JP6863027B2 - Three-dimensional object detection processing device - Google Patents

Description

The present invention relates to a three-dimensional object detection processing device.

There is an object detection device that uses LIDAR (light detection and ranging, laser imaging detection and ranging) to detect a three-dimensional object existing in front of a vehicle when traveling on a road. For example, a laser radar is provided on the front surface of the passenger compartment of an automobile to irradiate the road surface in front of the vehicle and measure the distance by the reflected light.

Detects the distance from the vehicle to the measurement points in various directions ahead, and if there are multiple measurement points with the same distance, the presence of a three-dimensional object in front is detected by calculating the height dimension. be able to. As a result, in the cell indicated by the polar coordinate grid with the position of the laser radar as the origin, when the cell in which the three-dimensional object exists is detected for the first time when the cells are searched in order from the origin side in the radial direction, the cell on the origin side of the three-dimensional object is detected. It can also be determined that the cell is a road surface.

Then, after that, the cells of the three-dimensional object and the cells of the road surface detected corresponding to the polar coordinate grid are added to the data obtained in the past by converting the data into the cells of the Cartesian coordinate grid corresponding to the road surface, and are probabilistic. Judge the existence of three-dimensional objects on the road.

However, the process of converting a cell in a polar coordinate grid to a cell in Cartesian coordinates is a process that involves coordinate conversion, so that the calculation cost is generally high. Therefore, considering the processing power of the CPU of the control device and the demand for real-time performance of the device, there is no choice but to use a grid with coarse intervals, which causes a problem that the resolution is lowered.

Japanese Unexamined Patent Publication No. 2013-140515 Japanese Unexamined Patent Publication No. 2011-123551 Japanese Unexamined Patent Publication No. 2011-150633

The present invention has been made in consideration of the above circumstances, and an object of the present invention is to reduce the processing load when converting the cell information detected in the polar coordinate grid into the Cartesian coordinate grid, whereby the equivalent is achieved. It is an object of the present invention to provide a three-dimensional object detection processing apparatus capable of improving the detection resolution with the processing capacity of the above.

The three-dimensional object detection processing device according to claim 1 sets a polar coordinate grid that divides into a plurality of cells at a predetermined angle and a distance range, and a three-dimensional distance based on polar coordinates acquired by a three-dimensional distance detection unit for a detection range around the vehicle. From the information, a three-dimensional object detection unit that detects a three-dimensional object cell in which a three-dimensional object exists, and a Cartesian coordinate grid that divides the cell information of the polar coordinate grid into a plurality of cells at predetermined distances in two dimensions are set, and the polar coordinate grid is set. A coordinate conversion unit that sets a three-dimensional object cell by associating the three-dimensional object cell included in the cell with a cell of a Cartesian coordinate grid, and a road surface cell corresponding to a road surface when each cell of the orthogonal coordinate grid corresponds to a three-dimensional object cell. The road surface cell setting table is read out from the storage unit corresponding to the position of the storage unit that stores the road surface cell setting table for specifying the above and the position of the three-dimensional object cell converted into orthogonal coordinates by the coordinate conversion unit, and the road surface cell is stored. It is equipped with a road surface cell setting unit for setting.

By adopting the above configuration, the three-dimensional distance detection unit acquires the three-dimensional distance information in polar coordinates in the detection range around the vehicle, and the three-dimensional object detection unit displays the three-dimensional object cell in which the three-dimensional object exists on the polar coordinate grid. Detect with. The coordinate conversion unit sets the three-dimensional object cell by associating the three-dimensional object cell included in the polar coordinate grid with the cell of the orthogonal coordinate grid. The road surface cell setting unit reads the road surface cell setting table from the storage unit and sets the road surface cell corresponding to the position of the three-dimensional object cell converted into orthogonal coordinates by the coordinate conversion unit.

As a result, the processing load can be reduced because the target cells of the polar coordinate grid on which the coordinate transformation processing is performed are only the three-dimensional object cells. Further, since the road surface cell setting table stored in the storage unit is read out and the road surface cell is set for the three-dimensional object cell on the Cartesian coordinate grid, the road surface cell can be associated and set only by the reading and setting process. It can reduce the processing load. In other words, if the processing load is the same, the grid can be divided more and the resolution can be improved.

Functional block configuration diagram showing the first embodiment Elevation view showing the detection range of the 3D distance detector and the polar coordinate grid Plan view showing the detection range of the 3D distance detection unit and the polar coordinate grid Diagram showing an example showing pointed distance information on a polar coordinate grid Arrangement of Cartesian coordinate grid Flow chart of data conversion process Flowchart of road surface cell setting process Explanatory drawing of the action of height detection of a three-dimensional object in a polar coordinate grid Explanatory drawing of the action of height detection of a three-dimensional object Explanatory drawing of operation when converting a three-dimensional object cell into a Cartesian coordinate grid Explanatory drawing of the operation of the search path when detecting a three-dimensional object cell in the Cartesian coordinate grid Operation explanatory diagram showing how to create a road surface cell setting table Operation explanatory diagram shown in the map of the road surface cell setting table Explanatory drawing of operation when setting road surface cell and unknown cell based on three-dimensional object cell Operation explanatory diagram of the search path for setting the road surface cell in the Cartesian coordinate grid of the second embodiment Operation explanatory diagram showing the method of creating a search route table (Part 1) Operation explanatory diagram showing the method of creating a search route table (Part 2) Diagram showing a setting example of the search route table Operation explanatory diagram showing the method of creating the search route table of the third embodiment (No. 1) Operation explanatory diagram showing the method of creating a search route table (Part 2) Operation explanatory diagram showing the method of creating a search route table (Part 3) Diagram showing a setting example of the search route table Operation explanatory diagram showing the process of creating the search route table Functional block configuration diagram showing a fourth embodiment Flow chart of data conversion process Explanatory drawing of operation when converting a three-dimensional object cell into a Cartesian coordinate grid Operation explanatory diagram The figure which shows the three-dimensional object cell in the polar coordinate grid which shows 5th Embodiment (the 1) The figure which shows an example of the recording method of a three-dimensional object cell (the 1) Figure showing a three-dimensional object cell on a polar coordinate grid (Part 2) The figure which shows an example of the recording method of a three-dimensional object cell (part 2)

(First Embodiment)
Hereinafter, the first embodiment will be described with reference to FIGS. 1 to 14.
The object detection device 1 shown in FIGS. 2 and 3 is mounted on, for example, the upper part of the front surface of the vehicle A, and is provided on the inner upper part of the windshield in the vehicle interior. As shown in FIG. 1, the object detection device 1 includes a three-dimensional distance detection unit 3, a data processing unit 4, and an output unit 5 with the control device 2 as the main body.

