WO2021226852A1 - 车位检测方法、装置、计算机设备和存储介质 - Google Patents

车位检测方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2021226852A1
WO2021226852A1 PCT/CN2020/089901 CN2020089901W WO2021226852A1 WO 2021226852 A1 WO2021226852 A1 WO 2021226852A1 CN 2020089901 W CN2020089901 W CN 2020089901W WO 2021226852 A1 WO2021226852 A1 WO 2021226852A1
Authority
WO
WIPO (PCT)
Prior art keywords
parking space
ultrasonic
image
probability
probability map
Prior art date
Application number
PCT/CN2020/089901
Other languages
English (en)
French (fr)
Inventor
金娜
余一徽
Original Assignee
上海欧菲智能车联科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 上海欧菲智能车联科技有限公司 filed Critical 上海欧菲智能车联科技有限公司
Priority to PCT/CN2020/089901 priority Critical patent/WO2021226852A1/zh
Publication of WO2021226852A1 publication Critical patent/WO2021226852A1/zh

Links

Images

Classifications

    • 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
    • G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
    • G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
    • G01S5/08—Position of single direction-finder fixed by determining direction of a plurality of spaced sources of known location

Definitions

  • This application relates to a parking space detection method, device, computer equipment and storage medium.
  • a parking space detection method, device, computer equipment, and storage medium are provided.
  • a parking space detection method includes:
  • the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain a parking space map
  • a parking space detection device includes:
  • An ultrasonic distance data acquisition module for acquiring ultrasonic distance data around the vehicle, the ultrasonic distance data carrying a first time stamp;
  • An ultrasonic parking space probability map generating module configured to generate an obstacle probability map according to the ultrasonic distance data by Bayesian principle, and obtain an ultrasonic parking space probability map according to the obstacle probability map;
  • An image data collection module for collecting image data around the vehicle, the image data carrying a second time stamp
  • An image data processing module for processing the image data to obtain image parking spaces and image passable areas
  • the fusion module is configured to fuse the ultrasonic parking space probability map, the image parking space and the image passable area according to the first time stamp and the second time stamp to obtain a parking space map;
  • the output module is used to determine the location of the parking space according to the parking space map.
  • a computer device including a memory and one or more processors, the memory stores computer readable instructions, and when the computer readable instructions are executed by the processor, the one or more processors execute The following steps:
  • the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain a parking space map
  • One or more non-volatile computer-readable storage media storing computer-readable instructions.
  • the computer-readable instructions When executed by one or more processors, the one or more processors perform the following steps:
  • the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain a parking space map
  • Computer readable instructions computer readable instructions computer readable instructions computer readable instructions computer readable instructions computer readable instructions
  • Fig. 1 is an application environment diagram of a parking space detection method according to one or more embodiments.
  • Fig. 2 is a schematic flowchart of a parking space detection method according to one or more embodiments.
  • Fig. 3 is a schematic diagram of an ultrasonic parking space probability map according to one or more embodiments.
  • Fig. 4 is a schematic diagram of image parking spaces and image passable areas according to one or more embodiments.
  • Fig. 5 is a schematic diagram of a merged parking space map according to one or more embodiments.
  • Fig. 6 is a schematic diagram of the ultrasonic probe according to one or more embodiments when no obstacle is scanned.
  • Fig. 7 is a schematic diagram of the ultrasonic probe when scanning an obstacle for the first time according to one or more embodiments.
  • Fig. 8 is a schematic diagram of the ultrasonic probe when scanning an obstacle for the last time according to one or more embodiments.
  • Fig. 9 is a schematic flowchart of a parking space detection method according to another or more embodiments.
  • Fig. 10 is a schematic flowchart of a parking space detection method according to still one or more embodiments.
  • Fig. 11 is a block diagram of a parking space detecting device according to one or more embodiments.
  • Figure 12 is a block diagram of a computer device according to one or more embodiments.
  • the parking space detection method provided in this application can be applied to the application environment as shown in FIG. 1.
  • the vehicle terminal 102 communicates with various controllers 104 installed on the vehicle.
  • the controller 104 may include, but is not limited to, cameras installed on the vehicle, ultrasonic sensors, and various controllers for collecting vehicle operating data.
  • the vehicle terminal 102 can collect ultrasonic distance data around the vehicle through an ultrasonic sensor and image data around the vehicle through a camera.
  • the ultrasonic distance data carries a first time stamp, and the image data carries a second time stamp.
  • the Si principle generates an obstacle probability map based on the ultrasonic distance data, and obtains an ultrasonic parking probability map based on the obstacle probability map.
  • the image parking space and image passable area are obtained by processing the image data, so that the vehicle terminal 102 finally uses the first timestamp and In the second time stamp, the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain the parking space map; the parking space location is determined according to the parking space map.
  • the obstacle probability map is generated through the ultrasonic distance data, and the ultrasonic parking probability map is obtained according to the obstacle probability map.
  • the image parking space and the image passable area are obtained from the image data, so that the first time stamp and the second time stamp will be
  • the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain the parking space map, so as to combine the two to improve the accuracy of parking space detection.
  • the vehicle terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, and tablet computers installed on the vehicle.
  • a parking space detection method is provided. Taking the method applied to the vehicle terminal in FIG. 1 as an example for description, the method includes the following steps:
  • S202 Acquire ultrasonic distance data around the vehicle, where the ultrasonic distance data carries a first time stamp.
  • the ultrasonic distance data is collected by an ultrasonic sensor installed on the vehicle.
  • the transmitter of the ultrasonic sensor emits ultrasonic waves and records the transmission time of the transmitted waves, so that the ultrasonic waves are reflected when they meet obstacles, so the receiver of the ultrasonic sensor When the reflected wave is received, the time of receiving the reflected wave, the transmission time of the transmitted wave, and the transmission rate of the light wave are ultrasonic distance data.
  • the vehicle terminal can obtain the obstacle relative to the vehicle based on the ultrasonic distance data, the speed of the vehicle, etc. s position.
  • S204 Generate an obstacle probability map according to the ultrasonic distance data by Bayesian principle, and obtain an ultrasonic parking space probability map according to the obstacle probability map.
  • the obstacle probability map refers to a map that characterizes the probability of an obstacle at each location on the map
  • the ultrasonic parking space probability map is a map that characterizes the probability of a parking space at each location on the map.
  • the vehicle terminal obtains the above-mentioned ultrasonic distance data in real time, and obtains the position of the obstacle according to the ultrasonic distance data, so that an obstacle probability map can be generated according to the detection range of the ultrasonic sensor and the position of the obstacle.
  • the ultrasonic data is the direct reading of the sensor.
  • the reading is the distance m, which means that there is an obstacle m from the ultrasonic sensor within the ultrasonic detection range.
  • the vehicle terminal can only determine that there is an obstacle on a section of the arc edge (radius m) from the ultrasonic sensor.
  • the specific part of the obstacle cannot be determined, so the distance from the ultrasonic sensor is m
  • the probability of obstacles in the detection range of the ultrasonic sensor increases, and the probability of obstacles at a distance of less than m from the ultrasonic sensor remains unchanged. In this way, the obstacle probability map can be obtained after multiple scans.
  • the ultrasonic sensor's detection envelope sweeps the grid map points, and a single grid map point will be scanned multiple times to get the most accurate obstacle position. Therefore, if there is no obstacle, it is considered that there may be a parking space. Therefore, the vehicle terminal can obtain an ultrasonic parking space probability map according to the obstacle probability map. The smaller the value, the greater the probability that there is a parking space. Specifically, the ultrasonic parking space probability map can be referred to as shown in FIG. 3.
  • S206 Collect image data around the vehicle, where the image data carries a second time stamp.
  • the image data refers to the image data collected by cameras around the vehicle.
  • the surrounding area of the vehicle may include multiple cameras, for example, 4 or more cameras.
  • the image data may be generated from the images collected by the multiple cameras.
  • the ring view can be generated based on a single image with the same time stamp, so that the same time stamp is the second time stamp of the ring view.
  • the step of collecting image data around the vehicle can be performed in parallel with the step of collecting ultrasonic distance data around the vehicle, so as to ensure the efficiency of data processing.
  • the vehicle terminal collects data through the first thread.
  • Ultrasonic distance data is used to obtain an ultrasonic parking space probability map based on the ultrasonic distance data.
  • the vehicle terminal collects image data through a second thread, and processes the image data to obtain image parking spaces and image passable areas.
  • S208 Process the image data to obtain image parking spaces and image passable areas.
  • the image parking space refers to the position of the parking space obtained by processing the image data
  • the image passable area refers to the position of the passable area obtained by processing the image data.
  • the way the vehicle terminal processes the image data may be to process the image data through a deep learning model to obtain image parking spaces and image passable areas.
  • Image parking space detection can use a variety of methods, such as using deep learning to train a learning network for parking space detection, manually marking a large number of training pictures in the early stage (marking parking spaces and non-parking pixels in the image), and obtaining a reliable recognition rate through the training network. Input when applying The picture can output the detected parking information. The same method can also be applied to the image passable area Freespace. The difference lies in the selection of the network structure and the marking of the previous identification.
  • the obtained image parking space and image passable area can be as shown in Figure 4.
  • S212 Determine the location of the parking space according to the parking space map.
  • the parking space map is a map that can characterize the location of the parking space, which is obtained by superimposing and fusing the ultrasonic parking space probability map, the image parking space, and the image passable area.
  • the aforementioned image parking space and image passable area are all in the map.
  • the vehicle terminal superimposes the ultrasonic parking space probability map, the image parking space, and the probability area of the image passable area, so as to obtain a parking space map that characterizes the probability of the parking space.
  • the probability of the image parking space and the image passable area is 100%, it is determined as the parking space or the communication area.
  • the corresponding deep learning model can be a two-class model; if the deep learning model is a probability model, the image parking space and image The communicable area is characterized by probability, and the probability of the image parking space and the image passable area needs to be superimposed with the probability in the ultrasonic parking space probability map to make a judgment.
  • the vehicle terminal judges by the preset probability threshold. For example, the vehicle terminal judges the position with the probability greater than the probability threshold as the parking space position, and finally filters the parking space position according to the shape of the conventional parking space to obtain the accurate parking space position, specifically, fusion
  • the following map can be seen in Figure 5.
  • the obstacle probability map is generated through the ultrasonic distance data, and the ultrasonic parking space probability map is obtained according to the obstacle probability map.
  • the image parking space and the image passable area are obtained through the image data, so that the first time stamp and the first time stamp
  • the two timestamps merge the ultrasonic parking space probability map, the image parking space and the image passable area to obtain the parking space map, so as to combine the two to improve the accuracy of parking space detection.
  • the obstacle probability map is generated based on the ultrasonic distance data by Bayesian principle, including: determining whether the ultrasonic parking space probability map has been generated; when the ultrasonic parking probability map has not been generated, reading from the ultrasonic distance data Take the ultrasonic detection range and the obstacle distance, mark the area within the ultrasonic detection range whose distance from the vehicle is less than the obstacle distance as the first probability, and mark the area within the ultrasonic detection range whose distance from the vehicle is greater than or equal to the obstacle distance as the first probability.
  • Two probabilities, and an obstacle probability map is generated according to the first probability and the second probability.
  • the first probability and the second probability here represent the parking space and the passable area
  • the first probability is 100%
  • the second probability is 0%. If the first probability and the second probability represent the obstacle area, Then the first probability is 0%, and the second probability is 100%.
  • the obstacle probability map is generated according to the ultrasonic distance data by Bayesian principle, including: when the ultrasonic parking space probability map has been generated, obtaining the last ultrasonic parking space probability map and the vehicle's traffic time period Travel distance.
