JP2020118575A - Inter-vehicle distance measurement device, error model generation device, learning model generation device, and method and program thereof - Google Patents

Abstract

To provide an inter-vehicle distance measurement device that can estimate an inter-vehicle distance to a preceding vehicle as well as an error of the distance.SOLUTION: An inter-vehicle distance measurement device comprises: a camera 10 installed on a vehicle; a GPS/IMU11; a distance calculation unit 12 for calculating an inter-vehicle distance to a preceding vehicle when the vehicle is in a first position by identifying a lower end of the preceding vehicle in an image of the preceding vehicle imaged when the vehicle is in the first position while obtaining a straight line connecting an optical center of the camera and the lower end, obtaining a shape of a road based on position information sequentially obtained by the vehicle's traveling, and obtaining a second position where the straight line crosses the road; an error model storage unit for storing a scale of an error using as variables length from the optical center of the camera to the lower end of the preceding vehicle reflected in the image and displacement in a pitch angle of the camera between the first position and the second position; and an error scale calculation unit 13 for calculating the scale of the error in the inter-vehicle distance based on a coordinate of the lower end of the preceding vehicle reflected in the image and the displacement in the pitch angle of the camera.SELECTED DRAWING: Figure 1

Description

The present invention relates to an inter-vehicle distance measuring device, an error model generating device, a learning model generating device, and the like.

The driving support function by Adaptive Cruise Control (hereinafter referred to as “ACC”) is currently installed in many vehicles and is an important system. A distance estimation system to a preceding vehicle is mentioned as an elemental technology of ACC, and a distance measuring method using a distance sensor or a stereo camera is used (Non-Patent Document 1). In recent years, distance estimation technology in monocular cameras has attracted attention from the viewpoint of cost, etc. Stein et al. proposes the distance to the preceding vehicle by triangulation by assuming the road surface from the own vehicle to the preceding vehicle as a plane. There is a method (Non-Patent Document 2). Also, a distance estimation method based on learning using Deep Neural Networks (hereinafter, “DNN”) has been proposed (Non-Patent Document 3), and shows good results as compared with the conventional geometric approach. ..

JP, 2005-79388, A

Saneyoshi, K. (1996, September) ``Drive assist system using stereo image recognition'' In Intelligent Vehicles Symposium, 1996., Proceedings of the 1996 IEEE (pp. 230-235).IEEE. Stein, GP, Mano, O., & Shashua, A. (2003, June) ``Vision-based ACC with a single camera: bounds on range and range rate accuracy'' In Intelligent vehicles symposium, 2003. Proceedings.IEEE (pp. 120125).IEEE. Laina, Iro, et al. ``Deeper depth prediction with fully convolutional residual networks.''3D Vision (3DV), 2016 Fourth International Conference on.IEEE, 2016.

The method using a distance sensor or a stereo camera has an industrial disadvantage that the cost is high. Therefore, the demand for a distance estimator using a monocular camera is currently increasing. Since a distance estimation method using a monocular camera based on a geometric constraint as in Non-Patent Document 2 is based on many assumptions, the estimation accuracy changes depending on the traveling environment. For example, since the road shape is assumed to be a plane, accurate estimation cannot be performed on a road surface having a curvature.

In addition, a method based on learning using a machine learning method has a possibility of performing estimation that does not depend on the gradient of the road or the surrounding environment. However, it is known that the learning-based estimator is affected by the bias existing in the learning data and the estimation accuracy is reduced in an environment different from the learning data. Furthermore, it is difficult to analyze errors and evaluate the stability of estimation results.

The present inventors proposed in Patent Document 1 a method for estimating an inter-vehicle distance with a monocular camera. With the method proposed in Patent Document 1, it is possible to perform distance estimation that does not depend on the shape of the road surface, but the accuracy of Visual Odometry used to restore the shape of the road surface depends on the accuracy of distance estimation.

The present invention has been made in view of the above background, and an object of the present invention is to provide an inter-vehicle distance measuring device capable of estimating an inter-vehicle distance to a preceding vehicle and an error thereof.

An inter-vehicle distance measuring device of the present invention includes a camera mounted on a vehicle, a position information detection unit that detects position information of the vehicle, and an image of a preceding vehicle when the vehicle is at the first position. A straight line connecting the optical center of the camera and the lower end is determined while identifying the lower end of the vehicle, the shape of the road is obtained based on the position information sequentially obtained by the traveling of the vehicle, and the straight line intersects with the road. A distance calculation unit that obtains an inter-vehicle distance between the vehicle and the preceding vehicle when the vehicle is at the first position, and the image from the optical center of the camera to the lower end of the preceding vehicle. An error model storage unit that stores a model of an error scale having a variable in length and a displacement of the pitch angle of the camera between the first position and the second position, and the error. Referring to the model of the error scale stored in the model storage unit, based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera, the inter-vehicle distance of An error scale calculation unit that calculates an error scale and an output unit that outputs the inter-vehicle distance and the error scale are provided. With this configuration, not only the inter-vehicle distance but also the scale of the error can be obtained. By obtaining the scale of the error, for example, appropriate processing can be performed in an application using the inter-vehicle distance.

In the inter-vehicle distance measuring device of the present invention, the error model storage unit includes an error of the inter-vehicle distance in a length from the optical center of the camera to a lower end of the preceding vehicle in the image and a displacement of a pitch angle of the camera. The model of the error scale may be stored in the form of a look-up table in which the scales are associated. With this configuration, the calculation load of the error scale can be reduced.

An inter-vehicle distance measuring device of the present invention includes a camera mounted on a vehicle, a position information detection unit that detects position information of the vehicle, and an image of a preceding vehicle when the vehicle is at the first position. A straight line connecting the optical center of the camera and the lower end is determined while identifying the lower end of the vehicle, the shape of the road is obtained based on the position information sequentially obtained by the traveling of the vehicle, and the straight line intersects with the road. And a second position relative to a road surface at the first position, and a distance calculation unit that obtains an inter-vehicle distance between the preceding vehicle when the vehicle is at the first position. An error scale calculator that calculates an error scale based on the inclination of the road surface and the length in the image from the optical center of the camera to the lower end of the preceding vehicle, and an output that outputs the inter-vehicle distance and the error scale. And a section. With this configuration, not only the inter-vehicle distance but also the scale of the error can be obtained. By obtaining the scale of the error, for example, appropriate processing can be performed in an application using the inter-vehicle distance.

