JP4533659B2 - Apparatus and method for generating map image by laser measurement - Google Patents

Description

  The present invention automatically generates and / or generates an image of the position of an obstacle in the environment around the moving sensor by measuring while moving a laser distance sensor capable of measuring the azimuth and distance to the obstacle. The present invention relates to a display apparatus and method.

As a method of measuring the environment around the laser distance sensor multiple times while moving the laser distance sensor and generating a map of the environment from the value of each measurement as an image, measurement of a certain area obtained by one measurement The displacement between the measured values that minimizes the evaluation value is calculated by using the square sum of the difference between the value and the measurement value of a certain area obtained by another measurement as an evaluation value (so-called A method of generating a map-like image by using this as the moving distance of the sensor and recording the measured values of each time in a common coordinate system according to this moving distance is proposed (by the least square method calculation method) ( For example, see Non-Patent Documents 1 and 2.)
Yoshiro Negishi, Jun Miura, Yoshiaki Shirai, “Map Generation for Mobile Robots by Integration of Omnidirectional Stereo and Laser Range Finder”, Journal of the Robotics Society of Japan, Vol.21, No.6, pp.110-116, 2003. D. Hahnel, W. Burgard, “Probabilistic Matching for 3D Scan Registration”, In Proc of the VDI-Conference Robotik 2002, 2002.

  However, when the displacement between measured values before and after the movement of the laser distance sensor in the prior art as described in Non-Patent Documents 1 and 2 described above is evaluated by the least square method calculation method, an optimum solution is always obtained. As a result, there has been a problem that a local solution may be obtained as a result. That is, there is an inconvenience that the matching of the obstacle in the environment evaluated by the measured values before and after the sensor movement is not accurately performed by the local solution (the matching of a certain obstacle does not match before and after the sensor movement).

  An object of the present invention is to provide a map image generation device that suppresses the generation of errors when accurately calculating the relative position and orientation between images before and after the movement of a sensor and generates or displays a map image with high accuracy. It is in.

In order to solve the above problems, the present invention mainly adopts the following configuration.
An image generation device that generates a map image around the distance sensor using an azimuth and distance to an obstacle,
Move the distance sensor to obtain the direction and distance measurements to the obstacle before and after the movement,
Based on the measurement value, the presence / absence of an obstacle around the distance sensor is assigned to each grid determined by the resolution of the distance sensor (the obstacle is occupied in the image), and is not occupied (the obstacle in the image). Is unoccupied) and unknown (it is unknown whether the obstacle is occupied in the image)
In each grid image obtained with the measurement values, a window indicating a certain range is provided around the grid occupied by the obstacle, and the state of the grid in the window is expressed as occupied, unoccupied, or unknown. the frequency of occurrence of each state upon when expressed in a histogram, the presence of the grid most obstacle occupancy extracts the window grids of the unknown states is small as feature points,
In current corresponding comparative of the feature points between the two grid image of the grid image obtained by the measurement of the grid image and the previous obtained by the measurement, the occupancy grid in the feature points of each grid image the extracted the distance between the several feature points by comparing between the grid image, and collating the feature points between the grid image,
Based on the result of the collation, the grid images are temporarily arranged by overlapping between the best matching grid images of the feature points,
The map image is generated by calculating the relative position and orientation between the grid images based on the measurement values between the temporarily arranged grid images.

