CN114581753A - Method, system and equipment for completing negative obstacles in blind area based on occupancy grid - Google Patents
Method, system and equipment for completing negative obstacles in blind area based on occupancy grid Download PDFInfo
- Publication number
- CN114581753A CN114581753A CN202210044255.2A CN202210044255A CN114581753A CN 114581753 A CN114581753 A CN 114581753A CN 202210044255 A CN202210044255 A CN 202210044255A CN 114581753 A CN114581753 A CN 114581753A
- Authority
- CN
- China
- Prior art keywords
- negative
- point cloud
- grid
- ground
- obstacle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1674—Programme controls characterised by safety, monitoring, diagnostic
- B25J9/1676—Avoiding collision or forbidden zones
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a method, a system and equipment for complementing negative obstacles in blind areas based on occupied grids, which belong to the technical field of environmental perception and are used for solving the technical problem of high cost of negative obstacle perception at present, wherein the method comprises the following steps: 1) acquiring a robot surrounding environment point cloud and rasterizing the point cloud; 2) extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and a corresponding ground point cloud; 3) performing hit information statistics from the negative obstacle boundary point cloud, and selecting an updating strategy of a corresponding target grid according to a statistical result; 4) updating the occupation probability according to the selected updating strategy; 5) and outputting negative obstacle information to the grids which are positioned in the laser radar blind area and contain the negative obstacle information, and outputting complete obstacle information by combining with the positive obstacle information. The invention has the advantages of low cost, comprehensive barrier information and the like.
Description
Technical Field
The invention mainly relates to the technical field of environment perception, in particular to a method, a system and equipment for completing negative obstacles in blind areas based on occupied grids.
Background
In the automatic navigation of the robot, the environment perception technology is a very critical technology which directly tells the robot which places around the robot can be driven and which places can not be driven. Obstacle detection techniques are an important component of environmental awareness techniques. In order to output obstacle information, a laser radar is generally configured at the top of the robot so that the robot can see farther as much as possible, and then the robot has a certain blind area nearby due to the configuration mode, so that corresponding auxiliary sensing equipment is generally installed in the blind area of the robot to extract obstacles. However, negative obstacles are generally not directly perceptible because they are recessed below the ground, while the installation of auxiliary sensors also inadvertently increases the cost of the robot, so that it is necessary to deal with negative obstacles in blind areas by other technical means.
The invention patent application with the publication number of CN109827585A and the invention name of 'a method for rapidly filling concave areas in a grid map' provides a method for filling concave areas in the grid map. The implementation mode is as follows: firstly, extracting an occupation grid map, then selecting a unit which is not an obstacle and has supporting force from the map, and then starting to expand in the horizontal direction from the left side to the right side until the expansion is stopped when the obstacles are arranged on the left side and the right side. In the expansion process, the invention needs to continuously check whether the obstacles exist in the up-down direction of the target grid or not, so that the expanded obstacles can correspond to the actual size. The invention is essentially a post-processing technology, so that the size of the processed negative obstacle can be consistent with the actual size, the problem of discontinuous fracture and dispersion of the negative obstacle can be solved, but the invention is only suitable for the scene that the negative obstacle is positioned in the visual field and is not suitable for the blind area.
The invention patent application with publication number CN113077551A and entitled "occupied grid map construction method, device, electronic device and storage medium" proposes a method for generating an occupied map. The method comprises the following specific steps: extracting obstacle point clouds and ground point clouds from the point cloud information, fitting a driving road surface from the ground point cloud information and ground physical size characteristics, and performing rasterization representation; and finally updating the occupation probability of the obstacles on the rasterized pavement. The invention uses the occupancy map to update the occupancy probability of the positive obstacle category, and the ground point cloud is only used for dividing the passable area in the invention. Although the idea of occupying the grid is utilized, the operation occupation probability is based on the point number of the point cloud, so when the point cloud of the obstacle is considerable (the situation is common in practice), the occupation probability is improved to be close to the limit by one observation, and the subsequent occupation probability updating does not bring practical benefit; furthermore, the method of the present invention is not capable of eliminating when false alarm positive obstacle point clouds are present (e.g., positive obstacle output caused by false alarm point clouds in a rainstorm situation). Therefore, the calculation method of the invention still has the disadvantages.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the technical problems in the prior art, the invention provides a method, a system and equipment for complementing negative obstacles in a blind area based on an occupation grid, which have the advantages of low cost and capability of guaranteeing the comprehensiveness of obstacle information.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a method for supplementing negative obstacles in blind areas based on occupation grids comprises the following steps:
1) acquiring a robot surrounding environment point cloud and rasterizing the point cloud;
2) extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and a corresponding ground point cloud;
3) performing hit information statistics from the negative obstacle boundary point cloud, and selecting an updating strategy of a corresponding target grid according to a statistical result;
4) updating the occupation probability according to the selected updating strategy;
5) and outputting negative obstacle information to the grids which are positioned in the laser radar blind area and contain the negative obstacle information, and outputting complete obstacle information by combining with the positive obstacle information.