The error model generation device of the present invention identifies the lower end of the preceding vehicle in an image obtained by photographing the preceding vehicle when the vehicle is at the first position, and obtains a straight line connecting the optical center of the camera and the lower end, By determining the shape of the road based on the position information sequentially obtained by the traveling of the vehicle and determining the second position where the straight line intersects the road, An inter-vehicle distance measuring apparatus for obtaining an inter-vehicle distance from a preceding vehicle, which is an error model generation apparatus for estimating an error in the inter-vehicle distance, and a road structure information input unit for inputting road structure information, The position of the vehicle and the position of the preceding vehicle is determined at random, and by using the road structure information, the true value of the inter-vehicle distance to the preceding vehicle is calculated, and the rotation angle of the camera determined by the road structure and the position of the vehicle are determined. A simulation unit that calculates a vehicle-to-vehicle distance by sequentially adding randomly generated errors to the true value to obtain the second position, and calculates an error of the calculated vehicle-to-vehicle distance with respect to the true value. An error analysis unit that analyzes the scale of the error in the inter-vehicle distance with the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera as variables.

According to the research conducted by the present inventors, the scale of the error of the inter-vehicle distance has a strong correlation between the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera among a plurality of variables. I understood. With the configuration of the present invention, it is possible to generate an error model for obtaining an error scale based on the length in the image and the displacement of the pitch angle from the optical center of the camera to the lower end of the preceding vehicle by simulation.

A learning model generation device of the present invention is a generation device that generates a learning model for estimating an inter-vehicle distance from an image captured by a camera, and is a video of a preceding vehicle and vehicle position information corresponding to each frame of the video. And an input section for inputting the vehicle, a lower end of the preceding vehicle is identified in an image obtained by photographing the preceding vehicle when the vehicle is at the first position, and a straight line connecting the optical center of the camera and the lower end is obtained. When the vehicle is at the first position, the shape of the road is obtained based on the position information sequentially obtained by driving the vehicle, and the second position where the straight line intersects the road is obtained. A distance calculation unit for obtaining an inter-vehicle distance, a length in the image from the optical center of the camera to the lower end of the preceding vehicle, and the camera between the first position and the second position. The error model storage unit that stores a model of the error scale having the displacement of the pitch angle of the variable and the model of the error scale stored in the error model storage unit are referred to, and the preceding vehicle is detected from the optical center of the camera. Based on the length in the image to the lower end of the and the displacement of the pitch angle of the camera, an error scale calculation unit that calculates the scale of the error of the inter-vehicle distance, and the inter-vehicle distance and the error of the plurality of images obtained. A model generation unit that generates a learning model for obtaining an inter-vehicle distance from an image using the scale as teacher data.

A learning model generation device of the present invention is a generation device that generates a learning model for estimating an inter-vehicle distance from an image captured by a camera, and is a video of a preceding vehicle and vehicle position information corresponding to each frame of the video. And an input section for inputting the vehicle, a lower end of the preceding vehicle is identified in an image obtained by photographing the preceding vehicle when the vehicle is at the first position, and a straight line connecting the optical center of the camera and the lower end is obtained. When the vehicle is in the first position, the shape of the road is obtained based on the position information sequentially obtained by driving the vehicle, and the second position where the straight line intersects the road is obtained. A distance calculation unit for obtaining an inter-vehicle distance to the vehicle, an inclination of the road surface at the second position with respect to the road surface at the first position, and a length in the image from the optical center of the camera to the lower end of the preceding vehicle. And an error scale calculation unit that calculates an error scale, and a model generation unit that generates a learning model that obtains the inter-vehicle distance from the images, using the inter-vehicle distance and the error scale obtained for the plurality of images as teacher data. ..

According to the present invention, since the teacher data is generated from the position information corresponding to the image of the preceding vehicle, a large amount of teacher data can be easily prepared. Further, since learning is performed using the inter-vehicle distance and the scale of the error as teacher data, the learning strength of the learning model can be changed according to the scale of the error, and an appropriate learning model can be generated. As one aspect, the model generation unit may perform learning with a smaller influence on the loss function as the scale of the error is larger.

An inter-vehicle distance measuring method of the present invention is a method for measuring an inter-vehicle distance with a preceding vehicle on the basis of an image captured by a camera mounted on a vehicle and position information of the vehicle. The step of inputting an image taken by the camera, the step of detecting positional information of the vehicle in the inter-vehicle distance measuring apparatus, and the inter-vehicle distance measuring apparatus when the vehicle is at the first position. The lower end of the preceding vehicle is identified in the image of the preceding vehicle, a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and based on the position information sequentially obtained by the traveling of the vehicle. Determining the shape of the road and determining the second position where the straight line intersects the road to determine the inter-vehicle distance with the preceding vehicle when the vehicle is in the first position; and the inter-vehicle distance. A measuring device varies the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera between the first position and the second position. With reference to an error model storage unit that stores a model of an error scale, the distance between the vehicles based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera. The method further includes a step of calculating a scale of a distance error, and a step of the inter-vehicle distance measuring device outputting the scale of the inter-vehicle distance and the error. Further, the step of calculating the scale of the error is preceded by the inclination of the road surface at the second position with respect to the road surface at the first position and the optical center of the camera instead of the method of referring to the data in the error model storage unit. The error scale may be calculated based on the length up to the lower end of the vehicle in the image.

The error model generation method of the present invention identifies a lower end of a preceding vehicle in an image obtained by photographing the preceding vehicle when the vehicle is at the first position, and connects a straight line connecting the optical center of the camera and the lower end in the world coordinate system. Is obtained, the shape of the road is obtained based on the position information sequentially obtained by the vehicle traveling, and the second position at which the straight line intersects the road is obtained. A method of generating an error model for estimating an error of the inter-vehicle distance in an inter-vehicle distance measuring apparatus for obtaining an inter-vehicle distance with a preceding vehicle when there is a situation, and inputting road structure information to the error model generating apparatus. Step, the error model generation device randomly determines the first position and the position of the preceding vehicle, and calculates the true value of the inter-vehicle distance to the preceding vehicle using the structure information of the road, and The inter-vehicle distance is calculated by adding the randomly generated error to the true value of the camera rotation angle and the vehicle position determined by the road structure, and calculating the second position. A step of calculating an error with respect to a true value of the inter-vehicle distance, the error model generation device, the displacement in the pitch angle of the camera and the length in the image from the optical center of the camera to the lower end of the preceding vehicle as variables. And analyzing the scale of the error of the inter-vehicle distance.

