A parking space state recognition method, device and equipment

Abstract

This application provides a parking space status identification method, apparatus, and device. The method includes: acquiring ultrasonic radar signals corresponding to a parking space to be identified via radar; constructing features of the parking space to be identified based on the ultrasonic radar signals to obtain a feature vector of the ultrasonic radar signals corresponding to the parking space; and identifying the feature vector of the ultrasonic radar signals using a parking space status identification model to obtain the status of the parking space, wherein the status of the parking space includes an vacant status and an occupied status. This method can accurately identify the status of parking spaces even when traditional ultrasonic radar signals are weak in reflecting weak objects such as pedestrians and have poor response to low obstacles.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to a parking space status recognition method, device, and equipment. Background Technology

[0002] Intelligentization is a key trend in the automotive industry today, and intelligent driving technologies and systems are rapidly developing worldwide. Ultrasonic radar and vision systems are currently the most widely used parking scene perception sensors in intelligent driving applications. Based on their physical characteristics, vision systems are more suitable for scene semantic analysis, while ultrasonic radar is more advantageous for detecting general obstacles at close range.

[0003] Currently, mass-produced ultrasonic radar is mainly used to provide obstacle location information within the detection range of corresponding sensors, and is applied to parking assistance systems to generate corresponding obstacle avoidance and parking space information. However, traditional ultrasonic radar signals are weak for weakly reflective objects such as pedestrians, and have poor response to low obstacles, resulting in an inability to accurately identify the status of parking spaces. Summary of the Invention

[0004] This application provides a parking space status identification method, device, and equipment that can accurately identify the status of parking spaces when traditional ultrasonic radar signals are weak for weakly reflective objects such as pedestrians and have poor response to low obstacles.

[0005] In a first aspect, embodiments of this application provide a parking space status recognition method, the method comprising:

[0006] The ultrasonic radar signal corresponding to the parking space to be identified is collected by radar.

[0007] Based on the ultrasonic radar signal, the features of the parking space to be identified are constructed to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0008] The feature vector of the ultrasonic radar signal is identified using a parking space status recognition model to obtain the status of the parking space, which includes an vacant status and an occupied status.

[0009] The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the first sample parking space status corresponding to the feature vector. The first sample parking space status is determined based on the second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal.

[0010] In the above embodiments, the parking space status recognition model is used to identify the feature vector of the ultrasonic radar signal, which is more robust than the traditional threshold filtering ultrasonic parking space detection method. The second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal is used as the first sample parking space status corresponding to the feature vector of the ultrasonic radar signal to supervise the training process of the parking space status recognition model, making it easier to generate ultrasonic radar training supervision signals.

[0011] In one possible implementation, the status of the first sample parking space is determined as follows:

[0012] For a target parking space, sample visual signals collected by a camera and sample ultrasonic radar signals collected by a radar are acquired at the same time. The sample visual signals include multiple frames of images, and the images include at least one parking space.

[0013] Analyze the sample visual signal to determine the state of the second sample parking space corresponding to the sample visual signal of the target parking space;

[0014] The second sample parking space status is used as the first sample parking space status corresponding to the sample ultrasonic radar signal associated with the sample visual signal.

[0015] In the above embodiments, training labels corresponding to ultrasonic radar signals are generated based on a visual semi-automatic annotation method, making it easier to generate ultrasonic radar training supervision signals.

[0016] In one possible implementation, the step of parsing the sample visual signal of the target parking space and determining the state of the second sample parking space corresponding to the sample visual signal includes:

[0017] The sample visual signal is analyzed to determine whether each frame of the sample visual signal contains the location information of the target parking space;

[0018] If the images in the sample visual signal for a consecutive preset number of frames all contain the location information of the target parking space, then the sample visual signal is taken as the visual signal corresponding to the target parking space, and the second sample parking space state corresponding to the visual signal is taken as the sample parking space state corresponding to the target parking space.

[0019] In the above embodiments, a visual signal includes multiple frames of images, and each frame of images may include the location information of multiple parking spaces. For a target parking space, it is necessary to determine that the visual signal is the visual signal corresponding to the target parking space only when it is determined that the location information of the target parking space is included in multiple consecutive frames of images. This step can filter out falsely detected visual signals.