Preferably, the specific process of step 2) is:
calculating a ground grid area from the rasterized point cloud by adopting a ground model algorithm to obtain ground information;
calculating the boundary position of the negative obstacle from the rasterized point cloud by adopting a negative obstacle detection algorithm to obtain negative obstacle information;
and extracting corresponding ground point clouds from the negative obstacle boundary grid by combining the ground information and the negative obstacle information.
Preferably, the ground grid region is calculated from the rasterized point cloud by using a ground model algorithm, and the specific process of obtaining the ground information is as follows:
converting the rasterized point cloud to a polar coordinate system; selecting a part of area from the area in front of the robot, wherein the heights of all point clouds in the selected area meet the following conditions: the absolute value of the height is not higher than a preset value;
and (3) taking the height average value of all point clouds in the selected area as a heuristic ground height empirical value, and taking the angle as a processing sequence to carry out ground height estimation on the point clouds in each angle direction: the ground height experience estimated value is added with the relaxation amount to obtain a ground height threshold value of a target sector corresponding to the angle, and a point cloud with the height not exceeding the ground height threshold value in the target sector is extracted as an effective ground point cloud to obtain the effective ground point cloud in the angle direction;
and performing polynomial curve fitting in the angle direction, and obtaining the ground information of the area by obtaining the effective ground point cloud in each angle direction, wherein the ground information is the ground height.
Preferably, the specific process of calculating the boundary position of the negative obstacle from the rasterized point cloud by using a negative obstacle detection algorithm to obtain negative obstacle information is as follows:
gradually traversing the point cloud of each scanning line, and projecting the point cloud to a grid to obtain the rasterization characteristics of the scanning lines;
carrying out continuity progressive identification by adopting the grid position, and selecting point cloud with the height below the grid ground height under a Cartesian coordinate system, wherein the grid ground height under the Cartesian coordinate system corresponds to the ground height threshold under a polar coordinate system;
when the positions of the ground point clouds on the grids are continuous, the ground point clouds are in a normal road surface area, and if the positions of the point clouds projected to the grids jump, the two projection grids are considered to have holes, the two projection positions are respectively marked as boundary areas of negative obstacles, and meanwhile, the corresponding ground point clouds, namely the point clouds in the boundary areas of the negative obstacles, are stored;
all the scanning lines are projected to the grids, the grids with the grid height lower than the ground height threshold value by more than a certain value are found, and the grids are marked as detectable negative obstacle areas.
Preferably, the specific process of extracting the corresponding ground point cloud from the negative obstacle boundary grid by combining the ground information and the negative obstacle information is as follows:
after a negative obstacle grid is observed, adding an initial observation confidence degree to the position, if a negative obstacle is arranged in the same position in a subsequent radar frame, improving the observation confidence degree, and if a negative obstacle area is not observed in the position in the subsequent radar frame, reducing the observation confidence degree; and when the observation confidence coefficient is greater than a certain value, extracting corresponding ground point cloud from the negative obstacle boundary grid.
Preferably, the specific process of step 3) is:
counting the times of hitting the grid by the ground point cloud for the grid with the negative obstacle, and selecting a first updating strategy, a second updating strategy and a third updating strategy according to the times, wherein the method specifically comprises the following steps:
when the point cloud hit times of the grid are the sparsity threshold or the hit times do not exist, a subtractive updating strategy is selected;
when the number of times of point cloud hit of the grid is normal, selecting a common additive updating strategy;
when the hitting times exceed a preset value, an active additive updating strategy is selected; the operation of the active additive updating strategy is as follows: multiple times of ordinary additive update strategy are used continuously.
Preferably, in step 4), when the target grid is a subtractive update strategy, the occupation probability of the target grid is updated to be reduced by a preset value; and when the target grid is in a common additive updating strategy, updating the occupation probability of the target grid to be increased by a preset value.
Preferably, the specific process of step 5) is:
setting a confidence threshold, and checking whether the occupation probability of the grids in the radar blind area of the robot is lower than the confidence threshold; if the occupation probability of the grid is lower than the confidence coefficient threshold value, the vacant state of the grid is reserved, otherwise, the grid is changed into a negative obstacle attribute;
after the grids in all the radar blind areas are checked, the negative obstacles in the blind areas can be obtained, and complete obstacle information is output after the negative obstacles are matched with the positive obstacle information.
The invention also discloses a system for complementing negative obstacles in blind areas based on the occupied grids, which comprises the following steps:
the point cloud generating module is used for acquiring the point cloud of the surrounding environment of the robot and rasterizing the point cloud;
the ground extraction module is used for extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and corresponding ground point cloud;
the negative obstacle detection module is used for counting hit information from the negative obstacle boundary point cloud and selecting a corresponding target grid updating strategy according to a counting result;
the negative obstacle state updating module is used for updating the occupation probability according to the selected updating strategy;
and the blind area result output module is used for outputting negative obstacle information of grids which contain the negative obstacle information and are positioned in the laser radar blind area, and outputting complete obstacle information by combining the positive obstacle information.
