization processing unit (not shown) of the CPU 31 performs navigation initialization processing, and the current location and direction sensor detected by the GPS sensor 15. 18 reads the vehicle direction detected by 18 and initializes various data. Next, the matching processing means (not shown) of the CPU 31 performs matching processing, and based on the trajectory of the read current location and the shape, arrangement, etc. of each road link constituting the road around the current location, The current location is identified by determining whether it is located on the road link.

  In addition, a feature recognition processing unit (not shown) of the CPU 31 performs a feature recognition process as an image recognition process, reads image data of a feature image captured by the camera 49, and reads the image data. Image processing is performed on the data to recognize features.

  Therefore, based on the recognized features, it is possible to detect road facilities such as intersections, railroad crossings, pedestrian crossings, and lanes on which vehicles are traveling (hereinafter referred to as “traveling lanes”). For example, when the recognized feature is a stop line, the intersection detection processing unit as the road facility detection processing unit of the feature recognition processing unit performs the intersection detection processing as the road facility detection processing, and the data recording unit 16 Then, the feature data is read out, and the intersection is detected by comparing the image data of the recognized stop line with the feature data. The crossing detection processing means as the road facility detection processing means performs a crossing detection process as the road facility detection processing, and detects a crossing by comparing the image data of the recognized crossing with the feature data. Similarly, the pedestrian crossing detection processing means as the road facility detection processing means performs a pedestrian crossing detection process as the road facility detection processing, and compares the image data of the recognized pedestrian crossing with the feature data to cross the pedestrian crossing. Is detected. Further, the travel lane detection processing means as the road facility detection processing means performs travel lane detection processing as the road facility detection processing, and image data and feature data such as recognized traffic band display lines, traffic classifications, manholes, etc. And the traveling lane is detected.

  The traveling lane detection processing means reads the sensor output of the geomagnetic sensor, and based on the sensor output, determines whether or not there is a detected object made of a ferromagnetic material such as a manhole in a predetermined lane on the road. The traveling lane can be detected based on the determination result. Furthermore, it is possible to detect the current location with high accuracy using the highly accurate GPS sensor 15 and to detect the traveling lane based on the detection result. Further, if necessary, it is possible to detect the traveling lane by combining the sensor output of the geomagnetic sensor, the current location, etc. simultaneously with performing image processing on the image data of the traffic band display line.

  Further, a distance calculation processing unit (not shown) of the CPU 31 performs a distance calculation process to calculate a distance from the current location to a detected intersection, railroad crossing, crosswalk, and the like. A current location correction processing unit (not shown) of the CPU 31 performs current location correction processing, reads the distance, and corrects the current location, for example, with respect to coordinates.

  Subsequently, a basic information acquisition processing unit (not shown) of the CPU 31 performs a basic information acquisition process and acquires the map data by reading it from the data recording unit 16 or receiving it from the information center 51 or the like via the communication unit 38. And get. In addition, when acquiring map data from the information center 51, the said basic information acquisition process means downloads the received map data to flash memory.

  A display processing unit (not shown) of the CPU 31 performs display processing and forms various screens on the display unit 35. For example, the map display processing means of the display processing means performs map display processing, forms a map screen on the display unit 35, and displays the vehicle position, a map around the vehicle position, and the vehicle direction on the map screen. . Therefore, the driver can drive the vehicle according to the vehicle position, the map around the vehicle position, and the vehicle direction.

  Further, when the driver operates the operation unit 34 to input a destination, a destination setting processing unit (not shown) of the CPU 31 performs a destination setting process to set the destination. Note that the departure place can be input and set as necessary. Moreover, a predetermined point can be registered in advance, and the registered point can be set as a destination. Subsequently, when the driver operates the operation unit 34 to input a search condition, a search condition setting processing unit (not shown) of the CPU 31 performs a search condition setting process to set the search condition.