The control device 2 includes a CPU, ROM, RAM, an interface, and the like, and may be incorporated in an in-vehicle ECU or may be separately provided. The control device 2 detects a three-dimensional object existing in front of the vehicle A from the distance information detected by the three-dimensional distance detection unit 3 based on the object detection program. FIG. 1 shows a configuration according to the function of the control device 2. As a functional component, it has a three-dimensional object determination unit 2a in addition to the data processing unit 4. The data processing unit 4 corresponds to a three-dimensional object detection processing device, and includes a three-dimensional detection unit 4a, a coordinate conversion unit 4b, a road surface cell setting unit 4c, and a storage unit 4d.

The three-dimensional distance detection unit 3 is a so-called laser radar (LIDAR). For example, the distance of the traveling range in front of the vehicle A is set to 3 by irradiating laser light for distance measurement in various directions in front of the vehicle A. It is detected dimensionally. The storage unit 4d may be provided inside the control device 2 or may be separately provided outside. The storage unit 4d stores the object detection program executed by the control device 2 and also stores the road surface cell setting table described later. The output unit 5 displays the detection result of a three-dimensional object, and can be displayed by using a display or the like provided in the vehicle interior.

As shown in FIG. 2, the three-dimensional distance detection unit 3 irradiates the front road surface RS of the vehicle A with a laser beam. Reflected light while moving the irradiation points P to the far point side at approximately equal intervals, such as distances r1, r2, ..., Rn on the road surface RS from the near point side with the position directly below the front surface of the vehicle A as the origin (O). The arrival time of is detected and the distance is calculated. Further, as shown in FIG. 3, the three-dimensional distance detecting unit 3 measures the distance in a fan-shaped detection area by changing the direction to the left and right around the front r direction and waving a laser beam.

In this case, when the laser beam is applied to the road surface RS, information on the distance corresponding to the irradiation point P of the road surface RS can be obtained. On the other hand, when an object such as a three-dimensional object exists on the road surface RS, the laser beam is irradiated to the three-dimensional object before reaching the road surface RS, and the reflected light is received to calculate the distance. Distance information at a position closer than the irradiation point P can be obtained.

As shown in FIG. 3, the distance information obtained by the three-dimensional distance detection unit 3 is obtained by a plurality of cells Gθn (for example, directions) divided by the polar coordinate grid G corresponding to the vehicle A according to the irradiation range of the laser beam. The three-dimensional object is determined by dividing it into θ: a to e and a distance n: 0 to 4). The polar coordinate grid G divides a range of a predetermined angle including the r-axis indicating the traveling direction of the vehicle A into cells Ga0 to Ga4 at predetermined distances from the side closer to the vehicle A. Similarly, the cells Gb0 to Gb4 are set in a predetermined angle range on the left side of the cells Ga0 to Ga4 with respect to the traveling direction of the vehicle A, and the cells Gc0 to Gc4 are further set on the left side thereof. Further, the cells Gd0 to Gd4 are set in a predetermined angle range on the right side of the cells Ga0 to Ga4 with respect to the traveling direction of the vehicle A, and the cells Ge0 to Ge4 are further set on the right side thereof.

As shown in FIG. 4, the distance information obtained by the reflection of the laser beam is associated with the cell Gθn of the polar coordinate grid G according to the distance. In the figure, the positions indicating the distance information in the corresponding cells are indicated by black circles (dots).

On the other hand, as shown in FIG. 5, a Cartesian coordinate grid S that is converted corresponding to the road surface RS is used for determining a three-dimensional object. The orthogonal coordinate grid S corresponds to the positions in the x direction and the y direction when the position of the three-dimensional object detection device 1 of the own vehicle is the origin O, the direction orthogonal to the traveling direction of the vehicle is x, and the traveling direction is y. Cells C (x, y) arranged in a rectangular matrix are set. In FIG. 5, the position of the object detection device 1 of the vehicle A is shown on the Cartesian coordinate grid S, and a part of the first quadrant and the second quadrant is shown.

Next, the object detection process by the object detection device 1 will be described with reference to FIG. The flowchart of this object detection process is stored in the storage unit 4d as an object detection program. The control device 2 reads the object detection program from the storage unit 4d when starting the object detection process.

When the control device 2 starts the program, first, in step S1, the data on the polar coordinate grid G and the orthogonal coordinate grid S are initialized, and the orthogonal coordinates on the road surface side in the region including the road surface RS in front corresponding to the position of the vehicle A. Set the grid. Next, when the control device 2 proceeds to step S2, a predetermined angle range is centered on the front of the vehicle A (r direction in the polar coordinate grid G; y direction in the orthogonal coordinate grid S) by the laser radar of the three-dimensional distance detection unit 3. The distance information is acquired corresponding to the position of the polar coordinate grid G by irradiating the laser beam toward. In this case, the control device 2 acquires the point cloud information of the distance information obtained by the three-dimensional distance detection unit 3. The point cloud information is information on the irradiation angle of the laser beam and the distance, and is information on polar coordinates centered on the position of the three-dimensional distance detection unit 3.