The invention further discloses a computer device comprising a memory and a processor, the memory having stored thereon a computer program which, when executed by the processor, performs the steps of the method as described above.
Compared with the prior art, the invention has the advantages that:
according to the invention, by means of the negative obstacle information of the blind area where the robot resides, the negative obstacle information is supplemented in the blind area, so that the number of auxiliary sensors can be reduced, the application cost of the robot is reduced, pain spots of the robot which are high in price and impractical are solved, and great economic benefits are achieved; the negative obstacle information can be supplemented in the blind area only according to the main sensing laser radar, and the comprehensiveness of the obstacle information is ensured; the idea of multi-frame fusion is fully utilized, the contribution of a certain frame negative obstacle is not excessively relied on, and the stability of the negative obstacle information is ensured.
Drawings
FIG. 1 is a flow chart of an embodiment of the method of the present invention.
FIG. 2 is a block diagram of an embodiment of the system of the present invention.
FIG. 3 is a schematic diagram of rasterization of scan results in front of a lidar in accordance with the present invention.
FIG. 4 is a diagram of the effective field of view of a prior art robot equipped with a ring scan radar, wherein (a) is a diagram of the effective field of view of a top radar; (b) the effective view field map after the front radar and the top radar are combined.
FIG. 5 is a schematic view of the negative obstacle area handling in the present invention; wherein (a) is an initial observation; (b) is a negative obstacle area diagram; (c) marking the position of the negative obstacle boundary with higher confidence after 3 times of observation; (d) the schematic diagram is that the negative obstacle area is not in the robot vision field; (e) is a graph of occupying a negative obstacle area in the graph; (f) the result of the negative obstacle area in the blind area is supplemented by the occupancy map; the upper square frame is the physical position of the robot, and the peripheral points are output near-vehicle-body negative obstacle areas.
Detailed Description
The invention is further described below with reference to the figures and the specific embodiments of the description.
As shown in fig. 1, the method for completing negative obstacles in blind areas based on occupancy grids in the embodiment of the present invention includes the steps of:
1) acquiring a robot surrounding environment point cloud and rasterizing the point cloud to obtain a rasterized point cloud;
2) extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and a corresponding ground point cloud;
3) performing hit information statistics from the negative obstacle boundary point cloud, and selecting an updating strategy of a corresponding target grid according to a statistical result;
4) updating the occupation probability according to the selected updating strategy;
5) and outputting negative obstacle information to grids which are positioned in the laser radar blind area and contain negative obstacle information, and outputting complete obstacle information by combining the positive obstacle information.
According to the invention, by means of the negative obstacle information of the blind area where the robot resides, the negative obstacle information is supplemented in the blind area, so that the number of auxiliary sensors can be reduced, the application cost of the robot is reduced, the problem that the robot is high in price and not practical is solved, and great economic benefits are achieved; and the negative obstacle information can be supplemented in the blind area only according to the main sensing laser radar, so that the comprehensiveness of the obstacle information is ensured.
In addition, the idea of multi-frame fusion is fully utilized, the contribution of a certain frame negative obstacle is not excessively relied on, and the stability of the negative obstacle information is ensured; specifically, it is assumed that a certain grid is determined as a negative obstacle region by one of the obstacle detection results, but the grid is not detected as a negative obstacle in the results of the next two frames (or other frames, for example only). Therefore, the occupation probability is only an initial value, the next frames consider that the grid does not belong to the negative obstacle, and the initial probability is updated by the non-occupation probability, so that the occupation probability is eliminated to the barrier-free property; so at the output, the negative barrier property has been cancelled by the next few frame observations to become free space; similarly, assuming that a certain grid is determined as a negative obstacle area by one frame of obstacle detection result, obtaining an initial occupation probability; and the following two frames (or other frames, for example only) have the result that the grid is detected as a negative obstacle, so that the occupation probability is continuously improved: on the basis of the initial probability, the occupation probabilities are continuously superposed, and the occupation probability is improved; so at the time of output, it possesses an occupancy probability sufficient to determine a negative obstacle grid, eventually labeled as a negative obstacle region. The above processing scheme using the occupancy probability method naturally has the idea of multi-frame fusion.
In a specific embodiment, the specific process of step 2) is: calculating a ground grid area from the rasterized point cloud by adopting a ground model algorithm to obtain a ground calculation result; calculating the boundary position of the negative obstacle from the rasterized point cloud by adopting a negative obstacle detection algorithm to obtain a negative obstacle calculation result; and extracting corresponding ground point clouds from the negative obstacle grid by combining the ground calculation result and the negative obstacle calculation result.