[0020] In one possible implementation, the sample ultrasonic radar signal associated with the sample visual signal is determined in the following manner:

[0021] Synchronize the camera clock with the radar clock, and register the camera coordinate system with the radar coordinate system.

[0022] If the sample visual signal acquired by the camera and the sample ultrasonic radar signal acquired by the radar are acquired at the same time, and the parking space location information contained in the sample visual signal matches the parking space location information contained in the sample ultrasonic radar signal, then it is determined that the sample visual signal and the sample ultrasonic radar signal are associated.

[0023] In the above embodiments, synchronizing the clocks of the camera and the radar, and registering the coordinate systems of the camera and the radar, is to facilitate subsequent determination that the sample visual signal and the sample ultrasonic radar signal were acquired at the same time, and that the parking space location information contained in the sample visual signal matches the parking space location information contained in the sample ultrasonic radar signal.

[0024] In one possible implementation, the step of constructing the feature vector of the parking space to be identified based on the ultrasonic radar signal to obtain the feature vector of the ultrasonic radar signal corresponding to the parking space to be identified includes:

[0025] Based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to each unit area.

[0026] The ultrasonic radar signal feature vectors corresponding to each unit area are combined to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0027] Using the method described in the above embodiments, feature construction is performed on the parking space to be identified. This can clearly identify the location information of the parking space contained in the ultrasonic radar signal corresponding to the parking space to be identified, as well as the time information corresponding to the ultrasonic radar signal. This facilitates subsequent determination that the sample visual signal and the sample ultrasonic radar signal were acquired at the same time, and that the location information of the parking space contained in the sample visual signal matches the location information of the parking space contained in the sample ultrasonic radar signal.

[0028] In one possible implementation, the step of constructing features based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, to obtain the ultrasonic radar signal feature vector corresponding to each unit area, includes:

[0029] Based on the location information of the vehicles waiting to park at the current acquisition time, the attitude information of the vehicles waiting to park, and the ultrasonic radar signal corresponding to any unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to any unit area of ​​the parking space.

[0030] In the above embodiment, a parking space is divided into multiple unit areas. The ultrasonic radar signal of any unit area is used to construct features based on the location information of the vehicle waiting to park at the current acquisition time, the attitude information of the vehicle waiting to park, and the ultrasonic radar signal corresponding to any unit area of ​​the parking space to be identified. This allows the feature vector to better describe the state of the parking space.

[0031] Secondly, embodiments of this application provide a parking space status recognition device, the device comprising:

[0032] The acquisition module is used to acquire ultrasonic radar signals corresponding to the parking spaces to be identified via radar.

[0033] The construction module is used to construct the features of the parking space to be identified based on the ultrasonic radar signal, and obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0034] The identification module is used to identify the feature vector of the ultrasonic radar signal using a parking space status identification model to obtain the status of the parking space, which includes an vacant status and an occupied status.

[0035] The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the first sample parking space status corresponding to the feature vector. The first sample parking space status is determined based on the second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal.

[0036] Thirdly, embodiments of this application provide a vehicle, the vehicle comprising:

[0037] Radar is used to collect ultrasonic radar signals corresponding to the parking space to be identified and send the ultrasonic radar signals to the electronic control unit.

[0038] Cameras are used to collect sample visual signals from parking spaces;

[0039] The electronic control unit is used to receive ultrasonic radar signals sent by the radar, construct features of the parking space to be identified based on the ultrasonic radar signals, and identify the status of the parking space to be identified.

[0040] Fourthly, embodiments of this application provide a parking space status recognition device, the device comprising:

[0041] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of the first aspect described above.

[0042] Fifthly, embodiments of this application provide a computer storage medium storing a computer program for causing a computer to perform the method described in the first aspect. Attached Figure Description

[0043] Figure 1 This is a schematic diagram illustrating an application scenario of a parking space status recognition method according to an exemplary embodiment of the present invention;

[0044] Figure 2 This is a schematic flowchart illustrating a parking space status recognition method according to an exemplary embodiment of the present invention.

