JP6851111B2 - Railroad crossing obstacle detection device - Google Patents

Description

The present invention relates to a railroad crossing obstacle detection device that irradiates an aerial propagating wave such as infrared rays into a railroad crossing to detect a railroad crossing passage (obstacle) such as a person or a vehicle on the railroad crossing. Regarding the level crossing obstacle detection device of the level crossing radar type that receives the reflected wave and measures the position while sweeping the level crossing, more specifically, the presence or absence of obstacles is judged based on the measurement data obtained by such a method. Regarding railroad crossing obstacle detection device performed by safe computer.

The level crossing obstacle detection device of the plane sweep radar type (see, for example, Patent Document 1) covers the area of the railroad crossing crossing the railroad track located between the blocking rods on both sides of the railroad track and the surface area above the railroad crossing. As a detection area, it is based on a plane sweep measurement unit that receives reflected waves while sweeping and transmitting airborne waves toward the obstacle detection area and measures the distance and direction, and the measurement data obtained by the measurement. It is equipped with a determination unit that determines the presence or absence of an obstacle in the obstacle detection area.

Among them, the plane sweep measurement unit (see, for example, Patent Documents 1 and 2) includes an aerial propagating wave transmitter / receiver that transmits an aerial propagating wave and a reflected wave toward an obstacle detection region, and an aerial propagating wave transmitter / receiver. A rotation mechanism that normally rotates within a predetermined angle, a rotation control unit that controls the rotational movement and measures or enables direction measurement of transmission / reception of airborne waves, and transmission / reception signals of the airborne wave transmission / reception unit. It is equipped with a signal processing unit that measures the distance from the transmission position to the reflection position based on.

Further, the determination unit is provided with a computer as a means for executing software for determining the contents of the data processing and the determination processing of determining the presence or absence of an obstacle based on the measurement data. Then, in the software processing, after tracing the sequence of data to grasp the object shape, the object shape is collated with the storage shape to determine whether or not the object is an obstacle. Since it is necessary to ensure the reliability necessary for detecting a crossing obstacle, a fail-safe computer that can manifest a hardware failure is adopted as the computer responsible for execution (see, for example, Patent Documents 3 and 4).

Since the aerial propagating wave transmitter / receiver is installed at the same height as the blocking rod at the time of descent in such a level crossing obstacle detection device of the plane sweep radar type, it is installed at a much higher place to cross the railroad crossing. Compared to a three-dimensional railroad crossing obstacle detection device that gives a bird's-eye view, it is easier to reduce costs.
In addition, compared to the beam type that has been used for a long time, the number of installations can be reduced because the transmitter / receiver can be integrated, and the resolution is good, so not only large ones such as automobiles but also small ones such as individual passers-by and wheelchairs. It has the advantage of being able to detect even things.

Further, in the data processing and the discrimination processing, each time a group of plane sweeps is performed, the object shape is grasped by tracing the sequence of measurement point data (pair of distance and direction) included in the measurement data. If the measurement data is grouped for each obstacle candidate image by tracking in the image, the grouping process is performed in the image, and even a part of the detected object is on the railroad crossing between the blocking rods. It is determined that there is an obstacle in the railroad crossing (see, for example, Patent Document 1). The grouping process in the image related to such an obstacle candidate image raises the original detection target such as a car or a passerby while avoiding undesired false detection by measurement data related to low-density disturbance such as snowfall or rainfall. Since the detection is performed based on the density measurement data, it is useful for improving the detection accuracy.

Japanese Unexamined Patent Publication No. 2006-007818 Japanese Unexamined Patent Publication No. 2006-214961 Japanese Unexamined Patent Publication No. 2006-338094 Japanese Patent Application No. 2016-163679 (Application)

In this way, the detection performance of railroad crossing obstacle detection devices has been improved by various improvements ranging from hardware to software, but further improvement is required in terms of practicality, so processing based on measurement data is required. It cannot be limited to the grouping process described above. Specifically, in the case of a railroad crossing, just because an object exists in the monitoring area (obstacle detection area) does not mean that it immediately issues an alarm or stops the train as an obstacle, but exceeds the specified time. A function is also required to accurately detect the accumulated matter, while excluding high-speed moving objects that pass the railroad crossing early from the detection target in order to prevent over-detection.

Then, as one of the means for responding to the request, it is conceivable to introduce a time-dependent tracking process between a plurality of images related to the obstacle candidate image. In this tracking process, on the premise that grouping processing in the image related to the obstacle candidate image is performed every time a group of plane sweeps is performed, the measurement point data group grouped as the obstacle candidate image is subjected to plane sweep and grouping. In a series of multiple measurement data that are repeatedly processed, information on how it has changed, especially how it has moved, is extracted over time, so the above functions can be realized by software. .. Moreover, by combining the grouping process with the tracking process, it can be expected to suppress false positives that could not be achieved by the grouping process alone.

On the other hand, in the tracking process, it is indispensable to introduce the element when the tracking disappears and to increase the number of tracking.
Among them, the tracking extinction element is a predetermined time for which the tracking information and the like are retained after the measurement point data group of the obstacle candidate to be tracked disappears. This is necessary even when tracking is temporarily interrupted, for example, when another object passes in front of the object being tracked or the airborne waves reflected from the object being tracked are weakened. This is so that tracking can be resumed promptly after the temporary factor disappears. An appropriate time as the restart waiting time is predetermined as the element when the tracking disappears, and the data is held in a state where it can be referred to during the tracking process. Then, even after the measurement point data group disappears, the tracking data and the like (tracking information) are maintained without being deleted until the progress of the tracking disappearance.