The control device 2 maps the acquired point cloud information to the polar coordinate grid G in step S3. At this time, when the distance information is shorter than the distance on the road surface RS, it is mapped to the cell Gθn according to the distance. For example, as shown in FIG. 8, assuming that the distance of the point Pb corresponding to a position farther than the point Pa near the road surface RS is almost the same, a three-dimensional object that blocks and reflects the laser beam here. Can be presumed to exist.

When the height T1 at this time is equal to or higher than a predetermined value set as a threshold value, for example, 5 to 10 cm or more, it may become an obstacle to ride on when the vehicle A advances. In such a case, it can be detected that there is a three-dimensional object. In this way, the points P of the distance information belonging to the same cell Gθn are associated with each cell Gθn of the polar coordinate grid G as shown in FIG.

Next, in step S4, the control device 2 performs a three-dimensional object detection process by the three-dimensional object detection unit 4a. Here, as described with reference to FIG. 7, for the distance information entering one cell Gθn, when a plurality of distance information exists, the height H is calculated, and the height is equal to or higher than the threshold value. Detects the presence of a three-dimensional object. Specifically, the control device 2 obtains the height of the measurement point based on the distance information in the cell Gθn when a plurality of distance information exists in the cell Gθn.

Here, the control device 2 calculates the difference between the point Pa of the lowest height HPa and the Pb of the highest height HPb from the height information of the plurality of measurement points, and the height T1 (= HPb−). HPa) is calculated. The control device 2 determines that the three-dimensional object exists in the cell Gθn when the obtained value of the height T1 is equal to or higher than the threshold value Tth for determining the three-dimensional object.

Specifically, as shown in FIG. 9, when the point Pa having the lowest height HPa is on the road surface RS, for example, the detection distance by the three-dimensional distance detection unit 3 is the detection distance ra of the road surface RS. Further, when the point Pb of the highest height HPb is above the point Pa, the detection distance by the three-dimensional distance detection unit 3 is close to the detection distance ra due to the existence of the three-dimensional object. When there is no three-dimensional object, the position of the detection distance rb of the road surface RS is detected.

As a result, the height T1 of the three-dimensional object can be calculated by the following equation (1) based on the distance information ra and rb when the point of the road surface RS is detected by the three-dimensional distance detecting unit 3.
T1 = T0 × (rb-ra) / rb… (1)
However, T0: the height, ra, and rb of the three-dimensional distance detection unit 3 are the detection distance at the road surface RS of the point Pa when there is no three-dimensional object, and the detection distance at the road surface Rs of the point Pb.

In this way, the control device 2 calculates the presence / absence of a three-dimensional object for all the cells of the polar coordinate grid G, and maps the cell in which the three-dimensional object is detected as a three-dimensional object cell (marked with a circle). Here, for example, as shown in the upper part of FIG. 10, among the cells of the polar coordinate grid G, cells Ga3, Gb3, Gc3, Gd3, Gd4, and Ge4 are detected as three-dimensional object cells.

Next, in step S5, the control device 2 performs a process of converting the cell determined to be a three-dimensional object cell by the polar coordinate grid G into the orthogonal coordinate grid S by the coordinate conversion unit 4b. Here, each of the cells Ga3, Gb3, Gc3, Gd3, Gd4, and Ge4 determined to be three-dimensional objects cells as described above is subjected to coordinate conversion processing, and the cells on which the cells are superimposed on the orthogonal coordinate grid S are three-dimensional. Judge as a thing cell. As a result, for example, as shown in the middle part of FIG. 10, each of the following cells is mapped as a three-dimensional object cell (marked with a circle).

In the Cartesian coordinate grid S shown in the middle of FIG. 10, in the first quadrant, cell C (1,4), cell C (2,4), cell C (2,5), cell C (3,3), cell C (3,4) and cell C (4,3) are three-dimensional object cells. In the second quadrant, cell C (-1,4), cell C (-2,4), cell C (-3,2), cell C (-4,2), cell C (-4,3) It is a three-dimensional object cell.

In this case, in this coordinate conversion process, the processing load of the control device 2 is larger than that in the normal process, but the number of cells to be converted is smaller than the total number of cells in the orthogonal coordinate grid S, so that the total number of cells is small. The processing load is reduced.

Next, the control device 2 proceeds to step S6 and performs a process of setting the road surface cell based on the three-dimensional object cell in the orthogonal coordinate grid S by the road surface cell setting unit 4c. The lower part of FIG. 10 shows a state in which road surface cells and unknown cells are mapped on the Cartesian coordinate grid S. In the illustrated state, the Cartesian coordinate grid S shows a part of the first quadrant and the second quadrant with respect to the position O of the object detection device 1, but the three-dimensional object cell is also detected behind the vehicle A. In that case, the same processing is performed for the third quadrant and the fourth quadrant.

The control device 2 sets the road surface cell according to the flowchart shown in FIG. 7 for the road surface cell setting process. Further, in this process, the control device 2 refers to the road surface cell setting table stored in the storage unit 4d. The road surface cell setting table is stored in the storage unit 4d in advance as described later, and the road surface cell and the unknown cell on the orthogonal coordinate grid S can be set by referring to this data.

The control device 2 first performs a process of searching for a three-dimensional object cell of the orthogonal coordinate grid S. In this case, the control device 2 designates the first quadrant of the plurality of quadrants as the search area in step S61. Next, in step S62, as shown in FIG. 11, the control device 2 first searches along the search path A1 in the first quadrant, for example, as a search path. The search path A1 starts from the origin cell C (1,1), moves one cell at a time in the x direction, and searches to the rightmost cell C (8,1). Subsequently, in the search path A2, the control device 2 goes up one step in the y direction and moves one cell at a time in the x direction from the cells C (1, 2) to search. In this way, when the control device 2 searches along the search path A8 in the x direction from the uppermost cell C (1,8), the search in the first quadrant ends.

Similarly, in the second quadrant, the search path B1 moves one cell at a time in the negative direction of x, starting from the origin cell C (-1,1) in the second quadrant, and moves to the leftmost cell C (-8). , 1). Subsequently, in the search path B2, the control device 2 goes up one step in the y direction, moves one cell at a time in the negative direction of x from cells C (-1 and 2), and searches. Search paths C1 to C8 and D1 to D8 are also set for the third quadrant and the fourth quadrant.