[0045] Figure 3 This is a schematic diagram illustrating a process for determining the status of a parking space according to an exemplary embodiment of the present invention;

[0046] Figure 4 This is a schematic diagram illustrating a specific process of a parking space status recognition method according to an exemplary embodiment of the present invention;

[0047] Figure 5 A schematic diagram of an automobile system as an example of an exemplary embodiment of the present invention;

[0048] Figure 6 This is a schematic diagram of a parking space status recognition device according to an exemplary embodiment of the present invention;

[0049] Figure 7 This is a schematic diagram of a parking space status recognition device according to an exemplary embodiment of the present invention. Detailed Implementation

[0050] The technical solutions in the embodiments of this application will now be described clearly and in detail with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0051] like Figure 1 The diagram illustrates a parking space status recognition application scenario, which includes at least one parking space. Figure 1The parking spaces 1, 2, and 3 are shown in the diagram. The car can travel from parking space 3 to parking space 1, or from parking space 1 to parking space 3; the specific direction is not limited here. The car is equipped with at least one camera (four surround-view cameras are shown in the diagram, indicated by white circles) and ultrasonic radar (twelve ultrasonic radars are shown in the diagram, indicated by black circles). This embodiment does not limit the number of cameras and radars. The surround-view cameras are used to collect visual signals corresponding to the parking spaces, and the ultrasonic radars are used to collect ultrasonic radar signals corresponding to the parking spaces.

[0052] To address the problem that traditional ultrasonic radar signals are weak in detecting weakly reflective objects such as pedestrians and have poor response to low obstacles, thus failing to accurately identify the status of parking spaces, this application provides a parking space status identification method, such as... Figure 2 As shown, the method includes:

[0053] S201: Collect ultrasonic radar signals corresponding to the parking space to be identified via radar.

[0054] The vehicle includes multiple radars. In this embodiment of the application, the number of radars used to collect ultrasonic radar signals of parking spaces is not specifically limited. The sampling frequency of the radar can be 20 Hz or other frequencies.

[0055] Currently, vehicle-mounted radar is mainly divided into three types: lidar, millimeter-wave radar, and ultrasonic radar. In this embodiment, ultrasonic radar is taken as an example.

[0056] Radar eliminates the confusion caused by drivers needing to look around when parking, reversing, and starting the vehicle, helping them overcome blind spots and blurred vision, thus improving driving safety. The function of radar varies depending on its installation location. For example, front and rear parking sensors greatly assist drivers when parking or following other vehicles in congested traffic, preventing unnecessary accidents. Therefore, front and rear parking sensors are important safety aids. When the gear lever is engaged in reverse, the reversing radar automatically activates. When the sensor detects an object behind, a buzzer sounds a warning. As the vehicle continues to reverse, the frequency of the warning sound gradually increases, eventually becoming a long beep. Another example is front parking assist, a safety aid for parking or reversing, composed of ultrasonic sensors, a controller, and a display. It informs the driver of surrounding obstacles through sound, eliminating the confusion caused by drivers needing to look around when parking, reversing, and starting the vehicle, and helping them overcome blind spots and blurred vision, thus improving driving safety.

[0057] S202: Based on the ultrasonic radar signal, construct the features of the parking space to be identified to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0058] First, feature construction is performed based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, resulting in the ultrasonic radar signal feature vector corresponding to each unit area.

[0059] The feature construction is based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, including:

[0060] Based on the location information of the vehicles waiting to park at the current acquisition time, the attitude information of the vehicles waiting to park, and the ultrasonic radar signal corresponding to any unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to any unit area of ​​the parking space.

[0061] The location information of vehicles waiting to be parked can be determined through the following methods:

[0062] (1) The attitude information of a vehicle waiting to park can be obtained through the formula: ψ t =ψ t-1 +ω t Δt is determined, where ψ t-1 For the attitude information of the vehicles waiting to park at the previous time, ω t Let Δt be the yaw rate of the vehicle waiting to park, and let Δt be the time difference between the current data acquisition time and the previous data acquisition time.

[0063] (2) The location information of the waiting vehicle includes the horizontal coordinate information and the vertical coordinate information of the waiting vehicle. The horizontal coordinate information can be obtained by formula: X t =X t-1 +v×Δt×cosψ t -ΔL is determined, X t Let X be the x-coordinate of the vehicle waiting to park at the current time of data collection. t-1 Let ψ be the x-coordinate of the vehicle waiting to park at the previous data collection time. t Here is the attitude information of the vehicle waiting to park (as in the formula in (1) above), v is the vehicle speed, which can be obtained by the vehicle speed sensor, Δt is the time difference between the current acquisition time and the previous acquisition time, and ΔL is the longitudinal offset of the ultrasonic radar installation in the vehicle coordinate system. The vehicle coordinate system can be the radar coordinate system.

