WO2024259938A1 - 一种可行驶区域分割方法、装置、可读存储介质及机器人 - Google Patents
一种可行驶区域分割方法、装置、可读存储介质及机器人 Download PDFInfo
- Publication number
- WO2024259938A1 WO2024259938A1 PCT/CN2023/141525 CN2023141525W WO2024259938A1 WO 2024259938 A1 WO2024259938 A1 WO 2024259938A1 CN 2023141525 W CN2023141525 W CN 2023141525W WO 2024259938 A1 WO2024259938 A1 WO 2024259938A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- prior
- drivable area
- point cloud
- cloud data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Definitions
- the present application belongs to the field of robotics technology, and in particular, relates to a method and device for segmenting a drivable area, a computer-readable storage medium, and a robot.
- embodiments of the present application provide a drivable area segmentation method, device, computer-readable storage medium and robot to solve the problem of high cost and low efficiency of existing drivable area segmentation methods.
- a first aspect of an embodiment of the present application provides a drivable area segmentation method, which may include:
- the drivable area segmentation model is a deep learning model trained based on prior images and prior point cloud data;
- the prior image is an image collected before the target image;
- the prior point cloud data is point cloud data collected synchronously with the prior image;
- the prior image is used as the training input of the deep learning model, and the prior segmented image is used as the expected training output of the deep learning model;
- the prior segmented image is a drivable area segmentation image determined based on the prior point cloud data.
- the method before performing region segmentation on the target image using a preset drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image, the method may further include:
- the deep learning model is trained according to the prior image and the prior segmented image to obtain the drivable area segmentation model.
- acquiring the prior image and the prior point cloud data may include:
- the previous point cloud data collected by the radar in response to the collection signal is obtained.
- converting the prior point cloud data from a preset radar coordinate system to a preset camera coordinate system to obtain converted point cloud data may include:
- the prior point cloud data is transformed from the radar coordinate system to the camera coordinate system according to the translation matrix and the rotation matrix to obtain the transformed point cloud data.
- the training of the deep learning model according to the prior image and the prior segmented image to obtain the drivable area segmentation model may include:
- the model parameters of the deep learning model are adjusted according to the training loss value to obtain the drivable area segmentation model.
- calculating the training loss value of the deep learning model according to the output image and the prior segmented image may include:
- the training loss value is calculated according to the intersection-and-union ratio of the drivable areas.
- the method before performing region segmentation on the target image using a preset drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image, the method may further include:
- the step of performing region segmentation on the target image using a preset drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image may include:
- the drivable area segmentation model is used to perform area segmentation on the normalized target image to obtain a region corresponding to the target image.
- the drivable area segmentation image corresponding to the landmark image.
- a second aspect of an embodiment of the present application provides a drivable area segmentation device, which may include:
- An image acquisition module used for acquiring a target image to be segmented
- a region segmentation module used to perform region segmentation on the target image using a preset drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image;
- the drivable area segmentation model is a deep learning model trained based on prior images and prior point cloud data;
- the prior image is an image collected before the target image;
- the prior point cloud data is point cloud data collected synchronously with the prior image;
- the prior image is used as the training input of the deep learning model, and the prior segmented image is used as the expected training output of the deep learning model;
- the prior segmented image is a drivable area segmentation image determined based on the prior point cloud data.
- the drivable area segmentation device may further include:
- a data acquisition module used to acquire the previous image and the previous point cloud data
- a data conversion module used to convert the previous point cloud data from a preset radar coordinate system to a preset camera coordinate system to obtain converted point cloud data
- An image determination module used to determine the previously segmented image according to the converted point cloud data
- a model training module is used to train the deep learning model according to the previous image and the previous segmented image to obtain the drivable area segmentation model.
- the data acquisition module may include:
- a signal sending unit used to send a synchronous acquisition signal to the camera and the radar through a preset synchronous trigger, so that the camera and the radar perform synchronous data acquisition;
- an image acquisition unit configured to acquire the previous image acquired by the camera in response to the acquisition signal
- a data acquisition unit is used to acquire the previous point cloud data acquired by the radar in response to the acquisition signal.