An image generation method for generating, as an image, a map indicating the presence or absence of an obstacle in the environment from a plurality of measurement values obtained by a laser distance sensor,
An acquisition step of acquiring a measured value of the azimuth and distance to the obstacle by a laser;
Occupancy of obstacles around the laser distance sensor based on the measurement value acquired from the acquisition step, and assigned to each grid determined by the resolution of the distance sensor (obstacles are occupied in the image) Generating a grid image represented by unoccupied (unoccupied obstacles in the image), unknown (unknown whether the obstacles are occupied in the image),
In each grid image obtained in the generation step, a window indicating a certain range is provided around the grid occupied by the obstacle, and the state of the grid in the window is expressed as occupied, unoccupied, or unknown. the frequency of occurrence of each state at the time of the time expressed in histogram, an extraction step of the presence or absence of grid most obstacle occupancy extracts the grid is small window of unknown states as feature points,
In a corresponding comparison of the feature points between the two grid image of the grid image obtained by the measurement of the grid image and the previous obtained by the current measurement, in the feature points of each grid image obtained by the extraction step the distance between the feature points and the number of occupied grid and collates and compared between each grid image, an identification step for identifying the feature point matching between each grid image,
The temporary disposed by overlapping each grid image to a common coordinate system in accordance with the relative position and orientation at which the feature points between the obtained verification result based grid image best matches with the identification process, are temporarily placed A calculation step of calculating an overlapping area between the grid images;
A calculation step of calculating a relative position and orientation between grid images using only the measurement value of the laser distance sensor included in the overlap region obtained in the calculation step;
An image generation method comprising: a generation step of combining a plurality of images into one image according to the relative position and orientation between the grid images obtained in the calculation step.

  According to the present invention, common feature points are extracted and collated in an image showing an area occupied by an obstacle (occupied grid) in an environment obtained for each measurement by a moving laser distance sensor. It is possible to automatically synthesize and / or display as an image of a single environmental map.

  More specifically, the feature points of the histogram are extracted from each of a plurality of images generated for each measurement, and the histograms are compared with each other by comparing the histograms with their relative positions. Measure the position and orientation, calculate the approximate overlap area between the images obtained from this relative position and orientation, and use only the orientation and distance measurement values included in the overlap area to measure between the images It is possible to calculate the relative position and orientation between images with higher accuracy, and to automatically generate and / or display a highly accurate environmental map image.

  Embodiments of an image generation apparatus for generating a map image according to the present invention will be described below in detail with reference to the accompanying drawings. The outline of the present invention is to generate and / or display a grid image for each measurement for the purpose of accurately generating and / or displaying a map indicating the presence or absence of an obstacle in the surrounding environment to which the laser distance sensor has moved. Pre-processing to detect the points and determine the relative position and orientation between the images (specifically, provisional alignment between the grid images before and after the movement of the sensor) Suppresses the occurrence of errors when the relative position and orientation are accurately calculated (for example, prevents a drop in a local solution 904 (not the optimal solution 903) shown in FIG. 9 described later), and generates an environmental map image with high accuracy. And / or display.

  An embodiment according to the map image generating apparatus of the present invention will be described with reference to the drawings. Here, in particular, a laser distance sensor capable of measuring the azimuth and distance to obstacles by scanning a horizontal plane is mounted on a wheel-moving robot, moved, and measured multiple times. An apparatus that generates and / or displays as an image the presence or absence of an obstacle in the environment where the sensor has moved will be described as an example.

  FIG. 1 shows an outline of the flow of processing in an embodiment according to the map image generating apparatus of the present invention. When the processing is started (101), processing for generating and / or displaying an environment map from the laser distance sensor (102 to 109) and processing for robot control (111) are started. In the map generation and / or display processing by the laser distance sensor, first, the direction and distance measurement values are input by controlling the laser distance sensor (102), and the direction of the obstacle in the environment is determined from the input measurement values. And a grid image (an entire image shown by a grid shown in FIG. 5 described later), which is an image indicating the distance, is generated (103).

  Next, a point having a large area occupied by obstacles (occupied grid) in the image and a small area unknown to the occupied area (unknown grid) is extracted as a feature point (104), and a histogram of the feature points between the images is extracted. And the relative position of the feature point is held between the images as a reference, and the relative position and orientation of the image that holds the most feature point among the images among the matching results are recorded ( 105) An overlapping area is calculated when the images are arranged in a common coordinate system according to the relative position and orientation (106).