Among them, the tracking number is the number of measurement point data groups of obstacle candidates that are tracked in parallel over time.
The reason why the pluralization is important is that when the number of tracking is one, another obstacle candidate object enters the monitoring area immediately after the obstacle being tracked leaves the monitoring area (obstacle detection area). In such a case, the start of tracking the approaching object is delayed not immediately after the obstacle advances, but from that time until after the time when the tracking disappears, which delays the detection and determination of the obstacle, which is inconvenient. Is.

In addition, it is ideal that the measurement point data group of obstacle candidates related to one object is collected into one, but it may be divided into a plurality of measurement point data groups when the measurement fluctuation factors such as the surface shape and reflectance of the object are large. Therefore, it is useful to have multiple tracking numbers even in such a case.
Furthermore, even for the effects of undesired factors that could not be eliminated by the grouping process, for example, undesired effects due to spatial disturbance factors such as snowfall and interference from oncoming radar, mitigation / suppression of the effects is the tracking number. It is expected to proceed by the pluralization of.

Then, in addition to the grouping process in the image related to the obstacle candidate image described above, the tracking process over time between a plurality of images related to the obstacle candidate image described above is also executed by the software. The hardware required to do so requires a large amount of processing power.
However, as described above, a fail-safe computer is used for the hardware of the railroad crossing obstacle detection device, and since ensuring high reliability is prioritized, the fail-safe computer is used in many consumer devices and the like. It is powerless because it sacrifices cost performance compared to general computers that are used.
Therefore, it is an important issue to reduce the load of software processing.

Specifically, in the tracking process between a plurality of images related to an obstacle candidate image over time, it is assumed that the number of tracking is not limited to a single number but a plurality of tracking numbers as described above, but if the number of tracking is large, the processing load is also high. Since it tends to be heavy, it is a first technical issue to suppress an increase in the number of tracking without sacrificing tracking performance.
Further, in the tracking process, position estimation and position prediction using a Kalman filter or the like are frequently used. In such position estimation, the data of the measurement point data group of the obstacle candidate is not used as it is, but a representative point is determined. An estimation calculation related to the point coordinates is performed.

Therefore, it is not very compatible with the above-mentioned method for determining the presence or absence of an obstacle at a railroad crossing, which determines that there is an obstacle in the railroad crossing if even a part of the detected object is between the blocking rods on the railroad crossing.
That is, although the position of the representative point of the measurement point data group of the obstacle candidate is calculated using the representative point in the tracking process after the grouping process following the measurement data acquisition, the tracking process is further performed in the subsequent determination process. Since the process is repeated by returning to the data of a plurality of points related to the measurement point data group of the obstacle candidate obtained in the earlier grouping process, the processing load is heavy.

However, after the measurement data acquisition, grouping processing and tracking processing are performed, and then the position of the representative point simply enters the monitoring area (obstacle detection area) without returning to the measurement point data group of multiple data. Although the processing load is reduced by the judgment method of determining whether or not there is, the representative point is the monitoring area even if the end of the measurement point data group of the obstacle candidate remains in the monitoring area (obstacle detection area). If it is out of the (obstacle detection area), the measurement point data group of the obstacle candidate is not detected as an obstacle, so that the accuracy of the determination may be sacrificed.
Therefore, it is a second technical problem to improve so that the burden of existence / non-existence determination is not heavy and the accuracy is not impaired even if the tracking process using the representative point, which has a light processing load, is performed.

The crossing obstacle detection device of the present invention (Solution 1) was devised to solve such a problem, and is a plane sweep measurement that sweeps an aerial propagating wave in a plane and acquires measurement data of a reflection position. It is equipped with a unit and a fail-safe computer equipped with a detection area defining data that defines an obstacle detection area of a crossing and a determination means that determines the presence or absence of an obstacle based on the measurement data and the detection area defining data. The crossing obstacle detection device, which is narrowed down to the data related to the position belonging to the obstacle detection area among the measurement data based on the detection area specified data, is adopted as the data for subsequent processing. Data limitation processing is performed, and the grouping process for identifying the measurement point data group of the obstacle candidate using the subsequent processing data and the position of the representative point of the measurement point data group are tracked over the time when the tracking disappears. It is characterized in that the tracking process is performed and the presence or absence of an obstacle is determined according to the presence or absence of the tracking target of the tracking process.

Further, the railroad crossing obstacle detection device of the present invention (solution means 2) is the railroad crossing obstacle detection device of the solution means 1, and the speed of the tracking extinction element in the tracking process is related to the measurement point data group. It is characterized in that it increases or decreases depending on whether it is slow or fast.

Further, the railroad crossing obstacle detection device of the present invention (solving means 3) is a railroad crossing obstacle detecting device of the solving means 1 and 2, and the speed related to the measurement point data group in the tracking process is faster than a predetermined speed. Occasionally, the tracking extinction element is set to zero time or ignored.