- the data conversion module may include:
- a matrix acquisition unit used to acquire a translation matrix and a rotation matrix from the radar coordinate system to the camera coordinate system;
- a data conversion unit is used to convert the previous point cloud data from the radar coordinate system to the camera coordinate system according to the translation matrix and the rotation matrix to obtain the converted point cloud data.
- model training module may include:
- An image acquisition unit configured to input the prior image into the deep learning model and acquire an output image of the deep learning model
- a loss calculation unit configured to calculate a training loss value of the deep learning model according to the output image and the previously segmented image
- a parameter adjustment unit is used to adjust the model parameters of the deep learning model according to the training loss value to obtain the drivable area segmentation model.
- the loss calculation unit may include:
- intersection-and-union ratio calculation subunit used to calculate the intersection-and-union ratio of the drivable area between the output image and the previous segmented image
- the loss value calculation subunit is used to calculate the training loss value according to the intersection-and-union ratio of the drivable area.
- the drivable area segmentation device may further include:
- a standardization processing module used for performing standardization processing on the target image to obtain a standardized target image
- a normalization processing module used for performing normalization processing on the standardized target image to obtain the normalized target image
- the region segmentation module may further include:
- a region segmentation unit is used to perform region segmentation on the normalized target image using the drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image.
- a third aspect of an embodiment of the present application provides a computer-readable storage medium, which stores a computer program.
- the computer program is executed by a processor, the steps of any of the above-mentioned drivable area segmentation methods are implemented.
- a fourth aspect of an embodiment of the present application provides a robot, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the above-mentioned drivable area segmentation methods when executing the computer program.
- a fifth aspect of an embodiment of the present application provides a computer program product.
- the computer program product runs on a robot, the robot executes the steps of any one of the above-mentioned drivable area segmentation methods.
- the embodiments of the present application have the following beneficial effects: the embodiments of the present application can obtain a drivable area segmentation model based on the synchronously collected previous images and the previous point cloud data, and use the drivable area segmentation model to perform regional segmentation on the target image to obtain a drivable area segmentation image corresponding to the target image.
- the previous image when the robot is in a new road environment, the previous image can be used as input and the synchronously collected previous point cloud data can be used as the expected output for model training without the need for a large amount of sample collection and manual labeling, thereby effectively reducing the training cost of the drivable area segmentation model and improving the training efficiency of the drivable area segmentation model.
- FIG1 is a schematic diagram of the use process of the drivable area segmentation method
- FIG2 is a schematic flow chart of a training process of a drivable area segmentation model in an embodiment of the present application
- FIG3 is a schematic diagram of the external parameter calibration process of the radar and camera
- FIG4 is a schematic diagram of a drivable area segmentation model
- FIG5 is a flow chart of an embodiment of a method for segmenting a drivable area in an embodiment of the present application
- FIG6 is a structural diagram of an embodiment of a drivable area segmentation device in an embodiment of the present application.
- FIG. 7 is a schematic block diagram of a robot in an embodiment of the present application.
- the term “if” may be interpreted as “when” or “upon” or “in response to determining” or “in response to detecting,” depending on the context.
- the phrases “if it is determined” or “if [described condition or event] is detected” may be interpreted as meaning “upon determination” or “in response to determining” or “upon detection of [described condition or event]” or “in response to detecting [described condition or event],” depending on the context.
- embodiments of the present application provide a drivable area segmentation method, device, computer-readable storage medium and robot to solve the problem of high cost and low efficiency of existing drivable area segmentation methods.
- the executor of the method of the present application is a robot, which may include but is not limited to a sweeping robot, an inspection robot, a food delivery robot, a guiding robot, and any other common robots in the prior art.