  • the travel time period is calculated based on the first time stamp carried by the ultrasonic distance data and the previous time stamp adjacent to the first time stamp; the ultrasonic detection range and obstacle distance are read from the ultrasonic distance data, and the ultrasonic The area within the detection range whose distance from the vehicle is less than the obstacle distance is marked as the third probability, and the area within the ultrasonic detection range whose distance from the vehicle is greater than or equal to the obstacle distance is marked as the fourth probability; An ultrasonic parking space probability map locates the ultrasonic detection range and obstacle distance; according to the third probability and the fourth probability, the probability of the area within the ultrasonic detection range and the obstacle distance is updated to obtain the obstacle probability map.
  • the third probability and the fourth probability here represent the parking space and the passable area, the third probability is 100%, and the fourth probability is 0%. If the third probability and the fourth probability represent the obstacle area, Then the third probability is 0%, and the fourth probability is 100%.
  • Figure 6 is a schematic diagram of the ultrasonic detector in an embodiment when the obstacle is not scanned
  • Figure 7 is the first time the ultrasonic detector in an embodiment scans the obstacle
  • Fig. 8 is a schematic diagram of the ultrasonic probe in an embodiment when the obstacle is scanned for the last time.
  • the black filled circle represents the actual obstacle, which is marked as a black dot
  • the black unfilled circle is a point around the actual obstacle, which should be free, that is, the non-obstacle area , Marked as white.
  • the probability of the white spot being non-obstacle has been increasing, and the probability of the black spot has not been scanned because it has not been scanned.
  • the probability of them being occupied that is, the probability of the obstacle will increase simultaneously.
  • the occupation probability of the white point that is, the probability of the obstacle
  • the probability of the black dot being occupied that is, the probability of the obstacle continues to increase, until the scanning is completed.
  • the construction of the above-mentioned obstacle probability map overcomes the following disadvantages of the previous ultrasonic ranging: the distance probability represented by the original ultrasonic ranging is 100%, which cannot be effectively eliminated.
  • the unique characteristics of the ultrasonic envelope make it necessary to repeatedly test a large number of scenes to calibrate the margin when using the ultrasonic parking process.
  • the construction of the obstacle map in this embodiment does not require the above operations, that is, the edge information of the obstacle with higher accuracy can be obtained at one time.
  • processing the image data to obtain the image parking space and the image passable area includes: inputting the image data into the pre-trained parking detection model to obtain the image parking space; inputting the image data to the pre-trained parking space In the passable area detection model, the passable area of the image is obtained.
  • the image parking space and the image passable position are obtained through deep learning respectively, and a relatively reliable recognition rate can be obtained.
  • processing the image data to obtain the image parking space and the image passable area includes: processing the image data to obtain a probability map representing the image parking space and a probability map representing the image passable area; according to the first time stamp And the second time stamp, the ultrasonic parking space probability map, the image parking space and the image passable area are merged to obtain the parking space map, including: according to the first time stamp and the second time stamp, the ultrasonic parking space probability map and the probability map representing the image parking space Fusion with the probability map that characterizes the passable area of the image to obtain the second parking space probability map.
  • the first parking space probability map representing the parking space can be obtained, because they are all probability maps. Only three maps need to be superimposed for the parking space, and the position with a probability higher than the probability threshold must be a parking space.
  • the probability threshold may be preset, such as 90%, etc. There is no specific restriction on this.
  • the vehicle terminal can adopt a multi-threaded processing method, where the first thread collects ultrasonic distance data through an ultrasonic sensor, then generates an obstacle probability map based on the ultrasonic distance data, and then obtains an ultrasonic parking probability map based on the obstacle probability map.
  • the second thread collects image data through the camera, which can stitch the data collected by the camera to get the ring view, and then the second thread starts another third thread, where the second thread can process the ring view or the original image data to generate vision
  • the passable area is the passable area of the image above.
  • the third thread processes the ring view or original image data to generate visual parking spaces, that is, the image parking spaces mentioned above.
  • the processing of the second thread and the third thread can be carried out through deep learning.
  • the vehicle terminal inputs image data into the pre-trained parking space detection model to obtain the image parking space; in the second thread, the vehicle terminal transfers the image The data is input into the pre-trained passable area detection model to obtain the passable area of the image.
  • the vehicle terminal fuses the obtained ultrasonic parking space probability map, the probability map representing the image parking space, and the probability map representing the passable area of the image, that is, the corresponding map is determined according to the timestamp, and then each corresponding image area is determined according to the image coordinates.
  • the probabilities of the corresponding image regions are added together for fusion, so that the second parking space probability map can be obtained, and the vehicle terminal can determine the parking space location according to the second parking space probability map.
  • the vehicle terminal can also set a fourth thread, so that the speed and driving direction of the vehicle can be read through the fourth thread, so that the vehicle can be located according to the speed and driving direction of the vehicle, and the probability map of the second parking space can be displayed
  • the probability map of the second parking space can be displayed
  • the direction determines the compensation displacement, and then the accurate position of the vehicle is determined according to the positioning position of the vehicle and the compensation displacement.
  • the first parking space probability map representing the parking space can be obtained, because they are all probability maps. Only three maps need to be superimposed for the parking space. For example, the probabilities in the three maps are added to obtain the comprehensive probability, and the comprehensive probability is compared with a preset threshold. The position with the probability higher than the threshold must be a parking space.
  • processing image data to obtain image parking spaces and image passable areas includes: inputting image data into a pre-trained parking space and passable area detection model to obtain image parking spaces and image passable areas.
  • the image parking space and the image passable position are obtained through deep learning, and a relatively reliable recognition rate can be obtained.
  • processing the image data to obtain the image parking space and the image passable area includes: processing the image data to obtain a probability map representing the image parking space and the image passable area; according to the first time stamp and the second time Stamp, merge the ultrasonic parking space probability map, the image parking space and the image passable area to obtain the parking space map, including: according to the first time stamp and the second time stamp, the ultrasonic parking space probability map and the probability of the image parking space and the image passable area are characterized
  • the map is fused to obtain the probability map of the first parking space.
  • the first parking space probability map representing the parking space can be obtained.
  • Two maps need to be superimposed. For example, the probabilities in the three maps at this location are added to obtain a comprehensive probability, and the comprehensive probability is compared with a preset threshold. The location where the probability is higher than the threshold must be a parking space.
  • the vehicle terminal can adopt a multi-threaded processing method, where the first thread collects ultrasonic distance data through an ultrasonic sensor, then generates an obstacle probability map based on the ultrasonic distance data, and then obtains an ultrasonic parking probability map based on the obstacle probability map.
  • the second thread collects image data through the camera, where the data collected by the camera can be spliced to obtain a ring view, and then the second thread inputs the image data into the pre-trained parking space and passable area detection model to obtain the image parking space and image availability. Passable area.
  • the vehicle terminal fuses the obtained ultrasonic parking space probability map and the probability map representing the image parking space and the image passable area to obtain the first parking space probability map, that is, the corresponding map is determined according to the timestamp, and then each corresponding image area is determined according to the image coordinates Finally, the probabilities of the corresponding image areas are added together for fusion, so that the first parking space probability map can be obtained, and the vehicle terminal can determine the parking space location according to the first parking space probability map.
  • the vehicle terminal can also set up another thread, so that the speed and driving direction of the vehicle can be read through another thread, so that the vehicle can be located according to the speed and driving direction of the vehicle, and the probability map of the first parking space can be displayed
  • the probability map of the first parking space can be displayed
  • the direction determines the compensation displacement, and then the accurate position of the vehicle is determined according to the positioning position of the vehicle and the compensation displacement.
  • the first parking space probability map representing the parking space can be obtained. Two maps need to be superimposed, and the location with high probability must be a parking space.
  • the method further includes: reading the driving distance of the vehicle during the transit time period from the vehicle's odometer, the transit time period being the first time carried according to the ultrasonic distance data The stamp and the previous timestamp adjacent to the first timestamp are calculated; the vehicle location is displayed on the parking map.
  • the vehicle terminal also displays the location of the vehicle on the parking map.
  • the vehicle terminal can start a separate thread in which the vehicle terminal reads the vehicle from the vehicle’s odometer.
  • the travel distance of the travel time period can be read in real time, and the travel time period is calculated according to the first time stamp carried by the ultrasonic distance data and the previous time stamp adjacent to the first time stamp , Obtain the historical location corresponding to the previous timestamp, and obtain the vehicle location based on the historical location and driving distance; that is, the difference between the collection time of adjacent ultrasonic distance data, so that the vehicle location can be displayed on the parking map through mileage accumulation ,
  • the time of the adjacent timestamp is related to the collection frequency of the ultrasonic distance data, such as 1ms or 10ms, etc., which can be set by the user.
  • the vehicle terminal can determine the compensation displacement according to the vehicle speed and driving direction, and will determine the compensation displacement according to the first and second timestamps.
  • the determined vehicle location and the compensation displacement are used to update the vehicle location, so that the obtained vehicle location is accurate, so that the accurate vehicle location is displayed on the parking map.
  • the vehicle terminal can determine the location of the vehicle based on the ultrasonic distance data, but the location is the location corresponding to the first time stamp. It will take a certain amount of time to calculate the ultrasonic obstacle probability map. In this way, the location of the vehicle will change.
  • the vehicle terminal determines the compensation displacement according to the vehicle speed and driving direction, and updates the vehicle position according to the vehicle position determined by the first and second timestamps and the compensation displacement to ensure the vehicle position Accuracy.
  • the location of the vehicle is also displayed on the map, so that it is convenient to find a parking space close to the vehicle, thereby facilitating parking.
  • a parking space detection device including: an ultrasonic distance data acquisition module 100, an ultrasonic parking space probability map generation module 200, an image data acquisition module 300, an image data processing module 400, Fusion module 500 and output module 600, of which:
  • the ultrasonic distance data acquisition module 100 is used to collect ultrasonic distance data around the vehicle, and the ultrasonic distance data carries a first time stamp;
  • the ultrasonic parking space probability map generating module 200 is used to generate an obstacle probability map according to the ultrasonic distance data by Bayesian principle, and obtain the ultrasonic parking space probability map according to the obstacle probability map;
  • the image data collection module 300 is used to collect image data around the vehicle, and the image data carries a second time stamp;
  • the image data processing module 400 is used to process image data to obtain image parking spaces and image passable areas;
  • the fusion module 500 is used for fusing the ultrasonic parking space probability map, the image parking space and the image passable area according to the first time stamp and the second time stamp to obtain a parking space map;
  • the output module 600 is used to determine the position of the parking space according to the parking space map.
  • the aforementioned ultrasonic parking space probability map generating module 200 may include:
  • the judging unit is used to judge whether the ultrasonic parking space probability map has been generated
  • the first generating unit is used to read the ultrasonic detection range and obstacle distance from the ultrasonic distance data when the ultrasonic parking space probability map is not generated, and mark the area within the ultrasonic detection range whose distance to the vehicle is less than the obstacle distance as With the first probability, the area within the ultrasonic detection range whose distance from the vehicle is greater than or equal to the obstacle distance is marked as the second probability, and the obstacle probability map is generated according to the first probability and the second probability.