Further, the railroad crossing obstacle detection device of the present invention (solution 4) is the railroad crossing obstacle detection device of the solution 3, and the predetermined speed is set to a value corresponding to the average walking speed of a healthy person. It is characterized by that.

In such a railroad crossing obstacle detection device of the present invention (solution 1), two-dimensional measurement data relating to a railroad crossing passage or the like can be obtained each time the airborne propagating wave is swept in a plane. Does not perform grouping processing or the like using the measurement data as it is, but narrows down the measurement data to the subsequent processing data prior to the grouping processing or the like, and then performs the grouping processing or the like using the subsequent processing data. Then, by narrowing down the data, the distribution range of the measurement point data group of the obstacle candidate is reduced, the amount of data processing is reduced / reduced, and the number of tracking is also reduced or at least the increase is prevented, while the obstacle. Since the data related to the position belonging to the object detection area is used for subsequent processing such as grouping without being removed, the increase in the number of tracking is suppressed without sacrificing the tracking performance.

Further, since the grouping process is performed using the subsequent processing data corresponding to the obstacle detection area, the measurement point data group of the obstacle that has advanced from the obstacle detection area disappears quickly, but the measurement point data group disappears. After that, the tracking process that keeps the corresponding tracking information and continues the tracking is also performed until the predetermined tracking disappearance time elapses, so that the measurement point data group of the obstacle candidate is temporarily tracked over time. Because it is done accurately without interruption due to various factors, this also suppresses the increase in the number of tracking without sacrificing the tracking performance.

Furthermore, by making it possible to determine the presence or absence of an obstacle according to the presence or absence of a tracking target in such tracking processing, it relates to a measurement point data group of obstacle candidates obtained in a grouping process prior to the tracking process. Without performing the heavy processing of returning to the data of multiple points and repeating the existence / non-existence judgment, and without performing the affiliation / non-existence judgment processing of the representative point position in the obstacle detection area which tends to sacrifice accuracy. The presence or absence of obstacles can be determined easily and quickly. Therefore, in this railroad crossing obstacle detection device, the burden of determining the existence or nonexistence does not become heavy even if the tracking process using the representative point, which has a light processing load, is performed. Accuracy is not compromised.

Therefore, according to the present invention, the above-mentioned first and second technical problems can be solved together. Moreover, it also exerts the further effects described below. That is, since the image data to be processed is narrowed down to those related to the obstacle detection area prior to the grouping process or the like, when the obstacle enters or advances into the obstacle detection area, it is in the middle of the process. Since the measurement point data group of the obstacle candidate is limited to the part belonging to the obstacle detection area instead of the whole obstacle, the measurement point data group is expanded or contracted while sticking to the boundary line of the obstacle detection area. Looking at the moving speed of each part of the image, the part farthest from the boundary line moves at the speed of the obstacle or a speed close to it, while the part along the boundary line almost keeps stopping, so it is an obstacle candidate. Internal points such as the center point in the measurement point data group of the above move at a slower speed than the obstacle.

On the other hand, position estimation and position prediction using a Kalman filter or the like calculate the next position from the previous position and speed, and are generally vulnerable to discontinuous jumping speed changes. The slower the speed change, the more accurately it can be tracked. Therefore, by adopting the inner point of the measurement point data group as the representative point of the measurement point data group of the obstacle candidate to be tracked by the tracking process using the representative point, it is easy to calculate the center position and the position of the center of gravity, especially among the inner points. By adopting the center point obtained by calculation or the like, it is possible to easily and accurately enhance the tracking ability in the tracking process.

Further, in the railroad crossing obstacle detection device of the present invention (solution 2), when tracking the measurement point data group by the tracking process, when the speed related to the measurement point data group is slow, the time when the tracking disappears. When the speed of the measurement point data group is high, the time required for tracking disappearance is shortened, so that it takes time to cross the railroad crossing because it moves slowly like an elderly person, for example. For, while ensuring safety by tracking closely, it is possible to reduce the burden of data processing by quickly rounding up tracking for items that move quickly and do not have the risk of staying inside the railroad crossing, such as automobiles. ..

Further, in the railroad crossing obstacle detection device of the present invention (solutions 3 and 4), for safety, for safety, the tracking is extended based on the time when the tracking disappears, such as an automobile or the like. For fast railroad crossing passages, the safety of railroad crossing passage and the reduction of data processing burden can be achieved at a high level by omitting the tracking extension based on the tracking disappearance.
Moreover, it is simplified by setting the tracking extinction element to zero time or ignoring the tracking extinction element.

Regarding Example 1 of the present invention, the structure of the railroad crossing obstacle detection device is shown, (a) is a schematic plan view of the railroad crossing at the installation destination, (b) is a schematic block diagram of the hardware of the railroad crossing obstacle detection device, ( c) is an outline block diagram of the software of the railroad crossing obstacle detection device. (A) is a schematic diagram of sweeping the airborne wave in a plane, (b) is an image diagram of measurement data, (c) is an image diagram of data for subsequent processing, and (d) is an image image diagram of measurement point data group. is there. (A) is an image diagram of tracking information, (b) is an image diagram of measurement point data group, (c) is an image diagram of tracking information, (d) is an image diagram of measurement point data group, and (e) is an image diagram of tracking information. Is. (A) to (f) are image diagrams of measurement point data groups and tracking information, and (g) is a graph of an example of speed change.