- a preset drivable area segmentation model may be used to segment an image into regions, thereby obtaining a drivable area image corresponding to the image, and the drivable area image may be used for auxiliary navigation, wherein the drivable area segmentation model is a deep learning model trained based on prior images and prior point cloud data.
- FIG. 2 shows a training process of the drivable area segmentation model, which may include the following steps:
- Step S201 Acquire a previous image and a previous point cloud data.
- a previous image containing the driving environment can be collected by a preset camera, and at the same moment, previous point cloud data of the same driving environment can also be collected by a preset radar.
- radars and cameras may be pre-installed on the robot for navigation, exploration, motion control and other related applications.
- a preset synchronization trigger can be used here to send a synchronized acquisition signal to the camera and the radar; thereafter, the camera collects the prior image in response to the acquisition signal, and the radar collects the prior point cloud data in response to the acquisition signal; the robot can obtain the prior image and the prior point cloud data.
- Step S202 converting the previous point cloud data from the preset radar coordinate system to the preset camera coordinate system to obtain converted point cloud data.
- the prior point cloud data can be mapped to the prior image according to a preset data mapping relationship, wherein the data mapping relationship can be a translation matrix and a rotation matrix from the radar coordinate system to the camera coordinate system. According to the translation matrix and the rotation matrix, the prior point cloud data can be transformed from the radar coordinate system to the camera coordinate system to obtain the transformed point cloud data.
- the data mapping relationship can be a translation matrix and a rotation matrix from the radar coordinate system to the camera coordinate system.
- the prior point cloud data can be transformed from the radar coordinate system to the camera coordinate system to obtain the transformed point cloud data.
- FIG. 3 shows a schematic diagram of external calibration.
- the camera can calculate the representation of the calibration plane in the camera coordinate system through the QR code or checkerboard of the calibration plane.
- the beam emitted by the radar falls on the calibration plane (black dot in the figure).
- the coordinates of the radar point in the radar coordinate system and the coordinates of the plane equation in the camera coordinate system are used to construct the constraints of the point on the plane to solve the translation matrix and rotation matrix.
- the point cloud data can be converted to the camera coordinate system according to the following formula:
- (x 1 ,y 1 ) is the converted point cloud data after conversion
- (x 2 ,y 2 ,z 2 ) is the previous point cloud data
- I is the unit matrix
- t is the translation matrix
- R is the rotation matrix
- Rx is the rotation matrix around the x-axis
- Ry is the rotation matrix around the y-axis
- Rz is the rotation matrix around the z-axis
- ⁇ x is the Euler angle of rotation around the x-axis
- ⁇ y is the Euler angle of rotation around the y-axis
- ⁇ z is the Euler angle of rotation around the z-axis.
- Step S203 determining a previously segmented image according to the converted point cloud data.
- the previously segmented image may be determined specifically according to the height information of the converted point cloud data, wherein the previously segmented image may be an image of the drivable area segmented from the converted point cloud data.
- the prior point cloud data may be three-dimensional data (x 2 , y 2 , z 2 ), and the corresponding converted point cloud data may be two-dimensional data (x 1 , y 1 ), and the z 2 value is the height information of the converted point cloud data.
- the drivable area image may be determined.
- the point cloud data within a preset height range may be determined as obstacle data, and the area included in the height range may be excluded to obtain the drivable area, thereby obtaining the prior segmented image.
- Step S204 training the deep learning model according to the previous image and the previous segmented image to obtain a drivable area segmentation model.
- the prior image can be used as the input of the deep learning model, and the prior segmented image can be used as the expected output of the deep learning model to train the deep learning model to obtain a drivable area segmentation model.
- the prior image can be input into the deep learning model, and the output image of the deep learning model can be obtained; the training loss value of the deep learning model can be calculated based on the output image and the prior segmented image; the model parameters of the deep learning model can be adjusted based on the training loss value, and the drivable area segmentation model shown in Figure 4 can be obtained, wherein the drivable area segmentation model can include a backbone network, a neck network, and a drivable area segmentation head.