  Further, the relative position and orientation between the images are precisely calculated by using the measured values of the orientation and distance of each image included in the overlapping region (107), and each image is calculated according to the calculated relative position and orientation between the images. An image of the environmental map is generated by arranging the images in a common coordinate system (108), and the generated environmental map image is displayed (109).

  The robot moves along the route designated by the operator or in advance by the robot control (111) process in parallel with the process (102 to 109) for generating and / or displaying the map of the environment from the above laser distance sensor. . The program is terminated after generation and / or display of the target environment map is completed (110).

  Based on the outline of the processing flow described above, FIG. 2 shows a more detailed processing flow of FIG. 2 in FIG. 1, 103 in FIG. 1 in 203 in FIG. 2, 104 in FIG. 1 in 204 to 206 in FIG. 2, 105 in FIG. 1 in 207 in FIG. 2, 106 in FIG. 2 corresponds to 208 to 209 in FIG. 2, 107 in FIG. 1 corresponds to 210 in FIG. 2, 108 in FIG. 1 corresponds to 211 in FIG. 2, and 109 in FIG. 1 corresponds to 212 in FIG.

  FIG. 3 shows details of the process 207 in FIG. Further, FIG. 4 shows how the robot moves in the environment and scans the direction and position of the obstacle in the environment by the laser distance sensor in performing the above processing.

  In FIG. 2, when the program is started (201), the azimuth and distance from the sensor to the obstacle are measured by the laser distance sensor, and the measured value is taken in (202). Here, at the position (405) shown in FIG. 4, the robot scans by a laser distance sensor in the direction of 180 degrees as shown by (404), and in the environment shown by the hatched area of (401). Assuming that the range indicated by the bold line in the figure is detected as an obstacle for the obstacle as shown in (402), the robot moves to the position of (403) and repeats the same measurement, and the map Is displayed as an image. In the illustrated example shown in FIG. 4, the obstacles at the top are not marked with a thick line because they are remote from the sensor.

  A grid image is generated from the measurement values captured by the laser distance sensor on the moving robot in this way (203). The measured value here is assumed to be a set of distances to obstacles in each direction obtained by scanning a range of a certain angle as indicated by (404). Based on this measurement value, a grid image (one grid shown in FIG. 5) determined based on the measurement error of the sensor is set in the coordinate system based on the sensor position, and measurement is performed on this grid. It is generated by assigning three kinds of colors indicating the state of the grid estimated from the values.

  FIG. 5 shows an example of processing (203) of grid image generation (103). FIG. 5 shows an obstacle measured within a scanned range by scanning with a laser beam (505) in a range of 180 degrees on a horizontal plane for one measurement in a state where a sensor is arranged at a position (507). The image is based on the azimuth and distance. In FIG. 5, it is assumed that (501) is an actual measurement value (direction and distance in a straight line 505 connecting 507 and 501).

  Each grid of the grid image has a ternary state, and has values of an occupied grid (503), an unoccupied grid (504), and an unknown grid (502). The occupied grid indicates an area where it is known that there is an obstacle. The unoccupied grid is an unoccupied grid on the assumption that there is no obstacle in the grid on the straight line (505) connecting the sensor position and the occupied grid. The grid in the direction away from the sensor on the straight line connecting the sensor position and the occupied grid is set as an unknown grid because it is an unknown area where the presence or absence of an obstacle is not known. The size of the grid is determined from the measurement error of the sensor. Here, the measurement error of the sensor is assumed to be ± d in the front-rear and left-right directions, and the size of one grid is 2d × 2d. As a result, it is assumed that collation is performed in consideration of errors when collating feature points of an image, which will be described later. Each grid holds one of three values (occupied, non-occupied, unknown) and an actual measurement value (direction and distance) of the sensor indicated by (501).