  When the destination and search conditions are set in this way, the route search processing means (not shown) of the CPU 31 performs route search processing, reads the current location, destination, search conditions, and the like from the data recording unit 16. Based on the current location, the destination, and the search data, a route from the departure point to the destination represented by the current location is searched based on the search condition, and route data representing the searched route is output. In this case, the route having the smallest total link cost assigned to each road link is set as the searched route. In addition, when a plurality of lanes are formed by a road and a traveling lane is detected, the route search processing means searches for a searched route for each lane, and the route data includes a traveling lane. The lane number and the like are included.

  Subsequently, a guidance processing means (not shown) of the CPU 31 performs guidance processing and provides route guidance to the driver. For this purpose, the route display processing means of the guidance processing means performs route display processing, reads the route data, and displays the searched route on the map screen according to the route data. When a search route for each lane is searched, route guidance for each lane is performed at a predetermined point, for example, a guidance intersection, and a travel lane for which route guidance is being performed is displayed in the enlarged intersection view. In this case, if necessary, the voice output processing means of the guidance processing means performs voice output processing and outputs a search route by voice from the voice output unit 37 to perform route guidance.

  The information center 51 can perform route search processing. In that case, the CPU 31 transmits the current location, the destination, search conditions, and the like to the information center 51. When the information center 51 receives the current location, destination, search conditions, etc., the route search processing means (not shown) of the CPU 54 performs route search processing similar to that of the CPU 31 and reads the search data from the database 58 to obtain the current location, destination. Based on the location and the search data, a route from the departure point to the destination is searched under the search conditions, and route data representing the searched route is output. Subsequently, a transmission processing unit (not shown) of the CPU 54 performs transmission processing and transmits the route data to the navigation device 14. Therefore, in the navigation device 14, when the basic information acquisition processing means receives the route data from the information center 51, the guidance processing means performs route guidance as described above.

  And when a guidance intersection exists on the searched route, when the vehicle reaches a route guidance point that is a predetermined distance L1 (for example, X [m]) from the guidance intersection, the intersection enlarged map display process of the guidance processing means is performed. The means performs an enlarged intersection map display process, forms an enlarged intersection map as described above in a predetermined area of the map screen, performs route guidance by the enlarged intersection map, and displays the vicinity of the guidance intersection in the enlarged intersection map. A map, a search route, a landmark such as a facility that serves as a landmark at a guidance intersection, and a route lane or the like are displayed when route guidance is provided for each lane. The voice output processing means instructs the voice output unit 37 to perform route guidance by voice, for example, “This is X [m] to the left”.

  By the way, a stop line, a pedestrian crossing, a traffic zone display line, a traffic classification, a road sign, and the like among the features are formed by painting paint or the like on the road to constitute a painted feature. Therefore, when the vehicle repeatedly travels on the painted feature, the painted feature may be rubbed, and a part thereof may be faint and lost. In that case, if it is attempted to recognize the painted feature based on the image of the painted feature that is partially missing, the recognition accuracy is lowered.

  As a result, the accuracy of detecting intersections, railroad crossings, pedestrian crossings, travel lanes, etc., and calculating the distance from the current location to intersections, railroad crossings, pedestrian crossings, etc. will be low.

  Therefore, in the present embodiment, before the feature recognition process is performed, an image correction processing unit (not shown) of the CPU 31 performs an image correction process to correct an image of a portion where the painted feature is likely to be lost, The recognition accuracy of painted features is increased.

  Next, the operation of the image correction processing means will be described in the case where a part of the stop line in the painted feature is missing.

  FIG. 2 is a first flowchart showing the operation of the image correction processing means in the embodiment of the present invention, FIG. 3 is a second flowchart showing the operation of the image correction processing means in the embodiment of the present invention, and FIG. The figure which shows the intersection in embodiment of invention, FIG. 5 is the 1st figure which shows the method of estimating the driving | running locus in the intersection in embodiment of this invention, FIG. 6 is driving | running locus in the intersection in embodiment of this invention It is a 2nd figure which shows the method of estimating.