A specific embodiment for carrying out such a railroad crossing obstacle detection device of the present invention will be described with reference to Examples 1 to 3 below.
The first embodiment shown in FIGS. 1 to 4 embodies the above-mentioned solution 1 (claim 1 at the time of filing), and the above-mentioned solutions 2 and 3 are omitted from the illustration. It embodies 4 (claims 2 to 4 at the time of filing).
For the sake of simplicity, these illustrations include housings, frames, fasteners such as bolts, drive sources such as electric motors, transmission members such as gears, electric circuits such as motor drivers, and electronic devices such as controllers. Circuits and the like are omitted from the illustrations, and block diagrams and the like are shown focusing on those necessary for explaining the invention and related ones.

The specific configuration of the first embodiment of the railroad crossing obstacle detection device of the present invention will be described with reference to the drawings.
1A is a schematic plan view showing the installation status of the railroad crossing obstacle detection device 10 on the railroad crossing road 4, and FIG. 1B is a schematic block diagram showing the hardware configuration of the railroad crossing obstacle detection device 10; ) Is a schematic block diagram showing a software configuration of the railroad crossing obstacle detection device 10. Further, FIG. 2 (a) is a schematic view of sweeping the airborne wave in a plane, FIG. 2 (b) is an image diagram of measurement data, FIG. 2 (c) is an image diagram of data for subsequent processing, and FIG. 2 (d). 3 (b) and FIG. 3 (d) are image images of the measurement point data group, and FIGS. 3 (a), 3 (c) and 3 (e) are image diagrams of the tracking information.

The railroad crossing obstacle detection device 10 (see FIGS. 1A and 1B, Patent Document 4) first describes a hardware configuration, which is a spatial area above the railroad crossing 4 and sandwiched between blocking rods. Plane sweep measurement unit 11 to the plane sweep radar type that receives reflected waves and measures the distance and direction while sweeping and transmitting airborne waves such as infrared rays (see the two-point chain line) toward the obstacle detection area 7. It includes 13 and a fail-safe computer 14 in which a determination program 20 for determining the presence or absence of an obstacle in the obstacle detection area 7 is installed from the measurement data obtained by the measurement.

The plane sweep measurement units 11 to 13 set the transmission direction of the air-propagated wave transmission / reception unit 12 that transmits the air-propagated wave and the reflected wave toward the obstacle detection region 7 and the transmission direction of the air-propagated wave transmission / reception unit 12, for example, 130. For the transmission / reception signals of the rotation control unit 11 and the airborne wave transmission / reception unit 12 that control the rotational movement of the rotation mechanism that sweeps within an angle range such as ° or 190 ° and measure or enable the direction measurement of the transmission / reception of airborne waves. It is provided with a signal processing unit 13 that measures the distance from the transmission position to the reflection position based on the signal processing unit 13. In this example, two sets (plurality) of plane sweep measurement units 11 to 13 are provided in order to measure the entire area of the obstacle detection area 7 by dividing the area.

Since a known product is sufficient for the fail-safe computer 14 (see, for example, Patent Documents 3 and 4), the fail-safe computer 14 is adopted, and the data memory has the detection area specified data which is a preset constant and the tracking disappears. It holds measurement data, data for subsequent processing, data of measurement point data group, and tracking information, which are variables and arrays that are changed with the execution of the determination program 20.

The determination program 20 (see FIG. 1C), in short, determines the presence or absence of an obstacle based on the measurement data and the detection area definition data. For that purpose, the data input program 21 and the data limitation program 22 are grouped. A program 23, a tracking program 24, and a final determination program 25 are provided, and these are repeatedly executed in that order, for example, at a predetermined cycle or executed each time a predetermined event occurs.

More specifically, the detection area definition data defines the obstacle detection area 7 related to the crossing road 4, for example, the two-dimensional coordinates of the positions of each corner of the obstacle detection area 7 are arranged in the order of rotation. , The coordinates may be Cartesian coordinates, but polar coordinates are more compatible with the plane sweep radar method.
As described above, the tracking extinction time element is a predetermined time during which tracking information and the like are retained after the measurement point data group of the obstacle candidate to be tracked disappears, and is a predetermined time for a plurality of railroad crossing passages. It is decided in consideration of the time from the coalescence to the separation of the measurement point data group due to the exchange.

The data input program 21 sets the set of polar coordinate values of the reflection positions obtained by the plane sweep measurement units 11 to 13 each time the plane sweep measurement units 11 to 13 perform the measurement by the plane sweep measurement units 11 to 13 (see FIG. 2A). It is input from 13 and used as measurement data (see FIGS. 1 (c) and 2 (b)), but in this example, two sets of plane sweep measurement units 11 to 13 correspond at the same time. The two sets of data obtained at the time are merged into one set of measurement data (see FIGS. 2 (a) and 2 (b)).