The inter-vehicle distance measuring program of the present invention is a program for measuring an inter-vehicle distance to a preceding vehicle based on an image taken by a camera mounted on the vehicle and position information of the vehicle, and a computer In the step of inputting an image taken by the camera, the step of inputting position information of the vehicle, and the lower end of the preceding vehicle in the image of the preceding vehicle taken when the vehicle is in the first position. Identify and determine a straight line that connects the optical center of the camera and the lower end in the world coordinate system, and determine the shape of the road based on the position information that is sequentially obtained by traveling of the vehicle, and the straight line is the road. Determining a second position at which the vehicle intersects with the preceding vehicle when the vehicle is at the first position; and in the image from the optical center of the camera to the lower end of the preceding vehicle. With reference to an error model storage unit that stores a model of an error scale having a variable of the length at, and a displacement of the pitch angle of the camera between the first position and the second position, Based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera, calculating the scale of the error of the inter-vehicle distance, and the scale of the inter-vehicle distance and the error. Perform the output step and. Further, the step of calculating the scale of the error is preceded by the inclination of the road surface at the second position with respect to the road surface at the first position and the optical center of the camera instead of the method of referring to the data in the error model storage unit. The error scale may be calculated based on the length up to the lower end of the vehicle in the image.

The error model generation program of the present invention identifies the lower end of the preceding vehicle in the image of the preceding vehicle when the vehicle is at the first position, and connects the optical center of the camera and the lower end in the world coordinate system. By obtaining the straight line, the shape of the road is obtained based on the position information sequentially obtained by the traveling of the vehicle, and the second position at which the straight line intersects the road is obtained, whereby the vehicle is In a vehicle-to-vehicle distance measuring device that obtains the vehicle-to-vehicle distance when the vehicle is in a position, the program is for generating an error model for estimating an error in the vehicle-to-vehicle distance. A step of inputting, randomly determining the first position and the position of the preceding vehicle, calculating the true value of the inter-vehicle distance to the preceding vehicle using the road structure information, and determining the camera according to the road structure. The rotation angle and the true value of the vehicle position are sequentially added with errors randomly generated to obtain the second position, thereby calculating the inter-vehicle distance, and calculating the true value of the inter-vehicle distance. And a step of analyzing the scale of the error of the inter-vehicle distance with the displacement in the pitch angle of the camera and the length in the image from the optical center of the camera to the lower end of the preceding vehicle as variables. And execute.

According to the present invention, not only the inter-vehicle distance but also the scale of the error can be obtained.

It is a figure which shows the structure of the vehicle distance measuring apparatus of 1st Embodiment. It is a figure which shows the installation position of a camera. It is a figure which shows the locus|trajectory of the own vehicle and the locus|trajectory of the ground point. (A) It is a figure which shows the straight line L which connects the optical center of a camera and the lower end of a preceding vehicle. (B) It is a figure which shows the straight line L which connects the optical center of a camera and the lower end of a preceding vehicle. (A) It is the figure which plotted the locus|trajectory of the ground point, and the straight line L in the world coordinate system. (B) It is the figure which plotted the locus|trajectory of the ground point, and the straight line L in the world coordinate system. It is a flow chart which shows operation of an inter-vehicle distance measuring method of a 1st embodiment. It is a figure which shows the example of calculation of the coordinate of the left edge, right edge, and lower edge of the vehicle detected from the image. It is a figure which shows the structure of the error model production|generation apparatus which produces|generates the error model used with an inter-vehicle distance measuring apparatus. It is a figure which shows the main sensor value used when calculating the inter-vehicle distance D. It is a flowchart which shows operation|movement of an error model production|generation apparatus. It is a figure which shows the structure of the vehicle distance measuring apparatus of 2nd Embodiment. It is a figure which shows the principle which calculates|requires the lower end of a preceding vehicle in 2nd Embodiment. It is a figure which shows the structure of the learning model production|generation apparatus of 3rd Embodiment. It is a flowchart which shows operation|movement of the learning model production|generation apparatus of 3rd Embodiment.

Hereinafter, an inter-vehicle distance measuring device according to an embodiment of the present invention will be described in detail with reference to the drawings.
(First embodiment)
[Configuration of inter-vehicle distance measuring device]
FIG. 1 is a diagram showing a configuration of an inter-vehicle distance measuring device 1 according to the first embodiment. The inter-vehicle distance measuring device 1 includes a camera 10 mounted on a vehicle, a GPS/IMU 11 that detects the current position of the vehicle and acquires position information, an image captured by the camera 10, and position information from the GPS/IMU 11. The distance calculation unit 12 that estimates the inter-vehicle distance to the preceding vehicle, the error scale calculation unit 13 that calculates the scale of the inter-vehicle distance error estimated by the distance calculation unit 12, and the error used to calculate the error scale. An error model storage unit 14 that stores a model and an output unit 15 that outputs the obtained inter-vehicle distance and error scale are provided. The camera 10 is, for example, a CCD camera 10.

FIG. 2 is a diagram showing the installation position of the camera 10. The camera 10 is installed at the tip (front bumper) of the vehicle. The height of the camera 10 from the ground is H, and in this document, the ground vertically below the optical center of the camera 10 is called the ground point.

When the camera 10 is installed in the vehicle, the optical axis (Z axis) of the camera 10 matches the traveling direction of the vehicle, and the horizontal axis (X axis) of the image captured by the camera 10 is parallel to the ground. The installation angle and the like of the camera 10 are adjusted. As a result, the vertical axis (Y axis) of the image coincides with the direction perpendicular to the ground. The camera 10 captures an image while the vehicle is traveling and inputs the captured image to the distance calculation unit 12.

The distance calculation unit 12 obtains the movement trajectory of the vehicle based on the vehicle position information input from the GPS/IMU 11. As a result, the undulation of the road on which the vehicle has passed can be obtained.

FIG. 3 is a diagram showing the trajectory of the host vehicle and the trajectory of the ground point. In FIG. 3, the position of the camera 10 at the initial time is the origin, the optical axis direction is the Z axis, the ground direction is the Y axis, and the X axis is defined by the right-hand coordinate system (world coordinate system). The locus of the ground point described on the YZ plane shows the undulations of the ground, and thus the undulations of the road on which the vehicle has passed can be known.

The distance calculation unit 12 also identifies the lower end of the preceding vehicle from the image captured by the camera 10 and obtains the image coordinates thereof. The origin of the image coordinates is the optical center of the camera. Using these image coordinates, as shown in FIG. 4A, a straight line L connecting the optical center of the camera 10 and the lower end of the preceding vehicle is obtained. This straight line L is a straight line in the world coordinate system.