  4 and 5, c1 is an intersection, r1 and r2 are roads intersecting at the intersection c1, h1 and h2 are center lines of the roads r1 and r2, k1 to k3 are lanes on the driving lane side of the road r1, and k4 to k6 is a lane on the opposite lane side of the road r1, m1 to m3 are lanes on the traveling lane side of the road r2, m4 to m6 are lanes on the opposite lane side of the road r2, and w1 is a vehicle that is the own vehicle. The vehicle w1 is traveling on a lane k3 dedicated to right turn, travels on the road r1, turns right on the intersection c1, and travels on the road r2. The road r1 is an approach road and the road r2 is an exit road. Further, q1 is a stop line formed before the intersection c1, and s1 is a traffic section indicated by an arrow indicating that the lane k3 is dedicated to a right turn.

  In the present embodiment, the stop line q1 is a painted feature that is a recognition target, that is, a target feature. However, at the same time as the stop line q1, another painted feature is used as a target feature. Other painted features can be targeted.

  First, the image correction condition determination processing means of the image correction processing means performs image correction condition determination processing, and determines whether or not the first image correction condition is satisfied depending on whether matching processing is performed. When the first image correction condition is satisfied and the matching process is performed, the image correction condition determination processing means reads the vehicle position detected by the GPS sensor 15 (FIG. 1) and the target feature information. The feature data is acquired by reading from the data recording unit 16.

  Next, the image correction condition determination processing means determines whether or not the target feature exists based on the feature data as to whether or not the second image correction condition is satisfied. In the present embodiment, the image correction condition determination processing means refers to the feature data, and determines whether or not the target feature exists in the vicinity of the vehicle position, that is, in a predetermined range in front of the vehicle w1. The In the present embodiment, there is a stop line q1 in front of the vehicle w1.

  When the second image correction condition is satisfied and the stop line q1 exists as in the present embodiment, the road shape information acquisition processing means of the image correction processing means performs road shape information acquisition processing, and the data recording unit The map data is read from 16 and the shape of the road around the stop line q1, in this embodiment, the shape of the roads r1 and r2 before and after the stop line q1 and the structure of the lanes k1 to k3 and m1 to m3 are roads. Acquired as shape information. Subsequently, the traveling locus estimation processing means of the image correction processing means performs traveling locus estimation processing, and estimates a steady traveling locus when the vehicle travels on the stop line q1 based on the road shape information. .

  As shown in FIG. 5, when the vehicle w1 travels on a lane k3 dedicated to a right turn and turns right at an intersection c1, it is assumed that the vehicle r1 travels on the lane m3 among the lanes m1 to m3 on the road r2, and the lane k3 A curved traveling locus Rt that reaches m3 is estimated.

  Next, the correction part calculation processing means of the image correction processing means performs a correction part calculation process, and based on the travel locus Rt, a part where the stop line q1 is missing is a part that needs to be corrected, that is, , Calculated as a corrected part. In FIG. 6, t1 and t2 are trajectories through which the front wheels pass when the vehicle travels on the travel trajectory Rt, and the stop line q1 is lost at the parts p1 and p2 as the front wheels pass. The parts p1 and p2 are calculated as corrected parts. In this case, although the trajectory through which the rear wheel passes is not shown, when the rear wheel passes, the stop line q1 is lost in a part adjacent to the parts p1 and p2, and the part is also calculated as a corrected part. . Note that the parts p1 and p2 are calculated based on the travel locus Rt, so that they are calculated to the right in the lane when making a right turn and to the left in the lane when making a left turn.

  Subsequently, the rubbing estimation parameter calculation processing means of the image correction processing means performs rubbing estimation parameter calculation processing, and calculates a rubbing estimation parameter representing the degree of defect in the correction portion based on the vehicle running index. The estimation parameter is larger as the degree of defect is larger, and is smaller as the degree of defect is smaller. The vehicle travel index is an index representing the actual travel degree of the vehicle on the stop line q1 or an assumed travel degree.

  For this purpose, first, the rubbing estimation parameter calculation processing means reads road data from the data recording unit 16, obtains a road type by referring to the road data, and calculates a rubbing estimation parameter J based on the road type. The road type represents a type of road such as a national road, a prefectural road, a municipal road, a narrow street, an expressway, and the like, and is used for estimating a traffic amount representing an assumed degree of travel on the stop line q1 as the vehicle travel index. It becomes an indicator.