[0064] The ordinate information can be obtained through the formula: Y t Y represents the ordinate information of the vehicles waiting to park at the current time of data collection. t-1 Here, v represents the x-coordinate of the vehicle waiting to park at the previous data collection time, v is the vehicle speed (obtained from a vehicle speed sensor), Δt is the time difference between the current data collection time and the previous data collection time, and ψ is the horizontal coordinate of the vehicle waiting to park at the previous data collection time.t The attitude information of the vehicle waiting to be parked is (as in the formula in (1) above), where W is the vehicle width and a is a preset value (e.g., 2).

[0065] (3) Taking a radar acquisition frequency of 20Hz as an example, the radar outputs an ultrasonic radar signal every 50ms. When a car travels through different unit areas of the parking space to be identified, its speed can vary. When the car travels through any two unit areas at different speeds, the number of ultrasonic radar signals output will differ. For example, if the car takes 100ms to travel through the first unit area, two ultrasonic radar signals will be output; if the car takes 50ms, one ultrasonic radar signal will be output. In this case, either one ultrasonic radar signal corresponding to the first unit area can be selected as the ultrasonic radar signal for the first unit area, or the two ultrasonic radar signals can be integrated to ensure that the length of the ultrasonic radar signals corresponding to each unit area is the same. Using formula D... t =[D 1,t D 2,t ,…,D n,t [ ] describes the ultrasonic radar signal corresponding to any unit area.

[0066] Using the formula: U t =[X t ,Y t ,ψ t D t ] T The attitude information of the vehicle waiting to park in (1), the horizontal and vertical coordinate information of the vehicle waiting to park in (2), and the ultrasonic radar signal of any unit area corresponding to the parking space to be identified in (3) are combined to obtain the ultrasonic radar signal feature vector (descriptor) corresponding to any unit area. t Let n be the feature vector of the ultrasonic radar signal corresponding to any unit area, and n be the signal output dimension of the ultrasonic radar signal (e.g., 4).

[0067] The ultrasonic radar signal feature vectors corresponding to each unit area are combined to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0068] Using the formula: P=[U1,U2,…,U c ] (n+b)*cThe ultrasonic radar signal feature vectors corresponding to each unit area are combined. Here, c represents the number of unit areas of the parking space to be identified, n represents the signal output dimension of the ultrasonic radar signal (e.g., 4), P represents the ultrasonic radar signal feature vector (descriptor) corresponding to the parking space to be identified, U represents the ultrasonic radar signal feature vector corresponding to any unit area, and b is a preset value (e.g., 3). If the length of the parking space to be identified is 3 meters, and U is sampled at fixed spatial intervals of 0.25 meters within 1.5 meters to the left and right of the center point of the parking space, then c is 12. This embodiment does not specifically limit the size of the parking space, the sampling interval, or the sampling start position; these can be set according to specific needs.

[0069] S203: The feature vector of the ultrasonic radar signal is identified using a parking space status recognition model to obtain the status of the parking space, which includes an vacant status and an occupied status.

[0070] The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the status of the first sample parking space corresponding to the feature vector. Due to the low gain of ultrasonic radar, the status of the parking space cannot be obtained directly by analyzing the ultrasonic radar signal. Therefore, a visual semi-automatic annotation method is used to generate ultrasonic radar signal training labels to solve the problem of difficulty in generating ultrasonic radar training supervision signals.

[0071] The status of the first sample parking space is determined using the following implementation method, such as... Figure 3 As shown:

[0072] S301: For a target parking space, acquire sample visual signals collected by a camera and sample ultrasonic radar signals collected by a radar at the same time. The sample visual signals include multiple frames of images, and the images include at least one parking space.