Specifically, the method comprises the following steps: calculating a ground grid area from the rasterized point cloud by adopting a ground model algorithm, wherein the specific process of obtaining the ground information comprises the following steps:
converting the rasterized point cloud to a polar coordinate system; selecting a part of area from the area in front of the robot, wherein the heights of all point clouds in the selected area meet the following conditions: the absolute value of the height is not higher than a preset value;
the height average value of all point clouds in a selected area is used as a heuristic ground height empirical value, angles are used as processing sequences, and ground height estimation is carried out on the point clouds in each angle direction: the ground height experience estimated value is added with the relaxation amount to obtain a ground height threshold value of a target sector corresponding to the angle, and a point cloud with the height not exceeding the ground height threshold value in the target sector is extracted as an effective ground point cloud to obtain the effective ground point cloud in the angle direction;
and performing polynomial curve fitting in the angle direction, and obtaining the ground information of the area by obtaining the effective ground point cloud in each angle direction, wherein the ground information is the ground height.
The specific process of calculating the boundary position of the negative obstacle from the rasterized point cloud by adopting a negative obstacle detection algorithm to obtain negative obstacle information comprises the following steps:
gradually traversing the point cloud of each scanning line, and projecting the point cloud to a grid to obtain the rasterization characteristics of the scanning lines;
carrying out continuity progressive identification by adopting the grid position, and selecting point cloud with the height below the grid ground height under a Cartesian coordinate system, wherein the grid ground height under the Cartesian coordinate system corresponds to the ground height threshold under a polar coordinate system;
when the positions of the ground point clouds on the grids are continuous, the ground point clouds are in a normal road surface area, and if the positions of the point clouds projected to the grids jump, the two projection grids are considered to have holes, the two projection positions are respectively marked as boundary areas of negative obstacles, and meanwhile, the corresponding ground point clouds, namely the point clouds in the boundary areas of the negative obstacles, are stored;
all scan lines are projected onto the grid, the grid height is found to be above a certain value below the ground height threshold, and these grids are marked as detectable negative obstacle regions.
Specifically, the specific process of extracting the corresponding ground point cloud from the negative obstacle boundary grid by combining the ground information and the negative obstacle information is as follows:
after a negative obstacle grid is observed, adding an initial observation confidence degree to the position, if a negative obstacle is arranged in the same position in a subsequent radar frame, improving the observation confidence degree, and if a negative obstacle area is not observed in the position in the subsequent radar frame, reducing the observation confidence degree; and when the observation confidence coefficient is greater than a certain value, extracting corresponding ground point cloud from the negative obstacle boundary grid.
In a specific embodiment, the specific process of step 3) is: counting the times of hitting the grid by the ground point cloud for the grid with the negative obstacle, and selecting a first updating strategy, a second updating strategy and a third updating strategy according to the times, wherein the method specifically comprises the following steps: when the point cloud hit times of the grid are a sparsity threshold or no hit times exist, a subtractive updating strategy is selected; when the number of times of point cloud hit of the grid is normal, selecting a common additive updating strategy; when the hitting times exceed a preset value, an active additive updating strategy is selected; the operation of the active additive updating strategy is as follows: multiple times of ordinary additive update strategy are used continuously.
In a specific embodiment, in step 4), when the target grid is a subtractive update strategy, the occupation probability of the target grid is updated to be reduced by a preset value; when the target grid is in a common additive updating strategy, updating the occupation probability of the target grid to be increased by a preset value; when the target grid is in the active additive updating strategy, updating the occupation probability of the target grid into a plurality of preset values, such as two preset values; of course, three, four or more preset values can be selected according to actual conditions.
In a specific embodiment, the specific process of step 5) is: setting a confidence threshold, and checking whether the occupation probability of the grids in the radar blind area of the robot is lower than the confidence threshold; if the occupation probability of the grid is lower than the confidence coefficient threshold value, the vacant state of the grid is reserved, otherwise, the grid is changed into a negative obstacle attribute; after the grids in all the radar blind areas are checked, the negative obstacles in the blind areas can be obtained, and complete obstacle information is output after the negative obstacles are matched with the positive obstacle information.
As shown in fig. 2, an embodiment of the present invention provides a system for completing negative obstacles in blind areas based on an occupancy grid, including:
the point cloud generating module is used for acquiring the point cloud of the surrounding environment of the robot and rasterizing the point cloud;
the ground extraction module is used for extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and corresponding ground point cloud;
the negative obstacle detection module is used for counting hit information from the negative obstacle boundary point cloud and selecting a corresponding target grid updating strategy according to a counting result;
the negative obstacle state updating module is used for updating the occupation probability according to the selected updating strategy;
and the blind area result output module is used for outputting negative obstacle information of grids which contain the negative obstacle information and are positioned in the laser radar blind area, and outputting complete obstacle information by combining the positive obstacle information.