  Next, grid image feature points are extracted (104). FIG. 5 shows an example of processing (204 to 205) of grid image feature point extraction (104). In the grid image generated from the measurement value input by the input of the laser distance sensor value (102), a window (506) centering on the occupied grid is provided, and a histogram (occupied, unoccupied, unknown) is displayed. The numerical table shown) is created, and this is used as a feature point. The feature points are extracted by comparing the histograms of the feature points, selecting feature points with many occupied grids (204), and selecting feature points with few unknown grids (205). Next, a histogram (table shown in FIG. 6 described later) and coordinates of each extracted feature point are recorded (206).

  FIG. 6 shows a tabulated value of the histogram of recorded feature points. A value indicating the appearance frequency of the grid state of each feature point when 1 to 5 are assigned as unique identifiers to the feature points is shown as an example.

  Next, feature point matching (105) of the grid image is performed. The feature point matching of the grid image is to calculate the corresponding feature point between two grid images of the grid image obtained by the current measurement and the grid image based on the previous measurement, and is performed by the process shown in FIG. Done. Here, as an example of the process of matching feature points (105) of the grid image, the process of FIG. 3 will be described with reference to FIG. 7A is a grid image when the environment is measured at the position (405) in FIG. 4, and FIG. 7B is a grid when measured at the position (403) in FIG. An example of an image is shown. It is assumed that a robot equipped with a laser distance sensor cannot recognize the absolute position of itself, and a map can be generated if a landscape that can be observed by the robot before and after the robot moves is drawn.

  First, two of the feature points extracted in each grid image are selected, and a coordinate system (701) passing through these feature points is set (301). The coordinates of other feature points in this coordinate system are obtained (302). Similarly, a reference coordinate system is set for other grid images (702), and the coordinates of other feature points are obtained. Next, it is considered to arrange the feature points of each grid image so that the orientation coincides with the origin of the reference coordinate system (303). At this time, the histogram and the relative position of the feature points of each grid image are compared (304). More specifically, a feature point (pointing to the window shown in FIG. 7) whose difference in the number of occupied grids in the histogram is equal to or smaller than a threshold is selected, and the relative distance (relative distance of the center grid position of the window) is equal to or smaller than the threshold. Only in the case of, it is counted that the feature points match. This process is performed for all combinations of coordinate systems including two feature points (305).

  The coordinate system selected when the matching rate of the feature points is the highest in the process of matching the feature points of the grid image (105) is recorded (306), and the origin and axis directions of the coordinate systems of each image are determined. Each image is arranged in the same coordinate system so as to coincide (superimpose coordinate systems 701 and 702 shown in FIG. 7) (208). The arrangement by the process (208) is a pre-process (temporary alignment process) for precisely calculating the relative position and orientation between images by the process (210) described later, and the relative position / attitude by the process (207). Is not adopted as the final relative position / posture between images (the final inter-image relative position / posture is the value of the processing (210) described later), and the processing (208). The image arrangement is a temporary arrangement (temporary alignment) for performing the processing (209).

  The overlapping area between the images temporarily arranged in the same coordinate system by the process (208) is calculated (209). FIG. 8 shows an image temporarily arranged by the processing (208) and its overlapping area (801). The relative position / orientation between the images is precisely calculated based on the actually measured values of the azimuth and the distance corresponding to each occupied grid included in the overlapping area (801). Here, the relative position and orientation between images are calculated by the Iterative Closest Point (ICP) (210).

  In FIG. 9, when an error evaluation function (902) is assumed in a coordinate system (901) in which the horizontal axis is the position / posture solution and the vertical axis is the error, the position / posture solution is simply obtained by the least square method or the like. However, if the grid images are temporarily arranged as described above and only the value of the overlapping region is used (see FIG. 8), the local solution (904) may be calculated as a solution. By not adopting the measured values of the windows other than the overlapping region 801 shown), it is easier to obtain the optimal position / posture solution (903) between the images by thinning out the extra measured values. In other words, if an overlapping area is temporarily secured, a solution is obtained at point P on the left side of point O on the horizontal axis in FIG. 9, and the optimum solution (903) is obtained without falling into the local solution (904). obtain.