The control device 2 determines whether or not each cell in each search path is a three-dimensional object cell in the next step S63. In step S63, when the target cell is a three-dimensional object cell, the control device 2 determines YES and reads out and refers to the road surface cell setting table for the three-dimensional object cell, and orthogonally crosses the road surface cell and the unknown cell. Set to the coordinate grid S.

Here, in the road surface cell setting table read from the storage unit 4d, when the corresponding cell is a three-dimensional object cell, the data of the cell corresponding to the road surface cell or the unknown cell is stored in advance. Therefore, the control device 2 can be set by applying the read data of the road surface cell and the unknown cell to the Cartesian coordinate grid. As a result, the control device 2 does not need to perform arithmetic processing on each cell to determine and set a road surface cell or an unknown cell, and can easily and quickly set the corresponding cell.

After executing step S64, or when the target cell is not a three-dimensional object cell and NO in step S63, the control device 2 proceeds to step SS65, and has the search completed for all the cells? Judge whether or not. If NO in step S65, the control device 2 returns to step S62 and repeats the above process.

In the setting process of the road surface cell or the unknown cell in step S64, when setting the repeating three-dimensional object cell, there may be data to be set for the cell that has already been set. In this case, the data may have settings different from the set state of the cell. In such a case, the priority is set as a rewriting rule. Here, the three-dimensional object cell, the road surface cell, and the unknown cell are designated in descending order of priority. Therefore, it is decided to rewrite only when the priority of the already set cell is higher than the priority of the data to be newly set.

In this way, by performing the road surface cell setting process for all the three-dimensional object cells in the orthogonal coordinate grid S according to the program of FIG. 7, all the cells become any of the three-dimensional object cell, the road surface cell, and the unknown cell. Can be set. When the search for all the cells is completed, the control device 2 determines YES in step S65 and proceeds to step S7 of the program of FIG.

In step S7, the control device 2 sets the three-dimensional object probability for each cell of the Cartesian coordinate grid S. Here, as the value of the three-dimensional object probability, for example, the three-dimensional object cell is set to "+1", the road surface cell is set to "-1", and the unknown cell is set to "0". As a result, it is possible to obtain a Cartesian coordinate grid S in which the probability of a three-dimensional object is set.

In the subsequent step S8, the control device 2 adjusts the map information of the three-dimensional object probability in the orthogonal coordinate grid S thus obtained to the position of the vehicle A with respect to the stationary orthogonal coordinate grid corresponding to the road surface RS. Performs the process of overlapping and adding. Next, since the vehicle A is moving, the control device 2 shifts the position of the own vehicle in step S9, returns to step S2 again, and repeatedly executes the above processing.

The control device 2 performs a three-dimensional object determination process in addition to the above process. As a result of repeating the above processing, the probabilities of three-dimensional objects are added to the cells of the stationary orthogonal coordinate grid corresponding to the road surface RS, so that if there is a stationary three-dimensional object on the road surface RS, The value of the three-dimensional object probability of the cell increases. As a result, the control device 2 can accurately estimate the stationary three-dimensional object existing on the road surface RS.

Also, when there is a moving three-dimensional object such as another vehicle, the three-dimensional object moves with the movement of the other vehicle, so the value of the three-dimensional object probability does not increase, but moves between cells. Become. Further, when the three-dimensional object does not exist in the pointless region, the existence of the three-dimensional object is not determined because the value of the three-dimensional object probability does not increase in the multiple addition processes.

Next, the method of creating the road surface cell setting table described above and the setting of the road surface cell by the search process will be described in detail with reference to FIGS. 12 to 14. FIG. 12 shows a procedure for setting a road surface cell and an unknown cell corresponding to one cell when one cell is a three-dimensional object cell in the first quadrant of the Cartesian coordinate grid S. In FIG. 12A, for example, when the cell C (3, 4) is a three-dimensional object cell, the cell data shown in the table C (3, 4) is mapped on the grid. In the figure, the three-dimensional object cell is indicated by a “◯” mark, the road surface cell is indicated by a “Δ” mark, and the unknown cell is indicated by an “x” mark.

As a method of making this distribution, as shown in FIG. 12B, when the three-dimensional object cell C (3, 4) is viewed from the origin O, the end points X1 and X2 are attached to both ends, which are the range of the cell. Next, as shown in FIG. 12 (c), region lines L1 and L2 connecting the origin O and the end point X1 and the origin O and the end point X2 are drawn. Subsequently, as shown in FIG. 12 (d), the target cell is set with reference to the area surrounded by the area lines L1 and L2.

Among the obtained target cells, a cell located closer to the origin O than the three-dimensional object cell C (3, 4) is designated as a road surface cell. This is because it can be determined that the cell is a road surface cell because the three-dimensional object is not detected up to the position of the three-dimensional object cell. Here, cells C (1,1), C (1,2), C (2,2), and C (2,3) are road surface cells.

Further, among the obtained target cells, a cell located on the side farther from the origin O than the three-dimensional object cell C (3, 4) is defined as an unknown cell. By identifying the three-dimensional object cell at the position closest to the vehicle A, it is not necessary to obtain information about the cell at a position farther than that, so the cell at a position farther than the three-dimensional object cell from the origin O is hidden by the three-dimensional object. If there is, or if there is a three-dimensional object at that position, the cell behind the three-dimensional object cell is regarded as an unknown cell.

Here, cells C (3,5), C (3,6), C (4,4), C (4,5), C (4,6), C (4,7), C (4, 8), C (5,5), C (5,6), C (5,7), C (5,8), C (6,6), C (6,7), C (6,8) ), C (7,7), C (7,8), C (8,8) are unknown cells.