  In the present embodiment, a rubbing estimation parameter map (not shown) is recorded in the data recording unit 16, and in the rubbing estimation parameter map, the rubbing estimation parameter J of the country road and the highway with the largest traffic volume is maximized. The rubbing estimation parameter J of the prefectural road with the largest traffic volume is then increased, and the rubbing estimation parameter J of the municipal road and the narrow street with the least traffic volume is minimized.

  The rubbing estimation parameter calculation processing means reads the travel history data from the data recording unit 16, refers to the travel history data, determines the traffic volume on the stop line q1 as the vehicle travel index, and records the vehicle performance. The rubbing estimation parameter J can be calculated based on the traffic volume representing the degree of travel. Further, the rubbing estimation parameter J can be calculated by determining the traffic volume for each type of vehicle. In this case, the friction estimation parameter J is increased as the vehicle type is larger, such as a truck, and the friction estimation parameter J is decreased as the vehicle type is smaller, such as a light vehicle. .

  Subsequently, the rubbing estimation parameter calculation processing means reads the traveling history data from the data recording unit 16, refers to the traveling history data, reads the actual traveling state of the vehicle on the stop line q1 as the vehicle traveling index, The rubbing estimation parameter S is calculated based on the actual traveling condition.

  The traveling actual state represents a link required time, an average vehicle speed, a type of the vehicle w1, and the like in a road link adjacent to the intersection c1, and serves as an index when estimating a load applied to the stop line q1 by another vehicle. In the present embodiment, in the rubbing estimation parameter map, the rubbing estimation parameter S is increased as the link required time is shorter and the average vehicle speed is higher, and the average vehicle speed is lower as the link required time is longer. As a result, the rubbing estimation parameter S is reduced.

  Subsequently, the rubbing estimation parameter calculation processing means reads statistical data from the data recording unit 16, refers to the statistical data, acquires a construction history such as repainting of the stop line q1 as the vehicle running index, A friction estimation parameter K is calculated based on the construction history.

  The construction history represents when the stop line q1 has been repaired and the estimated traveling degree of the vehicle thereafter, and serves as an index for estimating the life of the stop line q1. In the present embodiment, in the rubbing estimation parameter map, the rubbing estimation parameter K is increased as the repaired time is farther from the current time, and the rubbing estimation parameter K is decreased as the repaired time is closer from the current time.

  Thus, when the rubbing estimation parameters J, S, and K are calculated, the correction degree determination processing means of the image correction processing means performs the correction degree determination processing, and the image is calculated based on the rubbing estimation parameters J, S, and K. The degree of correction of the deficiency, i.e., the correction degree η representing the correction accuracy is determined. In the present embodiment, the total rubbing estimation parameter Y is calculated by adding the rubbing estimation parameters J, S, and K, and the correction degree determination processing means increases the total rubbing estimation parameter Y so that the image loss becomes larger. Since it is large, the correction degree η is increased to increase the correction accuracy. The smaller the total rubbing estimation parameter Y is, the smaller the image loss is. Therefore, the correction degree η is decreased to reduce the correction accuracy. In another embodiment, the correction degree η can be changed based on one or two of the rubbing estimation parameters J, S, and K.

  In this embodiment, the road type is used as an index for estimating the traffic volume on the stop line q1, but in other embodiments, the traffic volume based on the road traffic census is used. Can be used. Further, as an index for estimating the load applied to the stop line q1, the link required time of the travel history data, the average vehicle speed, and the like are used. In other embodiments, the current traffic information Link duration, average vehicle speed, etc. can be used.

  In this embodiment, the friction estimation parameter S is calculated based on the travel history data of the other vehicle. However, the friction estimation parameter S may be calculated based on the travel history data of the own vehicle. it can.