- the intersection-and-union ratio of the drivable area between the output image and the previous segmented image can be calculated, that is, the ratio of the intersection and the union of the drivable area between the output image and the previous segmented image can be calculated, and then the training loss value can be calculated according to the ratio.
- the ratio can be used as the training loss value.
- the parameters of the deep learning model can be adjusted according to the training loss value.
- the parameters of the deep learning model can be adjusted by reverse updating.
- An embodiment of a drivable area segmentation method in an embodiment of the present application may include:
- Step S501 Acquire a target image to be segmented.
- a target image to be segmented may be captured by a camera, wherein the target image to be segmented may be a target image for which a drivable area segmentation is to be performed.
- the camera can capture one or more images to be segmented in real time.
- the following will take any one of the captured images as an example of the target image to be segmented to illustrate the drivable area segmentation method in the embodiment of the present application.
- the target image in order to improve the segmentation efficiency of the drivable area segmentation model, the target image may be preprocessed before the drivable area segmentation model is used to segment the target image.
- the target image may be subjected to common image preprocessing operations in the prior art according to actual needs.
- the target image may be standardized using the following formula:
- xi is the image pixel value of the target image
- ⁇ is the mean of the image pixel values of the target image
- ⁇ is the standard deviation of the image pixel values of the target image
- N is the number of image pixels of the target image
- X' is the standardized target image.
- the standardized target image may be normalized to obtain a normalized target image.
- the standardized target image may be normalized using the following formula:
- x'i is the image pixel value of the standardized target image
- max(x') and min(x') are the maximum and minimum image pixel values of the standardized target image, respectively
- X" is the normalized target image
- the drivable area segmentation model can be used to perform region segmentation on the normalized target image.
- Step S502 Use a preset drivable area segmentation model to perform area segmentation on the target image to obtain a drivable area segmentation image corresponding to the target image.
- the standardized target image may be used as the input of the drivable area segmentation model, and the output of the drivable area segmentation model may be used as the corresponding drivable area segmentation image.
- the drivable area segmentation image can be used to indicate the drivable area in the road environment, while the non-drivable area can be an area where obstacles exist, or an area where there are obstacles between the robot and the drivable area.
- the feasible area segmentation image can be used for auxiliary navigation. For example, specific applications such as navigation path planning and obstacle avoidance planning can be performed based on the feasible area segmentation image.
- the above method can be used to train the drivable area segmentation model in real time, and the drivable area model can be used to segment the drivable area to achieve use in multiple scenarios and multiple environments.
- the embodiment of the present application can obtain a drivable area segmentation model based on the synchronously collected prior images and prior point cloud data training, and use the drivable area segmentation model to perform regional segmentation on the target image to obtain a drivable area segmentation image corresponding to the target image.
- the model can be trained with the prior image as input and the synchronously collected prior point cloud data as the expected output, without the need for a large amount of sample collection and manual labeling work, thereby effectively reducing the training cost of the drivable area segmentation model and improving the training efficiency of the drivable area segmentation model.
- FIG6 shows a structural diagram of an embodiment of a device for segmenting a drivable area provided in an embodiment of the present application.
- a drivable area segmentation device may include:
- An image acquisition module 601 is used to acquire a target image to be segmented
- a region segmentation module 602 is used to perform region segmentation on the target image using a preset drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image;
- the drivable area segmentation model is a deep learning model trained based on prior images and prior point cloud data;
- the prior image is an image collected before the target image;
- the prior point cloud data is point cloud data collected synchronously with the prior image;
- the prior image is used as the training input of the deep learning model, and the prior segmented image is used as the expected training output of the deep learning model;
- the prior segmented image is a drivable area segmentation image determined based on the prior point cloud data.
- the drivable area segmentation device may further include:
- a data acquisition module used to acquire the previous image and the previous point cloud data
- a data conversion module used to convert the previous point cloud data from a preset radar coordinate system to a preset camera coordinate system to obtain converted point cloud data
- An image determination module used to determine the previously segmented image according to the converted point cloud data
- a model training module is used to train the deep learning model according to the previous image and the previous segmented image to obtain the drivable area segmentation model.