The blind area internal negative obstacle completion system based on the occupation grid corresponds to the negative obstacle completion method, and has the advantages of the completion method.
The above invention is further described in detail with reference to a specific embodiment:
1. rasterizing the environment around the robot;
configuring the lidar so that the front negative obstacle is in a measurement field of view of the lidar; scanning the surrounding environment of the robot by using a laser radar and rasterizing an input point cloud;
fig. 3 illustrates the conversion of scanning in front of the lidar to rasterization, where different colors are used to distinguish the measured height of the grid, and the small box is the body area of the robot.
As can be seen from fig. 3, in this configuration, the real-time sensing range of the robot is in front of the body (effect is achieved without dead angle sensing), and the two sides and the rear of the robot near the body are blind areas, because no real-time sensing device is configured in this area; although the mainstream configuration is used to install a circular scanning radar at the top of the robot head, as shown in fig. 4 (a), even if the effective visual fields of the two are superimposed, a blind area still exists near the robot, as shown in fig. 4 (b); in order to effectively solve the problem that the robot possibly falls into negative obstacle areas such as pot holes, cliffs and the like when moving forwards, the negative obstacle compensation can be carried out on the position close to a vehicle body by utilizing the effective sensing information of the robot to the front at the past moment, the purpose of effectively avoiding obstacles is achieved, meanwhile, the cost requirement of deploying real-time sensors is saved, and the commercial performance is improved.
2. Extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle (namely extracting a grid with negative obstacle edge information in a grid image) and a corresponding ground point cloud;
calculating ground area information from the point cloud by adopting a ground model algorithm; calculating the boundary position of the negative obstacle from the point cloud by adopting a negative obstacle detection algorithm; extracting corresponding ground point clouds from the negative obstacle grids by combining the ground calculation result and the negative obstacle calculation result;
the ground model extraction method comprises the following steps: and converting the rasterized map point cloud into a polar coordinate system, wherein the angular resolution is 2 degrees, and the radial distance resolution is 50 cm. Selecting a part of area from the area in front of the robot, wherein the heights of all point clouds in the selected area meet the following conditions: the absolute value of the height is not higher than 30 cm; adopting the point cloud height average value of the selected area as a heuristic ground height empirical value; after obtaining the height empirical value, taking the angle as a processing sequence, and performing ground height estimation on the point cloud in each angle direction: firstly, calculating a ground experience threshold value, wherein the rule is that the relaxation amount is increased by 10cm per 2m according to the radial distance, the ground height threshold value of a target sector can be obtained by adding the relaxation amount to the ground height experience estimation value, and the point cloud with the point cloud height not exceeding the ground height threshold value in the sector is extracted as effective ground point cloud; obtaining all effective ground point clouds in the angle direction; finally, performing polynomial curve fitting (the fitting coefficient is 3 times) in the angle direction by using RANSAC; the ground height of all areas can be obtained by the operation method for each angle direction.
After the ground height value is obtained, the negative obstacle region extraction (negative obstacle extraction algorithm) is performed, and two characteristics are used: (1) a cavity area of a single scanning line is a negative obstacle candidate area, and edge point cloud of the negative obstacle area is extracted for marking; (2) the point clouds of which the heights measured by the multi-line scanning lines in the same sector are lower than the ground height of the sector by less than 30cm are point clouds falling into a negative obstacle area, and the positions of the point clouds are directly marked. The method comprises the following specific steps:
gradually traversing the point cloud of each scanning line, projecting the point cloud to a grid to obtain the rasterization characteristics of the scanning lines, and the rasterization characteristics comprise: point cloud density, point cloud height, and grid location information for the grid. The problem of overlarge result change is caused by only carrying out continuity judgment according to the positions of two adjacent point clouds, so that continuity progressive identification is carried out by adopting a grid position. Selecting point clouds with the height below the ground height of the grids (the ground estimation in the front is converted to the ground height of a polar coordinate system to obtain a forward view field, so that the result is converted to the ground height of the grids under a Cartesian coordinate system, the ground height of the grids can be obtained in the same way), and if the positions of the ground point clouds on the grids are continuous, the ground point clouds are in a normal road surface area (the projection position of the front point cloud and the projection position of the back point cloud are within 2 pixels, the resolution ratio of the grids is 10 cm/pixel), and if the positions of the point clouds projected on the grids are greatly jumped (the difference between the projection positions of the grids of the front point cloud and the back point cloud is more than 2 pixels), considering that the two projection grids have holes, respectively marking the boundary areas with the projection positions as negative obstacles, and simultaneously storing the corresponding ground point clouds (negative obstacle boundary area point clouds);
all the scanning lines are projected to grids, grids with the height more than 30cm lower than the ground estimated height are found, and the grids are marked as detectable negative obstacle areas. To eliminate false detection of individual grid detection, which causes false alarm, the process also introduces a spatial filtering technique, and the filtering window size is set to 3 × 1. And introducing space dimension verification into the grid left after filtering (the transverse size is at least 2 pixels, the longitudinal size is at least 2 pixels, and at least 3 negative obstacle marks can be found in the grid by using a 5-by-2 inspection window (the mark in the first step is effective), and improving the point cloud in the grid as a detectable negative obstacle point cloud area.