  • the aforementioned ultrasonic parking space probability map generating module 200 may further include:
  • the last ultrasonic parking space probability map acquisition unit is used to obtain the last ultrasonic parking space probability map and the driving distance of the vehicle during the passage time period when the ultrasonic parking space probability map has been generated.
  • the passage time period is carried based on the ultrasonic distance data Is calculated from the first timestamp of and the previous timestamp adjacent to the first timestamp;
  • the reading unit is used to read the ultrasonic detection range and obstacle distance from the ultrasonic distance data, and mark the area within the ultrasonic detection range whose distance from the vehicle is less than the obstacle distance as the third probability, and the area within the ultrasonic detection range and The area where the distance of the vehicle is greater than or equal to the distance of the obstacle is marked as the fourth probability;
  • the positioning unit is used to locate the ultrasonic detection range and obstacle distance in the last ultrasonic parking probability map according to the driving distance of the vehicle;
  • the updating unit is used to update the probability of the area within the ultrasonic detection range and outside the obstacle distance according to the third probability and the fourth probability to obtain the obstacle probability map.
  • the above-mentioned image data processing module 400 is also used to process the image data to obtain a probability map that characterizes the parking space and the passable area of the image;
  • the aforementioned fusion module 500 is further configured to merge the ultrasonic parking space probability map and the probability map representing the image parking space and the image passable area according to the first time stamp and the second time stamp to obtain the first parking space probability map.
  • the above-mentioned image data processing module 400 is further configured to process the image data to obtain a probability map representing the parking space of the image and a probability map representing the passable area of the image;
  • the aforementioned fusion module 500 is further configured to merge the ultrasonic parking space probability map, the probability map representing the image parking space, and the probability map representing the passable area of the image according to the first time stamp and the second time stamp to obtain a second parking space probability map.
  • the above-mentioned parking space detection module further includes:
  • the driving distance acquisition unit is used to read the driving distance of the vehicle during the travel time period from the vehicle's odometer.
  • the travel time period is based on the first time stamp carried by the ultrasonic distance data and the previous time adjacent to the first time stamp Calculated by stamping;
  • the display unit is used to obtain the historical position corresponding to the previous time stamp, and obtain the vehicle position according to the historical position and the driving distance; and to display the vehicle position on the parking map.
  • Each module in the above parking space detection device can be implemented in whole or in part by software, hardware and a combination thereof.
  • the above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.
  • a computer device is provided.
  • the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 12.
  • the computer equipment includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus.
  • the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system and computer readable instructions.
  • the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the non-volatile storage medium.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer readable instruction is executed by the processor to realize a parking space detection method.
  • the display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, or it can be a button, a trackball or a touchpad set on the housing of the computer equipment , It can also be an external keyboard, touchpad, or mouse.
  • FIG. 12 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
  • the specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.
  • a computer device includes a memory and one or more processors.
  • the memory stores computer-readable instructions.
  • the one or more processors perform the following steps: Distance data, the ultrasonic distance data carries the first time stamp; the obstacle probability map is generated according to the ultrasonic distance data through the Bayes principle, and the ultrasonic parking probability map is obtained according to the obstacle probability map; the image data around the vehicle is collected, and the image data is carried There is a second time stamp; the image data is processed to obtain the image parking space and the image passable area; according to the first time stamp and the second time stamp, the ultrasonic parking space probability map, the image parking space and the image passable area are merged to obtain a parking space map; Determine the location of the parking space according to the parking map.
  • the Bayesian principle is used to generate the obstacle probability map based on the ultrasonic distance data, including: determining whether the ultrasonic parking space probability map has been generated; when the ultrasonic parking space is not generated For the probability map, the ultrasonic detection range and obstacle distance are read from the ultrasonic distance data. The area within the ultrasonic detection range whose distance to the vehicle is less than the obstacle distance is marked as the first probability. The area whose distance is greater than or equal to the obstacle distance is marked as the second probability, and the obstacle probability map is generated according to the first probability and the second probability.
  • the Bayesian principle is used to generate the obstacle probability map based on the ultrasonic distance data, including: when the ultrasonic parking space probability map has been generated, obtaining the previous one The ultrasonic parking space probability map and the driving distance of the vehicle during the passage time period.
  • the passage time period is calculated according to the first time stamp carried by the ultrasonic distance data and the previous time stamp adjacent to the first time stamp; from the ultrasonic distance data Read the ultrasonic detection range and obstacle distance in the middle, and mark the area within the ultrasonic detection range whose distance from the vehicle is less than the obstacle distance as the third probability, and the area within the ultrasonic detection range whose distance from the vehicle is greater than or equal to the obstacle distance Marked as the fourth probability; according to the driving distance of the vehicle, locate the ultrasonic detection range and obstacle distance in the previous ultrasonic parking probability map; update the ultrasonic detection range and the area outside the obstacle distance according to the third probability and the fourth probability The probability of obtaining the obstacle probability map.
  • the processing of image data to obtain image parking spaces and image passable areas realized when the processor executes computer-readable instructions includes: processing the image data to obtain the probability of characterizing image parking spaces and image passable areas Figure;
  • the processor executes computer-readable instructions executes computer-readable instructions, the ultrasonic parking space probability map, the image parking space and the image passable area are merged to obtain the parking space map according to the first time stamp and the second time stamp, including: according to the first time stamp With the second time stamp, the ultrasonic parking space probability map and the probability map representing the image parking space and the image passable area are fused to obtain the first parking space probability map.
  • the processing of image data to obtain image parking spaces and image passable areas realized when the processor executes the computer-readable instructions includes: processing the image data to obtain a probability map representing the image parking spaces and the representation image availability. Probability map of the passable area; when the processor executes the computer-readable instructions, the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain the parking space map according to the first and second timestamps, including: The first time stamp and the second time stamp are used to fuse the ultrasonic parking space probability map, the probability map representing the image parking space, and the probability map representing the passable area of the image to obtain the second parking space probability map.
  • the method further includes: reading the travel distance of the vehicle during the travel time period from the vehicle's odometer, and the travel time period is It is calculated according to the first time stamp carried by the ultrasonic distance data and the previous time stamp adjacent to the first time stamp; the historical position corresponding to the previous time stamp is obtained, and the vehicle position is obtained according to the historical position and driving distance; The location of the vehicle is displayed on the parking map.
  • One or more non-volatile computer-readable storage media storing computer-readable instructions.
  • the one or more processors perform the following steps: Ultrasonic distance data, the ultrasonic distance data carries the first time stamp; the obstacle probability map is generated according to the ultrasonic distance data through the Bayes principle, and the ultrasonic parking probability map is obtained according to the obstacle probability map; the image data around the vehicle is collected, the image data Carrying a second time stamp; processing the image data to obtain image parking spaces and image passable areas; according to the first and second timestamps, the ultrasonic parking space probability map, image parking spaces and image passable areas are fused to obtain a parking space map ; Determine the location of the parking space according to the parking map.
  • the Bayesian principle is used to generate the obstacle probability map based on the ultrasonic distance data, including: judging whether the ultrasonic parking probability map has been generated; For the parking space probability map, the ultrasonic detection range and obstacle distance are read from the ultrasonic distance data. The area within the ultrasonic detection range that is less than the obstacle distance from the vehicle is marked as the first probability. The area with a distance greater than or equal to the obstacle distance is marked as the second probability, and an obstacle probability map is generated according to the first probability and the second probability.
  • the Bayesian principle is used to generate the obstacle probability map based on the ultrasonic distance data, including: when the ultrasonic parking space probability map has been generated, obtaining the previous The probability map of ultrasonic parking space and the driving distance of the vehicle during the passage time period.
  • the passage time period is calculated according to the first time stamp carried by the ultrasonic distance data and the previous time stamp adjacent to the first time stamp; from the ultrasonic distance
  • the ultrasonic detection range and obstacle distance are read in the data, and the area within the ultrasonic detection range whose distance to the vehicle is less than the obstacle distance is marked as the third probability, and the distance to the vehicle within the ultrasonic detection range is greater than or equal to the obstacle distance
  • the area is marked as the fourth probability; according to the driving distance of the vehicle, locate the ultrasonic detection range and obstacle distance in the previous ultrasonic parking probability map; update the ultrasonic detection range and the distance outside the obstacle according to the third probability and the fourth probability Probability of the area, get the obstacle probability map.
  • the processing of the image data to obtain the image parking space and the image passable area realized when the computer-readable instruction is executed by the processor includes: inputting the image data into a pre-trained parking space detection model to obtain the image Parking space: Input the image data into the pre-trained passable area detection model to obtain the passable area of the image.
  • the processing of image data to obtain image parking spaces and image passable areas realized when the computer-readable instructions are executed by the processor includes: inputting image data into pre-trained parking spaces and passable area detection Model, get image parking space and image passable area.
  • the processing of the image data to obtain the image parking space and the image passable area realized when the computer-readable instruction is executed by the processor includes: processing the image data to obtain the characterizing the image parking space and the image passable area Probability map; when the processor executes the computer readable instructions, the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain the parking space map according to the first time stamp and the second time stamp, including: according to the first time The first parking space probability map is obtained by fusing the ultrasonic parking space probability map and the probability map representing the image parking space and the passable area of the image through the stamp and the second time stamp.
  • the processing of the image data to obtain the image parking space and the image passable area realized when the computer-readable instructions are executed by the processor includes: processing the image data to obtain the probability map representing the image parking space and the representation image Probability map of the passable area; when the processor executes the computer-readable instructions, the ultrasonic parking space probability map, the image parking space and the image passable area are fused to obtain the parking space map according to the first time stamp and the second time stamp, which includes: According to the first time stamp and the second time stamp, the ultrasonic parking space probability map, the probability map representing the image parking space and the probability map representing the passable area of the image are fused to obtain the second parking space probability map.
  • the computer readable instruction after the computer readable instruction is executed by the processor to determine the parking space location based on the parking space map, it further includes: reading the vehicle's travel distance during the transit time period from the vehicle's odometer, and the transit time period It is calculated according to the first timestamp carried by the ultrasonic distance data and the previous timestamp adjacent to the first timestamp; the historical position corresponding to the previous timestamp is obtained, and the vehicle position is obtained according to the historical position and the driving distance; Show the location of the vehicle on the parking map.
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Traffic Control Systems (AREA)