FIG. 13 is a map showing a road surface cell and a table C (x, y) in which unknown cells are set for each cell C (x, y) in the first quadrant of the Cartesian coordinate grid S as described above. is there. Although the map of the table C (x, y) is shown here for the first quadrant, it can be dealt with by setting the second and subsequent quadrants in the same manner.

FIG. 14 shows a specific example in which a road surface cell and an unknown cell are mapped by searching for a three-dimensional object cell on the Cartesian coordinate grid S. In the first quadrant of the Cartesian coordinate grid S, in FIG. 14A, the three-dimensional object cells detected by the polar coordinate grid G are cells C (4,3), C (4,4), C (5,1). , C (5,2), C (5,3).

As described above, when the search path A1 moves from the origin cell C (1,1) in the x-axis direction, the control device 2 detects the cell C (5,1) as a three-dimensional object cell. Corresponding to the three-dimensional object cell C (5, 1), the control device 2 reads the data of the table C (5, 1) from the road surface cell setting table shown in FIG. 13 and sets the road surface cell and the unknown cell. .. FIG. 14B shows a state in which road surface cells and unknown cells are mapped with symbols.

Next, as the search path A2 moves from the cells C (1, 2) in the x-axis direction, the control device 2 detects the cells C (5, 2) as a three-dimensional object cell. Corresponding to the three-dimensional object cell C (5, 2), the control device 2 reads the data of the table C (5, 2) from the road surface cell setting table, and displays the road surface cell and the unknown cell as shown in FIG. 14 (c). Set to. At this time, if the cell settings are duplicated, the cells are rewritten in the above-mentioned priority order.

That is, when the cell is already a three-dimensional object cell, or the cell is set as a road surface cell or an unknown cell, in the case of a three-dimensional object cell or a road surface cell, the setting is utilized without being rewritten as it is. If an unknown cell is encountered, the cell to be set is rewritten if it is a road surface cell.

By repeating the process with the search paths A3 and A4 in the same manner, the settings are made as shown in FIG. 14 (d), and finally the settings are completed for all the cells in the first quadrant as shown in FIG. 14 (e). To do. In addition, the same search process is performed for other quadrants to complete the setting for all cells.

According to the first embodiment as described above, the control device 2 detects only the three-dimensional object cell on the polar coordinate grid G and converts the coordinates into the orthogonal coordinate grid S. After that, by setting the road surface cell and the unknown cell specified in the road surface cell setting table for the three-dimensional object cell on the Cartesian coordinate grid S, the three-dimensional object cell, the road surface cell, and the unknown cell are quickly identified for all the cells. can do.

Since the number of cells to be subjected to the coordinate conversion processing by the control device 2 is limited to the three-dimensional object cell, the processing load can be significantly reduced, which enables the real-time processing to be performed quickly, and further, the cell. Even if the number is increased to improve the detection accuracy, the processing load will not be increased.

In addition, when rewriting occurs in the setting of the road surface cell and the unknown cell by the three-dimensional object cell, the three-dimensional object cell, the road surface cell, and the unknown cell are set in descending order of priority. You will be able to perform the setting process of.

When a simulation was performed on the effect when this embodiment was applied, the following specific effects could be confirmed. For example, the resolution of the laser radar is 10,000 (three-dimensional object 50%, road surface 25%, unknown point 25%), the resolution of the polar coordinate grid G is 50,000, and the resolution of the orthogonal coordinate grid S is 40,000. When coordinate conversion processing is performed on all cells of the polar coordinate grid for 10,000 observation points by the laser radar, 50,000 points are targeted, whereas in this embodiment, the solid of the polar coordinate grid is converted to Cartesian coordinates. Since the object cells are limited to the cells in which the three-dimensional object is detected, the number of object cells is, for example, about 5,000. Therefore, regarding the coordinate conversion process, the processing load could be reduced to about 1/10. Therefore, the processing load can be significantly reduced even if other processing is included.

(Second Embodiment)
15 to 18 show the second embodiment, and the parts different from the first embodiment will be described below. In this embodiment, the search path is set as a path that moves radially around the origin O.

As described above, in the road surface cell setting table, as shown in FIG. 13 above, the road surface cell is set on the side closer to the position of the origin O and the unknown cell is set on the far side according to the position of the three-dimensional object cell. Therefore, a search path close to this setting principle is set as search paths R1 to R15 for the first quadrant. That is, the search path along the light projection direction by the laser radar is set.

In FIG. 15, the basic search paths R1 to R15 in the Cartesian coordinate grid S are indicated by dashed arrows. Specifically, the cell through which the broken line arrow indicating the search route passes is set as the cell of the search route. Further, the search paths R1 to R15 are cells C (8,1), C (8,2), and C (8,3) located in the periphery from the origin O in the first quadrant in the Cartesian coordinate grid S, respectively. ), C (8,4), C (8,5), C (8,6), C (8,7), C (8,8), C (7,8), C (6,8) , C (5,8), C (4,8), C (3,8), C (2,8), C (1,8) are set to move the cell.

16 and 17 show specific examples thereof. For example, in the search path R3 from the origin O to the cell C (8,3), as shown in FIG. 16, the cell through which the arrow of the search path R3 passes from the origin cell C (1,1) is C (2,1). ), C (3,1), C (3,2), C (4,2), C (5,2), C (6,2), C (6,3), C (7,3) , C (8,3). FIG. 17 shows an example in which these passing cells are described in the search route table as the search route R3.

Hereinafter, the search route table shown in FIG. 18 is set as search routes R1 to R15 for other peripheral cells in the same manner. Among these search route tables, the search routes R1 to R7 are cells whose cell movements are adjacent in the x-direction or y-direction, but the search route R8 is exceptionally moved to the diagonally adjacent cells. The search route is shown individually as exception handling. Further, since the search paths R9 to R15 are folded back with the search path R8 as the axis of symmetry, the x and y values of the cells set in the search paths R7 to R1 at the symmetrical positions are exchanged.