After the 2 steps of processing, the next question is naturally how to perform multi-frame superposition of these negative obstacle boundaries (1 st step result) and regions (2 nd step result) to identify true negative obstacles. The idea of a probability occupancy map is introduced here, with an initial occupancy probability of 0.5 (i.e. uncertain state) for all grids. When a negative obstacle grid is observed, an observation is added to the position (the probability value is 0.56 given the initial occupancy probability), the observation confidence is improved if a negative obstacle is co-located in the radar frame described later (an occupancy observation is added to the previous value, the probability value is 0.65), and the observation confidence is reduced if no negative obstacle area is observed in the position in the subsequent observation (the probability value is 0.58 subtracted from the previous occupancy probability). The processing object is a negative obstacle, the negative obstacle is derived from the statistical characteristics of the grid, so that the data volume of the point cloud contained in the grid is very important information, and if more point clouds are contained, the observation confidence rate can be increased to be larger (the observation confidence rate is increased by 2 times or 3 times conventionally, namely the observation effect is considered to be equal to the observation effect observed for 2 times or 3 times); if the number of the point clouds of the grid is small, the point clouds are added once. Because the system operates in a global coordinate system, precise coordinate values are needed, which is why the negative obstacle calculation results of the previous two steps output point clouds instead of grid positions.
3. Counting information hit by ground point clouds from the negative obstacle boundary point clouds, and judging an updating strategy of the target grid;
counting the times of hitting the grid by the ground point cloud for the grid with the negative obstacle, and respectively selecting a first updating strategy, a second updating strategy and a third updating strategy to update the occupation probability:
(1) when the point cloud hit times of the grid are the sparsity threshold or the hit times do not exist, a subtractive updating strategy is selected (when the number of the point clouds contained in the grid is less than 3, the detection is considered to be invalid, so that one observation is eliminated, namely subtractive);
(2) when the number of point cloud hits of the grid is normal count (the number of point clouds of the grid is between 4 and 10), selecting a common additive updating strategy;
(3) when the hitting times exceed a preset value (when the point cloud number of the grid is more than 10), selecting an active additive updating strategy; the operation of the active additive updating strategy is as follows: two general additive update strategies are used in succession.
4. And updating the occupation state by using a corresponding updating strategy according to the judgment result of the grid.
The method specifically comprises the following steps: when the target grid is a subtractive update strategy, updating the probability of the target grid to be lower by a preset value (for example, the probability value is 0.58); when the target grid is an additive update strategy, the probability of the target grid is increased by a preset value (e.g., the probability value is 0.65).
Note: the values are not equal because: 2 or 3 observations can determine whether the target is present, while eliminating the need for more observations to indicate that the target is not present.
Fig. 5 shows two operation effects, in which the circle in fig. 5 (b) is a mark boundary of a negative obstacle region with a small white dot inside, and the states of these values at the time of the first observation are shown as (a); after 3 times of observation, the position of the negative obstacle boundary with higher confidence of the red mark is shown in (c); when the negative obstacle area is not in the field of view of the robot, as shown in (d); occupancy of negative obstacle area in the graph is shown as (e), with the boxes representing high confidence locations; (f) the result graph of the negative obstacle area in the blind area is completed by using the occupation graph. The upper square frame is the physical position of the robot, and the gray points are output near-vehicle-body negative obstacle areas.
The main idea is as follows: the detection and confirmation are performed when the target (negative obstacle region) is within the visual field, and the determined target is output to the outside when the target is outside the visual field.
5. Setting a negative obstacle confidence threshold, outputting negative obstacle information to grids which are positioned in the blind area of the laser radar and contain the negative obstacle information, and completing the negative obstacle information in the blind area to an obstacle map;
and setting a confidence threshold, checking whether the occupation probability of the grids in the robot radar blind area is lower than the confidence threshold, if so, keeping the vacant state of the grids, and otherwise, changing the grids into negative obstacle attributes. After the grid in all blind areas is checked, the obstacle graph in the blind areas can be obtained, and complete obstacle information can be output after being matched with the positive obstacle information.