Abstract

一种车位检测方法，包括：采集车辆周围的超声距离数据，超声距离数据携带有第一时间戳（S202）；通过贝叶斯原理根据超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图（S204）；采集车辆周围的图像数据，图像数据携带有第二时间戳（S206）；对图像数据进行处理得到图像车位和图像可通行区域（S208）；根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图（S210）；根据车位地图以确定车位位置（S212）。

Description

车位检测方法、装置、计算机设备和存储介质 技术领域
本申请涉及一种车位检测方法、装置、计算机设备和存储介质。
背景技术
随着智能驾驶技术的发展，出现了全自动泊车(APA)技术，极大地提高了自动驾驶技术的发展。
然而，发明人意识到，从前期的纯超声泊车到现如今的纯视觉泊车都有其特有缺点，无法满足现有APA场景的需求。例如：纯超声泊车在全为空地的场景下，就失去了其基本功能；而纯视觉泊车在光线较暗，没有车位线或者车位线不清晰的情形下也失去了其基本功能。这样导致全自动泊车过程中车位检测不准确。
发明内容
根据本申请公开的各种实施例，提供一种车位检测方法、装置、计算机设备和存储介质。
一种车位检测方法，包括：
采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
对所述图像数据进行处理得到图像车位和图像可通行区域；
根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
根据所述车位地图以确定车位位置。
一种车位检测装置，包括：
超声距离数据采集模块，用于采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
超声车位概率地图生成模块，用于通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
图像数据采集模块，用于采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
图像数据处理模块，用于对所述图像数据进行处理得到图像车位和图像可通行区域；
融合模块，用于根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、 所述图像车位和图像可通行区域进行融合得到车位地图；及
输出模块，用于根据所述车位地图以确定车位位置。
一种计算机设备，包括存储器和一个或多个处理器，所述存储器中储存有计算机可读指令，所述计算机可读指令被所述处理器执行时，使得所述一个或多个处理器执行以下步骤：
采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
对所述图像数据进行处理得到图像车位和图像可通行区域；
根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
根据所述车位地图以确定车位位置。
一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行以下步骤：
采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
对所述图像数据进行处理得到图像车位和图像可通行区域；
根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
根据所述车位地图以确定车位位置。
计算机可读指令计算机可读指令计算机可读指令计算机可读指令
本申请的一个或多个实施例的细节在下面的附图和描述中提出。本申请的其它特征和优点将从说明书、附图以及权利要求书变得明显。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附图。
图1为根据一个或多个实施例中车位检测方法的应用环境图。
图2为根据一个或多个实施例中车位检测方法的流程示意图。
图3为根据一个或多个实施例中的超声车位概率地图的示意图。
图4为根据一个或多个实施例中的图像车位和图像可通行区域的示意图。
图5为根据一个或多个实施例中的融合后的车位地图的示意图。
图6为根据一个或多个实施例中的超声波探测器未扫描到障碍物时的示意图。
图7为根据一个或多个实施例中的超声波探测器初次扫描到障碍物时的示意图。
图8为根据一个或多个实施例中的超声波探测器最后一次扫描到障碍物时的示意图。
图9为根据另一个或多个实施例中车位检测方法的流程示意图。
图10为根据再一个或多个实施例中车位检测方法的流程示意图。
图11为根据一个或多个实施例中车位检测装置的框图。
图12为根据一个或多个实施例中计算机设备的框图。
具体实施方式
为了使本申请的技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请提供的车位检测方法，可以应用于如图1所示的应用环境中。其中，车辆终端102与车辆上安装的各个控制器104进行通信，该控制器104可以包括但不限于安装在车辆上的摄像头、超声波传感器以及用于采集车辆运行数据的各个控制器。车辆终端102可以通过超声波传感器采集车辆周围的超声距离数据以及通过摄像头采集车辆周围的图像数据，超声距离数据携带有第一时间戳，图像数据携带有第二时间戳，其中车辆终端102通过贝叶斯原理根据超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图，通过对图像数据进行处理得到图像车位和图像可通行区域，这样车辆终端102最后根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图；根据车位地图以确定车位位置。这样，通过超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图，此外还通过图像数据得到图像车位和图像可通行区域，这样根据第一时间戳和第二时间戳将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，从而将两者进行结合，可以提高车位检测的准确性。其中，车辆终端102可以但不限于是各种安装在车辆上的个人计算机、笔记本电脑、智能手机、平板电脑。
在一个实施例中，如图2所示，提供了一种车位检测方法，以该方法应用于图1中的车辆终端为例进行说明，包括以下步骤：
S202：采集车辆周围的超声距离数据，超声距离数据携带有第一时间戳。
具体地，超声距离数据是通过车辆上安装的超声波传感器所采集的，该超声波传感器的发射器发射超声波，并记录发射波的发射时间，从而超声波遇见障碍物则会反射，这样超声波传感器的接收器接收到反射波，通过接收到反射波的时间、发射波的发射时间以及光波的传输速率则为超声距离数据，车辆终端可以根据超声距离数据、车辆的行驶速度等可以得到障碍物的相对于车辆的位置。
S204：通过贝叶斯原理根据超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图。
具体地，障碍物概率地图是指表征地图上各个位置存在障碍物的概率的地图，进而超声车位概率地图则是表征地图上各个位置存在车位的概率的地图。
车辆终端实时获取到上述超声距离数据，并根据该超声距离数据得到障碍物的位置，从而可以根据超声波传感器的探测范围以及该障碍物的位置来生成障碍物概率地图。
具体地，超声数据为传感器直接读数，对超声波传感器来讲，读数为距离m，表示在该超声探测范围内，距离超声波传感器m处有障碍物。因读数没有角度信息，只有距离信息，因此车辆终端只能确定在距离超声波传感器的一段圆弧边(半径为m)上有障碍物，具体是哪个部分有障碍物无法确定，因此距离超声波传感器m处至超声波传感器探测范围内具有障碍物的概率增加，距离超声波传感器的距离小于m处具有障碍物的概率不变，这样经过多次扫描即可以得到障碍物概率地图。在车辆行驶过程中，超声波传感器探测包络扫过栅格地图点，单个栅格地图点会被多次扫描，从而得出最准确的障碍物位置。从而没有障碍物的区域，则认为可能存在车位，因此车辆终端可以根据障碍物概率地图得到超声车位概率地图，即存在障碍物的概率越大，则存在车位的概率越小，存在障碍物的概率越小，则存在车位的概率越大，具体地，超声车位概率地图可以参见图3所示。
S206：采集车辆周围的图像数据，图像数据携带有第二时间戳。
具体地，图像数据是指车辆周围的摄像头采集的图像数据，其中车辆周围可以包括多个摄像头，例如4个或者是更多的摄像头，图像数据可以是根据该多个摄像头所采集的图像生成环视图，该环视图可以是根据时间戳相同的单张图像来生成的，这样该相同的时间戳即为环视图的第二时间戳。
此外，可选地，该采集车辆周围的图像数据的步骤可以与上述的采集车辆周围的超声距离数据的步骤是并行进行的，这样保证数据的处理效率，例如，车辆终端通过第一线程来采集超声距离数据，并根据超声距离数据来得到超声车位概率图，车辆终端通过第二线程来采集图像数据，并对图像数据进行处理得到图像车位和图像可通行区域。
S208：对图像数据进行处理得到图像车位和图像可通行区域。
具体地，图像车位是指通过对图像数据进行处理得到的车位位置，图像可通行区域是指通过对图像数据进行处理得到的可通行区域的位置。车辆终端对图像数据进行处理的方式可以是通过深度学习模型来对图像数据进行处理以得到图像车位和图像可通行区域。
图像车位检测可以采用多种方法，例如使用深度学习训练车位检测的学习网络，前期人工标识大量训练图片(标记出图像中车位与非车位像素)，通过训练网络获得可靠的识别率，应用时输入图片即可输出检测到的车位信息。图像可通行区域Freespace也可以应用相同的方法，区别在于网络结构的选用以及前期标识时进行的标记不同图像可通行区域需要标记可通行区域和不可通行区域，具体地，通过对图像数据进行处理所得到的图像车位和图像可通行区域可以如图4所示。
S210：根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图。
S212：根据车位地图以确定车位位置。
具体地，车位地图是是可以表征车位位置的地图，其是通过将超声车位概率地图、图像车位和图像可通行区域进行叠加融合得到的，例如上述的图像车位和图像可通行区域均是地图中的概率区域，则车辆终端将超声车位概率地图、图像车位和图像可通行区域的概率区域进行叠加，这样得到一张表征车位的概率的车位地图。其中图像车位和图像可通行区域的概率为100％时，则确定为车位或可通信区域，此时对应的深度学习模型可以为二分类模型；若深度学习模型是概率模型，则图像车位和图像可通信区域是以概率进行表征的，则需要将图像车位和图像可通行区域的概率与超声车位概率地图中的概率进行叠加以进行判断。
然后车辆终端通过预设的概率阈值进行判断，例如车辆终端将概率大于概率阈值的位置判定为车位位置，最后根据常规车位的形状等对车位位置进行滤波处理得到准确的车位位置，具体地，融合后的地图可以参见图5所示。
上述车位检测方法中，通过超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图，此外还通过图像数据得到图像车位和图像可通行区域，这样根据第一时间戳和第二时间戳将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，从而将两者进行结合，可以提高车位检测的准确性。
在其中一个实施例中，通过贝叶斯原理根据超声距离数据生成障碍物概率地图，包括：判断是否已生成了超声车位概率地图；当未生成超声车位概率地图的，则从超声距离数据中读取超声探测范围以及障碍物距离，对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第一概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第二概率，并根据第一概率和第二概率生成障碍物概率地图。其中此处的第一概率和第二概率若是表征车位和可通行区域的，则第一概率为100％，第二概率为0％，若第一概率和第二概率若是表征障碍物区域的，则第一概率为0％，第二概率为100％。
在其中一个实施例中，通过贝叶斯原理根据超声距离数据生成障碍物概率地图，包括：当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离。通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；从超声距离数据中读取超声探测范围以及障碍物距离，并对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第三概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第四概率；根据车辆的行驶距离，在上一张超声车位概率地图中定位超声探测范围以及障碍物距离；根据第三概率和第四概率更新超声探测范围内、障碍物距离外的区域的概率，得到障碍物概率地图。其中此处的第三概率和第四概率若是表征车位和可通行区域的，则第三概率为100％，第四概率为0％，若第三概率和第四概率若是表征障碍物区域的，则第三概率为0％，第四概率为100％。