By searching for a three-dimensional object cell along the search paths R1 to R15 set in this way, the search process has the following advantages. That is, in the road surface cell setting table, as described above, the road surface cell and the unknown cell are defined by the rule for setting the road surface cell and the unknown cell based on the direction connecting the origin O to the three-dimensional object cell. Therefore, when the search process is sequentially advanced in the search paths R1 to R15, when the three-dimensional object cell is detected, the cell before that can be regarded as the road surface cell, and the cells after that can be regarded as the unknown cell. The search process on the search route can be terminated. That is, it is not necessary to search all cells, the time can be shortened, and the processing load can be reduced.

The same effect as that of the first embodiment can be obtained by such a second embodiment, and since the search paths R1 to R15 are set along the direction from the origin O toward the outer peripheral cell, the three-dimensional object cell can be formed. When it is detected, it is not necessary to search for the cells after that, and the processing time of the search process can be shortened to reduce the processing load.

(Third Embodiment)
19 to 23 show the third embodiment, and the parts different from the second embodiment will be described below. In this embodiment, the amount of data in the search route table in which the search routes R1 to R15 are set is reduced.

19 to 21 show a description method for the search paths R3 and R13. For example, in the second embodiment, as shown in FIGS. 16 and 17, the search path R3 sequentially describes the cells through which the arrow R3 shown in FIG. 16 passes as shown in FIG. On the other hand, in the third embodiment, for the cell through which the arrow R3 of the search path R3 shown in FIG. 19 passes, a cell in which the direction of movement between the cells changes as shown in FIG. 20 is described. I have to.

In the search path R3, the case of moving in the x direction from the origin cell C (1,1) in the first quadrant is indicated by "→", and the case of moving in the y direction is indicated by "↑", and the cell number in the x direction is indicated. In, "3" and "6" are cells whose moving direction is changing. In FIG. 21, the search path R3 is described by describing the number of the cell moving in the y direction in this way.

Using this description principle, with respect to the search route table shown in FIG. 18 described above, in each search route R1 to R7, as shown in FIG. 23, in the x direction from the origin cell C (1,1) in the first quadrant. The cells that have changed from the moving state to the moving state in the y direction are shaded. Then, when the x value of the cell moving in the y direction is taken out as a changing cell, a search route table as shown in FIG. 22 can be obtained. As described above, the search path R8 describes a cell to be searched as it is as exception handling.

Further, since the search paths R9 to R15 are folded back with the search path R8 as the axis of symmetry, here, the movement in the y direction is set as the basic direction, and the y value of the cell when moving in the x direction is described. To do. For example, in the search path 13, the case of moving in the y direction from the origin cell C (1,1) in the first quadrant is indicated by "↑", and the case of moving in the x direction is indicated by "→" in the y direction. In cell numbers, "3" and "6" are cells whose movement direction has changed.

In FIG. 21, the search path R13 is described by describing the number of the cell moving in the x direction in this way, but the data is exactly the same as the search path R3. Hereinafter, in the same manner, the search routes can be simply described for the remaining search routes R9 to R15 by describing the numbers of the cells moving in the x direction. As a result, the search routes R9 to R15 have the x and y values of the cells set in R1 to R7 exchanged, and when shown as a search route table, the same data as R7 to R1 are set respectively. It becomes. Therefore, by describing only the correspondence of the search route and the information of the search direction, it is not necessary to specifically describe the search routes R9 to R15.

According to such a third embodiment, the amount of data can be further reduced by using the method of describing the search route table using the x value or the y value of the cell whose movement direction changes.

(Fourth Embodiment)
24 to 27 show the fourth embodiment, and the parts different from the first embodiment will be described below. In this embodiment, as shown in FIG. 24, the object detection device 10 includes a three-dimensional distance detection unit 3, a data processing unit 12, and an output unit 5 with the control device 11 as the main body. As a functional configuration of the control device 11, the data processing unit 12 includes a three-dimensional object detection unit 4a, a coordinate conversion unit 4b, a road surface cell setting unit 4c, a storage unit 4d, and a three-dimensional object detection unit 4a and a coordinate conversion unit 4b. A noise three-dimensional object removing unit 4e is provided between them. The noise three-dimensional object removing unit 4e has a function of removing noise other than the three-dimensional object cells required by the vehicle from the three-dimensional object cells detected by the three-dimensional object detecting unit 4a.

FIG. 25 shows the flow of the three-dimensional object detection process in this embodiment, and step S4a is newly provided between steps S4 and S5. Step S4a is a step in which the control device 11 executes the processing content by the noise three-dimensional object removing unit 4e.

That is, the control device 11 performs the three-dimensional object detection process by the three-dimensional object detection unit 4a in step S4, and then proceeds to step S4a to perform the noise three-dimensional object removal process. At this time, in step S4, as shown in the upper part of FIG. 26, among the cells of the polar coordinate grid G, cells Ga3, Gb3, Gc3, Gd3, Gd4, and Ge4 are detected as three-dimensional object cells. Here, looking at the three-dimensional object cell for each radial direction Rn in polar coordinates, two cells Gd3 and Gd4 are detected in the same radial direction.

In step S4a, the control device 11 enables the cell Gd3 at the near position and removes the cell Gd4 at the distant position as a noise three-dimensional object with respect to the cells Gd3 and the cell Gd4 of the plurality of three-dimensional objects detected in the same radial direction as described above. It is determined that the cell is to be used, and the cell is set as a cell in which a three-dimensional object does not exist. As a result, the maximum number of three-dimensional object cells is one in each radial direction.

After that, the control device 11 proceeds to step S5 and performs a process of converting the cell determined to be a three-dimensional object cell by the polar coordinate grid G into the orthogonal coordinate grid S by the coordinate conversion unit 4b. Here, since the number of cells to be coordinate-transformed is smaller than that shown in the first embodiment, the processing load of the control device 11 is further reduced.

That is, as shown in the lower part of FIG. 26, since the cell Gd4 of the polar coordinate grid G has been removed, the conversion process of the orthogonal coordinate grid S into the cells C (2, 5) by the control device 11 is unnecessary here. It has become. As described above, when the number of three-dimensional object cells in the polar coordinate grid G is reduced, the coordinate conversion processing of the control device 11 is reduced, so that the conversion processing time for the effective cells can be shortened as a whole. ..