[0073] The installation location of vehicle cameras affects their function. Surround view cameras: Primarily installed around the vehicle's perimeter, typically using 4-8 cameras. These can be categorized as forward-facing fisheye cameras, left-side fisheye cameras, right-side fisheye cameras, and rear-facing fisheye cameras. They are used for panoramic surround view display, visual perception for parking functions, and target detection. RGB color matrix is ​​commonly used due to color accuracy requirements. Rear view cameras: Generally installed on the trunk, primarily for parking assistance. The field of view is between 120° and 140°, with a detection distance of approximately 50 meters. Side-front view cameras: Installed on the B-pillar or rearview mirror, these cameras typically have a field of view of 90°-100° and a detection distance of approximately 80 meters. Their main function is to detect vehicles and bicycles approaching from the side. Side-rear view cameras: Generally installed on the front fender, these cameras typically have a field of view of around 90° and a detection distance of approximately 80 meters. They are mainly used for lane changes and merging into other roads. Built-in camera: Primarily used to monitor the vehicle interior, enabling functions such as driver fatigue alerts and anti-theft measures. Since this embodiment focuses on detecting parking space status, the camera is either a surround-view camera or a rear-view camera.

[0074] S302: Analyze the sample visual signal to determine the second sample parking space state corresponding to the sample visual signal of the target parking space;

[0075] By analyzing the sample visual signal, it is determined whether each frame of the sample visual signal contains the location information of the target parking space.

[0076] The steps for analyzing the sample visual signal are as follows: First, the sample visual signal needs to be preprocessed. The preprocessing steps include, but are not limited to, distortion correction, region of interest cropping, scaling, and brightness normalization. Then, the preprocessed sample visual signal is input into a neural network model for feature extraction to obtain the parking space location information (u1, v1, u2, v2), where (u1, v1, u2, v2) are the coordinates of two points corresponding to the parking space entrance. The neural network model can be a target detection network model or other neural network models, without specific limitations here.

[0077] If the images in the sample visual signal for a consecutive preset number of frames all contain the location information of the target parking space, then the sample visual signal is taken as the visual signal corresponding to the target parking space, and the second sample parking space state corresponding to the visual signal is taken as the sample parking space state corresponding to the target parking space.

[0078] The sample visual signal comprises multiple frames of images, each frame potentially containing multiple parking spaces. In other words, one sample visual signal can serve as the corresponding sample visual signal for multiple parking spaces. If a sample visual signal contains a predetermined number of consecutive frames all containing the location information of the first target parking space and the second target parking space, and the first target parking space is in an vacant state while the second target parking space is in an occupied state, then the sample visual signal and the vacant state correspond to the first target parking space, and the sample visual signal and the occupied state correspond to the second target parking space.

[0079] S303: The second sample parking space state is taken as the first sample parking space state corresponding to the sample ultrasonic radar signal associated with the sample visual signal.

[0080] The sample ultrasonic radar signal associated with the sample visual signal is determined in the following manner:

[0081] The camera and radar clocks are synchronized, and their coordinate systems are registered. Clock synchronization is software-based; only the camera and radar clocks need to be synchronized, without needing to synchronize with network standard time. For example, if the network standard time is 8:00, the camera clock is 7:45, and the radar clock is 8:10, then the radar clock time is adjusted to 7:45, or the camera clock time is adjusted to 8:10. Coordinate system registration is performed as follows: first, the camera installation parameters are calibrated using a checkerboard pattern, and the ultrasonic radar installation parameters are calibrated using a tooling bracket. Then, the coordinate systems are registered based on the ultrasonic radar and camera installation parameters.

[0082] If the sample visual signal acquired by the camera and the sample ultrasonic radar signal acquired by the radar are acquired at the same time, and the parking space location information contained in the sample visual signal matches the parking space location information contained in the sample ultrasonic radar signal, then it is determined that the sample visual signal and the sample ultrasonic radar signal are associated.