Subsequently, a straight line L connecting the optical center of the camera 10 and the lower end of the preceding vehicle is similarly obtained from the image of the next frame taken by the camera 10. As shown in FIG. 4(b), since the host vehicle and the preceding vehicle are moving forward, a straight line different from the previously obtained straight line L can be obtained.

5A and 5B are diagrams in which the locus of the ground point and the straight line L are plotted in the world coordinate system. In the state shown in FIG. 5A, the direction in which the lower end of the preceding vehicle is located is known, but it is not possible to determine how many meters ahead the vehicle is ahead.

As shown in FIG. 5(b), if the locus of the ground point described up to time t1 and the straight line L at the past time t0 intersect, this intersection is the point of the preceding vehicle observed at time t0. It can be seen that it is the position where the lower end was. Therefore, the distance calculation unit 12 can calculate the inter-vehicle distance based on the position of the intersection obtained at the time t1 and the position of the optical center of the camera 10 at the time t0.

As can be seen from FIG. 5B, the principle of the inter-vehicle distance measurement of the present embodiment is that the inter-vehicle distance in the past (at time t0) is measured by proceeding while "surveying" the ground. The position of the host vehicle when the preceding vehicle is photographed at time t0 is referred to as a "first position", and the intersection of the straight line obtained at time t1 and the locus of the ground point is referred to as a "second position".

The process of obtaining the inter-vehicle distance is preferably performed while the vehicle is traveling in a straight section. This is because, in the straight section, the direction of “surveying” the undulations of the road coincides with the direction of the lower end of the preceding vehicle.

The error scale calculation unit 13 has a function of referring to the error model stored in the error model storage unit 14 and calculating the error of the inter-vehicle distance obtained by the distance calculation unit 12. The error model stored in the error model storage unit 14 includes the distance to the lower end of the preceding vehicle shown in the image (the y coordinate value of the lower end when the optical center is the origin) and the displacement of the pitch angle of the camera 10. It is a look-up table which linked|related the error scale. How to generate the error model stored in the error model storage unit 14 will be described later.

The error scale calculator 13 (i) the lower end coordinates of the preceding vehicle appearing in the image captured at the first position, and (ii) the pitch angle of the camera 10 between the first position and the second position. The error scale corresponding to the displacement of is read from the look-up table to obtain the error scale.

The output unit 15 has a function of outputting the inter-vehicle distance to the preceding vehicle obtained by the distance calculation unit 12 and the error scale data obtained by the error scale calculation unit 13. The output unit 15 may be, for example, a monitor that displays the inter-vehicle distance data to the user, or a communication interface that outputs the inter-vehicle distance data to other in-vehicle devices.

Although the inter-vehicle distance measuring apparatus 1 according to the embodiment has been described above, the inter-vehicle distance measuring apparatus 1 is composed of a microcomputer including a CPU, a ROM, a RAM, and the like (not shown). The inter-vehicle distance measuring device 1 calculates the inter-vehicle distance to the preceding vehicle based on the image captured by the camera 10 according to the program stored in the ROM, for example. Such a program is also included in the scope of the present invention.

[Operation of inter-vehicle distance measuring device]
FIG. 6 is a flowchart showing the operation of the inter-vehicle distance measuring device according to the present embodiment.
The inter-vehicle distance measuring device 1 captures an image in the traveling direction with the camera 10, and inputs the image to the distance calculation unit 12 (S10). At the same time, the inter-vehicle distance measuring unit acquires the position information of the own vehicle by GPS/IMU 11 and inputs it to the distance calculating unit 12 (S11).

The inter-vehicle distance measuring device 1 identifies the preceding vehicle from the input image. For example, a HOG feature is extracted from an image, and a pattern discriminator such as a support vector machine is applied to the obtained feature amount. The HOG features are introduced in detail in N. Dalal and B. Triggs "Histograms of oriented gradients for human detection" CVPR 2005.

Next, as shown in FIG. 7, the x coordinate of the left end, the x coordinate of the right end, and the y coordinate of the lower end of the vehicle detected from the image are calculated (S12). The image coordinates are x in the horizontal direction, y in the vertical direction, the unit is pixel, and the origin is the passing point of the optical axis. The method for obtaining the left edge may be, for example, a location in the vicinity of the left edge of the window to which the discriminator is applied, where the change in the brightness gradient is the largest. Alternatively, the coordinates of the left end of the previous frame may be moved to the current frame by obtaining the amount of movement (optical flow) on the image plane of the preceding vehicle between the previous frame and the current frame. The inter-vehicle distance measuring device 1 calculates a straight line L from the optical center of the camera 10 at the first position in the world coordinate system to the lower end of the preceding vehicle (S13).

Since the actual vehicle width of the preceding vehicle can be known by detecting the left edge and the right edge of the vehicle, after the vehicle width is determined, the inter-vehicle distance can be immediately obtained based on the vehicle width and the actual vehicle width of the preceding vehicle shown in the image. Like If such processing is not performed, the left and right ends of the vehicle may not be detected.

The inter-vehicle distance measuring device 1 obtains road undulation data based on the position information sequentially detected by the GPS/IMU 11 as the vehicle advances (S14), and the straight line L and the road intersect using the obtained data. It is determined whether or not (S15). When the road does not intersect the straight line L (NO in S15), the inter-vehicle distance measuring device 1 performs a process of obtaining road undulation data using the position information detected by the GPS/IMU 11.

When the straight line L and the road intersect, the distance between the preceding vehicle and the preceding vehicle is determined based on the intersecting position and the position of the camera 10 when the preceding vehicle is first identified (when the host vehicle is at the first position). The distance is calculated (S16). In the present embodiment, the position where the preceding vehicle is first detected (first position) is the origin of the world coordinates, so the Z coordinate value of the position where the straight line L and the road intersect is the inter-vehicle distance (FIG. 5). (See (b)).

Further, the inter-vehicle distance measuring device 1 corresponds to the lower end position Ly of the preceding vehicle detected at the first position and the displacement of the pitch angle Opitch of the camera 10 between the first position and the second position. The error scale is read from the look-up table stored in the error model storage unit 14 and the error scale is calculated (S17). The inter-vehicle distance measuring device 1 outputs the calculated inter-vehicle distance and the error scale (S18).