  Next, based on the relative positions and orientations between the images calculated by the process (210), one image is generated so that the images overlap each other, and this is used as the environment map image (211). Subsequently, the generated environment map image is displayed (212). If there is no new sensor value input (213), the program ends (213).

  Next, the structure of the Example which concerns on the map image generation apparatus of this invention is demonstrated using FIG. In this embodiment, a main processor (1001) for processing data based on a program, a main memory (1009) such as a memory for holding the program and data, an auxiliary memory (1008) such as a hard disk, a display (1002) and a graphic A board (1003), a laser distance sensor (1004) for inputting a measurement value of an azimuth / distance to an obstacle (102), a laser distance sensor control board (1005), and a robot serving as a carriage for moving the sensor (1006) and a robot control board (1007).

  The configuration shown in FIG. 10 can be realized in terms of hardware by a processor, memory, and other LSIs, and in terms of software, it is realized by a program loaded in the memory. The functional blocks to be performed are described. Therefore, these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.

  The program and data that operate in the main memory (1009) will be described. The operating system (1011) is basic software for managing execution of programs and the like. As a program for generating an environmental map image from the measured values of azimuth and distance, a main program (1012) for controlling the overall processing flow, and a laser distance sensor corresponding to the processing of laser distance sensor value acquisition (102) in FIG. A value acquisition program (1013), a grid image generation program (1014) corresponding to the processing of grid image generation (103), a grid image feature point extraction program (1015) corresponding to the processing of feature point extraction (103) of the grid image, Grid image feature point matching program (1016) corresponding to grid image feature point matching (105) processing, grid image overlapping region calculation program (1017) equivalent to processing of grid image overlapping region calculation (106), image Inter-image relative corresponding to the processing of relative position and orientation calculation (107) Position / attitude calculation program (1018), environment map image generation program (1019) corresponding to processing of environment map image generation (108), environment map image display program (1020) corresponding to processing of environment map image display (109) Exists.

  In addition, there is an image buffer (1021) that holds processing results in image processing. The auxiliary storage (1008) stores various programs (1011 to 1022) read into the main storage (1009) as programs (1010). These programs are read into a memory and processed by being executed by a processor.

  The processing of each program will be described. The main program (1012) controls the processing flow of the entire program from acquisition of the laser distance sensor value shown in FIGS. 1 to 9 to generation and / or display of the environmental map image. The laser distance sensor value acquisition program (1013) controls the laser distance sensor control board (1007) and the laser distance sensor (1004) as the laser distance sensor value acquisition (102), so that the orientation and distance obtained by the laser distance sensor are controlled. The measured value is taken into the buffer (1021) of the main memory (1009) (202 in FIG. 2). The buffer (1021) is also used as a work area for various programs described below.

  Next, the grid image generation program (1014) generates a grid image from the azimuth / distance measurement values acquired by the laser distance sensor value acquisition (102) (process 203 in FIG. 2), and the grid image feature point extraction program ( 1015) extracts a histogram in the window provided in the occupancy grid in the image as a feature point, and records the histogram and coordinates (processes 204 to 206 in FIG. 2).

  The grid image feature point matching program (1016) performs matching based on the comparison between the histogram of the feature points and the coordinates between the grid images (processing 207 in FIG. 2). More specifically, a coordinate system based on the two feature points selected in the grid image is set, and the coordinates of other feature points in the coordinate system and the histogram are compared between the grid images. If the difference between the histogram and the difference between the relative positions is equal to or smaller than the threshold value, it is considered that the features match and the number of matching feature points is counted. As a result, the relative position and posture relationship between the grid images with the highest matching rate are calculated.

  An inter-grid image overlap area calculation program (1017) uses a common coordinate system for each grid image according to the relative position and orientation when the matching rate of the feature points of the image obtained by the grid image feature point matching program (1016) is maximized. Are temporarily arranged, and a region overlapping between the temporarily arranged images is calculated. A relative position / orientation between grid images is accurately calculated by using a measured value of the azimuth and distance of the occupied grid of each grid image included in the overlapping area by an inter-image relative position / orientation calculation program (1018).