  Subsequently, the edge detection processing means of the image correction processing means performs edge detection processing, reads the image data sent from the camera 49, performs an operation on the image data using a predetermined edge detection operator, and stops. The edge of the line q1 is detected.

  FIG. 7 is a diagram showing a first example of edge detection processing in the embodiment of the present invention, and FIG. 8 is a diagram showing a second example of edge detection processing in the embodiment of the present invention.

  In the figure, Q1 is an original image composed of image data, q1 is a stop line, and the stop line q1 is horizontal, that is, edges qx1 and qx2 in the x-axis direction and vertical direction, that is, in the y-axis direction. Edges qy1 and qy2 are provided. Further, p11 to p13 are missing sites. The part p11 is missing in two axial directions, that is, the edge qx2 in the x-axis direction and the edge qy2 in the y-axis direction, and in the parts p12 and p13, one axial direction, that is, an edge in the x-axis direction. It is deficient in qx1 and qx2.

  Then, the edge detection processing means first reads a pixel of a 3 × 3 matrix in the original image Q1, sets the center as a target pixel for edge detection processing, performs a scanning operation on neighboring pixels according to the edge detection operator, Amplify edges (differential value between pixels).

  By the way, the edge detection operators include operators such as a Sobel operator, a Prewitt operator, and a Roberts operator. The Sobel operator is represented by a 3 × 3 matrix hx, hy expressed by Formula (1), and the Prewitt operator is represented by Formula ( The Roberts operator is represented by the 3 × 3 matrix hx, hy represented by the equation (3).

そして、いずれのオペレータを使用する場合も、前記エッジ検出処理手段は、マトリックスｈｘ、ｈｙを原画像Ｑ１のマトリックスに掛け合わせ、合計の絶対値を算出する。続いて、前記エッジ検出処理手段は、前記絶対値を所定の閾（しきい）値と比較して２値化を行い、絶対値が閾値より大きい対象画素をエッジとする。前記マトリックスｈｘ、ｈｙは、互いに９０〔°〕回転させたマトリックスであり、前記絶対値の大きい方のマトリックスが採用される。

When any operator is used, the edge detection processing unit multiplies the matrices hx and hy with the matrix of the original image Q1 to calculate the absolute value of the total. Subsequently, the edge detection processing unit compares the absolute value with a predetermined threshold (threshold value) to perform binarization, and sets a target pixel having an absolute value larger than the threshold as an edge. The matrices hx and hy are matrices rotated by 90 ° with respect to each other, and the matrix having the larger absolute value is employed.

  Each operator is used according to the type of edge. The Sobel operator and the Prewitt operator react strongly to the edges qy1 and qy2 in the y-axis direction by the matrix hx (the absolute value increases), and the x-axis direction by the matrix hy. It reacts strongly to the edges qx1 and qx2. In contrast, the Roberts operator reacts strongly to diagonally extending edges.

  In this way, when the edges qx1, qx2, qy1, qy2 of the stop line q1 are detected, the interpolation processing means of the image correction processing means performs an interpolation process, and corrects a defect in the image of the correction site. For this purpose, the interpolation processing means reads image data constituting the edges qx1, qx2, qy1, qy2 of the stop line q1, and performs edge interpolation according to the image data. There are various edge interpolation methods, and in this embodiment, Hough transform is used.

  Next, the Hough transform will be described.

  FIG. 9 is a diagram showing an XY parameter space in the embodiment of the present invention, and FIG. 10 is a diagram showing an ab parameter space in the embodiment of the present invention. In FIG. 7, the horizontal axis represents the Y coordinate, the vertical axis represents the X coordinate, and in FIG. 8, the horizontal axis represents the b coordinate and the vertical axis represents the a coordinate.

If the coordinates of each point of the straight line Ln1 in the XY parameter space (coordinate system) are (x1, y1), the slope is a1, and the intercept is b1,
y1 = a1 · x1 + b1 (4)
become.

