JP3723835B2 - Obstacle detection method on the road - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an obstacle detection device that autonomously detects an obstacle on a road such as an expressway.
[0002]
[Prior art]
Currently, traffic congestion and traffic accidents caused by obstacles on the road are frequent. Waste in traffic jams is enormous, and damage caused by traffic accidents is also becoming serious. It is extremely important to organically link sensors, information processing, and communication technologies and use them to build a system that integrates people, vehicles, and roads to improve efficiency and safety and reduce waste. is there.
[0003]
Conventionally, in order to detect the obstacle on the road as described above, the detecting device or the like has been attached to the traveling vehicle. However, only obstacles existing in the vicinity of the traveling vehicle can be detected. For example, obstacles existing 1 km away or on a curve with poor visibility cannot be detected.
[0004]
[Problems to be solved by the invention]
Accordingly, the present invention provides an obstacle detection device that solves the conventional problems as described above, can detect obstacles on the road over a wider area, and can improve road safety and efficiency. The purpose is to do.
[0005]
[Means for Solving the Problems]
In order to achieve the above object, the invention of claim 1 is a method for detecting an obstacle on a road using a millimeter wave sensor installed on a roadside, wherein the distance and speed of the object are detected by scanning with the millimeter wave sensor . A data input processing step for inputting measurement data for one scan at a time, a speed mask processing step for removing measurement data whose speed value exceeds a threshold value, and a background data free from obstacles from the measurement data. The background difference processing step to remove the part, the segmentation processing step to group the data whose distance and speed are almost the same, and the center of gravity position, average speed and width of each group are generated from the grouped measurement data The attribute generation process to be performed and whether the group obtained by the segmentation process has existed before. The history generation process that generates the history of appearance and the measurement data input when generating the history and the attribute of the history are compared to check whether there is a vehicle that is in front and not measured Occlusion check process, obstacle detection process that detects that there is an obstacle when there are always obstacle candidates in several millimeter wave sensor measurements, and background using the input measurement data A background data generation processing step for updating and generating data, and the background data generation processing step that cannot be measured when the background data is updated and generated in the background data generation processing step is not generated for the background data update generation. It is characterized by that.
[0007]
DETAILED DESCRIPTION OF THE INVENTION
An embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows an overall processing flow diagram of the obstacle detection algorithm. The obstacle detection algorithm is roughly divided into 12 processing steps. The outline of the whole process is as follows. That is, measurement data in the direction facing the distance and speed is input from the millimeter wave sensor, and an obstacle candidate is detected from the difference between this measurement data and background data without an obstacle. This process is repeated several times, and when there are always obstacle candidates, it is detected that an obstacle exists. This reduces the miss rate and false alarm rate. In addition, when detecting an obstacle candidate, it is checked whether there is a traveling vehicle or the like in front of the obstacle and copes with the influence of being shaded. After detection, the position / velocity / size of the obstacle is measured using the measurement data, and the data is transmitted to an alarm device or the like. In parallel with the obstacle detection, the background data is automatically updated using the measurement data.
[0008]
As shown in FIG. 2, the millimeter wave sensor 1 used here is composed of a single millimeter wave sensor 2, its driving device 3, a gantry 4 on which these are mounted, and a control computer 5. The millimeter wave sensor unit 2 outputs measurement data of distance, speed and direction. The driving device 3 scans the millimeter wave sensor unit 2 left and right. The control computer 5 receives distance, speed, and direction measurement data from the millimeter wave sensor unit 2 and controls the driving device 3. The control computer 5 receiving the measurement data in this way detects an obstacle using the measurement data and the obstacle algorithm shown in FIG.
[0009]
FIG. 3 shows a scene for explaining the operation status of the algorithm. As shown in FIG. 3A, this road has a guard rail 11, has three lanes, and a millimeter wave sensor 1 is installed on the road side of the road. There are traveling vehicles 12a and 12b in the center lane and the right lane, and two obstacles 13a and 13b are dropped in the left lane. The traveling vehicles 12a and 12b travel from left to right. 3 (B) and 3 (C) depict changes in the situation each time the millimeter wave sensor 1 scans, (A) is a scene with N-1 scans, and (B) is N times. Scene (C) is N + 1 scenes.
The measurement range of the millimeter wave sensor 1 is represented by a thick dotted line. The thick line is the distance measured by the millimeter wave sensor 1, and the thick arrow is the speed measured by the millimeter wave sensor 1. FIG. 4 shows an overview of background data. FIG. 5 shows only the distance and speed measured by the millimeter wave sensor 1 from FIG.