具体地，请参照图6至图8所示，其中图6为一个实施例中的超声波探测器未扫描到障碍物时的示意图，图7为一个实施例中的超声波探测器初次扫描到障碍物时的示意图，图8为一个实施例中的超声波探测器最后一次扫描到障碍物时的示意图。具体地，以图6中两个栅格地图点为例，黑色填充圆代表实际障碍物，标记为黑点，黑色未填充圆为实际障碍物周边一个点，应为空闲，即非障碍物区域，标记为白。当车辆还未扫到障碍物时，白点非障碍物概率一直在增加，黑点因为没有扫到，概率不变。当车辆扫描到障碍物时，因为白点与黑点距离传感器的距离相同，所以它们被占有的概率，即障碍物概率会同步增加。当车辆扫描范围离开白点时，白点的被占有概率，即障碍物概率不再改变，因其已不在探测范围内。而此时黑点的被占有概率，即障碍物概率持续增加，直到扫描完毕。可以看到，在整个扫描过程中，黑点只要在探测范围内，其被占有概率，即障碍物概率持续增加，而白点仅在部分时候会增加，其余时间空闲概率，即非障碍物概率会不变。这一现象在白点距离黑点更远时更加明显。因此可以确认，在扫描完毕后，实际障碍物处，被占有概率，也就是说障碍物概率最高，其周边栅格点的被占有概率，即障碍物概率逐渐减小，这样扫描完成即可以得到表征障碍物位置的障碍物概率图，从而车辆终端通过设定合理的阈值后即可提取障碍物边缘。
此外，上述过程中当扫单次描到某一障碍物时，是没有办法确定其具体位置的。如果不使用概率地图，则需要根据实际的包络形状，按照复杂的逻辑判断障碍物在何处。这个过程需要很多的标定值。而采用概率地图以后，按照上述方法，不再需要如此判断，也就不需要标定，从而可以提高效率。
上述实施例中，上述障碍物概率地图的构建，克服了之前超声测距的以下劣势：原超声测距所代表的距离概率都是100％，无法进行有效剔除。超声包络所特有的特性，这使得使用超声泊车过程中要根据场景反复大量测试进行余量的标定。而本实施例中的障碍物地图的构建不需要以上操作，即可以一次就获得较高精度的障碍物的边缘信息。
在其中一个实施例中，对图像数据进行处理得到图像车位和图像可通行区域，包括：将图像数据输入至预先训练得到的车位检测模型，得到图像车位；将图像数据输入至预先训练得到的可通行区域检测模型中，得到图像可通行区域。上述实施例中，分别通过深度学习得到图像车位和图像可通行位置，可以获得比较可靠的识别率。
在其中一个实施例中，对图像数据进行处理得到图像车位和图像可通行区域，包括：对图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，包括：根据第一时间戳和第二时间戳，将超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。上述实施例中，通过超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合，可以得到表征车位的第一车位概率图，因为均为概率地图，某一处是否为车位仅需要叠加三张地图，概率高于概率阈值的位置一定是可泊车位，其中该概率阈值可以是预先设定的， 例如90％等，对此不作具体的限制。
具体地，请参见图9，图9为另一个实施例中的车位检测方法的流程示意图。其中车辆终端可以采用多线程处理的方式，其中第一线程通过超声波传感器采集超声距离数据，然后根据超声距离数据生成障碍物概率地图，再根据障碍物概率地图得到超声车位概率地图。第二线程通过摄像头采集到图像数据，其中可以将摄像头采集的数据进行拼接得到环视图，然后第二线程开启另外一个第三线程，其中第二线程可以对环视图或原始图像数据进行处理生成视觉可通行区域，即上文中的图像可通行区域。第三线程对环视图或原始图像数据进行处理生成视觉车位，即上文中的图像车位。其中第二线程和第三线程的处理可以是通过深度学习来进行，例如第三线程中车辆终端将图像数据输入至预先训练得到的车位检测模型，得到图像车位；第二线程中车辆终端将图像数据输入至预先训练得到的可通行区域检测模型中，得到图像可通行区域。最后车辆终端将所得到上述超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合，即根据时间戳确定对应的图，然后根据图像坐标确定各个对应的图像区域，最后将对应图像区域的概率相加进行融合，这样可以得到第二车位概率图，进而车辆终端根据该第二车位概率图可以确定车位位置。且可选地，车辆终端还可以设置第四线程，这样通过第四线程读取车辆的速度和行驶方向，从而可以根据车辆的速度和行驶方向对车辆进行定位，并显示到第二车位概率图中，可选地，由于第二车位概率图的计算耗费时间，因此需要对车辆进行补偿，以确定车辆在第二车位概率图中的准确位置，例如通过预设耗费时间以及当前车速和车辆行驶方向来确定补偿位移，再根据车辆的定位位置以及该补偿位移确定车辆的准确位置。
上述实施例中，通过超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合，可以得到表征车位的第一车位概率图，因为均为概率地图，某一处是否为车位仅需要叠加三张地图，例如将该处的三张图中的概率进行相加得到综合概率，并将综合概率与预先设定的阈值进行比较，概率高于阈值的位置一定是可泊车位。
在其中一个实施例中，对图像数据进行处理得到图像车位和图像可通行区域，包括：将图像数据输入至预先训练得到的车位和可通行区域检测模型，得到图像车位和图像可通行区域。上述实施例中，通过深度学习得到图像车位和图像可通行位置，可以获得比较可靠的识别率。
在其中一个实施例中，对图像数据进行处理得到图像车位和图像可通行区域，包括：对图像数据进行处理得到表征图像车位和图像可通行区域的概率图；根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，包括：根据第一时间戳和第二时间戳，将超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图。上述实施例中，通过将超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合，可以得到表征车位的第一车位概率图，因为均为概率地图，某一处是否为车位仅需要叠加两张地图，例如将该处的三张图中的概率进行相加得到综合概率，并将综合概率与预先设定的阈值进行比较，概率高于阈值 的位置一定是可泊车位。
具体地，请参见图10，图10为再一个实施例中的车位检测方法的流程示意图。其中车辆终端可以采用多线程处理的方式，其中第一线程通过超声波传感器采集超声距离数据，然后根据超声距离数据生成障碍物概率地图，再根据障碍物概率地图得到超声车位概率地图。第二线程通过摄像头采集到图像数据，其中可以将摄像头采集的数据进行拼接得到环视图，然后第二线程将图像数据输入至预先训练得到的车位和可通行区域检测模型，得到图像车位和图像可通行区域。最后车辆终端将所得到超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图，即根据时间戳确定对应的图，然后根据图像坐标确定各个对应的图像区域，最后将对应图像区域的概率相加进行融合，这样可以得到第一车位概率图，进而车辆终端根据该第一车位概率图可以确定车位位置。且可选地，车辆终端还可以设置另外一个线程，这样通过另外一个线程读取车辆的速度和行驶方向，从而可以根据车辆的速度和行驶方向对车辆进行定位，并显示到第一车位概率图中，可选地，由于第一车位概率图的计算耗费时间，因此需要对车辆进行补偿，以确定车辆在第一车位概率图中的准确位置，例如通过预设耗费时间以及当前车速和车辆行驶方向来确定补偿位移，再根据车辆的定位位置以及该补偿位移确定车辆的准确位置。
上述实施例中，通过将超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合，可以得到表征车位的第一车位概率图，因为均为概率地图，某一处是否为车位仅需要叠加两张地图，概率高的位置一定是可泊车位。
在其中一个实施例中，根据车位地图以确定车位位置之后还包括：从车辆的里程计读取车辆在通行时间段的行驶距离，通行时间段是根据所述超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；在车位地图上显示车辆位置。
具体地，其中为了准确地对车辆进行控制，车辆终端还在车位地图上显示车辆的位置，具体地，车辆终端可以单独开启一个线程，在该线程中，车辆终端从车辆的里程计读取车辆的行驶距离，例如可以实时读取通行时间段的行驶距离，该通时间段是根据所述超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的，获取前一时间戳对应的历史位置，并根据历史位置以及行驶距离得到车辆位置；也就是相邻的超声距离数据的采集时间的差值，这样通过里程累加即可以在车位地图上显示车辆位置，其中相邻时间戳的时间是与超声距离数据的采集频率有关的，例如1ms或者是10ms等，用户可以自行设置。
且可选地，由于车位地图的生成需要耗费一定的时间，为了补偿该部分里程差，车辆终端可以根据车辆速度和行驶方向来确定补偿位移，并将根据第一时间戳和第二时间戳所确定的车辆位置以及该补偿位移来更新车辆位置，这样所得到的车辆的位置是准确的，从而在车位地图上显示准确的车辆位置。例如车辆终端可以根据超声距离数据来确定车辆的位置，但是该位置是对应第一时间戳的位置，在计算超声障碍物概率地图时会耗费一定的时间，这样车辆的位置会发生变化，因此车辆终端为了补偿该部分变化，车辆终端根据车 辆速度和行驶方向来确定补偿位移，并将根据第一时间戳和第二时间戳所确定的车辆位置以及该补偿位移来更新车辆位置，以保证车辆位置的准确性。
上述实施例中，还将车辆位置显示在地图上，这样便于查找距离车辆较近的车位，从而方便泊车。
应该理解的是，虽然图2、图9和图10的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，图2、图9和图10中的至少一部分步骤可以包括多个子步骤或者多个阶段，这些子步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些子步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
在其中一个实施例中，如图11所示，提供了一种车位检测装置，包括：超声距离数据采集模块100、超声车位概率地图生成模块200、图像数据采集模块300、图像数据处理模块400、融合模块500和输出模块600，其中：
超声距离数据采集模块100，用于采集车辆周围的超声距离数据，超声距离数据携带有第一时间戳；
超声车位概率地图生成模块200，用于通过贝叶斯原理根据超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图；
图像数据采集模块300，用于采集车辆周围的图像数据，图像数据携带有第二时间戳；
图像数据处理模块400，用于对图像数据进行处理得到图像车位和图像可通行区域；
融合模块500，用于根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图；
输出模块600，用于根据车位地图以确定车位位置。
在其中一个实施例中，上述的超声车位概率地图生成模块200可以包括：
判断单元，用于判断是否已生成了超声车位概率地图；
第一生成单元，用于当未生成超声车位概率地图的，则从超声距离数据中读取超声探测范围以及障碍物距离，对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第一概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第二概率，并根据第一概率和第二概率生成障碍物概率地图。
在其中一个实施例中，上述的超声车位概率地图生成模块200还可以包括：
上一超声车位概率地图获取单元，用于当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离，通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；
读取单元，用于从超声距离数据中读取超声探测范围以及障碍物距离，并对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第三概率，超声探测范围内的与车辆 的距离大于等于障碍物距离的区域标记为第四概率；
定位单元，用于根据车辆的行驶距离，在上一张超声车位概率地图中定位超声探测范围以及障碍物距离；
更新单元，用于根据第三概率和第四概率更新超声探测范围内、障碍物距离外的区域的概率，得到障碍物概率地图。
在其中一个实施例中，上述图像数据处理模块400还用于对图像数据进行处理得到表征图像车位和图像可通行区域的概率图；
上述融合模块500还用于根据第一时间戳和第二时间戳，将超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图。
在其中一个实施例中，上述图像数据处理模块400还用于对图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；