[0083] After clock synchronization and coordinate system registration, it can be determined whether the parking space information in the ultrasonic radar signal is the same as the parking space information in the visual signal. If they are the same, the parking space state corresponding to the visual signal is the parking space state corresponding to the ultrasonic radar signal. Then, a parking space state recognition model can be trained based on the feature vector of the sample ultrasonic radar signal and the first sample parking space state corresponding to the feature vector. During the training process, the feature vector of the sample ultrasonic radar signal can be input into the parking space state recognition model first, and then the sample visual signal associated with the sample ultrasonic radar signal can be determined to obtain the state of the parking space. Alternatively, the sample visual signal associated with the sample ultrasonic radar signal can be determined first to obtain the state of the parking space. That is, the sample ultrasonic radar signal is labeled first, and then the feature vector of the sample ultrasonic radar signal and the corresponding label are input into the parking space state recognition model for training. The specific order of the steps in the training process is not limited here. The parking space status recognition model can be a convolutional neural network. For example, during training, the feature vector of the sample ultrasonic radar signal is input into the network, and operations such as convolution, pooling and activation are performed. The classification loss function based on cross-entropy is solved using the ADAM solver through batch gradient descent (default batch size is 16), and the parking space status is output. This application does not specifically limit the method of solving the loss function or the type of loss function. Finally, the trained parking space status recognition model is compressed through pruning, quantization, and other operations before being deployed in the vehicle's electronic control unit.

[0084] This application provides a parking space identification method comprising two parts: one part is the process of training a parking space status identification model (offline), and the other part is the process of identifying the parking space status using the parking space status identification model (online). Specific implementation methods are as follows: Figure 4 As shown:

[0085] For the offline process: First, the clocks of the camera and the radar are synchronized, and the coordinate systems of the camera and the radar are registered. Simultaneously, sample visual signals are collected by the camera, and sample ultrasonic radar signals are collected by the radar. Then, the sample visual signals undergo preprocessing—determining the location information and status of the parking space corresponding to the sample visual signal—and tag generation, etc. The sample ultrasonic radar signals are preprocessed to obtain the location information of the parking space corresponding to the sample ultrasonic radar signal, and associated with sample visual signals that are related to the sample ultrasonic signals to obtain the tags corresponding to the sample ultrasonic radar signals. Finally, a parking space status recognition model is trained based on the sample ultrasonic radar signals and their corresponding tags. After pruning, quantization, and other compression operations, the parking space status recognition model is deployed in the electronic control unit.

[0086] For the online process: The ultrasonic radar signal corresponding to the parking space to be identified is collected by radar. Based on the ultrasonic radar signal, the feature of the parking space to be identified is constructed to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified. The parking space status recognition model is used to identify the ultrasonic radar signal feature vector to obtain the status of the parking space.

[0087] The parking space status recognition method provided in this application generates ultrasonic radar signal training labels based on a semi-automatic visual signal annotation method, solving the problem of difficulty in generating ultrasonic radar training and supervision signals. The neural network model trained in a good working environment can be generalized to harsh working environments such as dark scenes. Moreover, the neural network deep learning method is more robust than the traditional threshold filtering ultrasonic parking space detection method for missed detection of small obstacles such as cones and false detection of obstacles such as potholes, ensuring that false detections are suppressed while detecting obstacles.

[0088] Based on the same inventive concept, this application also provides a parking space status recognition device 500, such as... Figure 5 As shown, the device includes:

[0089] The acquisition module 501 is used to acquire ultrasonic radar signals corresponding to the parking space to be identified via radar.

[0090] The construction module 502 is used to construct the features of the parking space to be identified based on the ultrasonic radar signal, and obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0091] The identification module 503 is used to identify the feature vector of the ultrasonic radar signal using the parking space status identification model to obtain the status of the parking space, which includes an vacant status and an occupied status.

[0092] The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the first sample parking space status corresponding to the feature vector. The first sample parking space status is determined based on the second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal.

[0093] In one possible implementation, the device further includes a training module for determining the state of the first sample parking space in the following manner:

[0094] For a target parking space, sample visual signals collected by a camera and sample ultrasonic radar signals collected by a radar are acquired at the same time. The sample visual signals include multiple frames of images, and the images include at least one parking space.

[0095] Analyze the sample visual signal to determine the state of the second sample parking space corresponding to the sample visual signal of the target parking space;

[0096] The second sample parking space status is used as the first sample parking space status corresponding to the sample ultrasonic radar signal associated with the sample visual signal.