The details of each of the 12 processing steps will be described below using an assumed scene. This process is all performed on the control computer 5.
[0010]
(1) Data input processing step The measurement data of the millimeter wave sensor 1 is input in one scan. In the measurement data, as shown in FIG. 6, the azimuth angle (direction in which the sensor faces), the distance, and the speed obtained by the measurement of one scan are in one group.
[0011]
(2) Speed mask process The speed mask process is a process of removing measurement points having speed, that is, a process of removing the measurement data whose speed value exceeds the threshold value in FIG. By performing the speed mask process, the measurement data of the traveling vehicles 12a and 12b are removed as shown in FIGS.
[0012]
(3) Background Difference Processing Step Background difference processing is processing for removing a portion that matches the background data (distance) shown in FIG. 4 from the measurement data of the millimeter wave sensor 1 shown in FIGS.
[0013]
(4) Segmentation process step The segmentation process is a process for grouping measurement points at which azimuth angles, distances, and velocities substantially coincide. Giving a name such as 1, 2,... To the group is called labeling, and the name is called a label. By performing the segmentation process, the measurement data of the two obstacles 13a and 13b are divided into two groups in FIG. 5A, and labels 1 and 2 are given, respectively. In (B), since one obstacle 13a cannot be measured by the traveling vehicle 12a crossing the front, only the other obstacle 13b is grouped and given the label "1". In (C), as in (A), the measurement data of the two obstacles 13a and 13b are divided into two groups, which are labeled 1 and 2, respectively.
[0014]
(5) Attribute generation process The outline of the process will be described. The attribute generation process is a process for generating the centroid position, average speed, width, and the like of each group from the measurement data grouped in the segmentation process. A table as shown in FIG. 7 is prepared and stored for the generated center of gravity position, average speed, width, and the like. The position of the center of gravity, average speed, width, etc. are called attributes.
[0015]
(6) History generation process The process history generation process is a process for checking whether or not there has been a group obtained by the segmentation process before. As shown in FIG. 8, a virtual table in which the vertical axis called history data is a new history label and the horizontal axis is the number of scans, and a history attribute table storing attributes similar to those in attribute generation processing are prepared. . In fact, both tables are memory arrays. It is checked whether the group of measurement data obtained up to the attribute by the processing so far exists in the previous scene. Whether it exists or not is checked by checking whether the past attribute and the current attribute substantially coincide. If not, the label number of the group of the measurement data is written in the column of the current scan count in the new history label row of the history data.
Specifically, in the case of FIG. 5A, the history data generation process will be described. The attribute of the group of measurement data of the obstacle 13a called label 1 is compared with the previous attribute, and there is no match, so the history label Write label 1 in N-1 columns of 1 row. Similarly, for the obstacle 13b, the label 2 is written in the N-1th column of the row of the history label 2. In the case of FIG. 5B, the obstacle 13a is not processed at all because there is a traveling vehicle 12a in front and is not measured. When the attribute of the group of measurement data of the obstacle 13b called label 1 is compared with the previous attribute, it matches with that of history label 2, so label 1 is written in N columns of the row of history label 2. 5C is the same as FIG. 5A, and label 1 is written in N + 1 columns of the history label 1 row and label 2 is written in N + 1 columns of the history label 2 row.
[0016]
(7) Occlusion Check Process Step The occlusion check process is a process for comparing the input measurement data and the history attribute to check whether there is a vehicle that has not been measured in front of the traveling vehicles 12a, 12b, etc. It is checked whether or not the attributes of the measurement data match one point. If there is a match, a flag is written in the history label line having the history attribute on the history data. Specifically, the obstacle 13a in FIG. 5B corresponds to this case, and a flag OC is written in the Nth column of one row of history labels. The occlusion means that the traveling vehicles 12a, 12b, etc. are in front and cannot be seen.
[0017]
(8) Obstacle presence / absence detection process step The obstacle presence / absence detection process is a process of detecting that an obstacle is present when an obstacle candidate is always present by measurement of the millimeter wave sensor 1 several times. As shown in FIG. 8, the history data is searched, history label names that exist continuously are extracted, and this is recognized as an obstacle.
[0018]
(9) Obstacle position detection process step The obstacle position detection process is a process of reading the barycentric position of the attribute of the label from the label name obtained by the obstacle recognition process.