Embodiments of the present invention also provide a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, performs the steps of the method as described above. Embodiments further provide a computer device comprising a memory and a processor, the memory having stored thereon a computer program which, when executed by the processor, performs the steps of the method as described above. All or part of the flow of the method of the embodiments may be implemented by a computer program, which may be stored in a computer-readable storage medium and executed by a processor, to implement the steps of the embodiments of the methods. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like. The memory may be used to store computer programs and/or modules, and the processor may perform various functions by executing or executing the computer programs and/or modules stored in the memory, as well as by invoking data stored in the memory. The memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (10)
1. A method for complementing negative obstacles in blind areas based on occupation grids is characterized by comprising the following steps:
1) acquiring a robot surrounding environment point cloud and rasterizing the point cloud;
2) extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and a corresponding ground point cloud;
3) performing hit information statistics from the negative obstacle boundary point cloud, and selecting a corresponding target grid updating strategy according to a statistical result;
4) updating the occupation probability according to the selected updating strategy;
5) and outputting negative obstacle information to the grids which are positioned in the laser radar blind area and contain the negative obstacle information, and outputting complete obstacle information by combining with the positive obstacle information.
2. The method for completing negative obstacles in blind areas based on occupancy grids as claimed in claim 1, wherein the specific process of step 2) is as follows:
calculating a ground grid area from the rasterized point cloud by adopting a ground model algorithm to obtain ground information;
calculating the boundary position of the negative obstacle from the rasterized point cloud by adopting a negative obstacle detection algorithm to obtain negative obstacle information;
and extracting corresponding ground point cloud from the negative obstacle boundary grid by combining the ground information and the negative obstacle information.
3. The method for completing negative obstacles in blind areas based on occupied grids according to claim 2, wherein a ground model algorithm is adopted to calculate a ground grid area from a rasterized point cloud, and the specific process of obtaining the ground information is as follows:
converting the rasterized point cloud to a polar coordinate system; selecting a part of area from the area in front of the robot, wherein the heights of all point clouds in the selected area meet the following conditions: the absolute value of the height is not higher than a preset value;
and (3) taking the height average value of all point clouds in the selected area as a heuristic ground height empirical value, and taking the angle as a processing sequence to carry out ground height estimation on the point clouds in each angle direction: the ground height experience estimated value is added with the relaxation amount to obtain a ground height threshold value of a target sector corresponding to the angle, and a point cloud with the height not exceeding the ground height threshold value in the target sector is extracted as an effective ground point cloud to obtain the effective ground point cloud in the angle direction;
and performing polynomial curve fitting in the angle direction, and obtaining the ground information of the area by acquiring the effective ground point cloud in each angle direction, wherein the ground information is the ground height.
4. The method for completing negative obstacles in blind areas based on occupied grids according to claim 3, wherein the specific process of calculating the boundary position of the negative obstacle from the rasterized point cloud by adopting a negative obstacle detection algorithm to obtain the negative obstacle information comprises the following steps:
gradually traversing the point cloud of each scanning line, and projecting the point cloud to a grid to obtain the rasterization characteristics of the scanning lines;
carrying out continuity progressive identification by adopting the grid position, and selecting point cloud with the height below the grid ground height under a Cartesian coordinate system, wherein the grid ground height under the Cartesian coordinate system corresponds to the ground height threshold under a polar coordinate system;
when the positions of the ground point clouds on the grids are continuous, the ground point clouds are in a normal road surface area, and if the positions of the point clouds projected to the grids jump, the two projection grids are considered to have holes, the two projection positions are respectively marked as boundary areas of negative obstacles, and meanwhile, the corresponding ground point clouds, namely the point clouds in the boundary areas of the negative obstacles, are stored;
all scan lines are projected onto the grid, the grid height is found to be above a certain value below the ground height threshold, and these grids are marked as detectable negative obstacle regions.
5. The method for completing negative obstacles in blind areas based on occupancy grids as claimed in claim 4, wherein the concrete process of extracting the corresponding ground point cloud from the negative obstacle boundary grid by combining the ground information and the negative obstacle information is as follows:
after a negative obstacle grid is observed, adding an initial observation confidence degree to the position, if a negative obstacle is arranged in the same position in a subsequent radar frame, improving the observation confidence degree, and if a negative obstacle area is not observed in the position in the subsequent radar frame, reducing the observation confidence degree; and when the observation confidence coefficient is greater than a certain value, extracting corresponding ground point cloud from the negative obstacle boundary grid.
6. The method for completing negative obstacles in blind areas based on occupancy grids as claimed in any one of claims 2-5, wherein the specific process of step 3) is as follows:
counting the times of hitting the grid by the ground point cloud for the grid with the negative obstacle, and selecting a first updating strategy, a second updating strategy and a third updating strategy according to the times, wherein the method specifically comprises the following steps:
when the point cloud hit times of the grid are the sparsity threshold or the hit times do not exist, a subtractive updating strategy is selected;
when the number of times of point cloud hit of the grid is normal, selecting a common additive updating strategy;
when the hitting times exceed a preset value, an active additive updating strategy is selected; the operation of the active additive updating strategy is as follows: multiple times of ordinary additive update strategy are used continuously.
7. The method for completing negative obstacles in blind areas based on occupancy grids as claimed in claim 6, wherein in step 4), when the target grid is a subtractive update strategy, the occupancy probability of the target grid is updated to be lower by a preset value; and when the target grid is in a common additive updating strategy, updating the occupation probability of the target grid to be increased by a preset value.