[0097] In one possible implementation, the training module is further configured to parse the sample visual signal of the target parking space and determine the state of the second sample parking space corresponding to the sample visual signal, including:

[0098] The sample visual signal is analyzed to determine whether each frame of the sample visual signal contains the location information of the target parking space;

[0099] If the images in the sample visual signal for a consecutive preset number of frames all contain the location information of the target parking space, then the sample visual signal is taken as the visual signal corresponding to the target parking space, and the second sample parking space state corresponding to the visual signal is taken as the sample parking space state corresponding to the target parking space.

[0100] In one possible implementation, the training module is further configured to determine the sample ultrasonic radar signal associated with the sample visual signal in the following manner:

[0101] Synchronize the camera clock with the radar clock, and register the camera coordinate system with the radar coordinate system.

[0102] If the sample visual signal acquired by the camera and the sample ultrasonic radar signal acquired by the radar are acquired at the same time, and the parking space location information contained in the sample visual signal matches the parking space location information contained in the sample ultrasonic radar signal, then it is determined that the sample visual signal and the sample ultrasonic radar signal are associated.

[0103] In one possible implementation, the construction module 502 is used to construct features of the parking space to be identified based on the ultrasonic radar signal, obtaining an ultrasonic radar signal feature vector corresponding to the parking space to be identified, including:

[0104] Based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to each unit area.

[0105] The ultrasonic radar signal feature vectors corresponding to each unit area are combined to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

[0106] In one possible implementation, the construction module 502 is further configured to construct features based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, to obtain the ultrasonic radar signal feature vector corresponding to each unit area, including:

[0107] Based on the location information of the vehicles waiting to park at the current acquisition time, the attitude information of the vehicles waiting to park, and the ultrasonic radar signal corresponding to any unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to any unit area of ​​the parking space.

[0108] Based on the same inventive concept, this application also provides a vehicle, such as... Figure 6 As shown, the vehicle includes:

[0109] Radar 601 is used to collect ultrasonic radar signals corresponding to the parking space to be identified and send the ultrasonic radar signals to the electronic control unit.

[0110] Camera 603 is used to collect sample visual signals from parking spaces;

[0111] The electronic control unit 602 is used to receive ultrasonic radar signals sent by the radar, construct the features of the parking space to be identified based on the ultrasonic radar signals, and identify the status of the parking space to be identified.

[0112] Based on the same inventive concept, this application also provides a parking space status recognition device, the device comprising:

[0113] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a parking space status recognition method.

[0114] like Figure 7 As shown, the device includes a processor 701, a memory 702, and a communication interface 703; a bus 704. The processor 701, memory 702, and communication interface 703 are interconnected via the bus 704.

[0115] The processor 701 is used to read and execute instructions from the memory 702, so that the at least one processor can execute the parking space status recognition method provided in the above embodiments.

[0116] The memory 702 is used to store various instructions and programs for the parking space status recognition method provided in the above embodiments.

[0117] The communication interface 703 is used for data interaction between the transient smoke sensor and the electronic control unit.

[0118] The 704 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0119] The processor 701 can be a central processing unit (CPU), a network processor (NP), a graphics processing unit (GPU), or any combination of CPU, NP, and GPU. It can also be a hardware chip. The aforementioned hardware chip can be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0120] In addition, this application also provides a computer-readable storage medium storing a computer program for causing a computer to perform the method described in any of the above embodiments.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0123] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0124] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for recognizing the status of a parking space, characterized in that, The method includes: The ultrasonic radar signal corresponding to the parking space to be identified is collected by radar. Based on the ultrasonic radar signal, the features of the parking space to be identified are constructed to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified. The feature vector of the ultrasonic radar signal is identified using a parking space status recognition model to obtain the status of the parking space, which includes an vacant status and an occupied status. The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the first sample parking space status corresponding to the feature vector. The first sample parking space status is determined based on the second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal. The status of the first sample parking space is determined as follows: For a target parking space, sample visual signals collected by a camera and sample ultrasonic radar signals collected by a radar are acquired simultaneously. The sample visual signals include multiple frames of images, each containing at least one parking space. The sample visual signals are analyzed to determine a second sample parking space state corresponding to the sample visual signal of the target parking space. The second sample parking space state is then used as a first sample parking space state corresponding to the sample ultrasonic radar signal associated with the sample visual signal.