[0019]
(10) Obstacle speed detection process step The obstacle speed detection process is a process of reading the average speed of the attribute of the label from the label name obtained by the obstacle recognition process in the same manner as the position detection.
[0020]
(11) Obstacle Size Detection Process Step The obstacle size detection process is a process of reading the width of the label attribute in the same manner as the position detection.
[0021]
(12) Background Data Generation Processing Step Background data generation processing is processing for updating and generating background data using input measurement data. As shown in FIG. 9, first, a virtual graph of the horizontal axis azimuth angle and the vertical axis distance is created, and the input data is plotted on the corresponding coordinates. Actually, an array is prepared on the memory, and the number at the corresponding coordinates is written. Secondly, a blurring process is performed to expand it to a certain size area. Actually write the numbers in all the expanded areas. This is called blur data W. This blurring process is performed to absorb measurement errors of the azimuth angle and distance. Third, a certain operation is performed between the previous background data M ^t-1 and W ^t−1 to obtain new background data M ^t . Here, as an example, calculation is performed using the following equation, and the result is set as new background data M ^t .
[0022]
[Expression 1]
M ^t = (1−α) M ^t−1 + αW ^t−1
Here, f: measurement data name, M: background data, W: blur data, the number of scans on the subscript and the number of data at the time of scan on the subscript.
When the update coefficient α is small, as shown in FIG. 10, the influence of the measurement data input at one time is slight. In other words, if α is small, moving objects such as a traveling vehicle hardly affect the background data. However, this update process is not performed behind the measurement data of the movement of the traveling vehicle (the shaded portion).
[0023]
(13) External I / F transmission processing step Sends the result obtained in the obstacle position detection processing to an alarm device.
[0024]
【The invention's effect】
The invention of claim 1 is as described above, and the obstacle candidate is determined from the difference between the measurement data obtained by scanning with the millimeter wave sensor and the background data without the obstacle, and various processing steps including the background difference processing step. and the detection processing result, and detects the obstacle exists when there is always an obstacle candidates by repeating this process, can be detected more over a wide area of the obstacle on the road, safety and efficiency of road There is an excellent effect that can be increased.
[Brief description of the drawings]
FIG. 1 is an overall process flow diagram of an obstacle detection algorithm showing an embodiment of the present invention.
FIG. 2 is a diagram showing a configuration of a millimeter wave sensor.
FIGS. 3A to 3C are conceptual diagrams of a scene and measurement data for explaining an operation state of an algorithm.
FIG. 4 is a conceptual diagram of background data.
FIGS. 5A to 5C are conceptual diagrams of measurement data and processes. FIG.
FIG. 6 is a conceptual diagram of measurement data.
FIG. 7 is an attribute table.
8A is a conceptual diagram showing a history plane, and FIG. 8B is a conceptual diagram showing a history attribute table.
FIG. 9 is a conceptual diagram of background data generation processing.
FIG. 10 is a diagram illustrating details of a background data update unit.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Millimeter wave sensor 2 Millimeter wave sensor single-piece | unit 3 Drive apparatus 4 Base 5 Control computer
11 Guardrail
12a, 12b traveling vehicle
13a, 13b Obstacle

Claims (1)

An obstacle detection method on a road using a millimeter wave sensor installed on a roadside,
A data input processing step of inputting measurement data in a direction facing the distance and speed by scanning with a millimeter wave sensor in one scan;
A speed mask process that removes data whose speed value exceeds the threshold,
A background difference processing step for removing a portion corresponding to background data without obstacles from measurement data;
A segmentation process that groups together the distances and speeds of measurement data that are almost the same,
Attribute generation processing step for generating and processing the center of gravity position, average speed, and width of each group from the grouped measurement data;
A history generation processing step that checks whether a group obtained by the segmentation process has existed before, and generates a history of the information that has been examined and when it appeared,
An occlusion check processing step that compares the measurement data input when generating the history with the attributes of the history to check whether there is anything that is not measured in front of the traveling vehicle,
Obstacle presence / absence detection processing step for detecting that there is an obstacle when there are always obstacle candidates in several millimeter wave sensor measurements,
A background data generation process that updates and generates background data using the input measurement data,
An obstacle detection method on a road characterized by not performing background data update generation behind a measurement data of travel of a traveling vehicle that cannot be measured when background data is updated and generated in the background data generation processing step.

Images