Next, with reference to FIG. 27, the above-mentioned processing load reducing effect will be specifically described.
FIG. 27 shows a state in which the own vehicle X is traveling in the P direction (upward direction in the figure) on a road in an urban area as viewed from above. A median strip S, guardrails G1 and G2 are arranged on the road, and buildings B1 and B2 are present. Further, another vehicle Y is traveling in the Q direction (lower direction in the figure) in the lane in the opposite direction across the median strip S.

In the detection area Z, two or more three-dimensional objects are detected in the areas Za and Zc in the range of areas Za to Zd. In the regions Zb and Zd, "none" or one three-dimensional object is detected.

For example, in the region Za where two or more three-dimensional objects are detected, three three-dimensional objects r1a, r1b, and r1c are detected in the R1 direction. r1a is a median strip S, r1b is another vehicle Y, and r1c is a guardrail G1. Further, three three-dimensional objects, r2a, r2b, and r2c, are detected in the R2 direction. r2a is a median strip S, r2b is a guardrail G2, and r2c is a building B1 guardrail G1.

Further, in the region Zb where one or less three-dimensional objects are detected, the median strip S is detected as r3a in the R3 direction. Then, in the region Zc where two or more three-dimensional objects are detected, two three-dimensional objects r4a and r4b are detected in the R4 direction. r4a is a guardrail G2, and r4b is a building B2.

When processing is carried out by the method of this embodiment, when a plurality of three-dimensional objects are detected in each direction, the cell at the closest position is left, and the cell at a position farther than that is removed as noise. ing. As a result, in the coordinate conversion process, the conversion process is reduced by the amount removed.

The effect in this case can be roughly calculated as follows. Now, in the above-mentioned detection state, when the control device 11 performs processing by the method of the first embodiment, it is assumed that the processing load is 100%. On the other hand, as shown in FIG. 27, two or more three-dimensional objects are detected in one direction at an angle of view (angle range) of 50% or more of the detection area while traveling in an urban area. It is assumed that there is.

In this case, as shown in FIG. 27, since the regions Za and Zc are regions in which two or more three-dimensional objects are detected in one direction, the angle of view is 50% or more. Since at least two three-dimensional objects are detected in this range, when this embodiment is applied, the number of objects to be coordinate-converted can be one, so that the calculation processing load of the control device 11 is 50%. It can be:

Approximately based on this, the processing load required for coordinate transformation can be obtained as shown in the following equation (1). That is, since 50% of the whole is an area where one three-dimensional object is present, the processing load is 100% (1) as it is, and the remaining 50% is an area where two or more three-dimensional objects are detected in one direction. The processing load is reduced by at least 50% (0.5).

50% x 1 + 50% x 0.5 = 75% ... (1)
As a result, it can be seen that the processing load of the coordinate conversion process can be reduced by at least 25% in an area where many three-dimensional objects exist, such as an urban area. Further, the effect of reducing the processing load can be increased as the number of three-dimensional objects detected in the same direction is larger, and can be increased as the area where two or more three-dimensional objects are detected is wider.

According to this embodiment, the control device 11 is provided with the noise three-dimensional object removing unit 4e, and among the three-dimensional object cells detected by the three-dimensional object detection unit 4a in the polar coordinate grid G, a plurality of three-dimensional object cells in the radial direction. When is present, the remaining three-dimensional object cells excluding the three-dimensional object cell having the shortest radial distance are removed as noise cells.

As a result, the processing load of converting the three-dimensional object cell of the polar coordinate grid G into the orthogonal coordinate grid S by the coordinate conversion unit 4b can be reduced. As a result, the rate at which a plurality of three-dimensional objects are detected increases in places such as urban areas where there are many buildings and three-dimensional objects, so that the effect of reducing the processing load can be further enhanced.

(Fifth Embodiment)
28 to 31 show the fifth embodiment, and the parts different from the first embodiment or the fourth embodiment will be described below. In this embodiment, a data holding format in the polar coordinate grid G is shown. In the first embodiment or the fourth embodiment described above, the data retention format of the three-dimensional object cell in the polar coordinate grid G is not particularly shown, but the number of memory accesses may be increased depending on the data retention format, and the processing The load will increase.

Therefore, in this embodiment, for example, a case where a three-dimensional object cell (marked with a circle) is detected as shown in FIG. 28 in the three-dimensional object cell detection process performed in step S4 in the first embodiment will be described. Here, among the cells of the polar coordinate grid G, the cells Gc3 in the θ1 direction as the angular direction, the cells Gb3, Gb4 in the θ2 direction, the cells Ga3, Ga4 in the θ3 direction, and the cells Gd3, Gd4 in the θ4 direction, and the cell Ge4 in the θ5 direction. Is detected as a three-dimensional object cell.

For example, since the three-dimensional object cell Gc3 has a radial length r1 of "4" in the θ1 direction, the r-direction cell index value r_idx of the angular cell index θ1 is set to “4” as shown in FIG. be able to. Similarly, for all angular cell indexes, the radial length of the three-dimensional cell can be recorded in the table as the r-direction cell index value.

As a result, in the coordinate conversion process in step S5, the three-dimensional object cell on the polar coordinate grid G can be recognized by reading the index value of the r-direction cell with respect to the angular direction cell index.

Further, by recording the three-dimensional object cell of the polar coordinate grid G by the method as described above, in the case shown in the fourth embodiment, only the three-dimensional object cell at the closest position in the radial direction is left, so that each angle. Direction cell index There is one index value for each r-direction cell.

FIG. 30 shows a state in which the three-dimensional object cell detected as shown in FIG. 28 is left at the closest position obtained by performing the process performed in step S4a of the fourth embodiment. Therefore, in this case, for all the angular cell indexes, the shortest length r_min in the radial direction of the three-dimensional object cell can be recorded in the table as the r-direction cell index value r_min_idx as shown in FIG.