8. The method for completing negative obstacles in blind areas based on occupancy grids as claimed in claim 7, wherein the specific process of step 5) is as follows:
setting a confidence threshold, and checking whether the occupation probability of the grids in the radar blind area of the robot is lower than the confidence threshold; if the occupation probability of the grid is lower than the confidence coefficient threshold value, the vacant state of the grid is reserved, otherwise, the grid is changed into a negative obstacle attribute;
after the grids in all the radar blind areas are checked, negative obstacles in the blind areas can be obtained, and complete obstacle information is output after the negative obstacles are matched with the positive obstacle information.
9. A system for compensating for negative obstacles in blind areas based on an occupancy grid, comprising:
the point cloud generating module is used for acquiring the point cloud of the surrounding environment of the robot and rasterizing the point cloud;
the ground extraction module is used for extracting ground information and negative obstacle information from the rasterized point cloud to obtain a boundary grid position of a negative obstacle and corresponding ground point cloud;
the negative obstacle detection module is used for counting hit information from the negative obstacle boundary point cloud and selecting a corresponding target grid updating strategy according to a counting result;
the negative obstacle state updating module is used for updating the occupation probability according to the selected updating strategy;
and the blind area result output module is used for outputting negative obstacle information of grids which contain the negative obstacle information and are positioned in the laser radar blind area, and outputting complete obstacle information by combining the positive obstacle information.
10. A computer arrangement comprising a memory and a processor, the memory having stored thereon a computer program, characterized in that the computer program, when executed by the processor, performs the steps of the method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210044255.2A CN114581753A (en) | 2022-01-14 | 2022-01-14 | Method, system and equipment for completing negative obstacles in blind area based on occupancy grid |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210044255.2A CN114581753A (en) | 2022-01-14 | 2022-01-14 | Method, system and equipment for completing negative obstacles in blind area based on occupancy grid |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114581753A true CN114581753A (en) | 2022-06-03 |
Family
ID=81772615
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210044255.2A Pending CN114581753A (en) | 2022-01-14 | 2022-01-14 | Method, system and equipment for completing negative obstacles in blind area based on occupancy grid |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114581753A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116990832A (en) * | 2023-08-04 | 2023-11-03 | 长沙行深智能科技有限公司 | Dangerous road edge perception method, dangerous road edge perception device, terminal equipment and storage medium |
-
2022
- 2022-01-14 CN CN202210044255.2A patent/CN114581753A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116990832A (en) * | 2023-08-04 | 2023-11-03 | 长沙行深智能科技有限公司 | Dangerous road edge perception method, dangerous road edge perception device, terminal equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5820774B2 (en) | Road boundary estimation apparatus and program | |
| US11709058B2 (en) | Path planning method and device and mobile device | |
| CN111448478B (en) | System and method for correcting high-definition maps based on obstacle detection | |
| US10429193B2 (en) | Method and apparatus for generating high precision map | |
| CN106951879B (en) | Multi-feature fusion vehicle detection method based on camera and millimeter wave radar | |
| CN112464812B (en) | Vehicle-based concave obstacle detection method | |
| WO2021179983A1 (en) | Three-dimensional laser-based container truck anti-hoisting detection method and apparatus, and computer device | |
| WO2017042224A1 (en) | Method for generating an environmental map of an environment of a motor vehicle based on an image of a camera, driver assistance system as well as motor vehicle | |
| CN110298311A (en) | Surface gathered water detection method and device | |
| EP3029602A1 (en) | Method and apparatus for detecting a free driving space | |
| CN114581753A (en) | Method, system and equipment for completing negative obstacles in blind area based on occupancy grid | |
| CN114384491B (en) | Point cloud processing method and device for laser radar and storage medium | |
| WO2021051736A1 (en) | Method and apparatus for determining sensing area, and storage medium and vehicle | |
| CN114842166A (en) | Negative obstacle detection method, system, medium, and apparatus applied to structured road | |
| CN111273314A (en) | Point cloud data processing method and device and storage medium | |
| CN113777622A (en) | Method and device for identifying rail obstacle | |
| JP2017117105A (en) | Visibility determination device | |
| CN116051777B (en) | Super high-rise building extraction method, apparatus and readable storage medium | |
| Pfeiffer et al. | Ground truth evaluation of the Stixel representation using laser scanners | |
| CN115077563A (en) | Vehicle positioning accuracy evaluation method and device and electronic equipment | |
| CN114661051A (en) | Front obstacle avoidance system based on RGB-D | |
| JP5915728B2 (en) | Three-dimensional object detection device | |
| CN112020722A (en) | Road shoulder identification based on three-dimensional sensor data | |
| CN114078326B (en) | Collision detection method, device, visual sensor and storage medium | |
| WO2024042607A1 (en) | External world recognition device and external world recognition method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |