JP5478419B2 - Axle detection system - Google Patents

Description

  The present invention relates to an axle detection system that captures an image of a vehicle using a 3D image capturing apparatus and detects an axle from acquired 3D data.

In general, in an axle detection system arranged to face a vehicle, the axle is detected by detecting a tire of the vehicle in contact with the road surface.
As a conventional axle detection system, a technique using a camera (for example, see Patent Document 1) and a technique using a laser (for example, see Patent Document 2) have been proposed.
In the conventional system described in Patent Document 1, the irradiation pattern is acquired as data by receiving irradiation light with a camera, and the presence or absence of a tire (wheel) is detected from the parallelism of the acquired data.

  On the other hand, in the conventional system described in Patent Document 2, distance information is acquired based on a time difference from pulse laser irradiation to light reception, and three-dimensional data is acquired using a scanning angle at the time of laser irradiation. In this case, if even one piece of acquired data exists within the height range above the road surface and below the vehicle body (vehicle body), the tire is detected.

JP-A-8-293090 JP 2002-183881 A

In the conventional axle detection system, in Patent Document 1, the tire is detected from the parallelism of the irradiation pattern acquired by the camera, but when the parallelism of the irradiation pattern cannot be measured due to the influence of the background light or the like, Since tire detection is impossible, there is a problem that erroneous detection is caused.
Further, in Patent Document 2, if there is at least one three-dimensional data within the height range above the road surface and below the vehicle body in the acquired data, it is detected as a tire, so the acquired three-dimensional data There has been a problem that a tire is erroneously detected when erroneous detection data is included in.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to obtain an axle detection system that suppresses erroneous detection of tires and improves the detection probability of an axle.

An axle detection system according to the present invention includes a first three-dimensional image pickup device that acquires first three-dimensional data of a vehicle that is an object, and an axle detection device that counts the number of axles based on the first three-dimensional data. The axle detection system includes: a first three-dimensional data threshold processing device that performs a false detection data reduction process on the first three-dimensional data; and a post-false detection data reduction process. A first height / distance gate device that performs gate processing for narrowing down the first detection region on the first three-dimensional data by a height / distance gate, and extracts the first three-dimensional data after the gate processing. A first height histogram calculation device that divides the first detection region in the height direction to create a total of N bins, and creates a height histogram of the first three-dimensional data after gate processing; The height of the first 3D data A data existence number calculating device for calculating, as a data existence number, the number of bins in which a predetermined number or more of first three-dimensional data after gate processing is stored based on a tomgram, and a threshold value using a maximum likelihood estimation method for the data existence number A threshold calculation device that automatically sets and an axle number counting device that counts the number of axles using the number of data present and the threshold, the threshold calculation device is a moving average processing device that performs a moving average process on the number of data present And a sorting device that performs processing for rearranging the number of existing data in ascending order based on the moving average processing result from the moving average processing device, and a model for setting a threshold calculation model based on the sorting processing result from the sorting device A setting device, and a threshold value calculation unit that calculates a threshold value based on the model set by the model setting device. The threshold value calculation unit uses a maximum likelihood estimation method. Te, and it calculates a threshold value for separating the data present number other than the data existing number and the axle of the axle.

  According to the axle detection system of the present invention, even when erroneous detection data is included in the acquired three-dimensional data, the detection region is narrowed down by the height / distance gate by reducing the erroneous detection data. Later, by calculating the data existence probability, the detection probability in the acquired data can be improved and the axle can be detected, so that erroneous detection of the axle can be suppressed.

It is a block diagram which shows the whole structure of the axle shaft detection system which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows notionally the three-dimensional data acquired by Embodiment 1 of this invention, and its process. It is explanatory drawing which shows the operation example of the height and distance gate apparatus by Embodiment 1 of this invention. It is explanatory drawing which shows the operation example of the detection flag generator by Embodiment 1 of this invention. It is explanatory drawing which shows an example of the detection flag by Embodiment 1 of this invention. It is a block diagram which shows the whole structure of the axle shaft detection system which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows the operation example of the height and distance gate apparatus by Embodiment 2 of this invention. It is explanatory drawing which shows the operation example of the detection flag generator by Embodiment 2 of this invention. It is a block diagram which shows the whole structure of the axle shaft detection system which concerns on Embodiment 3 of this invention. It is a block diagram which shows the whole structure of the axle shaft detection system which concerns on Embodiment 4 of this invention. It is a block diagram which shows the whole structure of the axle shaft detection system which concerns on Embodiment 5 of this invention. It is a block diagram which shows the structure of the threshold value calculation apparatus in FIG. It is explanatory drawing which shows the threshold value calculation process by Embodiment 5 of this invention. It is a block diagram which shows the structure of the axle number counting apparatus in FIG. It is explanatory drawing which shows the axle number count process by Embodiment 5 of this invention. It is a block diagram which shows the structure of the threshold value calculation apparatus by Embodiment 6 of this invention. It is explanatory drawing which shows the threshold value calculation process by Embodiment 6 of this invention.

Embodiment 1 FIG.
1 is a block diagram showing an overall configuration of an axle detection system according to Embodiment 1 of the present invention.
In FIG. 1, the axle detection system includes a first three-dimensional image pickup device (hereinafter simply referred to as “three-dimensional image pickup device”) 1 that acquires three-dimensional data A of a vehicle (object) 3, and a detection device 2. It is comprised by.

The three-dimensional image capturing apparatus 1 is a laser image measuring apparatus, which scans a beam in a one-dimensional direction in the height direction, and determines a vehicle from a plurality of distance measuring and scanning angles in a detection region (scanning range, imaging region) R. Acquire 3 two-dimensional sections. Further, as the vehicle 3 passes, the three-dimensional image capturing apparatus 1 acquires the three-dimensional data A of the vehicle.
The three-dimensional data A includes an electric signal corresponding to the distance of the vehicle 3 and an electric signal corresponding to the reflection intensity.

  The detection device 2 includes a three-dimensional data threshold processing device 21, a height / distance gate device 22, a height histogram calculation device 23, a data existence probability calculation device 24, and a detection flag generation device 25, and includes three-dimensional data. Based on A, an axle detection flag F (first detection flag) is output.

In the detection device 2, the three-dimensional data threshold processing device 21 performs a false detection data reduction process for the three-dimensional data A.
That is, the three-dimensional data threshold processing device 21 provides a threshold for the three-dimensional data (electrical signal) A corresponding to the intensity and distance acquired by the three-dimensional image capturing device 1 to reduce data that is assumed to be erroneously detected. To do.
The three-dimensional data A ′ after the erroneous detection data reduction process output from the three-dimensional data threshold processing device 21 is an electrical signal corresponding to the intensity and distance after the false detection data reduction process.

The height / distance gate device 22 extracts the three-dimensional data B after the gate processing in the detection region R from the three-dimensional data A ′ after the erroneous detection data reduction processing.
In other words, the height / distance gate device 22 performs gate processing relating to height and distance in order to subdivide and narrow down the detection region R with respect to the three-dimensional data A ′ corresponding to the intensity and distance after the erroneous detection data reduction processing. Apply. The three-dimensional data B after gate processing output from the height / distance gate device 22 is an electrical signal corresponding to the intensity and distance after the false detection data reduction processing in the detection region R.

The height histogram calculation device 23 divides the detection area R into N in the height direction to create a total of N height bins Bh (see FIG. 3), and the distance after the erroneous detection data reduction processing in the detection area R For the three-dimensional data B corresponding to, a height histogram C of the three-dimensional data B after gate processing (after erroneous detection data reduction processing) is created.
From the height histogram calculation device 23, an electrical signal corresponding to the height histogram C of the three-dimensional data A ′ after the erroneous detection data reduction processing is output.

Based on the height histogram C, the data existence probability calculation device 24 calculates the number of bins n in which three (a predetermined number, for example, two) or more three-dimensional data B after gate processing (after false detection data reduction processing) is stored. The ratio n / N with the total number N of bins divided in the height direction is calculated as the data existence probability P.
The data existence probability calculation device 24 outputs an electrical signal corresponding to the data existence probability P of the number of bins n in which the three-dimensional data B is stored.

  The detection flag generator 25 sets a detection threshold Th for the data existence probability P of the number of bins n in which the three-dimensional data B is stored, and for the highly reliable three-dimensional data B based on the data existence probability P. A detection flag (electric signal) F is generated.

Next, operations of the 3D image capturing apparatus 1 and the detection apparatus 2 according to the first embodiment of the present invention shown in FIG. 1 will be described with reference to FIGS.
FIG. 2 is an explanatory diagram conceptually showing the three-dimensional data A acquired by the three-dimensional image acquisition device 1 and its processing. The captured image of the vehicle 3 by the three-dimensional image imaging device 1 and the acquired three-dimensional data A are shown in FIG. Show.

2A shows the relationship between the vehicle 3 and the detection region R in the three-dimensional image acquisition device 1. FIG.
The three-dimensional imaging device 1 acquires the three-dimensional data A of the vehicle 3 when the vehicle 3 crosses the detection region R.
FIG. 2B shows a cross-sectional view of the three-dimensional data A acquired when the viewpoint is placed on the side of the vehicle 3. In FIG. 2B, the horizontal axis is time, and the vertical axis is height.
In general, on the side surface of the vehicle 3, since a sufficient reception signal-to-noise ratio cannot be obtained at a location where the reflectance is low, the distance measurement accuracy is low, so that the three-dimensional data A as shown in FIG. Contains a lot of false positive data.

Therefore, in the detection device 2, the three-dimensional data threshold processing device 21 obtains three-dimensional data A ′ in which erroneous detection data is reduced.
FIG. 2C shows a cross-sectional view of the three-dimensional data A ′ after the erroneous detection data reduction process. In FIG.2 (c), a horizontal axis is time and a vertical axis | shaft is height.

  Here, as an example, for the intensity signal, a threshold value is a signal value corresponding to a low reception SN ratio or a low reception intensity that provides a fixed distance measurement accuracy, and a lane width or road surface height for the distance signal. A signal value corresponding to the following distance is set as a threshold value.

FIG. 3 is an explanatory view showing an operation example of the height / distance gate device 22, and shows a cross-sectional view of the three-dimensional data A ′ after the false detection data reduction processing when the viewpoint is placed in the traveling direction of the vehicle 3. Yes.
FIG. 3A shows an example of the three-dimensional data A ′ (see black dots) of the vehicle 3 when the tire is detected. The three-dimensional data A ′ includes the roof portion, the door portion, the tire portion, and the vehicle. Includes floor and road surface.
FIG. 3B shows an example of the three-dimensional data A ′ of the vehicle 3 when there is no tire and only the vehicle body is detected. The three-dimensional data A ′ includes the roof portion, the door portion, and the vehicle floor. Including parts and road surfaces.

  In FIG. 3, the height / distance gate device 22 sets an upper limit value H1 and a lower limit value H2 in the height direction, and an upper limit value S1 and a lower limit value S2 in the distance direction, and is an area surrounded by these four values. The three-dimensional data B after gate processing for only (corresponding to the tire portion) is extracted.

For example, the lower limit value H2 in the height direction is set to the standard deviation value of the altitude relative to the road surface, and the upper limit value H1 is 120 mm which is the lowest ground altitude of a domestic vehicle (Automobile Handbook 2007-2008, vol. 54, Nissan, Fairlady) Z VersionST).
Further, in order to detect only the rubber part of the tire, the upper limit value H1 in the height direction is set to 82.5 mm (the automobile handbook 2007-2008, vol. 54) which is the minimum value of the thickness of the tire sold in Japan. , Move Custom RS2WD).

On the other hand, the lower limit value S2 in the distance direction is set to the minimum distance between the three-dimensional image capturing apparatus 1 and the vehicle 3 obtained from the three-dimensional data A ′ after the erroneous detection data reduction process.
The upper limit value S1 in the distance direction is 295 mm (the automobile handbook 2007-2008, vol. 54, Mitsubishi Fuso Super Great FP-R4 × 2), which is the maximum value of the tire width sold in the country, and false detection data. It is set to the sum of the minimum distance between the three-dimensional image pickup apparatus 1 and the vehicle 3 obtained from the three-dimensional data A ′ after the reduction processing.

Thereafter, the height histogram calculation device 23 divides the detection region R into N pieces in the height direction to set the height bin Bh, and the three-dimensional data B after the gate processing (false detection within the detection region R). The height histogram C is calculated for the electrical signal corresponding to the distance after the data reduction processing.
For example, the width of the height bin Bh is set to a value corresponding to the spatial resolution in the height direction of the three-dimensional data A acquired by the three-dimensional image capturing apparatus 1.

  Next, the data existence probability calculation device 24 divides the number n of bins in which a or more three-dimensional data B after the false detection data reduction processing and the gate processing is stored based on the height histogram C and the height direction. A ratio n / N with the total number N of bins is calculated, and an electric signal corresponding to the data existence probability P is calculated.

Here, the upper limit value H1 of the height set by the height / distance gate device 22 is the minimum ground altitude of the domestic vehicle, and the lower limit value H2 is the standard deviation value of the altitude relative to the road surface.
The lower limit value S2 of the distance is the minimum distance between the three-dimensional image capturing apparatus 1 and the vehicle 3 obtained from the three-dimensional data A ′ after the erroneous detection data reduction process.
Further, the upper limit value S1 in the distance direction is the minimum value between the three-dimensional imaging device 1 and the vehicle 3 obtained from the maximum value of the tire width sold in the country and the three-dimensional data A ′ after the erroneous detection data reduction processing. Sum with distance.

As shown in FIG. 3A, when a tire is present in the detection region R, a histogram C in which the three-dimensional data B exists is present at any position of the set height bin Bh. P is 100%.
On the other hand, as shown in FIG. 3B, when there is no tire in the detection region R, since the three-dimensional data B exists only in a part of the set height bin Bh, the histogram C is biased. The data existence probability P is, for example, 25%.

FIG. 4 is an explanatory view showing an operation example of the detection flag generator 25, and shows an example of the detection flag F output when the vehicle 3 passes in the detection region R of the three-dimensional image pickup apparatus 1.
FIG. 4A shows a cross-sectional view of the three-dimensional data B after gate processing (after false detection data reduction processing) when the viewpoint is placed on the side of the vehicle 3, and the horizontal axis represents time, vertical The axis is height.

FIG. 4B shows the data existence probability P corresponding to the ratio n / N of the number of height bins n storing the three-dimensional data B with respect to the bin total number N of the set height bin Bh.
As shown in FIG. 3A, when only the three-dimensional data B in the vicinity of the tire is extracted, the data existence probability P is high when the tire is detected.

  In addition, regarding the detection threshold Th, the case where the detection threshold Th is 25% is taken as an example assuming the tire outer diameter Dw = 40 cm and the tire cross-section height H = 5 cm. However, the value of the detection threshold Th is not particularly limited, and can be arbitrarily set according to the design specifications of the vehicle 3, detection accuracy requirements, and the like.

  In general, the reflection intensity from the tire is constant with respect to the vehicle 3, but the reflection intensity from the wheel portion depends on the vehicle 3, and the ranging accuracy may be deteriorated. Therefore, regarding the detection threshold Th, for example, the ratio 2H / Dw of the tire cross-sectional height H with respect to the tire outer diameter Dw is set, and the detection threshold Th is set so as to detect the tire reliably.

FIG. 4C shows an example of the detection flag F (detection trigger).
When the data presence probability P calculated by the data presence probability calculation device 24 is higher than the detection threshold Th, the detection flag generation device 25 sets the detection flag F to the H level and generates a signal corresponding to axle detection.

Furthermore, in order to reduce false detection, as shown in FIG. 5, two types of detection threshold values, an upper limit detection threshold value and a lower limit detection threshold value, may be set.
FIG. 5 is an explanatory diagram showing an example of a detection flag according to the first embodiment of the present invention.
In general, the data existence probability is low on the ground and the body head portion, but depending on the type of vehicle, the altitude of the body portion may be lower than the driver seat portion, such as a large truck.

In this case, depending on the height gate to be set, the data existence probability may be high even in the body portion, and if a single detection threshold is set, there is a possibility of erroneous detection.
However, since there is a gap between the tire and the vehicle body, paying attention to the temporary decrease in the data existence probability, in order to detect this, as shown in FIG. Set.

In FIG. 5, the state with respect to each detection threshold value of the existence probability of each data is shown as a time-series detection state p, q, r, s (see black circles) as an example.
In FIG. 5, when the lower limit detection threshold is exceeded for the first time, state p is set, then when the upper limit detection threshold is exceeded, state q is set, and when the upper limit detection threshold is exceeded, state r is set, and finally the lower limit detection threshold is set. The case where it falls below is referred to as state s, and the time for transitioning between states p to s is referred to as St.