  The environment map image generation program (1019) arranges each grid image in a common coordinate system according to the obtained relative position and orientation, and generates an environment map image. Next, the environment map image display program (1020) displays the environment map image generated by the environment map image generation program (1019) (process 212 in FIG. 2).

  A recording medium recording a software program for realizing the functions of the above-described embodiments of the present invention is supplied to a system or apparatus, and a processor stored in the recording medium is read and executed by a processor of the system or apparatus. However, it goes without saying that the object of the present invention is achieved.

  In the above description, as an embodiment of the present invention, a configuration example in which a laser distance sensor is used as a measuring device for azimuth and distance has been shown. However, measurement values are obtained from a storage device on a network that stores measurement values. Even when it is input, it is possible to obtain the same result as in the configuration example described above. In addition, the method of moving the laser distance sensor that accompanies multiple measurements of the environment is similar when the sensor is attached to a vehicle other than a robot, or when a tripod with a sensor attached is changed by a human. The same result can be obtained if measurement is performed by changing the measurement point in the range where the characteristic is observed. In this embodiment, an example of a laser distance sensor that horizontally scans within a range of 180 degrees is shown. However, similar results can be obtained even when the scanning range is wider or narrower.

  In addition, as an example of industrial application of the map image generation device of the present invention, an obstacle is occupied from a plurality of azimuth and distance measurements taken by a laser distance sensor mounted on a moving object such as an automobile or a ship. An image showing a region to be generated, and automatically generating and / or displaying a single environmental map image, or a single environmental map image from a plurality of images obtained by arranging a plurality of laser distance sensors. It can be automatically generated and / or displayed, and labor saving can be expected. Thus, according to the embodiment of the present invention, in an environment where the arrangement of obstacles is unknown, by measuring a certain area while moving the laser distance sensor by a moving body such as human power or a vehicle, It is possible to accurately generate and display a map of the surrounding environment to which the sensor has moved.

  As described above, when the features of the present invention are summarized, the orientation and distance to the obstacle by the laser distance sensor are measured a plurality of times, and the feature points by the histogram are obtained from the image representing the position of the obstacle obtained by each measurement. By extracting and collating the feature points of each image generated by each measurement between the images before and after the measurement, the common overlapping area between the images is calculated, and the relative position between the images is calculated based on this overlapping area calculation. By calculating the posture, the image is generated and displayed as a single image showing the position of the obstacle in the environment where the laser distance sensor has moved.

  The features of the present invention will be explained in detail. Measurement values are obtained by controlling a laser distance sensor or similar device capable of measuring the azimuth and distance to an obstacle, or a storage device that already stores measurement values. Is generated, an image showing the occupied position of the obstacle in the environment is generated from the input measurement value, and a certain range (window) centered on the occupied area (occupied grid) in each image corresponding to multiple measurements ) As feature points, and the feature point histogram and position are recorded, and the relative positional relationship is maintained between images having feature points whose feature point histograms and feature points have similar relative positions. Only the feature points that are present, calculate the approximate relative position and orientation between images with the same feature point, and the images at the calculated relative position and orientation When overlapping, the overlapping area is calculated, and only the occupied area (occupied grid) in the image included in the calculated overlapping area is used, and the square of the difference in the distance between the occupied areas (occupied grid) in each image is minimized. The relative position and orientation are precisely calculated, and the relative position and orientation are used as the relative position and orientation between the images to superimpose the images to generate and display the environment map image.