As a result, in the coordinate conversion process in step S5, the three-dimensional object cell on the polar coordinate grid G can be recognized by reading the index value r_min of the cell having the shortest distance in the r direction with respect to the angular cell index.

As a result, in the coordinate conversion process of step S5, the control device 11 can reduce the number of memory accesses required for the operation of referencing and searching for the three-dimensional object information from the grid by the coordinate conversion unit 4b. Further, this makes it possible to reduce the processing cost of the control device 11 as a whole.

Therefore, according to the present embodiment, the control device 11 stores the recording method of the three-dimensional object cell in the polar coordinate grid G as an index value in the r direction with respect to the angular index in the storage unit 4d as a table, thereby increasing the number of memory accesses. It will be possible to reduce the processing load and the processing load.

(Other embodiments)
The present invention is not limited to the above-described embodiment, and can be applied to various embodiments without departing from the gist thereof. For example, the present invention can be modified or extended as follows.

In each of the above embodiments, the polar coordinate grid G is set to a range directed to the front of the vehicle A, but a range directed to the side or the rear of the vehicle A can also be set. In this case, a plurality of three-dimensional distance detection units can be provided corresponding to the detection direction.

An example of converting a three-dimensional object cell of the polar coordinate grid G to a Cartesian coordinate grid S and then converting it to a static orthogonal coordinate grid corresponding to the road surface RS has been shown. A process of converting to a Cartesian coordinate grid may be performed.

In the drawings, 1, 10 are object detection devices, 2 and 11 are control devices, 2a is a three-dimensional object determination unit, 3 is a three-dimensional distance detection unit, 4 and 12 are data processing units, 4a is a three-dimensional object detection unit, and 4b is. The coordinate conversion unit, 4c is a road surface cell setting unit, 4d is a storage unit, 4e is a noise three-dimensional object removal unit, and 5 is an output unit.

Claims (8)

A three-dimensional object exists based on the three-dimensional distance information based on the polar coordinates acquired by the three-dimensional distance detection unit (3) for the detection range around the own vehicle by setting a polar coordinate grid that divides the cells into multiple cells at a predetermined angle and distance range. A three-dimensional object detection unit (4a) that detects an object cell,
A Cartesian coordinate grid that divides the cell information of the polar coordinate grid into a plurality of cells at predetermined distances in two dimensions is set, and the three-dimensional object cell included in the polar coordinate grid is associated with the cells of the orthogonal coordinate grid to form a three-dimensional object cell. Coordinate conversion unit (4b) to set
A storage unit (4d) for storing a road surface cell setting table for designating a road surface cell corresponding to the road surface when the cell corresponds to a three-dimensional object cell for each cell of the Cartesian coordinate grid.
A solid provided with a road surface cell setting unit (4c) for reading the road surface cell setting table from the storage unit and setting the road surface cell corresponding to the position of the three-dimensional object cell converted into orthogonal coordinates by the coordinate conversion unit. Object detection processing device. The road surface cell setting table is provided as data for designating a road surface cell and an unknown cell when the cell is a three-dimensional object cell corresponding to each cell of the orthogonal coordinate grid.
The road surface cell setting unit searches for a three-dimensional object cell in order from the cell corresponding to the origin position with the position of the three-dimensional distance detection unit as the origin, and when the three-dimensional object cell is detected, the road surface cell data is obtained from the road surface cell setting table. The three-dimensional object detection processing device according to claim 1, wherein the road surface cell and the unknown cell are read out and set in the Cartesian coordinate grid. In the road surface cell setting table, when a group of cells in which search lines having a certain inclination passing through the origin, which is the position of the three-dimensional distance detection unit, overlap on a Cartesian coordinate grid is set as a search cell group, the inclination of the search line is defined as It is stored as the search cell group obtained for each slope when changed in the search range of the Cartesian coordinate grid.
When the road surface cell setting unit changes the inclination of the search straight line in the search range, reads out the corresponding search cell group in the road surface cell setting table, searches from the origin side, and reaches the three-dimensional object cell. The three-dimensional object detection process according to claim 1, wherein a cell located on the origin side of the three-dimensional object cell is set as a road surface cell, and a cell located on a side farther from the origin than the three-dimensional object cell is set as an unknown cell. apparatus. The road surface cell setting table moves with respect to the two axial directions of the orthogonal coordinate grid when moving along the search straight line starting from the cell at the origin position as information for storing the search cell group. The three-dimensional object detection processing device according to claim 3, wherein the data specifies a cell at a position where the axial direction changes. In the road surface cell setting table, as information for storing the search cell group, in each quadrant delimited by the two axes of the Cartesian coordinate grid, the search straight line changes in inclination by 45 ° from one axis. Request that data for specifying a cell at a position where the movement axis direction changes is set for one range, and data for the remaining second range is specified by exchanging the relationship between the data of the first range and the two axes. Item 4. The three-dimensional object detection processing apparatus according to Item 4. When the road surface cell setting unit (4c) sets a road surface cell or an unknown cell for a three-dimensional object cell based on the road surface cell setting table, when the cell to be set is a cell that has already been set, The three-dimensional object detection processing apparatus according to any one of claims 1 to 5, wherein the three-dimensional object cell is set as the first priority and the road surface cell is rewritten as the second priority and set. Among the three-dimensional object cells detected by the three-dimensional object detection unit (4a) in the polar coordinate grid, when a plurality of three-dimensional object cells exist in the radial direction, the three-dimensional object cell having the shortest radial distance is excluded. The three-dimensional object detection processing apparatus according to any one of claims 1 to 6, further comprising a noise three-dimensional object removing unit (4e) for removing the remaining three-dimensional object cells as noise cells. The three-dimensional object according to any one of claims 1 to 7, wherein the three-dimensional object detection unit (4e) holds data of the detected three-dimensional object cell on the polar coordinate grid at a distance index with respect to a direction index. Detection processing device.

Images