[Error model generator]
FIG. 8 is a diagram showing a configuration of an error model generation device 20 that generates an error model used in the inter-vehicle distance measuring device 1. First, the outline of the error model generation device 20 will be described. FIG. 9 is a diagram showing main sensor values used when obtaining the inter-vehicle distance D. There are the position (X, Y, Z) of the host vehicle, the pitch angle φ of the camera 10, and the position coordinate y of the lower end of the preceding vehicle shown in the image. Note that f is the focal length, and H is the height of the camera 10.

The error factors for the above-described measurement of the inter-vehicle distance include the error (Mx, My, Mz) in the vehicle position (X, Y, Z), the roll angle, the pitch angle, and the yaw angle error (Oroll, Opitch, Oyaw) and the error (Lx, Ly) in the lower end detection position (x, y) of the preceding vehicle can be considered as an eight-element error. The error model generation device 20 repeats the error estimation by pseudo giving each error, the curvature of the road surface, the vehicle speed, etc. by Monte Carlo simulation, and performs the error analysis based on the obtained error estimation result.

The present inventors have found that, among these eight elements, the governing term of the error with respect to the inter-vehicle distance is the error between the pitch angle Opitch of the host vehicle and the lower end detection position Ly. Therefore, when performing the error analysis, the error analysis is performed using the pitch angle Opitch and the lower end detection position Ly of the preceding vehicle as variables. Thereby, the scale of the error is modeled by the pitch angle Opitch and the lower end detection position Ly of the preceding vehicle. The configuration of the error model generation device 20 will be specifically described below.

As shown in FIG. 8, the error model generation device 20 includes a road structure information input unit 21 for inputting road structure information, a vehicle and a preceding vehicle at random on the input road structure, and an inter-vehicle distance between them. And a simulation unit 22 for simulating the error, an error analysis unit 28 for performing an error analysis using the inter-vehicle distance and the error data obtained by the simulation, and an error model generated based on the analysis result by the error analysis unit 28 are stored. It has a storage unit 29 for performing.

The simulation unit 22 includes a vehicle placement unit 23, a true value calculation unit 24, an error setting unit 25, a distance calculation unit 26, and an error scale calculation unit 27. The vehicle arranging unit 23 has a function of randomly arranging the own vehicle and the preceding vehicle on the road structure. The true value calculation unit 24 has a function of calculating the true value of the inter-vehicle distance between the own vehicle and the preceding vehicle that are randomly arranged. The distance calculation unit 26 has a function of performing a simulation of moving the host vehicle to the position of the preceding vehicle, similarly to the distance calculation unit 12 of the inter-vehicle distance measuring device 1, and calculating the inter-vehicle distance between the own vehicle and the preceding vehicle.

In the calculation of the inter-vehicle distance, the distance calculation unit 26 determines the position of the host vehicle (X, Y, Z) while the host vehicle moves to the position of the preceding vehicle, the attitude (Oroll, Opitch, Oyaw) of the host vehicle, and the lower end of the preceding vehicle. A random error is set for the true value of the detection position (Lx, Ly), and the inter-vehicle distance is calculated using the value with the error added. The error setting unit 25 has a function of generating a random error and inputting it to the distance calculation unit 12. When the inter-vehicle distance is determined based on the speed of the host vehicle and the inter-vehicle time, the error setting unit 25 may randomly set not only the error of each sensor value but also the speed of the host vehicle and the inter-vehicle time. ..

The error scale calculation unit 27 has a function of calculating an error based on the difference between the inter-vehicle distance calculated by the distance calculation unit 26 and the true value. The simulation unit 22 changes the added error and repeatedly calculates the inter-vehicle distance by the distance calculation unit 26 to obtain a large number of simulation results. In addition, the vehicle placement unit 23 randomly changes the position of the vehicle to obtain a large number of simulation results.

The error analysis unit 28 has a function of generating an error model based on the inter-vehicle distance and the simulation result of the error for the inter-vehicle distance. The error analysis unit 28 uses the displacement of the pitch angle Opitch of the vehicle and the detection position Ly of the lower end as input variables for the obtained simulation result, as shown in the following equation, from the error dispersion regression problem to the error dispersion. To model.

FIG. 10 is a flowchart showing the operation of the error model generation device 20. The error model generation device 20 first receives an input of information on a road structure to be simulated (S20). As the road structure information, information on the structure conforming to the regulations of the Road Structure Ordinance is input. The error model generation device 20 randomly determines the positions of the own vehicle and the preceding vehicle (S21), and calculates the true value of the inter-vehicle distance (S22).

Next, the error model generation device 20 calculates the inter-vehicle distance by performing a simulation of moving from the position of the own vehicle to the position of the preceding vehicle. Specifically, a random error is added to the true value of each sensor value while moving from the position of the own vehicle to the position of the preceding vehicle, the inter-vehicle distance is calculated, and the calculated inter-vehicle distance is calculated. An error from the value is calculated (S23).

The error model generation device 20 determines whether or not to end the simulation with the placement randomly set by the vehicle placement unit 23 (S24). When the simulation with the set arrangement is not ended (NO in S24), the error model generation device 20 sets a random error and repeatedly calculates the inter-vehicle distance and the error. When the simulation with the set arrangement is ended (YES in S24), it is determined whether or not the own vehicle and the preceding vehicle are set to different positions and calculation is performed (S25).

When performing a simulation with another arrangement (YES in S25), the error model generation device 20 randomly determines the positions of the own vehicle and the preceding vehicle (S21), and performs the above-described simulation to perform the inter-vehicle distance and the error. Is obtained (S22 to S24). If the simulation is not performed in another arrangement, it is determined whether or not the simulation is performed in another road structure (S26). When a simulation is performed with another road structure (YES in S26), the process returns to the input of road structure information (S20), the input of new road structure information is accepted, and the above-described simulation is performed (S21 to S25). .. When the simulation with another road structure is not performed (NO in S26), the inter-vehicle distance and the error obtained by the simulation are analyzed to generate an error model, and the generated error model is stored (S27).

Here, as an example, an example of polynomial regression of an error model will be shown.

As described above, in the inter-vehicle distance measuring device 1, the error model storage unit 14 is a lookup table format in which the scale of the inter-vehicle distance error is associated with the lower end position of the preceding vehicle and the pitch angle displacement of the camera 10 shown in the image. It stores the error model of. In accordance therewith, the error model generation device 20 may process the error model into a look-up table. The inter-vehicle distance measuring apparatus 1 of the first embodiment, the error model generating apparatus 20 that generates an error model used in the inter-vehicle distance measuring apparatus 1, and their operations have been described above.