Here, a threshold value Tht is set in the time direction, and the case of St <Tht is detected as an axle. For example, the length of the threshold Tht is set to 0.04 seconds that can be calculated as the time for a tire having a diameter of 1 m traveling at 80 km / h to pass through the same point.
Further, the dimension of the set threshold value Tht and the acquired time St may be a distance value instead of the time.

  As described above, the axle detection system according to Embodiment 1 (FIGS. 1 to 4) of the present invention includes the three-dimensional image pickup device 1 that acquires the three-dimensional data A of the vehicle (target) 3, and the three-dimensional image. Three-dimensional data threshold processing device 21 that performs false detection data reduction processing on data A, and three-dimensional data B after gate processing in detection region R are extracted from three-dimensional data A ′ after false detection data reduction processing. The height / distance gate device 22 and the detection area are divided in the height direction to create a total of N height bins Bh, and the height of the three-dimensional data B after gate processing (after false detection data reduction processing) A height histogram calculation device 23 that creates a histogram C, a bin number n in which a predetermined number a or more of three-dimensional data B after gate processing is stored based on the height histogram C, and a bin total number N divided in the height direction The ratio n / N of the data exists A data existence probability calculation unit 24 for calculating a ratio P, and a detection flag generator 25 to generate a detection flag F on the basis of the data existence probability P.

  As a result, even when false detection data is included in the acquired three-dimensional data A, the false detection data is reduced, and after the detection region R is narrowed down by the height / distance gate, the data existence probability P is calculated and acquired. Since the detection probability in the data can be improved and the axle can be detected, erroneous detection of the axle can be suppressed.

  Further, in the data existence probability calculation device 24, the height histogram C in the height direction and the bin number n in which a (= 2) or more of three-dimensional data B after gate processing (after false detection data reduction processing) are stored. By calculating the ratio n / N with the total number N of divided bins, even when erroneous detection data is stored in the height bin Bh, bins including erroneous detection data can be eliminated. The detection probability can be improved and the axle can be detected.

  Although not particularly mentioned in the above description, the t-line portion in the time axis direction with respect to the three-dimensional data A acquired by the three-dimensional image capturing apparatus 1 or the three-dimensional data A ′ from which erroneous detection data has been removed. After performing the moving average process, the detection region R may be narrowed down to calculate the data existence probability P, and the detection flag F may be generated. Thereby, the ranging accuracy of the vehicle 3 (object) can be improved, and the axle can be detected with higher accuracy.

Further, the case where the three-dimensional image capturing apparatus 1 is a laser image measuring apparatus has been described as an example, but the present invention is not limited to this, and an apparatus based on various passive and active non-contact three-dimensional measuring systems. If it is.
For example, a stereo method using camera parallax or an active stereo method, a focus method using a difference in focus position, or a time-of-flight method using a time difference between irradiation light and reception light can be applied.

  Further, although the case where the three-dimensional data A acquired by the three-dimensional image capturing apparatus 1 is an electric signal corresponding to intensity data and distance data has been described as an example, it may be an electric signal corresponding only to distance data. .

  In addition, the thresholds for the intensity and distance set by the three-dimensional data threshold processing device 21 are, for the intensity signal, a signal value corresponding to a low reception S / N ratio or a low reception intensity that provides a fixed distance measurement accuracy. For the distance signal, the threshold value is the signal value corresponding to the distance below the lane width or road height, but the threshold value for strength and distance is variably set as a function that depends on the distance measurement accuracy (distance). Also good.

  The upper limit value H1 and lower limit value H2 in the height direction and the upper limit value S1 and lower limit value S2 in the distance direction set by the height / distance gate device 22 are fixed values. You may variably set as a function depending on the distance from 1.

The width of the height bin Bh is set as a variable for the distance because it is a value corresponding to the spatial resolution in the height direction, but may be a fixed width.
Furthermore, although the detection threshold Th set by the detection flag generator 25 is a fixed value, it may be variably set as a function depending on the distance.

Embodiment 2. FIG.
In the first embodiment (FIG. 1), only the detection device 2 for detecting the axle is provided. However, as shown in FIG. 6, the detection device 2a for detecting the vehicle is added, and the axle number counting device 4 is provided. It may be provided.
FIG. 6 is a block diagram showing the overall configuration of the axle detection system according to Embodiment 2 of the present invention. The same components as those described above (see FIG. 1) are denoted by the same reference numerals as those described above. A detailed description will be omitted with “a” added later.

In FIG. 6, the axle detection system includes a three-dimensional image pickup device 1 that acquires three-dimensional data A of the vehicle 3, two systems of detection devices 2, 2 a arranged side by side, and an axle number counting device 4. Yes.
The detection device 2a has the same configuration as the detection device 2 (see FIG. 1), and generates a detection flag Fa (second detection flag) for the vehicle body of the vehicle 3.

  The axle number counting device 4 counts the number of axles G of the detected vehicle based on the detection flags F and Fa obtained from the detection devices 2 and 2a. From the axle number counting device 4, an electrical signal corresponding to the number G of axles of the detected vehicle is output.

Next, the operation of the second embodiment of the present invention shown in FIG. 6 will be described with reference to FIGS.
As described above, the three-dimensional image capturing apparatus 1 acquires the three-dimensional data A of the vehicle 3 when the vehicle 3 crosses the detection region R.

  As described above (FIGS. 2 to 4), the detection device 2 divides only the detection region near the tire by the height bin Bh, and detects the axle from the data existence probability P of the data stored in the height bin Bh. Processing is performed to generate a detection flag F.

  On the other hand, the detection device 2a divides the height bin Bha using the vehicle body (the roof portion and the door portion) as a detection region, performs vehicle detection processing from the data existence probability of the data stored in the height bin, and detects the detection flag. Fa is generated.

FIG. 7 is an explanatory view showing an operation example of the height / distance gate device of the detection device 2a according to the second embodiment of the present invention. The detailed description is omitted.
FIG. 7 shows a setting example of the height / distance gate for vehicle detection in relation to the cross-sectional view of the three-dimensional data A ′ after the erroneous detection data reduction processing.

  In FIG. 7, a lower limit value H2a in the height direction in the height / distance gate device of the detection device 2a is set to, for example, a minimum ground altitude of 120 mm of a domestic vehicle, and the upper limit value H1a in the height direction is a three-dimensional image pickup device. 1 is set to the maximum obtainable height.

On the other hand, the lower limit value S2a in the distance direction is the minimum distance between the three-dimensional image pickup device 1 and the vehicle 3 obtained from the three-dimensional data A ′ after the erroneous detection data reduction process, and the upper limit value S1 in the distance direction is It is the width of the traveling lane. The number of divisions of the height bin Bha is two.
When the vehicle 3 enters the detection region R, the data existence probability Pa is increased in the detection device 2a regardless of the presence or absence of tires, and vehicle detection is performed.

FIG. 8 is an explanatory view showing an operation example of the detection flag generating device in the detection device 2a. The same components as those described above (see FIG. 4) are denoted by “a” after the reference numerals, and detailed description thereof is omitted. .
In FIG. 8, an example of the detection flag Fa that is output when the vehicle 3 passes in the detection region R of the three-dimensional image capturing apparatus 1 is shown together with the detection flag F described above.

FIG. 8A shows a cross-sectional view of the three-dimensional data A ′ after the erroneous detection data reduction processing when looking at the side of the vehicle 3, where the horizontal axis is time and the vertical axis is height. is there.
FIG. 8B shows the data existence probability Pa corresponding to the ratio na / Na of the bin number na storing the three-dimensional data A ′ with respect to the total number Na of height bins set for vehicle detection. . Similarly to the above, when the vehicle body is detected, the data existence probability Pa increases.

FIG. 8C shows an example of a detection flag Fa for vehicle detection.
When the data existence probability Pa is higher than the detection threshold value Tha, the detection flag generator in the detection device 2a sets the detection flag Fa to H level and generates a signal corresponding to vehicle detection.

  FIGS. 8D and 8E are the same as those described above (FIGS. 4B and 4C), and FIG. 8D is a three-dimensional view of the total number N of height bins set for axle detection. FIG. 8E shows an example of the detection flag F for the axle detection, which shows the data existence probability P corresponding to the ratio n / N of the bin number n storing the data A ′.