It is a figure which shows the outline | summary of the flow of the process in the Example which concerns on the map image generation apparatus of this invention. It is a figure which shows the detailed flow in the whole process in a present Example. It is a figure which shows the flow of the detailed process regarding the feature point collation of a grid image among the flow of the whole process shown in FIG. It is a figure which shows a mode that the environment is scanned using the laser distance sensor mounted in the robot. It is explanatory drawing regarding the process in connection with a grid image generation, and grid image feature point extraction in the flow of the whole process shown in FIG. It is the figure which showed the histogram of the feature point of a grid image with the table | surface. It is explanatory drawing regarding a grid image feature point collation in the flow of the whole process shown in FIG. It is explanatory drawing regarding calculation of the temporary arrangement | positioning of a grid image and the overlap area | region in the flow of the whole process shown in FIG. It is explanatory drawing which shows the effect at the time of the relative position between grid images by the provisional arrangement shown in FIG. 8, and attitude | position calculation. It is a figure which shows the whole structure in the Example which concerns on the map image generation apparatus of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Program start process 102 Laser distance sensor value acquisition process 103 Grid image generation process 104 Grid image feature point extraction process 105 Grid image feature point collation process 106 Grid image overlap area calculation process 107 Grid image relative position / posture calculation process 108 Environmental map image generation processing 109 Environmental map image display processing 110 Program end processing 401 Obstacles in environment 402 Obstacles detected 403 Robot position after movement 404 Scanning range 405 Robot position before movement 501 Actual measurement value 502 Unknown Grid 503 Occupied grid 504 Non-occupied grid 505 Laser beam 506 Window 507 Sensor placement position

Claims (4)