- the data acquisition module may include:
- a signal sending unit used to send a synchronous acquisition signal to the camera and the radar through a preset synchronous trigger, so that the camera and the radar perform synchronous data acquisition;
- an image acquisition unit configured to acquire the previous image acquired by the camera in response to the acquisition signal
- a data acquisition unit is used to acquire the previous point cloud data acquired by the radar in response to the acquisition signal.
- the data conversion module may include:
- a matrix acquisition unit used to acquire a translation matrix and a rotation matrix from the radar coordinate system to the camera coordinate system;
- a data conversion unit is used to convert the previous point cloud data from the radar coordinate system to the camera coordinate system according to the translation matrix and the rotation matrix to obtain the converted point cloud data.
- the model training module may include:
- An image acquisition unit configured to input the prior image into the deep learning model and acquire an output image of the deep learning model
- a loss calculation unit configured to calculate a training loss value of the deep learning model according to the output image and the previously segmented image
- a parameter adjustment unit is used to adjust the model parameters of the deep learning model according to the training loss value to obtain the drivable area segmentation model.
- the loss calculation unit may include:
- intersection-and-union ratio calculation subunit used to calculate the intersection-and-union ratio of the drivable area between the output image and the previous segmented image
- the loss value calculation subunit is used to calculate the training loss value according to the intersection-and-union ratio of the drivable area.
- the drivable area segmentation device may further include:
- a standardization processing module used for performing standardization processing on the target image to obtain a standardized target image
- a normalization processing module used for performing normalization processing on the standardized target image to obtain the normalized target image
- the region segmentation module may further include:
- a region segmentation unit is used to perform region segmentation on the normalized target image using the drivable region segmentation model to obtain a drivable region segmentation image corresponding to the target image.
- FIG7 shows a schematic block diagram of a robot provided in an embodiment of the present application. For ease of explanation, only the parts related to the embodiment of the present application are shown.
- the robot 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70.
- the processor 70 executes the computer program 72, the steps in the above-mentioned embodiments of the method for segmenting the drivable area are implemented, such as steps S501 to S502 shown in FIG5 .
- the processor 70 executes the computer program 72, the functions of the modules/units in the above-mentioned embodiments of the device are implemented, such as the functions of modules 601 to 602 shown in FIG6 .
- the computer program 72 may be divided into one or more modules/units, which are stored in the memory 71 and executed by the processor 70 to complete the present application.
- One or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 72 in the robot 7.
- FIG. 7 is merely an example of the robot 7 and does not constitute a limitation on the robot 7.
- the robot 7 may include more or fewer components than shown in the figure, or a combination of certain components, or different components.
- the robot 7 may also include input and output devices, network access devices, buses, etc.
- the processor 70 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
- a general-purpose processor may be a microprocessor or any conventional processor, etc.
- the memory 71 may be an internal storage unit of the robot 7, such as a hard disk or memory of the robot 7.
- the memory 71 may also be an external storage device of the robot 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), etc. equipped on the robot 7.
- the memory 71 may also include both an internal storage unit and an external storage device of the robot 7.
- the memory 71 is used to store the computer program and other programs and data required by the robot 7.
- the memory 71 may also be used to temporarily store data that has been output or is to be output.
- the technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration.
- the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
- the functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit.
- the above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.
- the disclosed devices/robots and methods can be implemented in other ways.
- the device/robot embodiments described above are merely schematic.
- the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.
- Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
- the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
- each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
- the above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
- the integrated module/unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
- the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program.
- the computer program can be stored in a computer-readable storage medium.
- the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented.
- the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form.
- the computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium.
- ROM Read-Only Memory
- RAM Random Access Memory
- electric carrier signal telecommunication signal and software distribution medium.
- the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.