This measurement data includes three vehicles (three vehicles of FIGS. 2 (a) and 2 (b)) in the obstacle detection area 7 (see the alternate long and short dash line in FIGS. 2 (a) and 2 (b)). Image data of the contour of (see) is included (see the black dots in FIG. 2 (b)), but not only that, but also the portion of them protruding from the obstacle detection area 7 and the black dots in (FIG. 2 (b)). (See), although not shown, the image data of equipment outside the obstacle detection area 7 is included as long as it is within the plane sweep range.
In this embodiment, the measurement data is a series of measurement point data consisting of a pair of distance and direction, and the distance and direction are variable values obtained by measurement. Therefore, for example, two rows and N columns (N is a positive integer). ) Is held in the sequence area.

The data limiting program 22 executes data limiting processing and creates data for subsequent processing from the measurement data based on the detection area definition data (see FIG. 1C), and is included in the measurement data at that time. Of the points of the image data (see the black dots in FIG. 2 (b)), those located outside the obstacle detection area 7 (see the single-point chain line in FIG. 2 (b)) are excluded. As a result, among the points of the measurement data, those narrowed down to the data related to the position belonging to the obstacle detection area 7 are adopted as the data for subsequent processing (see the black dots in FIG. 2C).
In the case of an object straddling the boundary of the obstacle detection region 7, the data for subsequent processing includes only the image data of the portion within the region (see the left and right black spots in FIG. 2C).

The grouping program 23 executes an in-image grouping process for creating a measurement point data group of obstacle candidates from the subsequent processing data (see FIG. 1C), and each point of the image data of the subsequent processing data. (Refer to the black dot in Fig. 2 (c)), the short-distance data can be collected one after another into a measurement point data group, and the measurement point data can be obtained based on the conditions considering the characteristics of crossing passages such as people and vehicles. By combining the groups, identify only the measurement point data group that can be an obstacle candidate image (see the three solid lines in Fig. 2 (d) and the two solid lines in Fig. 3 (b)). ), The measurement point data group disappears immediately if there is no crossing passage that can be an obstacle candidate (see FIG. 3 (d)).

The tracking program 24 executes a tracking process of tracking the measurement point data group obtained by the in-image grouping process over a plurality of images over time (see FIG. 1C), and the measurement point data group. If there are a plurality of measurement point data groups, the same number of tracking information is generated and held and assigned to each measurement point data group, so that each measurement point data group is individually tracked.
Moreover, for each measurement point data group, at least the inner point of the center point of the measurement point data group can be obtained by an appropriate calculation such as calculation of the center point or calculation of the center of gravity, and this can be adopted as the representative point of the measurement point data group. (See the black dots in FIGS. 3 (a) and 3 (c)).

Further, the tracking program 24 performs temporal tracking processing between a plurality of images related to obstacle candidate images by estimating the next representative point position from the previous representative point position or the like by an estimation calculation of a known Kalman filter or the like. Specifically, for each tracking information, the next predicted position is calculated from the position and speed of the previous representative point by, for example, a linear Kalman filter, and the representative point is within a predetermined range from the predicted position. For the included measurement point data group, group specific information such as the position of the representative point is included in the corresponding tracking information.

Here, the above-mentioned predetermined range is determined by a predicted radius or the like in which a value is set in advance, but in the commonly used constant velocity linear motion model, the advantage that the traceable maximum speed also increases as the predicted radius increases. On the other hand, there is a disadvantage that the separation performance of the tracking target deteriorates when there are multiple tracking targets, and there is a possibility that even those that should not be tracked get inside the predicted radius and mistracking occurs. Since there is also the inconvenience of increasing the number of predicted radii, the set value of the predicted radius is determined in advance in consideration of the balance between them.

Further, when the tracking program 24 performs such a tracking process, the tracking program 24 refers to a preset tracking extinction time element (see FIG. 1 (c)) and tracks over the time of the time element. It has become. Specifically, when the measurement point data group to be tracked disappears by the grouping process, the tracking information related to the measurement point data group is not deleted immediately, but is kept until the tracking disappearance time elapses. There is. For example, even when the three measurement point data groups to be tracked move to the right (see FIGS. 2 (d) and 3 (a)) and the rightmost measurement point data group disappears (FIG. 3). (Refer to (b)), before the elapse of the tracking extinction element, the tracking information related to the representative point of the measurement point data group at the right end continues to exist (see FIG. 3C), and the deletion of the tracking information is the tracking extinction element. It is supposed to be done after the passage of.

The final determination program 25 determines the presence or absence of an obstacle in the obstacle detection area 7, and the determination process is performed according to the presence or absence of the tracking target of the tracking process (FIG. 1 (c)). reference). Specifically, if there is at least one tracking information created by the tracking program 24 (see FIGS. 3A and 3C, three in the figure), it is determined that an obstacle exists in the obstacle detection area 7. However, if there is no such tracking information (see FIG. 3E), it is determined that there is no obstacle in the obstacle detection area 7.

The usage mode and operation of the railroad crossing obstacle detection device 10 of the first embodiment will be described with reference to the drawings.

FIG. 2A shows a state in which the aerial propagating wave transmitting / receiving units 12 and 12 are sweeping the aerial propagating wave in a plane toward the obstacle detection area 7 above the crossing road 4 and is shown in FIG. 2B. ) Is an image diagram of the measurement data, FIG. 2 (c) is an image diagram of the data for subsequent processing, and FIGS. 2 (d) and 3 (a) are image diagrams of the initial measurement point data group and the tracking information. 3 (b) and 3 (c) are image diagrams of the subsequent measurement point data group and tracking information, and FIGS. 3 (d) and 3 (e) are image diagrams of the last measurement point data group and tracking information. .. Further, FIG. 4 is an image diagram in which measurement point data groups and tracking information are superimposed in each of (a) to (f), and is a graph showing a speed change state when (g) passes a railroad crossing 4. Yes, the horizontal axis is time and the vertical axis is speed.