When the equation (4) is subjected to parameter conversion (coordinate conversion) into the ab parameter space,
a = (b−y1) / x1 (5)
What is represented by a point (for example, point PA) in the XY parameter space is represented as a straight line (for example, straight line LA1) in the ab parameter space. For example, when the straight line Ln1 is the edge qx1 (FIG. 7) to be subjected to the interpolation processing, each point of the straight line Ln1 becomes each edge point constituting the edge qx1.

  When the parameter conversion is performed on all edge points on the edge qx1, the straight lines corresponding to the edge points intersect at the intersection (a1, b1) on the ab parameter space. Therefore, when each straight line corresponding to each edge point of the missing part p11 on the edge qx1 is assumed, the straight lines also intersect at the intersection (a1, b1).

  The Hough transform is an edge interpolation method using this geometric characteristic, and the interpolation processing means reversely converts an assumed straight line into an XY parameter space and converts each edge point of the part p11. Is calculated.

  In order to calculate the intersection (a1, b1), the ab parameter space is divided into a plurality of cells. Each time each straight line is formed on the ab parameter space, the count value (voting value) of each cell is incremented, and the coordinates of the cell whose count value is greater than the threshold are the intersection points (a1, b1). Calculated.

  In this way, the interpolation processing means calculates each edge point of the part p11 and interpolates the edge qx2.

  In this case, if the size of the cell, that is, the cell size is reduced, the speed at which the count value increases is reduced, and the time until the count value becomes greater than the threshold value increases. Can be easily specified, and the accuracy of image correction can be increased. On the other hand, when the cell size is increased, the speed at which the count value increases is increased, and the time until the count value becomes larger than the threshold value is shortened, but it is difficult to specify the lines constituting the edge. Thus, the accuracy of image correction is reduced.

  Therefore, in the present embodiment, a value obtained by multiplying the reciprocal of the cell size by a predetermined coefficient is set as the correction degree η. Therefore, the correction degree determination processing means increases the correction accuracy by decreasing the cell size as the total rubbing estimation parameter Y is larger, and increases the cell size as the total rubbing estimation parameter Y is smaller. Lower.

  In other embodiments, the threshold value can be changed without changing the cell size. In this case, if the threshold value is decreased, edge interpolation can be performed even if the number of straight lines formed on the ab parameter space is small, so that the possibility of performing edge interpolation is increased. be able to. In addition, for example, when the total rubbing estimation parameter Y is small and the image loss is small, increasing the threshold value can increase the number of straight lines formed on the ab parameter space. Can be picked up.

  By the way, in the XY parameter space, the slope a1 and the intercept b1 both take values of (−∞, + ∞), so that the conversion from the XY parameter space to the ρ-θ parameter space is performed. Is done.

  FIG. 11 is a diagram showing a ρ-θ parameter space in the embodiment of the present invention. In the figure, the horizontal axis represents the θ coordinate, and the vertical axis represents the ρ coordinate.

In this case, what is represented by each point (for example, point PA) of the straight line Ln1 (FIG. 9) in the XY parameter space is represented by a curve (for example, curve LA2) in the ρ-θ parameter space. The equation for each curve is
ρ = x · cos θ + y · sin θ
It is represented by

  By the way, in the part p11 as shown in FIG. 7, since it is missing at the edge qx2 in the x-axis direction and the edge qy2 in the y-axis direction, the x-axis direction and the y-axis direction are used using the matrices hx and hy. It is preferable to perform edge detection processing in step (b).

  On the other hand, the parts p12 and p13 are missing at the edges qx1 and qx2 only in the x-axis direction, and the parts p12 and p13 penetrate between the edges qx1 and qx2. Therefore, if edge detection processing is performed in the x-axis direction and the y-axis direction, an edge that does not exist may be erroneously detected unnecessarily.

  Therefore, before the edge detection processing by the edge detection processing means is started, the edge detection direction determination processing means of the image correction processing means performs the edge detection direction determination processing, reads the travel locus Rt, and reads the travel locus Rt. Based on this, the direction of the edge qx1, qx2, qy1, qy2 missing in the parts p11 to p13 is determined, and this direction is set as the edge detection direction.