The inter-vehicle distance measuring device 1 according to the first embodiment can obtain not only the inter-vehicle distance to the preceding vehicle but also the scale of the error. The error scale indicates the reliability of the inter-vehicle distance, and it is possible to obtain auxiliary information for distance estimation. For example, when applying the measured inter-vehicle distance to the Kalman filter, by inputting the error scale, it is possible to perform processing such that the measurement result with low reliability is not used for updating the Kalman filter.

The error model generation device 20 collects the samples of the measurement results of the inter-vehicle distance by repeating the pseudo distance estimation by simulation, and thereby the error by the polynomial regression using the dominant term of each sample and the error analytically specified. Can be modeled.

(Second embodiment)
FIG. 11 is a diagram showing a configuration of an inter-vehicle distance measuring device 2 according to the second embodiment. The basic configuration of the second embodiment is the same as that of the inter-vehicle distance measuring apparatus 1 of the first embodiment, and the inter-vehicle distance and its error scale are obtained. The method of measuring the inter-vehicle distance by the inter-vehicle distance measuring apparatus 2 of the second embodiment is the same as that of the inter-vehicle distance measuring apparatus 1 of the first embodiment. The inter-vehicle distance measuring apparatus 2 according to the second embodiment is different in that the error scale calculation unit 16 calculates an error scale using an error scale calculation formula.

First, the concept of error scale calculation in the second embodiment will be described. Even when the road surface between the host vehicle and the preceding vehicle has a curvature, the road surface on which the host vehicle and the preceding vehicle are mounted is a plane when viewed locally. The inventors of the present invention have noticed that the error in the inter-vehicle distance depends on the inclination of the road surface on which the preceding vehicle is mounted as viewed from the road surface on which the host vehicle is mounted.

これをＺについて解くと、

となる。先行車両の下端の観測誤差を

とすれば、１次のテイラー展開により、車間距離の誤差のスケールは、次式（４）で表される。

FIG. 12 is a diagram showing the principle of obtaining the lower end of the preceding vehicle in the second embodiment. The inclination of the plane on which the preceding vehicle is mounted with respect to the plane on which the host vehicle is mounted is “a”. The lower end of the preceding vehicle is the intersection of (i) the straight line connecting the optical center of the camera of the own vehicle and the lower end, and (ii) the road surface on which the preceding vehicle is mounted. (I) A straight line connecting the optical center and the lower end of the camera of the vehicle is represented by the following equation (1), where f is the focal length of the camera, and (ii) the road surface is represented by the following equation (2).

Solving this for Z,

Becomes The observation error at the bottom of the preceding vehicle

Then, the scale of the error of the inter-vehicle distance is expressed by the following equation (4) by the first-order Taylor expansion.

In the inter-vehicle distance measuring device 2 of the second embodiment, the storage unit 17 stores the formula of the error scale represented by the formula (4). The error scale calculation unit 16 reads the error scale formula from the storage unit 17 and calculates the error scale.

In the inter-vehicle distance measuring device 2 according to the second embodiment, the configuration of the distance calculation unit 12 that calculates the inter-vehicle distance to the preceding vehicle based on the camera image and the position information from the GPS/IMU 11, and the inter-vehicle distance and the error scale. The configuration of the output unit 15 that outputs is the same as that of the inter-vehicle distance measuring device 1 of the first embodiment.

The inter-vehicle distance measuring apparatus 2 of the second embodiment can obtain not only the inter-vehicle distance to the preceding vehicle but also the scale of the error thereof, as in the inter-vehicle distance measuring apparatus 1 of the first embodiment. Further, since the error scale can be calculated without performing the simulation to generate the error model, the inter-vehicle distance measuring device 2 can be easily configured.

In the present embodiment, an example in which the pinhole camera model is adopted has been described, but it is possible to obtain the error scale in a camera other than the pinhole camera in the same manner as above.

(Third Embodiment)
In the third embodiment, a learning model generation device 3 that generates a learning model used in a device that estimates an inter-vehicle distance from an image captured by a camera will be described. In order to generate a learning model, it is necessary to prepare a set of correct answer data for the vehicle image and the distance to the preceding vehicle in the camera image as teacher data. Generally, prepare a large amount of teacher data. Is not easy. The learning model generation device 3 according to the third embodiment generates a learning model using a video image of the front taken while traveling on a road and the position information of the own vehicle corresponding to each frame of the video image.

FIG. 13 is a diagram showing the configuration of the learning model generation device 3 according to the third embodiment. The learning model generation device 3 includes an input unit 30, a distance calculation unit 31, an error scale calculation unit 32, an error model storage unit 33, a model generation unit 34, and a learning model storage unit 35. The input unit 30 receives an input of a video image of the front taken when the vehicle travels on a road and position information of the host vehicle corresponding to each frame of the video image. The distance calculation unit 31 and the error scale calculation unit 32 have the same configurations as the distance calculation unit 31 and the error scale calculation unit 32 described in the vehicle measurement device according to the first embodiment, and the preceding vehicle shown in each frame of the moving image. Find the inter-vehicle distance to. As a result, the teacher data including the image of each frame of the moving image, the inter-vehicle distance and the error scale thereof is generated.

The model generation unit 34 generates a learning model using the image of the preceding vehicle, the inter-vehicle distance and its error scale as teacher data, and stores the generated learning model in the learning model storage unit 35. If the estimator having the learning parameter θ is represented by the function f _θ (·), the input variable is x, and the target value of the inter-vehicle distance is Z, the following formulation as a regression problem is possible.

Although the inter-vehicle distance calculated by the distance calculating unit 31 is treated as a correct answer, the inter-vehicle distance calculated by the distance calculating unit 31 includes an error as described above. The model generation unit 34 performs learning by reducing the influence on the loss function as the error scale increases. Specifically, in the present embodiment, learning is performed using the following equation so that the inter-vehicle distance having a large error scale is not considered important.

FIG. 14 is a flowchart showing the operation of the learning model generation device 3 according to the third embodiment. The learning model generation device 3 receives the input of the moving image captured by the vehicle traveling and the position information in each frame (S30). The learning model generation device 3 identifies the preceding vehicle in each frame, and calculates the inter-vehicle distance to the preceding vehicle and the error scale (S31). This processing is the same as the operation of measuring the inter-vehicle distance in the inter-vehicle distance measuring apparatus 1 of the first embodiment (see FIG. 6). As a result, the learning model generation device 3 generates teacher data. The learning model generation device 3 generates a learning model using the generated teacher data (S32).

The learning model generation device 3 according to the third embodiment calculates the inter-vehicle distance corresponding to the image by using the moving image and the position information corresponding to the moving image, and uses the calculated inter-vehicle distance as the teacher data. The cost can be reduced significantly.