To use the railroad crossing obstacle detection device 10 (see FIG. 2A), install an appropriate number of aerial propagating wave transmitters / receivers 12 and 12 facing the railroad crossing road 4 as much as possible. A plane sweep is performed from a plurality of directions (two facing directions in the figure) over a range wider than the entire area of the obstacle detection area 7.
Then, in the railroad crossing obstacle detection device 10, plane sweeping is repeatedly performed by two sets of plane sweep measuring units 11 to 13, 11 to 13, and each time, if there is a passerby on the railroad crossing road 4 (three in the figure). The reflected wave is detected, the reflected position is acquired by the polar coordinate value, and the measurement data relating to the area wider than the obstacle detection area 7 is completed as the aggregate data of the polar coordinate value (FIG. 2 (FIG. 2). See the black dots on almost all of the three vehicles in b)).

Then, every time a set or a set of measurement data is obtained by such a plane sweep, the fail-safe computer 14 executes the determination program 20.
To elaborate on the determination procedure, first, the measurement data is taken into the fail-safe computer 14 by executing the data input program 21, and then the subsequent processing data is created from the measurement data by executing the data limiting program 22. The image data to be processed is narrowed down to those belonging to the obstacle detection area 7 (in FIG. 2C, the black dots related to the rear end portion of the vehicle on the right side, the entire vehicle in the center, and the front end portion of the vehicle on the left side are defined. reference).

As described above, since the image data including the obstacle candidate is limited at the initial stage of the determination process, the load of the subsequent data processing is reduced.
Then, the in-image grouping process is performed by executing the grouping program 23, and the data of the measurement point data group of the obstacle candidate is created from the data for subsequent processing (in FIG. 2D, the rear end of the vehicle on the right side). See the 匚 -shaped solid line for the peripheral part, the mouth-shaped solid line for the entire circumference of the central vehicle, and the U-shaped solid line for the front end peripheral part of the vehicle on the left side).

Further, by executing the tracking program 24, a representative point is selected for tracking over time for each measurement point data group, and one (set) of tracking information used for tracking over time is created. (See the three sunspots in FIG. 3A).
Tracking information is added each time a new measurement point data group is added to the tracking target, and each tracking information includes, for example, position data of a representative point of the measurement point data group of the tracking target and speed data calculated from the change. Can be included. When the tracked object disappears, the corresponding tracking information is deleted. Therefore, when there is no tracking target at all, the tracking information is completely lost (see FIG. 3 (e)).

Then, by executing the final determination program 25, if there is even one (one set) of tracking information, it is determined that an obstacle exists on the railroad crossing 4. The presence or absence of tracking information can be confirmed quickly and accurately by simply referring to any one of the information such as the number of tracking information data and the presence / absence mark of the first tracking information in order to manage the tracking information. Can be done.
Therefore, for all the tracking information, it is necessary to determine whether or not the measurement point data group to be tracked and its representative point belong to the obstacle detection area 7, and then the determination of the absence of obstacles can be made. The railroad crossing obstacle detection device 10 is quickly put out with a light load. Then, this determination result is sent to a railroad crossing control device or the like and used for alarm output to a special signal light emitter or the like.

In this way, every time the plane sweep measurement units 11 to 13 perform plane sweep and reflected wave measurement on the railroad crossing 4, data limiting processing, grouping processing, tracking processing, and obstacle to the obstacle detection area 7 related to the measurement data are performed. The existence / non-existence determination process is performed by the fail-safe computer 14, and the process is repeated at a predetermined cycle or the like. Then, among those processes, from the plane sweeping process to the grouping process, the measurement point data group of the processing result is uniquely determined by the traffic state and the image data at that time, but in the subsequent tracking process, the time lapse between a plurality of images is determined. Since the position of the representative point of the measurement point data group is tracked over the time when the tracking disappears, the tracking information remains for a while even after the measurement point data group disappears.

For example, when tracking three measurement point data groups for three vehicles (see the black dots in the image on the right side of FIGS. 2 (d) and 3 (a)), the data of the measurement point data groups is also tracked. Information data is also held in three sets each (see the block on the left side in Fig. 2 (d) and Fig. 3 (a)), but three vehicles move from left to right and the vehicle on the right is in trouble. When the object detection area 7 is exited, the measurement point data group immediately becomes two for the central vehicle remaining in the obstacle detection area 7 and the left vehicle on which at least a part is hung (FIG. 3). (Refer to the solid line in the image on the right side of (b)), the tracking information related to the vehicle on the right side is maintained until the time when the tracking disappears (black dots in the image on the right side in FIG. 3C). And the block on the left).