- the computer-readable storage medium does not include electric carrier signal and telecommunication signal.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
本申请属于机器人技术领域,尤其涉及一种可行驶区域分割方法、装置、计算机可读存储介质及机器人。所述方法包括:可以根据同步采集的在先图像和在先点云数据训练得到可行驶区域分割模型,并利用该可行驶区域分割模型对目标图像进行区域分割,得到与目标图像对应的可行驶区域分割图像。通过上述方法,当机器人处于新的道路环境时,可以以在先图像作为输入,以同步采集的在先点云数据作为预期输出进行模型训练,而无需进行大量样本采集和人工标注工作,从而能够有效降低可行驶区域分割模型的训练成本,并提升可行驶区域分割模型的训练效率。
Description
本申请要求于2023年06月19日在中国专利局提交的、申请号为202310733916.7、发明名称为“一种可行驶区域分割方法、装置、可读存储介质及机器人”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本申请属于机器人技术领域,尤其涉及一种可行驶区域分割方法、装置、计算机可读存储介质及机器人。
随着道路的建设,行驶环境也越来越复杂。在现有技术中,当机器人部署到新的道路环境时,通常需要采集大量新的道路环境数据用于模型训练,以进行可行驶区域的分割。然而,现有方法的迭代时间较长,且数据标注过程依赖于人力,导致可行驶区域分割方法成本高而效率低。
有鉴于此,本申请实施例提供了一种可行驶区域分割方法、装置、计算机可读存储介质及机器人,以解决现有的可行驶区域分割方法成本高而效率低的问题。
本申请实施例的第一方面提供了一种可行驶区域分割方法,可以包括:
获取待分割的目标图像;
使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像;
其中,所述可行驶区域分割模型为根据在先图像和在先点云数据训练得到的深度学习模型;所述在先图像为在所述目标图像之前采集的图像;所述在先点云数据为与所述在先图像同步采集的点云数据;以所述在先图像作为所述深度学习模型的训练输入,以在先分割图像作为所述深度学习模型的预期训练输出;所述在先分割图像为根据所述在先点云数据确定的可行驶区域分割图像。
在第一方面的一种具体实现方式中,在使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像之前,还可以包括:
获取所述在先图像和所述在先点云数据;
将所述在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据;
根据所述转换点云数据确定所述在先分割图像;
根据所述在先图像和所述在先分割图像对所述深度学习模型进行训练,得到所述可行驶区域分割模型。
在第一方面的一种具体实现方式中,所述获取所述在先图像和所述在先点云数据,可以包括:
通过预设的同步触发器向相机和雷达发送同步的采集信号,以使所述相机和所述雷达进行同步数据采集;
获取所述相机响应于所述采集信号所采集的所述在先图像;
获取所述雷达响应于所述采集信号所采集的所述在先点云数据。
在第一方面的一种具体实现方式中,所述将所述在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据,可以包括:
获取由所述雷达坐标系至所述相机坐标系的平移矩阵和旋转矩阵;
按照所述平移矩阵和所述旋转矩阵将所述在先点云数据由所述雷达坐标系转换至所述相机坐标系中,得到所述转换点云数据。
在第一方面的一种具体实现方式中,所述根据所述在先图像和所述在先分割图像对所述深度学习模型进行训练,得到所述可行驶区域分割模型,可以包括:
将所述在先图像输入至所述深度学习模型,并获取所述深度学习模型的输出图像;
根据所述输出图像和所述在先分割图像计算所述深度学习模型的训练损失值;
根据所述训练损失值对所述深度学习模型进行模型参数调整,得到所述可行驶区域分割模型。
在第一方面的一种具体实现方式中,所述根据所述输出图像和所述在先分割图像计算所述深度学习模型的训练损失值,可以包括:
计算所述输出图像和所述在先分割图像之间的可行驶区域交并比;