上述融合模块500还用于根据第一时间戳和第二时间戳，将超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。
在其中一个实施例中，上述车位检测模块还包括：
行驶距离获取单元，用于从车辆的里程计读取车辆在通行时间段的行驶距离，通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；
显示单元，用于获取前一时间戳对应的历史位置，并根据历史位置以及行驶距离得到车辆位置；在车位地图上显示车辆位置。
关于车位检测装置的具体限定可以参见上文中对于车位检测方法的限定，在此不再赘述。上述车位检测装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在其中一个实施例中，提供了一种计算机设备，该计算机设备可以是终端，其内部结构图可以如图12所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口、显示屏和输入装置。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统和计算机可读指令。该内存储器为非易失性存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机可读指令被处理器执行时以实现一种车位检测方法。该计算机设备的显示屏可以是液晶显示屏或者电子墨水显示屏，该计算机设备的输入装置可以是显示屏上覆盖的触摸层，也可以是计算机设备外壳上设置的按键、轨迹球或触控板，还可以是外接的键盘、触控板或鼠标等。
本领域技术人员可以理解，图12中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备 可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
一种计算机设备，包括存储器和一个或多个处理器，存储器中储存有计算机可读指令，计算机可读指令被处理器执行时，使得一个或多个处理器执行以下步骤：采集车辆周围的超声距离数据，超声距离数据携带有第一时间戳；通过贝叶斯原理根据超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图；采集车辆周围的图像数据，图像数据携带有第二时间戳；对图像数据进行处理得到图像车位和图像可通行区域；根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图；根据车位地图以确定车位位置。
在其中一个实施例中，处理器执行计算机可读指令时所实现的通过贝叶斯原理根据超声距离数据生成障碍物概率地图，包括：判断是否已生成了超声车位概率地图；当未生成超声车位概率地图的，则从超声距离数据中读取超声探测范围以及障碍物距离，对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第一概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第二概率，并根据第一概率和第二概率生成障碍物概率地图。
在其中一个实施例中，处理器执行计算机可读指令时所实现的通过贝叶斯原理根据超声距离数据生成障碍物概率地图，包括：当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离，通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；从超声距离数据中读取超声探测范围以及障碍物距离，并对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第三概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第四概率；根据车辆的行驶距离，在上一张超声车位概率地图中定位超声探测范围以及障碍物距离；根据第三概率和第四概率更新超声探测范围内、障碍物距离外的区域的概率，得到障碍物概率地图。
在其中一个实施例中，处理器执行计算机可读指令时所实现的对图像数据进行处理得到图像车位和图像可通行区域，包括：对图像数据进行处理得到表征图像车位和图像可通行区域的概率图；处理器执行计算机可读指令时所实现的根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，包括：根据第一时间戳和第二时间戳，将超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图。
在其中一个实施例中，处理器执行计算机可读指令时所实现的对图像数据进行处理得到图像车位和图像可通行区域，包括：对图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；处理器执行计算机可读指令时所实现的根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，包括：根据第一时间戳和第二时间戳，将超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。
在其中一个实施例中，处理器执行计算机可读指令时所实现的根据车位地图以确定车位位置之后，还包括：从车辆的里程计读取车辆在通行时间段的行驶距离，通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；获取前一时间戳对应的历史位置，并根据历史位置以及行驶距离得到车辆位置；在车位地图上显示车辆位置。
一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行以下步骤：采集车辆周围的超声距离数据，超声距离数据携带有第一时间戳；通过贝叶斯原理根据超声距离数据生成障碍物概率地图，并根据障碍物概率地图得到超声车位概率地图；采集车辆周围的图像数据，图像数据携带有第二时间戳；对图像数据进行处理得到图像车位和图像可通行区域；根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图；根据车位地图以确定车位位置。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的通过贝叶斯原理根据超声距离数据生成障碍物概率地图，包括：判断是否已生成了超声车位概率地图；当未生成超声车位概率地图的，则从超声距离数据中读取超声探测范围以及障碍物距离，对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第一概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第二概率，并根据第一概率和第二概率生成障碍物概率地图。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的通过贝叶斯原理根据超声距离数据生成障碍物概率地图，包括：当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离，通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；从超声距离数据中读取超声探测范围以及障碍物距离，并对超声探测范围内的与车辆的距离小于障碍物距离的区域标记为第三概率，超声探测范围内的与车辆的距离大于等于障碍物距离的区域标记为第四概率；根据车辆的行驶距离，在上一张超声车位概率地图中定位超声探测范围以及障碍物距离；根据第三概率和第四概率更新超声探测范围内、障碍物距离外的区域的概率，得到障碍物概率地图。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的对图像数据进行处理得到图像车位和图像可通行区域，包括：将图像数据输入至预先训练得到的车位检测模型，得到图像车位；将图像数据输入至预先训练得到的可通行区域检测模型中，得到图像可通行区域。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的对图像数据进行处理得到图像车位和图像可通行区域，包括：将图像数据输入至预先训练得到的车位和可通行区域检测模型，得到图像车位和图像可通行区域。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的对图像数据进行处理 得到图像车位和图像可通行区域，包括：对图像数据进行处理得到表征图像车位和图像可通行区域的概率图；处理器执行计算机可读指令时所实现的根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，包括：根据第一时间戳和第二时间戳，将超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的对图像数据进行处理得到图像车位和图像可通行区域，包括：对图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；处理器执行计算机可读指令时所实现的根据第一时间戳和第二时间戳，将超声车位概率地图、图像车位和图像可通行区域进行融合得到车位地图，包括：根据第一时间戳和第二时间戳，将超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。
在其中一个实施例中，计算机可读指令被处理器执行时所实现的根据车位地图以确定车位位置之后，还包括：从车辆的里程计读取车辆在通行时间段的行驶距离，通行时间段是根据超声距离数据所携带的第一时间戳以及与第一时间戳相邻的前一时间戳计算得到的；获取前一时间戳对应的历史位置，并根据历史位置以及行驶距离得到车辆位置；在车位地图上显示车辆位置。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机可读指令计算机可读指令来指令相关的硬件来完成，所述的计算机可读指令计算机可读指令可存储于一非易失性计算机可读取存储介质中，该计算机可读指令计算机可读指令在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种车位检测方法，包括：
    采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
    通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
    采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
    对所述图像数据进行处理得到图像车位和图像可通行区域；
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
    根据所述车位地图以确定车位位置。
  2. 根据权利要求1所述的方法，其特征在于，所述通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，包括：
    判断是否已生成了超声车位概率地图；及
    当未生成超声车位概率地图的，则从所述超声距离数据中读取超声探测范围以及障碍物距离，对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第一概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第二概率，并根据所述第一概率和所述第二概率生成障碍物概率地图。
  3. 根据权利要求2所述的方法，其特征在于，所述通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，还包括：
    当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离，所述通行时间段是根据所述超声距离数据所携带的第一时间戳以及与所述第一时间戳相邻的前一时间戳计算得到的；
    从所述超声距离数据中读取超声探测范围以及障碍物距离，并对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第三概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第四概率；
    根据所述车辆的行驶距离，在所述上一张超声车位概率地图中定位所述超声探测范围以及障碍物距离；及
    根据所述第三概率和第四概率更新所述超声探测范围内、所述障碍物距离外的区域的概率，得到障碍物概率地图。
  4. 根据权利要求1至3任意一项所述的方法，其特征在于，所述对所述图像数据进行处理得到图像车位和图像可通行区域，包括：
    对所述图像数据进行处理得到表征图像车位和图像可通行区域的概率图；及
    所述根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图，包括：