Then, when the tracking disappearance time elapses for the tracking, the tracking information also becomes two related to the two vehicles related to the obstacle detection area 7 (not shown).
In this way, it is carefully confirmed that the railroad crossing passerby has left the obstacle detection area 7. In addition, although not shown, the time from merger to separation is tracked even when a merger or separation occurs in a plurality of measurement point data groups related to a plurality of railroad crossing passages depending on the traffic of the railroad crossing passages. If it is shorter than the extinction time element, the time-dependent tracking between a plurality of images related to the obstacle candidate image can be accurately performed without being disturbed by a large change in the image due to the merger.
After that, when all the railroad crossing passersby exit the obstacle detection area 7, the measurement point data group disappears (see FIG. 3D), and when the tracking disappearance time elapses, the tracking information also disappears (Fig. 3). 3 (e)), it is found that there is no obstacle in the obstacle detection area 7.

Furthermore (see Fig. 4), the inner point of the measurement point data group is adopted as the representative point of the measurement point data group of the obstacle candidate to be tracked by the tracking process using the representative point, and the Kalman filter using the constant velocity linear motion model. Regarding the combination with predicting the position of the measurement point data group by the above, the situation when only one vehicle as described above passes through the obstacle detection area 7 at a constant speed V from left to right can be seen. It will be described by giving a specific example (Note that in FIGS. 4A to 4F, the measurement point data group and the representative point are superimposed on the same image image diagram, but they are the data of the measurement point data group. It is an intuitive image of the tracking information data).

In this case, the measurement point data group and its representative points start from the front end of the vehicle entering from the left end of the obstacle detection area 7 (see FIG. 4A), and sequentially from the front end of the vehicle. It extends to the center (see FIG. 4 (b)), affects the entire vehicle (see FIG. 4 (c)), and remains as it is related to the entire vehicle in the obstacle detection area 7. It moves from left to right (see FIG. 4 (d)) and relates to the central and rear ends of the vehicle that has begun to advance from the right end of the obstacle detection area 7 (see FIG. 4 (e)). It is reduced only at the rear end (see FIG. 4 (f)) and disappears at the end (see FIGS. 3 (d) and 3 (e)).

Looking at the measurement point data group and the position change and speed change of the representative point according to the movement of such a vehicle, assuming that the representative point is always at the center of the measurement point data group for the sake of simplicity ( (See FIG. 4 (g)), the speed is “0” when there is no vehicle (see time t1 on the thick solid line graph in FIG. 4 (g)), and starting from there during the vehicle approach (see FIG. 4 (g)). (See a) and (b)), the speed of the representative point becomes the intermediate speed V / 2 between the speed "0" at the left end of the stationary obstacle detection area 7 and the front speed V of the vehicle (thick in FIG. 4 (g)). While the vehicle is within the obstacle detection area 7 (see FIGS. 4 (c) and 4 (d)), the speed of the representative point becomes the same speed V as the vehicle during the time t1 to t2 of the solid line graph). (Refer to the time t2 to t3 portion of the thick solid line graph in FIG. 4 (g)).

Further, while the vehicle is advancing (see FIGS. 4 (e) and 4 (f)), the speed of the representative point is the intermediate speed V between the rear end speed V of the vehicle and the stationary right end speed "0" of the obstacle detection area 7. It becomes / 2 (see the time t3 to t4 part of the thick solid line graph in Fig. 4 (g)), and the speed of the representative point returns to "0" after the vehicle advances (in the thick solid line graph in Fig. 4 (g)). See time t4).
In this way, even if the vehicle speed is a single speed V, the speed of the representative point is divided into the time when the area does not exist, the time when the area enters and exits, and the time when the vehicle exists in the area. Since the discontinuous change is limited to half V / 2, the tracking state such as the tracking speed is stable (see the alternate long and short dash line in FIG. 4 (g)).

On the other hand, unlike the railroad crossing obstacle detection device 10 in which the limitation within the obstacle detection area of the measurement data is completed before the grouping process in the image, it is time to limit the measurement data within the obstacle detection area. In the case of a slow railroad crossing obstacle detection device, the speed of the representative point in the tracking process and its discontinuous change are the same V as the vehicle speed (see the broken line graph in FIG. 4 (g)), so a Kalman filter or the like is used. Since the accuracy of the position estimation remains low (see the alternate long and short dash line in FIG. 4 (g)), it becomes difficult to stabilize the tracking state such as the tracking speed.
Therefore, the railroad crossing obstacle detection device 10 enhances the tracking ability in the tracking process by mitigating the adverse effect of the discontinuous jumping speed change by a simple method of limiting the data at an appropriate timing. It can be said that there is.

Although the illustration is omitted, the difference between the railroad crossing obstacle detection device of the second embodiment and the railroad crossing obstacle detection device 10 described above is that the part of the processing of the tracking program 24 that uses the element at the time of disappearance of tracking is changed. It is a point.
Specifically, a value obtained by converting the moving speed of an object on a cutway, which is generally regarded as the average walking speed of a healthy person, at or near 4 km / h into the moving speed of a representative point of a measurement point data group in the measurement data is set in advance as a predetermined speed. At the same time, when the speed related to the measurement point data group is slower than the above-mentioned predetermined speed during the tracking process, the tracking disappearance element even after the disappearance of the measurement point data group to be tracked in the grouping process is the same as in the first embodiment. The tracking information is retained before the lapse, but when the speed related to the measurement point data group is faster than the above-mentioned predetermined speed, unlike the case of Example 1, the measurement point data group to be tracked is ignored, ignoring the element at the time of tracking disappearance. Is disappeared by the grouping process, and the corresponding tracking information is promptly deleted.