2. The method according to claim 1, characterized in that, The step of parsing the sample visual signal to determine the second sample parking space state corresponding to the sample visual signal of the target parking space includes: The sample visual signal is analyzed to determine whether each frame of the sample visual signal contains the location information of the target parking space; If the images in the sample visual signal for a consecutive preset number of frames all contain the location information of the target parking space, then the sample visual signal is taken as the visual signal corresponding to the target parking space, and the second sample parking space state corresponding to the visual signal is taken as the sample parking space state corresponding to the target parking space.

3. The method according to claim 2, characterized in that, The sample ultrasonic radar signal associated with the sample visual signal is determined in the following manner: Synchronize the camera clock with the radar clock, and register the camera coordinate system with the radar coordinate system. If the sample visual signal acquired by the camera and the sample ultrasonic radar signal acquired by the radar are acquired at the same time, and the parking space location information contained in the sample visual signal matches the parking space location information contained in the sample ultrasonic radar signal, then it is determined that the sample visual signal and the sample ultrasonic radar signal are associated.

4. The method according to claim 1, characterized in that, The feature construction of the parking space to be identified based on the ultrasonic radar signal, resulting in the ultrasonic radar signal feature vector corresponding to the parking space to be identified, includes: Based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to each unit area. The ultrasonic radar signal feature vectors corresponding to each unit area are combined to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified.

5. The method according to claim 4, characterized in that, The feature construction based on the ultrasonic radar signals of each unit area of ​​the parking space to be identified yields the ultrasonic radar signal feature vector corresponding to each unit area, including: Based on the location information of the vehicles waiting to park at the current acquisition time, the attitude information of the vehicles waiting to park, and the ultrasonic radar signal corresponding to any unit area of ​​the parking space to be identified, feature construction is performed to obtain the ultrasonic radar signal feature vector corresponding to any unit area of ​​the parking space to be identified.

6. A parking space status recognition device, characterized in that, The device includes: The acquisition module is used to acquire ultrasonic radar signals corresponding to the parking spaces to be identified via radar. The construction module is used to construct the features of the parking space to be identified based on the ultrasonic radar signal, and obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified. The identification module is used to identify the feature vector of the ultrasonic radar signal using a parking space status identification model to obtain the status of the parking space, which includes an vacant status and an occupied status. The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the first sample parking space status corresponding to the feature vector. The first sample parking space status is determined based on the second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal. The status of the first sample parking space is determined as follows: For a target parking space, sample visual signals collected by a camera and sample ultrasonic radar signals collected by a radar are acquired simultaneously. The sample visual signals include multiple frames of images, each containing at least one parking space. The sample visual signals are analyzed to determine a second sample parking space state corresponding to the sample visual signal of the target parking space. The second sample parking space state is then used as a first sample parking space state corresponding to the sample ultrasonic radar signal associated with the sample visual signal.

7. A car, characterized in that, The vehicle includes: Radar is used to collect ultrasonic radar signals corresponding to the parking space to be identified and send the ultrasonic radar signals to the electronic control unit. Cameras are used to collect sample visual signals from parking spaces; The electronic control unit is used to receive ultrasonic radar signals sent by the radar and construct the features of the parking space to be identified based on the ultrasonic radar signals to obtain the ultrasonic radar signal feature vector corresponding to the parking space to be identified. The feature vector of the ultrasonic radar signal is identified using a parking space status recognition model to obtain the status of the parking space, which includes an vacant status and an occupied status. The parking space status recognition model is trained based on the feature vector of the sample ultrasonic radar signal of the parking space and the first sample parking space status corresponding to the feature vector. The first sample parking space status is determined based on the second sample parking space status corresponding to the sample visual signal associated with the sample ultrasonic radar signal. The status of the first sample parking space is determined as follows: For a target parking space, sample visual signals collected by a camera and sample ultrasonic radar signals collected by a radar are acquired simultaneously. The sample visual signals include multiple frames of images, each containing at least one parking space. The sample visual signals are analyzed to determine a second sample parking space state corresponding to the sample visual signal of the target parking space. The second sample parking space state is then used as a first sample parking space state corresponding to the sample ultrasonic radar signal associated with the sample visual signal.

8. A parking space status recognition device, characterized in that, The device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-5.

9. A computer storage medium, characterized in that, The computer storage medium stores a computer program that enables the computer to perform the method as described in any one of claims 1-5.

Images