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图以及表征图像车位 和图像可通行区域的概率图进行融合得到第一车位概率图。
  5. 根据权利要求1至3任意一项所述的方法，其特征在于，所述对所述图像数据进行处理得到图像车位和图像可通行区域，包括：
    对所述图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；及
    所述根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图，包括：
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。
  6. 根据权利要求1所述的方法，其特征在于，所述根据所述车位地图以确定车位位置之后，还包括：
    从车辆的里程计读取车辆在通行时间段的行驶距离，所述通行时间段是根据所述超声距离数据所携带的第一时间戳以及与所述第一时间戳相邻的前一时间戳计算得到的；
    获取所述前一时间戳对应的历史位置，并根据所述历史位置以及所述行驶距离得到车辆位置；及
    在所述车位地图上显示所述车辆位置。
  7. 一种车位检测装置，包括：
    超声距离数据采集模块，用于采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
    超声车位概率地图生成模块，用于通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
    图像数据采集模块，用于采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
    图像数据处理模块，用于对所述图像数据进行处理得到图像车位和图像可通行区域；
    融合模块，用于根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
    输出模块，用于根据所述车位地图以确定车位位置。
  8. 根据权利要求7所述的装置，其特征在于，所述超声车位概率地图生成模块包括：
    判断单元，用于判断是否已生成了超声车位概率地图；及
    第一生成单元，用于当未生成超声车位概率地图的，则从所述超声距离数据中读取超声探测范围以及障碍物距离，对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第一概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第二概率，并根据所述第一概率和所述第二概率生成障碍物概率地图。
  9. 根据权利要求8所述的装置，其特征在于，所述超声车位概率地图生成模块包括：
    上一超声车位概率地图获取单元，用于当已经生成了超声车位概率地图的，则获取上 一张超声车位概率地图以及车辆在通行时间段的行驶距离，所述通行时间段是根据所述超声距离数据所携带的第一时间戳以及与所述第一时间戳相邻的前一时间戳计算得到的；
    读取单元，用于从所述超声距离数据中读取超声探测范围以及障碍物距离，并对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第三概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第四概率；
    定位单元，用于根据所述车辆的行驶距离，在所述上一张超声车位概率地图中定位所述超声探测范围以及障碍物距离；及
    更新单元，用于根据所述第三概率和第四概率更新所述超声探测范围内、所述障碍物距离外的区域的概率，得到障碍物概率地图。
  10. 一种计算机设备，包括存储器及一个或多个处理器，所述存储器中储存有计算机可读指令，所述计算机可读指令被所述一个或多个处理器执行时，使得所述一个或多个处理器执行以下步骤：
    采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
    通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
    采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
    对所述图像数据进行处理得到图像车位和图像可通行区域；
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
    根据所述车位地图以确定车位位置。
  11. 根据权利要求10所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时所实现的所述通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，包括：
    判断是否已生成了超声车位概率地图；及
    当未生成超声车位概率地图的，则从所述超声距离数据中读取超声探测范围以及障碍物距离，对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第一概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第二概率，并根据所述第一概率和所述第二概率生成障碍物概率地图。
  12. 根据权利要求11所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时所实现的所述通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，还包括：
    当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离，所述通行时间段是根据所述超声距离数据所携带的第一时间戳以及与所述第一时间戳相邻的前一时间戳计算得到的；
    从所述超声距离数据中读取超声探测范围以及障碍物距离，并对所述超声探测范围内 的与车辆的距离小于所述障碍物距离的区域标记为第三概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第四概率；
    根据所述车辆的行驶距离，在所述上一张超声车位概率地图中定位所述超声探测范围以及障碍物距离；及
    根据所述第三概率和第四概率更新所述超声探测范围内、所述障碍物距离外的区域的概率，得到障碍物概率地图。
  13. 根据权利要求10至12任意一项所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时所实现的所述对所述图像数据进行处理得到图像车位和图像可通行区域，包括：
    对所述图像数据进行处理得到表征图像车位和图像可通行区域的概率图；及
    所述根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图，包括：
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图。
  14. 根据权利要求10至13任意一项所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时所实现的所述对所述图像数据进行处理得到图像车位和图像可通行区域，包括：
    对所述图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；及
    所述根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图，包括：
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。
  15. 根据权利要求10所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时所实现的所述根据所述车位地图以确定车位位置之后，还包括：
    从车辆的里程计读取车辆在通行时间段的行驶距离，所述通行时间段是根据所述超声距离数据所携带的第一时间戳以及与所述第一时间戳相邻的前一时间戳计算得到的；
    获取所述前一时间戳对应的历史位置，并根据所述历史位置以及所述行驶距离得到车辆位置；及
    在所述车位地图上显示所述车辆位置。
  16. 一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行以下步骤：
    采集车辆周围的超声距离数据，所述超声距离数据携带有第一时间戳；
    通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，并根据所述障碍物概率地图得到超声车位概率地图；
    采集所述车辆周围的图像数据，所述图像数据携带有第二时间戳；
    对所述图像数据进行处理得到图像车位和图像可通行区域；
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图；及
    根据所述车位地图以确定车位位置。
  17. 根据权利要求16所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时所实现的所述通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，包括：
    判断是否已生成了超声车位概率地图；及
    当未生成超声车位概率地图的，则从所述超声距离数据中读取超声探测范围以及障碍物距离，对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第一概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第二概率，并根据所述第一概率和所述第二概率生成障碍物概率地图。
  18. 根据权利要求17所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时所实现的所述通过贝叶斯原理根据所述超声距离数据生成障碍物概率地图，还包括：
    当已经生成了超声车位概率地图的，则获取上一张超声车位概率地图以及车辆在通行时间段的行驶距离，所述通行时间段是根据所述超声距离数据所携带的第一时间戳以及与所述第一时间戳相邻的前一时间戳计算得到的；
    从所述超声距离数据中读取超声探测范围以及障碍物距离，并对所述超声探测范围内的与车辆的距离小于所述障碍物距离的区域标记为第三概率，所述超声探测范围内的与车辆的距离大于等于所述障碍物距离的区域标记为第四概率；
    根据所述车辆的行驶距离，在所述上一张超声车位概率地图中定位所述超声探测范围以及障碍物距离；及
    根据所述第三概率和第四概率更新所述超声探测范围内、所述障碍物距离外的区域的概率，得到障碍物概率地图。
  19. 根据权利要求16至18任意一项所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时所实现的所述对所述图像数据进行处理得到图像车位和图像可通行区域，包括：
    对所述图像数据进行处理得到表征图像车位和图像可通行区域的概率图；及
    所述根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图，包括：
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图以及表征图像车位和图像可通行区域的概率图进行融合得到第一车位概率图。
  20. 根据权利要求16至18任意一项所述的存储介质，其特征在于，所述计算机可读 指令被所述处理器执行时所实现的所述对所述图像数据进行处理得到图像车位和图像可通行区域，包括：
    对所述图像数据进行处理得到表征图像车位的概率图和表征图像可通行区域的概率图；及
    所述根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、所述图像车位和图像可通行区域进行融合得到车位地图，包括：
    根据所述第一时间戳和所述第二时间戳，将所述超声车位概率地图、表征图像车位的概率图和表征图像可通行区域的概率图进行融合得到第二车位概率图。
PCT/CN2020/089901 2020-05-13 2020-05-13 车位检测方法、装置、计算机设备和存储介质 WO2021226852A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/089901 WO2021226852A1 (zh) 2020-05-13 2020-05-13 车位检测方法、装置、计算机设备和存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/089901 WO2021226852A1 (zh) 2020-05-13 2020-05-13 车位检测方法、装置、计算机设备和存储介质