  Finally, the axle number counting device 4 counts the number of times the detection flag F for the axle is at the H level during the period in which the detection flag Fa for the vehicle is at the H level, and outputs an axle detection signal corresponding to the axle number G of the detected vehicle. Output.

As described above, the axle detection system according to the second embodiment (FIGS. 6 to 8) of the present invention includes the axle number counting device 4 that counts the detected axle number G of the vehicle, and generates a detection flag. Based on the data existence probability P, the device (detection device 2, 2a) generates a detection flag F (first detection flag) for the axle and a detection flag Fa (second detection flag) for the vehicle.
The axle number counting device 4 counts the detected number of axles G of the vehicle by counting the number of occurrences of the detection flag F during the generation period of the detection flag Fa.

  Thereby, even when erroneous detection data is included in the three-dimensional data A, the detection probability in the acquired data can be improved and the vehicle and the axle can be detected with high detection accuracy, as described above, and the vehicle detection period The number of axles G can be counted by counting the number of axles G.

In FIG. 6, the detection devices 2, 2 a are two parallel devices, but a single detection device 2 includes a plurality of detection functions. Two detection flags F and Fa may be generated by detecting two of the axles.
In addition, although one 3D image capturing apparatus 1 is used, two 3D image capturing apparatuses may be provided in parallel, and individual 3D data may be input to the detection apparatuses 2 and 2a.

  In FIG. 7, as an example of setting the height / distance gate for vehicle detection, the lower limit value H2a in the height direction is set to the minimum ground altitude of 120 mm for domestic vehicles, and the upper limit value H1a in the height direction is set to the tertiary. The maximum height that can be acquired by the original image pickup device 1 is set, the lower limit value S2a in the distance direction is set as the minimum distance between the three-dimensional image pickup device 1 and the vehicle 3, and the upper limit value S1a in the distance direction is set as the lane width. Although set, the lower limit value H2a in the height direction may be set to the standard deviation value of the altitude relative to the road surface.

Similarly to the above, the upper limit value H1a and the lower limit value H2a in the height direction set by the height / distance gate device in the detection device 2a, and the upper limit value S1a and the lower limit value S2a in the distance direction are fixed values. However, it may be variably set as a function depending on the distance from the three-dimensional image capturing apparatus 1.
Furthermore, although the axle number counting device 4 outputs only the axle detection signal corresponding to the axle number G of the detected vehicle, it may output an electrical signal corresponding to axle detection and vehicle detection.

Embodiment 3 FIG.
In the second embodiment (FIG. 6), the axle number counting device 4 is provided. However, as shown in FIG. 9, the second three-dimensional image pickup device (hereinafter simply referred to as “three-dimensional image pickup device”). While providing 1b and the 2nd detection apparatus (henceforth a "detection apparatus" simply) 2b, you may provide the vehicle width measurement apparatus 5. FIG.
FIG. 9 is a block diagram showing the overall configuration of the axle detection system according to Embodiment 3 of the present invention. Components similar to those described above (see FIGS. 1 and 6) are denoted by the same reference numerals as described above, Alternatively, “b” is appended after the reference numerals and the detailed description is omitted.

  In FIG. 9, the axle detection system includes three-dimensional image pickup devices 1 and 1b that acquire three-dimensional data A and Ab of the vehicle 3, three-dimensional data B and Bb after gate processing (after erroneous detection data reduction processing), and Detection devices 2 and 2b that generate detection flags F and Fb, and a vehicle width measurement device 5 are provided.

  The three-dimensional image pickup devices 1 and 1b are arranged on both sides of the roadside belt so as to pick up both side surfaces of the vehicle 3, and are mutually in the vehicle width W (tread width Wt) direction of the vehicle 3 to be detected. They are spaced apart at an interval L1.

  The detection device 2b has the same configuration as the detection device 2 (see FIG. 1), and includes a three-dimensional data threshold processing device 21b, a height / distance gate device 22b, a height histogram calculation device 23b, a data existence probability calculation device 24b, and A detection flag generating device 25b (none of which is shown) is provided, thereby generating three-dimensional data Bb and detection flag Fb after gate processing (after erroneous detection data reduction processing).

The vehicle width measuring device 5 uses the three-dimensional data B, Bb after the gate processing (after the false detection data reduction processing) extracted in the detection area in the detection devices 2, 2b, to use the vehicle width W (or Tread width Wt) is measured.
An electric signal corresponding to the vehicle width W is output from the vehicle width measuring device 5.

Next, the operation according to the third embodiment of the present invention shown in FIG. 9 will be described.
The three-dimensional imaging devices 1 and 1b acquire the three-dimensional data A and Ab of the vehicle 3 when the vehicle 3 crosses the detection regions R and Rb.
As described above, the detection devices 2 and 2b divide only the detection area near the tire into height bins, detect the axle from the data existence probability of the data stored in the height bin, and perform gate processing (false detection). Three-dimensional data B and Bb after data reduction processing are generated.

  The vehicle width measuring device 5 is subjected to gate processing (after erroneous detection data reduction processing) in the detection region R output from the height / distance gate devices 22 and 22b (not shown) in the detection devices 2 and 2b. Using the three-dimensional data B and Ba, the minimum distances d1 and d1b from the respective three-dimensional image pickup devices 1 and 1b to the side surface of the vehicle 3 are measured.

  Further, the vehicle width measuring device 5 uses the mutual distance L1 between the 3D image capturing device 1 and the 3D image capturing device 1b, and the vehicle width W (or the vehicle width) of the vehicle 3 according to the following equation (1). The tread width Wt of the left and right tire side surfaces approximated to W is calculated, and an electric signal corresponding to the vehicle width W is output.

  W = L1-d1-d1b (1)

  As described above, the axle detection system according to Embodiment 3 (FIG. 9) of the present invention is a three-dimensional image pickup device that includes the vehicle (object) 3 in addition to the configuration described above (FIG. 1). 1, a three-dimensional image pickup device 1 b that acquires three-dimensional data Ab (second three-dimensional data) of the vehicle 3, a detection device 2 b, and a vehicle width measurement device that measures the vehicle width W of the vehicle 3. And 5.

  The detection device 2b includes a second three-dimensional data threshold processing device 21b that performs erroneous detection data reduction processing on the three-dimensional data Ab, and a detection region Rb (first) from the second three-dimensional data after the erroneous detection data reduction processing. And a second height / distance gate device 22b for extracting the three-dimensional data Bb (second three-dimensional data) after the gate processing in the second detection region).

The vehicle width measuring device 5 measures the vehicle width W by the equation (1) using the mutual distance L1 between the three-dimensional image pickup devices 1 and 1b and the three-dimensional data B and Bb after the gate processing.
As a result, even if erroneous detection data is included in the three-dimensional data A and Ab acquired by the two three-dimensional imaging devices 1 and 1b installed on both sides of the roadside belt so as to sandwich the vehicle 3, The data 3 is reduced, the detection areas R and Rb are narrowed down by the height / distance gate, the measurement accuracy of the acquired data is improved, and the minimum distances d1 and d1b to the vehicle 3 can be measured with high accuracy. The vehicle width W can be calculated.

  In the above description, in the height / distance gate devices 22 and 22b in the detection devices 2 and 2b, the gate is set so as to detect the vicinity of the tire, and the tread width approximated to the vehicle width W by the vehicle width measurement device 5 is set. Although Wt is calculated, in order to calculate the vehicle width W of the vehicle 3 with high accuracy, the gate may be set so as to detect the vehicle body as in the second embodiment (FIG. 7).

  Further, in the height / distance gate devices 22 and 22b in the detection devices 2 and 2b, the upper limit values H1 and H1b in the height direction are set to the height of the side mirror of the vehicle 3 so as to be lower than the side mirror. The minimum distances d1 and d1b to a certain vehicle door may be obtained. In this case, the vehicle width W of the vehicle 3 that does not include the side mirror can be measured.

In FIG. 9, only the electrical signal corresponding to the vehicle width W is output. However, as in the first embodiment (FIG. 1), the electrical signal corresponding to the detection of the axle and the vehicle 3 is output. Also good.
Further, as in the above-described second embodiment (FIG. 6), by providing the axle number counting device 4, an electrical signal corresponding to the axle number G may be output.

  Further, the vehicle width W of the vehicle 3 is calculated using the three-dimensional data B and Bb after gate processing (after erroneous detection data reduction processing) extracted in the detection regions R and Rb. The three-dimensional data A ′ may be used.

Further, the added detection device 2b only needs to include at least the three-dimensional data threshold processing device 21b and the height / distance gate device 22b (not shown).
In FIG. 9, the two detection devices 2 and 2 b are installed. However, a plurality of detection functions are included in one detection device 2, for example, two three-dimensional image imaging devices installed in the roadside belt. 1, 1b may be processed by one detection device 2.

  Further, FIG. 9 shows the case where the three-dimensional image pickup device 1b and the detection device 2b are added to the configuration of FIG. 1, but the three-dimensional image pickup device 1b and the detection device 2b are added to the configuration of FIG. Also good. In this case, not only the vehicle width W but also the axle number G can be measured.

Embodiment 4 FIG.
Although the vehicle width measuring device 5 is provided in the third embodiment (FIG. 9), a vehicle speed / vehicle length measuring device 6 may be provided as shown in FIG.
FIG. 10 is a block diagram showing the overall configuration of an axle detection system according to Embodiment 4 of the present invention. Components similar to those described above (see FIGS. 1, 6, and 9) are denoted by the same reference numerals. In addition, “c” is added after the reference numeral, and the detailed description is omitted.

  In FIG. 10, the axle detection system includes a three-dimensional image pickup device 1 that acquires three-dimensional data A of the vehicle 3 and a third three-dimensional image pickup device that acquires three-dimensional data Ac of the vehicle 3 (hereinafter simply “ 1c), a detection device 2 that generates a detection flag F, and a third detection device that generates a detection flag Fc (third detection flag) (hereinafter simply referred to as “detection device”). 2c and a vehicle speed / vehicle length measuring device 6 are provided.

The three-dimensional image pickup apparatuses 1 and 1c are installed on both sides of the roadside belt and spaced apart from each other with a distance L1 in the vehicle width direction, as described above (see FIG. 9).
Further, the added three-dimensional image capturing device 1c is spaced apart from the three-dimensional image capturing device 1 by a distance Lc with respect to the traveling direction of the vehicle 3, and images the vehicle 3 by the detection region Rc.

  The detection device 2c has the same configuration as the detection device 2 (see FIG. 1), and includes a three-dimensional data threshold processing device 21c, a height / distance gate device 22c, a height histogram calculation device 23c, a data existence probability calculation device 24c, and A detection flag generator 25c (none of which is shown) is provided, and a detection flag Fc is generated based on the three-dimensional data Ac.

The vehicle speed / vehicle length measuring device 6 measures at least one of the vehicle speed V and the vehicle length z of the vehicle 3 using the detection flags F and Fc for the vehicle 3 extracted in the detection regions R and Rc in the detection devices 2 and 2c. To do.
An electric signal corresponding to the vehicle speed V or the vehicle length z is output from the vehicle speed / vehicle length measuring device 6.

Next, the operation according to the fourth embodiment of the present invention shown in FIG. 10 will be described.
The three-dimensional imaging devices 1 and 1c acquire the three-dimensional data A and Ac of the vehicle 3 when the vehicle 3 crosses the detection regions R and Rc.
Similarly to the above, the detection devices 2 and 2c divide the vehicle body into height bins, perform vehicle detection based on the data existence probabilities P and Pc of the three-dimensional data A ′ stored in the height bin, and detect the detection flag. F and Fc are generated.

  The vehicle speed / vehicle length measurement device 6 uses the detection time difference Δt of the detection flags F and Fc for the vehicle 3 acquired by the detection devices 2 and 2c and the interval Lc of the three-dimensional image pickup devices 1 and 1c with respect to the vehicle traveling direction. The vehicle speed V is calculated by the following equation (2).

  V = Lc / Δt (2)

  The vehicle speed / vehicle length measuring device 6 calculates the vehicle length z by the following equation (3) using the detection time t for the vehicle 3 based on the detection flags F and Fc and the vehicle speed V.

  z = V × t (3)

  The vehicle speed / vehicle length measuring device 6 derives the vehicle speed V and the vehicle length z as shown in equations (2) and (3), and outputs an electrical signal corresponding to the vehicle speed V and the vehicle length z.

  As described above, the axle detection system according to Embodiment 4 (FIG. 10) of the present invention has a vehicle (object) 3 of the three-dimensional image pickup device 1 in addition to the configuration described above (FIG. 1). A third-dimensional imaging device 1c that is spaced apart in the moving direction and acquires the three-dimensional data Ac (third three-dimensional data) of the vehicle 3, and a third that performs false detection data reduction processing on the three-dimensional data Ac. The 3D data threshold processing device 21c and the third 3D data Bc after gate processing in the detection region Rc (third detection region) are extracted from the third 3D data after the false detection data reduction processing. 3 height / distance gate device 22c and detection region Rc are divided in the height direction to create a total of Nc bins, and a third height histogram Cc of third three-dimensional data Bc after gate processing Height histogram calculation device for creating 3c and a ratio nc / between the number of bins nc storing a predetermined number ac or more of the third three-dimensional data Bc after gate processing based on the third height histogram Cc and the total number of bins Nc divided in the height direction A third data existence probability calculation device 24c that calculates Nc as a third data existence probability Pc, and a third detection flag that generates a detection flag Fc (third detection flag) based on the third data existence probability Pc A generator 25c and a vehicle speed / vehicle length measuring device 6 for measuring at least one of the detected vehicle speed V and vehicle length z of the vehicle 3 are provided.

The vehicle speed / vehicle length measuring device 6 uses the detection time difference Δt of the vehicle 3 based on the detection flags F and Fc, and the interval Lc between the 3D image pickup device 1 and the 3D image pickup device 1c to determine the vehicle speed V or the vehicle. The length z is measured.
Thereby, even when erroneous detection data is included in the acquired three-dimensional data A and Ac, the erroneous detection data is reduced, the detection area is narrowed down by the height / distance gate, and the data existence probabilities P and Pc are calculated. The detection probability in the acquired data is improved, the vehicle can be detected with high detection accuracy, the vehicle speed V can be measured by measuring the detection time difference Δt of the vehicle 3, and the detection time t of the vehicle 3 and the vehicle speed V Can be used to measure the vehicle length z.

Further, the passing direction of the vehicle 3 can be grasped from the detection times t and tc of the vehicle 3 in the detection devices 2 and 2c.
When the traveling lane of the vehicle 3 is one-way, the passing direction of the vehicle 3 corresponds to forward / backward movement of the vehicle 3. For example, in FIG. 10, the detection device 2 c detects the vehicle 3 first. In this case, it can be predicted that the vehicle 3 is moving forward.
Conversely, when the detection device 2 detects the vehicle 3 first, it can be predicted that the vehicle 3 is moving backward (or running backward).

Further, the outer diameter Dt of the tire of the vehicle 3 can be measured from the vehicle speed V measured by the vehicle speed / vehicle length measuring device 6 and the detection time of the detection flag F when the detection device 2 detects the axle. .
As a result, in the vehicle 3 in which the mudguard is installed behind the tire, even if the mudguard is erroneously detected as a tire, it is possible to determine whether the tire or mudguard is detected based on the detection width, thereby improving the detection accuracy of the axle. can do.

  In the above description, the vehicle speed / vehicle length measuring device 6 outputs only the electrical signal corresponding to the vehicle speed V or the vehicle length z, but in addition to this, as in the first embodiment (FIG. 1) described above. An electric signal corresponding to detection of the axle or the vehicle 3 may be output.

  Further, as in the above-described second embodiment (FIG. 6), by providing the axle number counting device 4, an electrical signal corresponding to the number of axles G may be output, and in the third embodiment (FIG. 9). ), The electric signal corresponding to the vehicle width W may be output by providing the vehicle width measuring device 5.

In FIG. 10, the three-dimensional imaging devices 1 and 1 c are arranged on both sides of the roadside band, but may be arranged on one side of the roadside band.
Furthermore, although the two detection devices 2 and 2c are provided, a plurality of detection functions are included in the single detection device 2, and the two three-dimensional image pickup devices 1 and 1c installed in the roadside belt are The processing may be performed by only one detection device 2.

Embodiment 5 FIG.
In the second embodiment (FIG. 6), the detection device 1 and the axle number counting device 4 are provided separately. However, as shown in FIG. 11, an axle detection device 7 including the axle number counting device 4 is provided. Also good.
FIG. 11 is a block diagram showing an overall configuration of an axle detection system according to Embodiment 5 of the present invention. Components similar to those described above are denoted by the same reference numerals as those described above, or “d” after the reference numerals. A detailed description will be omitted.

In FIG. 11, the axle detection system includes a 3D imaging device 1 that acquires 3D data A, and an axle detection device 7 that counts the number of axles G based on the 3D data A.
In addition to the three-dimensional data threshold processing device 21, the height / distance gate device 22, and the height histogram calculation device 23, the axle detection device 7 includes a data existence probability calculation device 30, a threshold calculation device 40, and an axle number counting device. And 4d.

  Based on the height histogram C divided into N, the data existence number calculation device 30 calculates the number n of bins in which a predetermined number a or more of three-dimensional data after gate processing (after reduction of false detection) is stored. Yc (the number of bins in which data is stored) is calculated, and an electric signal corresponding to the data existence number Yc is input to the threshold value calculation device 40 and the axle number counting device 4d.

The threshold value calculation device 40 automatically sets a threshold value Thd for detecting an axle using the maximum likelihood estimation method for the data existence number Yc within an arbitrary time range (or during vehicle detection), and corresponds to the threshold value Thd. The electrical signal is input to the axle number counting device 4d.
The axle number counting device 4d counts the number of axles G using the data existence number Yc and the threshold value Thd, and outputs an electric signal corresponding to the axle number G as a detection result of the axle detecting device 7.

  Although not specifically shown here, as described above (FIG. 10), the three-dimensional image pickup device 1c and the vehicle speed / vehicle length measuring device 6 are added, and the vehicle speed V measured by the vehicle speed / vehicle length measuring device 6 and At least one of the vehicle length z may be added as input information of the axle number counting device 4d.

Next, the threshold value calculation device 40 will be specifically described with reference to FIGS. 12 and 13. FIG. 12 is a block diagram showing the configuration of the threshold value calculation device 40 in FIG.
In FIG. 12, the threshold value calculation device 40 includes a moving average processing device 41, a sorting device 42, a model setting device 43, and a threshold value calculation unit 44.

The moving average processing device 41 performs moving average processing on the data existence number Yc (x) (x is the number of pixels) input from the data existence number calculating device 30, and outputs an electric signal corresponding to the moving average processing result 41d. Input to the sorting device 42.
The sorting device 42 performs a process of rearranging the data existence number Yc (x) in ascending order based on the moving average processing result 41d, and generates an electric signal corresponding to the rearrangement processing result 42d of the data existence number Yc (x) as a model setting device. 43.

The model setting device 43 sets the model 43d used by the threshold value calculation unit 44 based on the rearrangement processing result 42d, and inputs an electrical signal corresponding to the model 43d to the threshold value calculation unit 44.
Based on the model 43d, the threshold value calculation unit 44 calculates a threshold value Thd for separating the number of data existing on the axle and the number of data existing outside the axle using the maximum likelihood estimation method. Output.

  FIG. 13 is an explanatory diagram showing the processing contents in the threshold value calculation device 40. FIG. 13 (a) schematically shows the number of existing data Yc input to the moving average processing device 41, and FIG. 13 (b) shows the sorting device. FIG. 13C schematically illustrates data processing by the threshold value calculation unit 44. FIG.

As shown in FIG. 13A, the data existence number Yc input to the threshold value calculation device 40 is high in the axle portion and is lower in the other portions than the axle portion.
In FIG. 13C, the threshold calculation unit 44 sets a boundary D (variable parameter) that separates the ground and the axle.
Further, the threshold value calculation unit 44 regards the left side of the boundary D as the ground area and the right side of the boundary D as the axle area, and the probability density function M1 of the data existence number of the ground area and the probability of the data existence number of the axle area. The density function M2 is obtained.

  Here, the probability density functions M1 and M2 represent the probability of acquiring bins of n tire candidates from N bins in the definition of the number of axle data. Therefore, in the model setting device 43, the following formula ( It shall follow a binomial distribution as shown in 1).

The probability density functions M1 and M2 are not limited to the binomial distribution of Expression (1), and may be defined by other functions.
In Expression (1), N represents the total number of tire presence counts, n represents the tire presence count, and P represents the probability of occurrence of each data. This time, in order to estimate the optimum values of the two models of the probability density functions M1 and M2, there are variable parameters P1 and P2 corresponding to the respective models.
Next, the log likelihood L (θ) can be described by the following equation (2).

In Equation (2), θ is a variable parameter (P1, P2, D) for searching for the maximum likelihood value, n is a tire presence count, and f * (n) is a weight.
Further, f (x, n: θ) is expressed as the following formula (3).

In Expression (3), u (x) represents a uniform distribution, d0 represents the data number of the ground area, and d1 represents the pixel number of the axle area.
Further, the weights f * (n), f ′ * (n), and f ″ * (n) are the number of occurrences of the data existence number in each area, and the following expressions (4), (5), and ( It is expressed as 6).

In Formula (4), Formula (5), and Formula (6), the greater the number of occurrences, the greater the value.
Further, in Expressions (5) and (6), L (n _i ) represents the number of occurrences of each tire presence count, and n _i represents the number of data existing for the pixel number i.
Substituting the above relationship into equation (2), the log likelihood L (θ) is expressed as in the following equation (7).

  If the first item of the equation (7) is g1 (x), the first item g1 (x) can be expanded as in the following equation (8).

  Similarly, when the second item of Expression (8) is g2 (x), the second item g2 (x) can be expanded as shown in Expression (9) below.

  From the equations (8) and (9), the log likelihood L (θ) is expressed as the following equation (10).

However, the first item of the equation (10) is a constant that does not depend on the probability density functions P1 and P2, and in this method, the purpose is to obtain the maximum log likelihood L (θ). can do.
Therefore, Expression (10) can be expressed as the following Expression (11).

The probability density functions P1 and P2 and the boundary D that maximize the log likelihood L (θ) of the equation (11) are derived by circular calculation. The threshold value is n corresponding to the boundary D when the log likelihood L (θ) is maximized.
Since the number of data in the axle portion is not always constant and varies depending on the environment, the use of the above method has an advantage of reducing erroneous detection of the axle.

Next, the threshold value calculation device 40 will be specifically described with reference to FIGS. 14 and 15. FIG. 14 is a block diagram showing a configuration of the axle number counting device 4d.
In FIG. 14, the axle number counting device 4d uses the threshold value Thd from the threshold value calculation device 40 and the data existence number Yc from the data existence number calculation device 30 as input information, and the threshold value processing device 45 (first threshold value processing). Device), a distance value calculation device 46, a window setting processing device 47, an in-window integration processing device 48, a threshold processing device 49 (second threshold processing device), and a labeling processing device 50.

  FIG. 15 is an explanatory diagram showing the processing contents of the axle number counting device 4d. FIG. 15 (a) schematically shows the threshold processing by the threshold processing device 45, and FIG. 15 (b) shows the window by the window setting processing device 47. FIG. 15C schematically shows the integration processing by the in-window integration processing device 48, and FIG. 15D schematically shows the labeling processing by the labeling processing device 50.

In the axle number counting device 4d, first, the threshold value processing device 45 on the input side uses the threshold value Thd input from the threshold value calculation device 40 to perform threshold processing on the data existence number Yc, and binarized pixels. Data (the number of existing data) 45d (see FIG. 15A) is input to the distance value calculation device 46.
The distance value calculation device 46 uses the binarized data existence number 45d and, for example, the vehicle speed V of the vehicle 3 obtained from the vehicle speed / vehicle length measurement device 6 (see FIG. 10), and the number of pixels per unit distance. Rp [m ⁻¹ ] is calculated by the following equation (12).

In Expression (12), fn represents a scan rate [Hz] for acquiring data of one line.
Hereinafter, the distance value calculation device 46 inputs an electric signal corresponding to the calculated pixel number Rp per unit distance to the window setting processing device 47.

The window setting processor 47 sets a window size (window width) Sd [m] used in the next window integration processor 48 based on the number of pixels Rp, and an electric signal corresponding to the window size Sd is set in the window. Input to the integration processor 48.
As a specific example of the window size Sd, an average axle interval (0.3 [m]) in a vehicle having a plurality of axles (such as a three-axis vehicle) is set.

The in-window integration processing device 48 performs integration processing for converting the window size Sd set by the window setting processing device 47 into a pixel value by using the number of pixels Rp obtained by the equation (12), and performs pixel unit The window value (pixel set value) Wp converted to is calculated as in the following equation (13).

That is, the in-window integration processing device 48 uses the window size Sd for the binarized data existence number in FIG. 15A to calculate the sum of the values in the window as shown in FIG. Movement integration using the value of each pixel is performed.
Thereafter, the in-window integration processing device 48 inputs an electric signal corresponding to the window value Wp in pixel units to the threshold processing device 49.

As shown in FIG. 15C, the threshold processing device 49 performs threshold processing using the threshold Tp on the window value Wp in pixel units (see the broken line).
The threshold value Tp used for the threshold value processing at this time can be calculated as in the following formula (14).

In equation (14), Ts represents the designated axle outer diameter [m], fs represents the scan rate [Hz], and V represents the vehicle speed [m / s]. In addition, a general axle outer diameter (0.6 [m]) or the like is used as an index of the axle outer diameter Ts.
Hereinafter, the threshold processing device 49 inputs an electrical signal corresponding to the threshold processing result K using the threshold Tp to the labeling processing device 50.

  As described above, the axle detection system according to Embodiment 5 (FIGS. 11 to 15) of the present invention includes the three-dimensional image capturing apparatus 1 that acquires the three-dimensional data A of the vehicle 3 (object), and the three-dimensional And an axle detection device 7 for counting the number of axles G based on the data A.

  The axle detection device 7 includes a three-dimensional data threshold processing device 21 that performs false detection data reduction processing on the three-dimensional data A, and three-dimensional data A ′ after erroneous detection data reduction processing, after gate processing in the detection region. A height / distance gate device 22 for extracting the three-dimensional data B, and a total of N bins are created by dividing the detection area in the height direction, and a height histogram C of the three-dimensional data B after the gate processing is created. A height histogram calculation device 23, a data existence number calculation device 30 that calculates, as the data existence number Yc, the bin number n in which the predetermined number a or more of the three-dimensional data B after gate processing is stored based on the height histogram C; A threshold calculation device 40 that automatically sets a threshold value Thd using the maximum likelihood estimation method for the data existence number Yc, and an axle number that counts the axle number G using the data existence number Yc and the threshold Thd. Since a few devices 4d, it can be improved in the same manner as described above, axle detection accuracy.

  Further, the threshold value calculation device 40 sets the threshold value Thd to the optimum value using the data existence number of the ground portion and the data existence number of the plurality of points assumed to be the axle portion with respect to the data existence number Yc. Further, the detection accuracy of the axle can be further improved.

Embodiment 6 FIG.
In the fifth embodiment (FIG. 12), the sorting device 42 and the model setting device 43 are used as the threshold calculation device 40. However, as shown in FIG. 16, the vehicle center pixel extraction device 51 and the vehicle center pixel data are used. The existence number extraction device 52 and the second peak data existence number extraction device 53 may be used.

FIG. 16 is a block diagram showing a threshold value calculation device 40e according to Embodiment 6 of the present invention. Components similar to those described above (FIG. 12) are denoted by the same reference numerals as those described above, or “e” after the reference numerals. The detailed description is omitted.
The overall configuration of the axle detection system according to Embodiment 6 of the present invention is as shown in FIG.

  In FIG. 16, the threshold value calculation device 40e uses the data existence number Yc as input information in order to determine the threshold value Thd for detecting the axle, and includes a moving average processing device 41, a vehicle center pixel extraction device 51, and a vehicle center pixel. A data existence number extraction device 52, a second peak data existence number extraction device 53, and a threshold value calculation unit 44e are provided.

Next, the operation of the threshold value calculation device 40e according to the sixth embodiment of the present invention shown in FIG. 16 will be described with reference to FIG.
FIG. 17 is an explanatory view schematically showing a threshold value calculation method according to Embodiment 6 of the present invention, in which the horizontal axis indicates the pixels and the vertical axis indicates the data existence number Yc.

  In FIG. 17, the first and second peaks PY1 and PY2 of the data existence number Yc, the vehicle center pixel data existence number Yc (Pm) corresponding to the vehicle center pixel Pm, the vehicle detection start pixel Pin, and the vehicle detection end pixel. Pout is shown.

In the threshold value calculation device 40e, first, the moving average processing device 41 inputs an electric signal corresponding to the aforementioned moving average processing result 41d to the vehicle center pixel extracting device 51.
The vehicle center pixel extraction device 51 uses the vehicle detection start pixel Pin and the vehicle detection end pixel Pout based on the moving average processing result 41d to extract the vehicle center pixel Pm as shown in the following equation (15).

FIG. 17 schematically shows the calculation process of the vehicle center pixel Pm based on the equation (15).
The vehicle center pixel extraction device 51 inputs an electrical signal corresponding to the vehicle center pixel Pm calculated by the equation (15) to the vehicle center pixel data existence number extraction device 52.

The vehicle center pixel data existence number extraction device 52 extracts the vehicle center pixel data existence number Yc (Pm) corresponding to the vehicle center pixel Pm from the data existence number Yc.
Hereinafter, the vehicle center pixel data existence number extraction device 52 inputs an electrical signal corresponding to the vehicle center pixel data existence number Yc (Pm) to the second peak data existence number extraction device 53.

  The second peak data existence number extracting device 53 extracts the data existence number Yc (PY2) of the second peak PY2 (second and subsequent peaks in the case of a three-axis vehicle) from the data existence number Yc.

Specifically, the second peak data existence number extraction device 53 first acquires the maximum value (first peak PY1) of the data existence number Yc and sets it as the first peak data existence number Yc (PY1).
Subsequently, it is assumed that the first peak PY1 and the second peak PY2 exist in regions opposite to each other with the vehicle center pixel Pm interposed therebetween, and the maximum value (second value) of the region opposite to the first peak PY1 (second The peak PY2) is acquired and set as the second peak data existence number Yc (PY2).

According to the above method, the amount of data used for threshold estimation can be reduced as compared with the above-described fifth embodiment, so that the effect of reducing the amount of calculation can be obtained.
Hereinafter, the second peak data existence number extraction device 53 inputs an electrical signal corresponding to the data existence number Yc (PY2) of the second peak PY2 to the threshold value calculation unit 44e.

  The threshold value calculation unit 44e sets the average value of the acquired second peak data existence number Yc (PY2) and the vehicle center pixel data existence number Yc (Pm) as the threshold value Thd, and counts the electric signal corresponding to the threshold value Thd as the number of axles. Input to device 4d.

  As described above, the threshold value calculation device 40e of the axle detection system according to Embodiment 6 (FIGS. 11, 16, and 17) of the present invention performs the moving average processing device that performs the moving average processing on the data existence number Yc. 41, a vehicle center pixel extraction device 51 that extracts the vehicle center pixel Pm based on the moving average processing result 41d, and the vehicle center pixel data existence number Yc (Pm) corresponding to the vehicle center pixel Pm among the data existence number Yc. The vehicle center pixel data existence number extracting device 52 to extract, the second peak data existence number extracting device 53 for extracting the second peak data existence number Yc (PY2) from the data existence number Yc, and the second peak data existence number. Since a threshold value calculation unit 44e that outputs an average value of Yc (PY2) and the vehicle center pixel data existence number Yc (Pm) as a threshold value Thd is provided, Similarly, it is possible to improve the detection accuracy of the axle.

  In the above description, the average value of the second peak data existence number Yc (PY2) and the vehicle center pixel data existence number Yc (Pm) is used as the threshold Thd, but it is not an average value but specified by the user. A value obtained by multiplying the coefficient may be used.

  DESCRIPTION OF SYMBOLS 1 3D image pick-up device, 2, 2a, 2b, 2c detection device, 21 3D data threshold processing device, 22 height / distance gate device, 23 height histogram calculation device, 24 data existence probability calculation device, 25 detection flag Generating device, 3 vehicle (object), 30 data existence number calculating device, 4, 4d axle number counting device, 40, 40e threshold value calculating device, 41 moving average processing device, 42 sorting device, 43 model setting device, 44, 44e Threshold calculation unit, 45 Threshold processing device, 46 Distance value calculation processing device, 47 Window setting processing device, 48 In-window integration processing device, 49 Threshold processing device, 5 Vehicle width measurement device, 50 Labeling processing device, 51 Vehicle center pixel extraction Device, 52 vehicle center pixel data existence number extraction device, 53 second peak data existence number extraction device , 6 Vehicle speed / length measurement device, 7 Axle detection device, A, Ab, Ac 3D data, A '3D data after false detection data reduction processing, B, Bb 3D data after gate processing, Bh, Bha Height bin, C Height histogram, P, Pa Data existence probability, F, Fa, Fb, Fc detection flag, G number of axles, R, Rb, Rc detection area, V vehicle speed, W vehicle width, Wt tread width, z Vehicle length, Δt Detection time difference.

Claims (7)

A first three-dimensional image capturing device that acquires first three-dimensional data of a vehicle that is an object;
An axle detection system comprising: an axle detection device that counts the number of axles based on the first three-dimensional data,
The axle detection device is
A first three-dimensional data thresholding apparatus for performing erroneous detection data reduction process with respect to the first three-dimensional data,
The first three-dimensional data after the erroneous detection data reduction process, performs a gating Filter first detection area by the height-range gate for extracting a first three-dimensional data after the gate processing A first height / distance gate device;
A first height histogram calculation device that divides the first detection region in the height direction to create a total of N bins and generates a height histogram of the first three-dimensional data after the gate processing; ,
A data existence number calculating device for calculating, as a data existence number, a bin number in which a predetermined number or more of the first three-dimensional data after the gate processing is stored based on a height histogram of the first three-dimensional data;
A threshold value calculation device that automatically sets a threshold value using a maximum likelihood estimation method for the number of data existing;
Look including the axle counting device which performs counting of the number of the axles with the data present number and said threshold value,
The threshold value calculation device includes:
A moving average processing device that performs a moving average process on the number of data existing;
Based on the moving average processing result from the moving average processing device, a sorting device that performs processing for rearranging the data existence number in ascending order;
A model setting device for setting a model for threshold calculation based on the sorting process result from the sorting device;
A threshold value calculation unit for calculating the threshold value based on the model set by the model setting device;
With
The threshold value calculation unit calculates the threshold value for separating the data existence number of axles from the data existence number other than the axles using a maximum likelihood estimation method . A first three-dimensional image capturing device that acquires first three-dimensional data of a vehicle that is an object;
An axle detection device for counting the number of axles based on the first three-dimensional data;
Axle detection system with
The axle detection device is
A first three-dimensional data threshold processing device for performing a false detection data reduction process on the first three-dimensional data;
A gate process for narrowing down the first detection area by a height / distance gate is performed on the first three-dimensional data after the false detection data reduction process, and the first three-dimensional data after the gate process is extracted. A first height / distance gate device;
A first height histogram calculation device that divides the first detection region in the height direction to create a total of N bins and generates a height histogram of the first three-dimensional data after the gate processing; ,
A data existence number calculating device for calculating, as a data existence number, a bin number in which a predetermined number or more of the first three-dimensional data after the gate processing is stored based on a height histogram of the first three-dimensional data;
A threshold value calculation device that automatically sets a threshold value using a maximum likelihood estimation method for the number of data existing;
An axle number counting device for counting the number of axles using the data existence number and the threshold;
Including
The axle number counting device is:
A first threshold value processing device for generating a binarized data existence number by performing threshold processing on the data existence number using the threshold;
A distance value calculating device that calculates the number of pixels per unit distance using a scan rate for acquiring data of one line and a vehicle speed of the vehicle;
A window setting processing device for setting a window size based on the number of pixels ;
An in-window integration processing device that performs integration processing for converting the window size into a pixel value using the number of pixels, and calculates a window value in units of pixels;
A second threshold processing device that performs threshold processing on the window value in pixel units;
An axle detection system comprising: a labeling processor that performs a labeling process on a threshold processing result from the second threshold processor and outputs the number of axles. The axle number counting device is:
A first threshold value processing device for generating a binarized data existence number by performing threshold processing on the data existence number using the threshold;
A distance value calculating device that calculates the number of pixels per unit distance using a scan rate for acquiring data of one line and a vehicle speed of the vehicle;
A window setting processing device for setting a window size based on the number of pixels;
An in-window integration processing device that performs integration processing for converting the window size into a pixel value using the number of pixels, and calculates a window value in units of pixels;
A second threshold processing device that performs threshold processing on the window value in pixel units;
A labeling processing device for performing a labeling process on the threshold processing result from the second threshold processing device and outputting the number of axles;
The axle detection system according to claim 1, further comprising: The threshold value calculation device sets the threshold value to an optimum value with respect to the data existence number, using the data existence number of the ground part and the data existence number of a plurality of points assumed to be the axle part. The axle detection system according to any one of claims 1 to 3 . A first three-dimensional image capturing device that acquires first three-dimensional data of a vehicle that is an object;
An axle detection device for counting the number of axles based on the first three-dimensional data;
Axle detection system with
The axle detection device is
A first three-dimensional data threshold processing device for performing a false detection data reduction process on the first three-dimensional data;
A gate process for narrowing down the first detection area by a height / distance gate is performed on the first three-dimensional data after the false detection data reduction process, and the first three-dimensional data after the gate process is extracted. A first height / distance gate device;
A first height histogram calculation device that divides the first detection region in the height direction to create a total of N bins and generates a height histogram of the first three-dimensional data after the gate processing; ,
A data existence number calculating device for calculating, as a data existence number, a bin number in which a predetermined number or more of the first three-dimensional data after the gate processing is stored based on a height histogram of the first three-dimensional data;
A threshold value calculation device that automatically sets a threshold value using a maximum likelihood estimation method for the number of data existing;
An axle number counting device for counting the number of axles using the data existence number and the threshold;
Including
The threshold value calculation device includes:
A moving average processing device that performs a moving average process on the number of data existing;
A vehicle center pixel extraction device that extracts a vehicle center pixel based on a moving average processing result from the moving average processing device;
A vehicle center pixel data existence number extraction device that extracts a vehicle center pixel data existence number corresponding to the vehicle center pixel in the data existence number;
A second peak data existence number extracting device for extracting a second peak data existence number from the data existence number;
An axle detection system comprising: a threshold value calculation unit that outputs an average value of the second peak data existence number and the vehicle center pixel data existence number as the threshold value. A second three-dimensional image pickup device that is arranged opposite to the first three-dimensional image pickup device so as to intervene the vehicle and acquires second three-dimensional data of the vehicle;
A second three-dimensional data threshold processing device for performing a false detection data reduction process on the second three-dimensional data;
A second height / distance gate device for extracting the second three-dimensional data after the gate processing in the second detection region from the second three-dimensional data after the erroneous detection data reduction processing;
A vehicle width measuring device for measuring the width of the vehicle;
The vehicle width measuring device includes an interval between the first three-dimensional image pickup device and the second three-dimensional image pickup device, the first three-dimensional data after the gate processing, and the second three-dimensional image. The axle detection system according to any one of claims 1 to 5, wherein the width of the vehicle is measured using data. Based on the height histogram of the first three-dimensional data, the ratio n / N between the number n of bins in which a predetermined number or more of the first three-dimensional data after the gate processing is stored and the total number N of bins divided in the height direction A first data existence probability calculating device for calculating as a first data existence probability;
A first detection flag generator for generating a first detection flag for the axle of the vehicle based on the first data existence probability;
A second three-dimensional image pickup device that is spaced apart from the first three-dimensional image pickup device in the moving direction of the vehicle and acquires second three-dimensional data of the vehicle;
A second three-dimensional data threshold processing device for performing a false detection data reduction process on the second three-dimensional data;
A second height / distance gate device for extracting the second three-dimensional data after the gate processing in the second detection region from the second three-dimensional data after the erroneous detection data reduction processing;
A second height histogram calculation device that divides the second detection region in the height direction to create a total number of Nc bins and generates a height histogram of the second three-dimensional data after the gate processing; ,
Based on the height histogram of the second three-dimensional data, a ratio nc / Nc between the number of bins nc storing a predetermined number or more of the second three-dimensional data after the gate processing and the total number of bins Nc divided in the height direction A second data existence probability calculating device that calculates the second data existence probability;
A second detection flag generator for generating a second detection flag for the axle of the vehicle based on the second data existence probability;
A vehicle speed / vehicle length measuring device for measuring at least one of a vehicle speed and a vehicle length of the vehicle;
With
The vehicle speed / vehicle length measurement device includes a difference in detection time of the vehicle based on the first detection flag and the second detection flag, the first three-dimensional image pickup device, and the second three-dimensional image pickup device. The axle speed detection system according to any one of claims 1 to 5, wherein the vehicle speed or the vehicle length is measured using a distance to the vehicle.

Images