  The edge detection processing means performs edge detection processing based on the determination result of the edge detection direction determination processing, reads the edge detection direction, and when the edge detection directions are the x-axis direction and the y-axis direction, the x-axis direction When the edge detection direction is the x-axis direction, the edge detection process is performed. When the edge detection direction is the y-axis direction, the edge detection process is performed in the y-axis direction. I do.

  The interpolation processing means interpolates edges in the x-axis direction and y-axis direction when edge detection processing is performed in the x-axis direction and y-axis direction, and when edge detection processing is performed in the x-axis direction. When the edge detection process is performed in the y-axis direction, the edge in the y-axis direction is interpolated.

  As described above, when the stop line q1 exists, the travel locus Rt of the vehicle is estimated, the corrected part is calculated, the image at the corrected part is corrected, and the image data and the feature data of the corrected image are obtained. Since collation is performed, the accuracy of image recognition can be increased. Therefore, the recognition accuracy of the stop line q1 can be increased.

  As a result, the accuracy of detecting the intersection c1 and calculating the distance to the intersection c1 can be increased.

  In the present embodiment, the travel locus Rt of the vehicle is estimated, the missing direction is determined based on the travel locus Rt, the edge detection process is performed in the missing direction based on the determination result, and the interpolation process is performed. Therefore, not only can the interpolation be performed stably, but also the processing time required for the image correction process can be shortened.

Next, a flowchart will be described.
Step S1: Wait for the matching process to be performed.
Step S2: Obtain feature data.
Step S3: Determine whether the target feature exists. If the target feature exists, the process proceeds to step S4. If the target feature does not exist, the process returns to step S1.
Step S4: Obtain road shape information around the target feature.
Step S5 Estimate the travel line of the vehicle.
Step S6: An interpolation part of the target feature is calculated.
Step S7: The road type is acquired, and the rubbing estimation parameter J is calculated.
Step S8 Read travel history data.
Step S9 The rubbing estimation parameter S is calculated from the link required time, the average vehicle speed, and the like.
Step S10: Obtain construction history information of the target feature and calculate a friction estimation parameter K.
Step S11: The correction degree η is determined based on each rubbing estimation parameter J, S, K.
Step S12: Interpolation processing is performed and the processing is terminated.

  In the present embodiment, a corrected part is calculated, and a defect in the image of the corrected part is corrected by a correction degree η determined based on the rubbing estimation parameters J, S, K, and image data of the corrected image However, in another embodiment, when the corrected portion is calculated by the corrected portion calculation processing means, the feature data (not shown) of the CPU 31 is recognized. The correction processing means can perform feature data correction processing to correct and delete a portion corresponding to the corrected portion in the feature data. In that case, the feature recognition processing means recognizes the feature by comparing the image data with the corrected feature data. Therefore, the recognition accuracy of the feature can be increased.

  In still another embodiment, when the corrected portion is calculated by the corrected portion calculation processing means, the feature recognition processing means determines the correction portion when comparing the image data with the feature data. The matching degree of matching is changed, and other parts are matched without changing the matching degree. In this case, the degree of coincidence represents the degree to which the image data and the feature data are regarded as coincident, and the degree of coincidence decreases as the total rubbing estimation parameter Y increases. Therefore, the feature recognition processing means lowers the matching degree of matching for the corrected part, and collates the other parts without changing the matching degree. Therefore, the recognition accuracy of the feature can be increased.

  In yet another embodiment, when the corrected portion is calculated by the corrected portion calculation processing means, the image data and the feature data can not be collated for the corrected portion. Therefore, the recognition accuracy of the feature can be increased.

  In addition, this invention is not limited to the said embodiment, It can change variously based on the meaning of this invention, and does not exclude them from the scope of the present invention.