Publications (1)

Publication Number Publication Date
WO2021226852A1 true WO2021226852A1 (zh) 2021-11-18

Family

ID=78526149

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/089901 WO2021226852A1 (zh) 2020-05-13 2020-05-13 车位检测方法、装置、计算机设备和存储介质

Country Status (1)

Country Link
WO (1) WO2021226852A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114648882A (zh) * 2022-02-09 2022-06-21 上海欧菲智能车联科技有限公司 一种车位检测方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130021171A1 (en) * 2011-07-22 2013-01-24 Automotive Research & Test Center Parking space detection device and method thereof
EP2557020B1 (de) * 2011-08-12 2014-10-01 Robert Bosch Gmbh Verfahren zur Unterstützung eines Fahrers beim Einparken
CN108281041A (zh) * 2018-03-05 2018-07-13 东南大学 一种基于超声波和视觉传感器相融合的泊车车位检测方法
CN109191889A (zh) * 2018-08-07 2019-01-11 西安理工大学 基于led显像的地下车场停车引导系统及其引导方法
CN110497901A (zh) * 2019-08-30 2019-11-26 的卢技术有限公司 一种基于机器人vslam技术的泊车位自动搜索方法和系统
CN110775052A (zh) * 2019-08-29 2020-02-11 浙江零跑科技有限公司 一种基于视觉与超声波感知融合的自动泊车方法
CN110861639A (zh) * 2019-11-28 2020-03-06 安徽江淮汽车集团股份有限公司 泊车信息融合方法、装置、电子设备及存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130021171A1 (en) * 2011-07-22 2013-01-24 Automotive Research & Test Center Parking space detection device and method thereof
EP2557020B1 (de) * 2011-08-12 2014-10-01 Robert Bosch Gmbh Verfahren zur Unterstützung eines Fahrers beim Einparken
CN108281041A (zh) * 2018-03-05 2018-07-13 东南大学 一种基于超声波和视觉传感器相融合的泊车车位检测方法
CN109191889A (zh) * 2018-08-07 2019-01-11 西安理工大学 基于led显像的地下车场停车引导系统及其引导方法
CN110775052A (zh) * 2019-08-29 2020-02-11 浙江零跑科技有限公司 一种基于视觉与超声波感知融合的自动泊车方法
CN110497901A (zh) * 2019-08-30 2019-11-26 的卢技术有限公司 一种基于机器人vslam技术的泊车位自动搜索方法和系统
CN110861639A (zh) * 2019-11-28 2020-03-06 安徽江淮汽车集团股份有限公司 泊车信息融合方法、装置、电子设备及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114648882A (zh) * 2022-02-09 2022-06-21 上海欧菲智能车联科技有限公司 一种车位检测方法及装置
CN114648882B (zh) * 2022-02-09 2023-05-09 上海欧菲智能车联科技有限公司 一种车位检测方法及装置

Similar Documents

Publication Publication Date Title
WO2022083402A1 (zh) 障碍物检测方法、装置、计算机设备和存储介质
CN111753649B (zh) 车位检测方法、装置、计算机设备和存储介质
US11092444B2 (en) Method and system for recording landmarks in a traffic environment of a mobile unit
CN110458112B (zh) 车辆检测方法、装置、计算机设备和可读存储介质
WO2020206708A1 (zh) 障碍物的识别方法、装置、计算机设备和存储介质
US20100259609A1 (en) Pavement marker recognition device, pavement marker recognition method and pavement marker recognition program
CN112633152B (zh) 停车位检测方法、装置、计算机设备和存储介质
US11204610B2 (en) Information processing apparatus, vehicle, and information processing method using correlation between attributes
CN104167109A (zh) 车辆位置的检测方法和检测装置
WO2021223116A1 (zh) 感知地图生成方法、装置、计算机设备和存储介质
JP5225542B2 (ja) 地図情報生成方法、地図情報生成プログラム及び地図情報収集装置
CN112432650B (zh) 高精地图数据的获取方法、车辆控制方法以及设备
CN113030990B (zh) 车辆的融合测距方法、装置、测距设备及介质
CN113936198A (zh) 低线束激光雷达与相机融合方法、存储介质及装置
CN111967396A (zh) 障碍物检测的处理方法、装置、设备及存储介质
CN111753639A (zh) 感知地图生成方法、装置、计算机设备和存储介质
CN112753038A (zh) 识别车辆变道趋势的方法和装置
CN114359334A (zh) 目标跟踪方法、装置、计算机设备和存储介质
CN114494618A (zh) 地图的生成方法、装置、电子设备及存储介质
WO2021226852A1 (zh) 车位检测方法、装置、计算机设备和存储介质
JP2007011994A (ja) 道路認識装置
CN114663598A (zh) 三维建模方法、装置和存储介质
KR20160125803A (ko) 영역 추출 장치, 물체 탐지 장치 및 영역 추출 방법
CN115407355B (zh) 库位地图的验证方法、装置及终端设备
CN115830073A (zh) 地图要素重建方法、装置、计算机设备和存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20935216

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 28.04.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20935216

Country of ref document: EP

Kind code of ref document: A1