Similarly, although the illustration is omitted, the railroad crossing obstacle detection device of the third embodiment is different from the railroad crossing obstacle detection device 10 described above in that the tracking processing portion using the tracking extinction element is also changed.
Specifically, not only one element at the time of tracking extinction but also a plurality of elements can be associated with each measurement point data group one by one. In addition, during the tracking process, the tracking extinction time element is individually changed for each measurement point data group to be tracked, but in the change process, the moving speed of the representative point of the corresponding measurement point data group is changed. If it becomes slower, the value of the element at the time of tracking disappearance will increase or at least maintain the current status, and if the movement speed of the representative point of the corresponding measurement point data group increases, the value of the element at the time of tracking disappearance will decrease or at least maintain the current status. Will be done. Further, for each measurement point data group, when the moving speed of the representative point is faster than the predetermined speed of Example 2, the value of the corresponding tracking extinction element is cleared to zero time, so that it can be easily and quickly. It is also designed to omit the tracking extension based on the tracking disappearance.

[Other]
In the above embodiment, only one obstacle detection area 7 is specified, but when the number of parallel running tracks is large and the obstacle detection area related to one railroad crossing is very long, the obstacle detection area may be defined. It may be divided into several subregions that do not overlap or overlap.
In that case, for example, by dividing the final determination program 25 into an intermediate determination program and a determination integration program, it is possible to deal with it relatively easily by software.

In the above embodiment, the situation where one measurement point data group is selected for one railroad crossing passage is described, but it is not rare that a plurality of measurement point data groups are selected for one railroad crossing passage. For example, when the front and rear parts of the railroad crossing passage are clearly detected but the middle part is unclear, or when the left and right sides of the railroad crossing passage are swept in a plane by different aerial propagating wave transmission / reception units 12 and 12, both images of the railroad crossing passage It is easy to have multiple measurement point data groups, such as when they are separated from each other in front of and behind the body. In addition, there are cases where only one measurement point data group is selected for a plurality of railroad crossing passages, such as when two railroad crossing passages approach each other. In any case, if there is even one measurement point data group, it is judged on the safe side that an obstacle exists, so there is no inconvenience apart from the amount of data processing.

In the above embodiment, the distance and the direction of the measurement point data are variable values, and the measurement data in which the measurement data are connected are held in the array area of 2 rows and N columns. If the direction of sweep measurement is determined in advance, for example, if the measurement start direction is a known fixed direction and the subsequent direction difference is also known, the part of the measurement data that is known for the direction is one line N. In addition to holding it in a column constant table, etc., for the portion of the measurement data that is variable data related to distance, distance measurement point data consisting of distance measurement values related to one known direction is measured in N consecutive known directions. The distance data may be used and stored in the array area of 1 row and N columns. As a result, the measurement point data group or the like that is data-processed as a two-dimensional matrix becomes a so-called focus-finding point group or the like that is data-processed as a one-dimensional vector, so that the burden on the fail-safe computer 14 is further reduced. It will be.

The application of the railroad crossing obstacle detection device of the present invention is not limited to those that detect only obstacles, and in addition to obstacles such as detection of railroad crossing passages in situations where there is no obstacle and detection of trains. It can also be applied to those that detect other things.

4 ... Railroad crossing, 7 ... Obstacle detection area (monitoring area),
10 ... Railroad crossing obstacle detection device,
11-13 ... Plane sweep measurement unit, 11 ... Rotation control unit,
12 ... Airborne wave transmitter / receiver, 13 ... Signal processing unit, 14 ... Fail-safe computer,
20 ... Judgment program,
21 ... Data input program, 22 ... Data limited program,
23 ... Grouping program (grouping process in image),
24 ... Tracking program (tracking process over time),
25 ... Final judgment program

Claims (4)

Based on the measurement data and the detection area regulation data, the plane sweep measurement unit that acquires the measurement data of the reflection position by sweeping the airborne wave in the plane and the detection area regulation data that defines the obstacle detection area of the crossing are held. It is a crossing obstacle detection device equipped with a fail-safe computer equipped with a determination means for determining the presence or absence of an obstacle, and the determination means inputs the measurement data acquired by the plane sweep measurement unit. After that, data limiting processing is performed in which the measurement data narrowed down to the data related to the position belonging to the obstacle detection area based on the detection area regulation data is adopted as the subsequent processing data, and the subsequent processing data is performed. The grouping process for identifying the measurement point data group of the obstacle candidate and the tracking process for tracking the position of the representative point of the measurement point data group over the time when the tracking disappears are performed, and the tracking process is tracked. A crossing obstacle detection device characterized in that the presence or absence of an obstacle is determined according to the presence or absence of an object. The railroad crossing obstacle detection device according to claim 1, wherein in the tracking process, the tracking extinction element is increased or decreased depending on whether the speed of the measurement point data group is slow or fast. The railroad crossing according to claim 1 or 2, wherein when the speed related to the measurement point data group is faster than a predetermined speed in the tracking process, the tracking extinction element is set to zero time or ignored. Obstacle detection device. The railroad crossing obstacle detection device according to claim 3, wherein the predetermined speed is set to a value corresponding to the average walking speed of a healthy person.

Images