Moreover, since learning is performed using not only the data of the inter-vehicle distance but also the scale of the error, a highly accurate learning model can be generated. By using the learning model generated in this way, when estimating the inter-vehicle distance, the inter-vehicle distance can be estimated immediately from the image showing the preceding vehicle.

In addition, in the third embodiment, an example in which the method described in the inter-vehicle distance measuring apparatus 1 of the first embodiment is adopted for calculation of the inter-vehicle distance and the error scale has been described, but the second embodiment is described. Of course, it is also possible to adopt the method described in the inter-vehicle distance measuring device 2 of FIG.

Although the inter-vehicle distance measuring device, the error model generating device, and the learning model generating device of the present invention have been described above, the present invention is not limited to the above-described embodiments.

In the above-described embodiment, an example in which the vehicle position information is acquired from the GPS/IMU has been described, but the vehicle position information may be measured by Visual Odometry as described in Patent Document 1.

INDUSTRIAL APPLICABILITY The present invention is useful as an inter-vehicle distance measuring device that measures an inter-vehicle distance from a preceding vehicle.

1 and 2 distance measurement device 3 learning model generation device 10 camera 11 GPS/IMU
12 distance calculation unit 13, 16 error scale calculation unit 14, 17 error model storage unit 15 output unit 20 error model generation device 21 road structure information input unit 22 simulation unit 23 vehicle placement unit 24 true value calculation unit 25 error setting unit 26 distance Calculation unit 27 Error scale calculation unit 28 Error analysis unit 29 Error model storage unit 30 Input unit 31 Distance calculation unit 32 Error scale calculation unit 33 Error model storage unit 34 Model generation unit 35 Learning model storage unit

Claims (13)