It is a figure which shows the navigation system in embodiment of this invention. It is a 1st flowchart which shows operation | movement of the image correction process means in embodiment of this invention. It is a 2nd flowchart which shows operation | movement of the image correction process means in embodiment of this invention. It is a figure which shows the intersection in embodiment of this invention. It is a 1st figure which shows the method of estimating the traveling locus in the intersection in embodiment of this invention. It is a 2nd figure which shows the method of estimating the driving | running | working locus | trajectory in the intersection in embodiment of this invention. It is a figure which shows the 1st example of the edge detection process in embodiment of this invention. It is a figure which shows the 2nd example of the edge detection process in embodiment of this invention. It is a figure which shows XY parameter space in embodiment of this invention. It is a figure which shows the ab parameter space in embodiment of this invention. It is a figure which shows (rho) -theta parameter space in embodiment of this invention.

Explanation of symbols

31 CPU
p1, p2 part q1 stop line Rt travel locus

Claims (10)

A photographing device for photographing features;
A travel locus estimation processing means for estimating a steady travel locus based on road shape information of a road around the target feature when the vehicle travels on the target feature to be corrected of the captured image; ,
A corrected part calculation processing means for calculating a corrected part of a captured image of the target feature based on the estimated traveling locus;
A rubbing estimation parameter calculation processing means for calculating a rubbing estimation parameter representing the degree of image loss at the calculated corrected portion based on the vehicle running index on the target feature;
An image correction apparatus comprising: an interpolation processing unit that corrects a defect in the image of the corrected region based on the calculated friction estimation parameter . Before Symbol vehicles traveling index image correction apparatus according to claim 1 which is traffic on the target feature. The image correction apparatus according to claim 1 , wherein the vehicle traveling index is a traveling actual state of the vehicle on the target feature. The image correction apparatus according to claim 1 , wherein the vehicle travel index is a construction history of the target feature. The image correction apparatus according to claim 1 , further comprising a correction degree determination processing unit that determines a correction degree of the interpolation processing based on the rubbing estimation parameter. While having an edge detection direction determination processing means for determining the edge detection direction of the target feature based on the travel locus,
The image correction apparatus according to claim 1, wherein the interpolation processing unit corrects a defect in the image of the correction site in an edge detection direction. A photographing device for photographing features;
A travel locus estimation processing means for estimating a steady travel locus based on road shape information of a road around the target feature when the vehicle travels on the target feature to be corrected of the captured image; ,
A corrected part calculation processing means for calculating a corrected part of a captured image of the target feature based on the estimated traveling locus;
A rubbing estimation parameter calculation processing means for calculating a rubbing estimation parameter representing the degree of image loss at the calculated corrected portion based on the vehicle running index on the target feature;
Based on the calculated rubbed estimated parameter, feature recognition apparatus characterized by having a feature recognition processing means for recognizing a feature of definitive to the modification site. A recording section in which the feature data is recorded;
A photographing device for photographing features;
A travel locus estimation processing means for estimating a steady travel locus based on road shape information of a road around the target feature when the vehicle travels on the target feature to be corrected of the captured image; ,
A corrected part calculation processing means for calculating a corrected part of a captured image of the target feature based on the estimated traveling locus;
Feature data correction processing means for correcting a location corresponding to the correction site in the feature data;
A feature recognition apparatus comprising: feature recognition processing means for recognizing a feature by comparing the corrected feature data with image data. A recording section in which the feature data is recorded;
A photographing device for photographing features;
A travel locus estimation processing means for estimating a steady travel locus based on road shape information of a road around the target feature when the vehicle travels on the target feature to be corrected of the captured image; ,
A corrected part calculation processing means for calculating a corrected part of a captured image of the target feature based on the estimated traveling locus;
A feature recognition apparatus comprising: a feature recognition processing means for recognizing a feature by changing the degree of coincidence of the correction portion and comparing the feature data with image data. Take a picture of a feature with a photographic device,
When a vehicle travels on a target feature that is a target of correction of a captured image , a steady traveling locus based on road shape information of a road around the target feature is estimated,
Based on the estimated travel trajectory, calculate the corrected part of the captured image of the target feature,
Based on the vehicle running index on the target feature, to calculate a rubbing estimation parameter representing the degree of image loss in the calculated correction site,
Based on the calculated rubbed estimated parameters, the feature recognition method and recognizes the feature of definitive to the modification site.

Images