An image generation device that generates a map image around the distance sensor using an azimuth and distance to an obstacle,
Move the distance sensor to obtain the direction and distance measurements to the obstacle before and after the movement,
Based on the measurement value, the presence / absence of an obstacle around the distance sensor is assigned to each grid determined by the resolution of the distance sensor (the obstacle is occupied in the image), and is not occupied (the obstacle in the image). Is unoccupied) and unknown (it is unknown whether the obstacle is occupied in the image)
In each grid image obtained with the measurement values, a window indicating a certain range is provided around the grid occupied by the obstacle, and the state of the grid in the window is expressed as occupied, unoccupied, or unknown. When the frequency of appearance of each state is represented by a histogram, a window with many occupied grids and few grids with unknown obstacles is extracted as a feature point,
When comparing the corresponding feature points between the two grid images of the grid image obtained by the current measurement and the grid image obtained by the previous measurement, the occupied grid at the feature point of each extracted grid image By comparing the number and distance between the feature points between the grid images, the feature points between the grid images are collated,
Based on the result of the collation, the grid images are temporarily arranged by overlapping between the best matching grid images of the feature points,
An image generation apparatus, wherein the map image is generated by calculating a relative position and orientation between the grid images based on the measurement values between the temporarily arranged grid images. An image generating device that generates an image of a map around the laser distance sensor from a measurement value of a laser distance sensor capable of measuring an azimuth and distance to an obstacle,
Laser distance sensor value acquisition means capable of acquiring a measured value of the direction and distance to the obstacle,
Based on the measurement value acquired from the laser distance sensor value acquisition means, the presence or absence of an obstacle around the distance sensor is determined by the occupancy assigned to each grid determined by the resolution of the distance sensor (the obstacle in the image). Grid image generating means for generating a grid image represented by states of (occupied), non-occupied (unoccupied by obstacles in the image), unknown (unknown whether the obstacle is occupied in the image),
In each grid image obtained with the measurement values, a window indicating a certain range is provided around the grid occupied by the obstacle, and the state of the grid in the window is expressed as occupied, unoccupied, or unknown. Grid image feature point extraction means for extracting, as a feature point, a window having a large number of occupied grids and a small number of grids in which the presence or absence of obstacles is unknown when the appearance frequency of each state is represented by a histogram,
When comparing the corresponding feature points between the two grid images of the grid image obtained by the current measurement and the grid image obtained by the previous measurement, the occupied grid at the feature point of each extracted grid image A grid image feature point matching means for matching the feature points between the grid images by comparing the number and the distance between the feature points between the grid images;
Based on the matching result obtained by the grid image feature point matching means, each grid image is superimposed and placed in a common coordinate system according to the relative position and orientation when the feature points between the grid images best match, A grid image overlap area calculation means for calculating an overlap area between the temporarily arranged grid images;
Grid image relative position / attitude calculation means for calculating the relative position and attitude between grid images using only the measurement value of the laser distance sensor included in the overlap area obtained by the overlap area calculation means between the grid images;
An image generating apparatus comprising: map image generating means for combining a plurality of images into one image according to the relative position and orientation between the grid images obtained by the grid image relative position / posture calculating means. An image generation method for generating, as an image, a map indicating the presence or absence of an obstacle in the environment from a plurality of measurement values obtained by a laser distance sensor,
An acquisition step of acquiring a measured value of the azimuth and distance to the obstacle by a laser;
Occupancy of obstacles around the laser distance sensor based on the measurement value acquired from the acquisition step, and assigned to each grid determined by the resolution of the distance sensor (obstacles are occupied in the image) Generating a grid image represented by unoccupied (unoccupied obstacles in the image), unknown (unknown whether the obstacles are occupied in the image),
In each grid image obtained in the generating step, a window indicating a certain range is provided around the grid occupied by the obstacle, and the state of the grid in the window is expressed as occupied, unoccupied, or unknown. When the appearance frequency of each state is represented by a histogram, an extraction step for extracting as a feature point a window with a large number of occupied grids and a small number of grids with unknown obstacles;
When comparing the corresponding feature points between two grid images of the grid image obtained by the current measurement and the grid image obtained by the previous measurement, in the feature points of each grid image obtained in the extraction step An identification step for identifying the feature points that match between the grid images by comparing and collating the number of occupied grids and the distance between the feature points between the grid images;
Based on the collation result obtained in the identification step, each grid image is temporarily placed on a common coordinate system according to the relative position and orientation when the feature points between the grid images best match, and temporarily placed. A calculation step of calculating an overlapping area between the grid images;
A calculation step of calculating a relative position and orientation between grid images using only the measurement value of the laser distance sensor included in the overlap region obtained in the calculation step;
An image generation method comprising: a generation step of combining a plurality of images into a single image according to a relative position and orientation between grid images obtained in the calculation step. A computer program that generates an image of a map around a sensor from measurement values of a sensor capable of measuring the direction and distance to the obstacle with a laser,
An acquisition step of acquiring a measured value of the azimuth and distance to the obstacle by the laser;
Occupancy of obstacles around the laser distance sensor based on the measurement value acquired from the acquisition step, and assigned to each grid determined by the resolution of the distance sensor (obstacles are occupied in the image) Generating a grid image represented by unoccupied (unoccupied obstacles in the image), unknown (unknown whether the obstacles are occupied in the image),
In each grid image obtained in the generation step, a window indicating a certain range is provided around the grid occupied by the obstacle, and the state of the grid in the window is expressed as occupied, unoccupied, or unknown. When the appearance frequency of each state is represented by a histogram, an extraction step for extracting as a feature point a window with a large number of occupied grids and a small number of grids with unknown obstacles;
When comparing the corresponding feature points between two grid images of the grid image obtained by the current measurement and the grid image obtained by the previous measurement, in the feature points of each grid image obtained in the extraction step An identification step for identifying the feature points that match between the grid images by comparing and collating the number of occupied grids and the distance between the feature points between the grid images;
Based on the collation result obtained in the identification step, each grid image is temporarily placed on a common coordinate system according to the relative position and orientation when the feature points between the grid images best match, and temporarily placed. A calculation step of calculating an overlapping area between the grid images;
A calculation step of calculating a relative position and orientation between grid images using only the measurement value of the laser distance sensor included in the overlap region obtained in the calculation step;
An image generation program that causes the computer to execute a generation step of combining a plurality of images into a single image according to the relative position and orientation between grid images obtained in the calculation step.

Images