A camera mounted on the vehicle,
A position information detection unit that detects the position information of the vehicle,
The lower end of the preceding vehicle is identified in the image of the preceding vehicle taken when the vehicle is in the first position, and a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. By determining the shape of the road on the basis of the position information that is sequentially obtained, the second position where the straight line intersects the road is obtained, and the preceding vehicle when the vehicle is at the first position is obtained. A distance calculation unit that calculates the following distance between
The error in which the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera between the first position and the second position are variables. An error model storage unit that stores a scale model,
Referring to the model of the error scale stored in the error model storage unit, based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera, An error scale calculator that calculates the scale of the distance error,
An output unit that outputs the inter-vehicle distance and the scale of the error,
An inter-vehicle distance measuring device equipped with. The error model storage unit stores a look-up table in which the scale in the inter-vehicle distance error is associated with the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera. The inter-vehicle distance measuring device according to claim 1, wherein the model of the error scale is stored in a format. A camera mounted on the vehicle,
A position information detection unit that detects the position information of the vehicle,
The lower end of the preceding vehicle is identified in the image of the preceding vehicle taken when the vehicle is in the first position, and a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. By determining the shape of the road on the basis of the position information that is sequentially obtained, the second position where the straight line intersects the road is obtained, and the preceding vehicle when the vehicle is at the first position is obtained. A distance calculation unit that calculates the following distance between
Error scale calculation for calculating a scale of error based on the inclination of the road surface at the second position with respect to the road surface at the first position and the length in the image from the optical center of the camera to the lower end of the preceding vehicle Department,
An output unit that outputs the inter-vehicle distance and the scale of the error,
An inter-vehicle distance measuring device equipped with. When the vehicle is in the first position, the lower end of the preceding vehicle is identified in the image of the preceding vehicle, a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. Thus, the shape of the road is obtained based on the position information that is sequentially obtained, and the second position at which the straight line intersects with the road is obtained, whereby the vehicle with the preceding vehicle when the vehicle is at the first position is obtained. In an inter-vehicle distance measuring device that obtains an inter-vehicle distance, a device for generating an error model for estimating an error in the inter-vehicle distance,
A road structure information input section for inputting road structure information;
The first position and the position of the preceding vehicle are randomly determined, the true value of the inter-vehicle distance to the preceding vehicle is calculated using the road structure information, and the rotation angle of the camera determined by the road structure, the vehicle The error generated at random is sequentially added to the true value of the position, and the inter-vehicle distance is calculated by obtaining the second position, and the error with respect to the true value of the calculated inter-vehicle distance is calculated. The simulation part,
An error analysis unit that analyzes the scale of the error in the inter-vehicle distance, with the variable length and displacement of the pitch angle of the camera in the image from the optical center of the camera to the lower end of the preceding vehicle,
An error model generation device comprising: A generation device for generating a learning model for estimating an inter-vehicle distance from an image captured by a camera,
An input unit for inputting an image of the preceding vehicle and vehicle position information corresponding to each frame of the image,
The lower end of the preceding vehicle is identified in the image of the preceding vehicle when the vehicle is at the first position, a straight line connecting the optical center of the camera and the lower end in the world coordinate system is obtained, and the vehicle travels. By determining the shape of the road on the basis of the position information that is sequentially obtained, the second position where the straight line intersects the road is obtained, and the preceding vehicle when the vehicle is at the first position is obtained. A distance calculation unit that calculates the following distance between
The error in which the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera between the first position and the second position are variables. An error model storage unit that stores a scale model,
Referring to the model of the error scale stored in the error model storage unit, based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera, An error scale calculator that calculates the scale of the distance error,
A model generation unit that generates a learning model for obtaining the inter-vehicle distance from the image, using the inter-vehicle distance and the scale of the error obtained for the plurality of images as teacher data,
A learning model generation device including. A generation device for generating a learning model for estimating an inter-vehicle distance from an image captured by a camera,
An input unit for inputting an image of the preceding vehicle and vehicle position information corresponding to each frame of the image,
When the vehicle is in the first position, the lower end of the preceding vehicle is identified in the image of the preceding vehicle, a straight line connecting the optical center of the camera and the lower end in the world coordinate system is obtained, and the vehicle travels. Thus, the shape of the road is obtained based on the position information that is sequentially obtained, and the second position at which the straight line intersects with the road is obtained, whereby the vehicle with the preceding vehicle when the vehicle is at the first position is obtained. A distance calculation unit that obtains an inter-vehicle distance,
Error scale calculation for calculating a scale of error based on the inclination of the road surface at the second position with respect to the road surface at the first position and the length in the image from the optical center of the camera to the lower end of the preceding vehicle Department,
A model generation unit that generates a learning model for obtaining the inter-vehicle distance from the image, using the inter-vehicle distance and the scale of the error obtained for the plurality of images as teacher data,
A learning model generation device including. The learning model generation device according to claim 5 or 6, wherein the model generation unit performs learning by reducing the influence on the loss function as the scale of the error increases. A method for measuring an inter-vehicle distance with a preceding vehicle based on an image taken by a camera mounted on the vehicle and position information of the vehicle,
A step of inputting an image taken by the camera to the inter-vehicle distance measuring device,
Detecting the position information of the vehicle in the inter-vehicle distance measuring device,
The inter-vehicle distance measuring device identifies the lower end of the preceding vehicle in the image of the preceding vehicle when the vehicle is at the first position, and defines a straight line connecting the optical center of the camera and the lower end in the world coordinate system. In addition to the above, the shape of the road is obtained based on the position information sequentially obtained by the traveling of the vehicle, and the second position where the straight line intersects the road is obtained, so that the vehicle is located at the first position. The step of obtaining the inter-vehicle distance with the preceding vehicle when there is,
The inter-vehicle distance measuring device has a length in the image from the optical center of the camera to the lower end of the preceding vehicle, and a displacement of the pitch angle of the camera between the first position and the second position. With reference to an error model storage unit that stores a model of error scale with and as variables, based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera. A step of calculating a scale of the error of the inter-vehicle distance,
The inter-vehicle distance measuring device outputs the inter-vehicle distance and an error scale,
An inter-vehicle distance measuring method including. A method for measuring an inter-vehicle distance with a preceding vehicle based on an image taken by a camera mounted on the vehicle and position information of the vehicle,
A step of inputting an image taken by the camera to the inter-vehicle distance measuring device,
Detecting the position information of the vehicle in the inter-vehicle distance measuring device,
The inter-vehicle distance measuring device identifies the lower end of the preceding vehicle in the image of the preceding vehicle when the vehicle is at the first position, and defines a straight line connecting the optical center of the camera and the lower end in the world coordinate system. In addition to the above, the shape of the road is obtained based on the position information sequentially obtained by the traveling of the vehicle, and the second position where the straight line intersects the road is obtained, so that the vehicle is located at the first position. The step of obtaining the inter-vehicle distance with the preceding vehicle when there is,
The inter-vehicle distance measuring device detects an error based on the inclination of the road surface at the second position with respect to the road surface at the first position and the length in the image from the optical center of the camera to the lower end of the preceding vehicle. A step of calculating a scale,
The inter-vehicle distance measuring device outputs the inter-vehicle distance and an error scale,
An inter-vehicle distance measuring method including. When the vehicle is in the first position, the lower end of the preceding vehicle is identified in the image of the preceding vehicle, a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. Thus, the shape of the road is obtained based on the position information that is sequentially obtained, and the second position at which the straight line intersects with the road is obtained, whereby the vehicle with the preceding vehicle when the vehicle is at the first position is obtained. A method for generating an error model for estimating an error of the inter-vehicle distance in an inter-vehicle distance measuring device for obtaining an inter-vehicle distance,
Inputting road structure information to the error model generator,
The error model generation device randomly determines the first position and the position of the preceding vehicle, calculates the true value of the inter-vehicle distance to the preceding vehicle using the road structure information, and An error generated at random is sequentially added to the determined rotation angle of the camera and the true value of the vehicle position, and the inter-vehicle distance is calculated by obtaining the second position. Calculating the error with respect to the true value,
The error model generation device, the displacement of the pitch angle of the camera and the length in the image from the optical center of the camera to the lower end of the preceding vehicle in the image as a variable, analyzing the scale of the error of the inter-vehicle distance, ,
An error model generation method comprising: A program for measuring an inter-vehicle distance to a preceding vehicle on the basis of an image taken by a camera mounted on the vehicle and position information of the vehicle, the computer including:
Inputting an image taken by the camera,
Inputting position information of the vehicle,
The lower end of the preceding vehicle is identified in the image of the preceding vehicle taken when the vehicle is in the first position, and a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. By determining the shape of the road on the basis of the position information that is sequentially obtained, the second position where the straight line intersects the road is obtained, and the preceding vehicle when the vehicle is at the first position is obtained. The step of obtaining the inter-vehicle distance of
The error in which the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera between the first position and the second position are variables. Referring to an error model storage unit that stores a model of the scale, based on the length in the image from the optical center of the camera to the lower end of the preceding vehicle and the displacement of the pitch angle of the camera, the error of the inter-vehicle distance A step of calculating a scale,
Outputting the inter-vehicle distance and error scale,
A program to execute. A program for measuring an inter-vehicle distance to a preceding vehicle on the basis of an image taken by a camera mounted on the vehicle and position information of the vehicle, the computer including:
Inputting an image taken by the camera,
Inputting position information of the vehicle,
The lower end of the preceding vehicle is identified in the image of the preceding vehicle taken when the vehicle is in the first position, and a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. By determining the shape of the road on the basis of the position information that is sequentially obtained, the second position where the straight line intersects the road is obtained, and the preceding vehicle when the vehicle is at the first position is obtained. The step of obtaining the inter-vehicle distance of
Calculating a scale of the error based on the inclination of the road surface at the second position with respect to the road surface at the first position and the length in the image from the optical center of the camera to the lower end of the preceding vehicle,
Outputting the inter-vehicle distance and error scale,
A program to execute. When the vehicle is in the first position, the lower end of the preceding vehicle is identified in the image of the preceding vehicle, a straight line connecting the optical center of the camera and the lower end is obtained in the world coordinate system, and the vehicle travels. Thus, the shape of the road is obtained based on the position information that is sequentially obtained, and the second position at which the straight line intersects with the road is obtained, whereby the vehicle with the preceding vehicle when the vehicle is at the first position is obtained. In a vehicle-to-vehicle distance measuring apparatus that obtains a vehicle-to-vehicle distance, a program for generating an error model for estimating an error of the vehicle-to-vehicle distance,
The step of inputting road structure information,
The first position and the position of the preceding vehicle are randomly determined, the true value of the inter-vehicle distance to the preceding vehicle is calculated using the road structure information, and the rotation angle of the camera determined by the road structure, the vehicle The error generated at random is sequentially added to the true value of the position, and the inter-vehicle distance is calculated by obtaining the second position, and the error with respect to the true value of the calculated inter-vehicle distance is calculated. Steps,
Displacement of the pitch angle of the camera and the length in the image from the optical center of the camera to the lower end of the preceding vehicle as a variable, analyzing the scale of the error of the inter-vehicle distance,
A program to execute.

Images