根据所述可行驶区域交并比计算所述训练损失值。
在第一方面的一种具体实现方式中,在使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像之前,还可以包括:
对所述目标图像进行标准化处理,得到标准化的所述目标图像;
对标准化的所述目标图像进行归一化处理,得到归一化的所述目标图像;
所述使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像,可以包括:
使用所述可行驶区域分割模型对归一化的所述目标图像进行区域分割,得到与所述目
标图像对应的可行驶区域分割图像。
本申请实施例的第二方面提供了一种可行驶区域分割装置,可以包括:
图像获取模块,用于获取待分割的目标图像;
区域分割模块,用于使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像;
其中,所述可行驶区域分割模型为根据在先图像和在先点云数据训练得到的深度学习模型;所述在先图像为在所述目标图像之前采集的图像;所述在先点云数据为与所述在先图像同步采集的点云数据;以所述在先图像作为所述深度学习模型的训练输入,以在先分割图像作为所述深度学习模型的预期训练输出;所述在先分割图像为根据所述在先点云数据确定的可行驶区域分割图像。
在第二方面的一种具体实现方式中,所述可行驶区域分割装置还可以包括:
数据获取模块,用于获取所述在先图像和所述在先点云数据;
数据转化模块,用于将所述在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据;
图像确定模块,用于根据所述转换点云数据确定所述在先分割图像;
模型训练模块,用于根据所述在先图像和所述在先分割图像对所述深度学习模型进行训练,得到所述可行驶区域分割模型。
在第二方面的一种具体实现方式中,所述数据获取模块可以包括:
信号发送单元,用于通过预设的同步触发器向相机和雷达发送同步的采集信号,以使所述相机和所述雷达进行同步数据采集;
图像采集单元,用于获取所述相机响应于所述采集信号所采集的所述在先图像;
数据采集单元,用于获取所述雷达响应于所述采集信号所采集的所述在先点云数据。
在第二方面的一种具体实现方式中,所述数据转化模块可以包括:
矩阵获取单元,用于获取由所述雷达坐标系至所述相机坐标系的平移矩阵和旋转矩阵;
数据转换单元,用于按照所述平移矩阵和所述旋转矩阵将所述在先点云数据由所述雷达坐标系转换至所述相机坐标系中,得到所述转换点云数据。
在第二方面的一种具体实现方式中,所述模型训练模块可以包括:
图像获取单元,用于将所述在先图像输入至所述深度学习模型,并获取所述深度学习模型的输出图像;
损失计算单元,用于根据所述输出图像和所述在先分割图像计算所述深度学习模型的训练损失值;
参数调整单元,用于根据所述训练损失值对所述深度学习模型进行模型参数调整,得到所述可行驶区域分割模型。
在第二方面的一种具体实现方式中,所述损失计算单元可以包括:
交并比计算子单元,用于计算所述输出图像和所述在先分割图像之间的可行驶区域交并比;
损失值计算子单元,用于根据所述可行驶区域交并比计算所述训练损失值。
在第二方面的一种具体实现方式中,所述可行驶区域分割装置还可以包括:
标准化处理模块,用于对所述目标图像进行标准化处理,得到标准化的所述目标图像;
归一化处理模块,用于对标准化的所述目标图像进行归一化处理,得到归一化的所述目标图像;
所述区域分割模块还可以包括:
区域分割单元,用于使用所述可行驶区域分割模型对归一化的所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像。
本申请实施例的第三方面提供了一种计算机可读存储介质,所述计算机可读存储介质存储有计算机程序,所述计算机程序被处理器执行时实现上述任一种可行驶区域分割方法的步骤。
本申请实施例的第四方面提供了一种机器人,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序,所述处理器执行所述计算机程序时实现上述任一种可行驶区域分割方法的步骤。
本申请实施例的第五方面提供了一种计算机程序产品,当计算机程序产品在机器人上运行时,使得机器人执行上述任一种可行驶区域分割方法的步骤。
本申请实施例与现有技术相比存在的有益效果是:本申请实施例可以根据同步采集的在先图像和在先点云数据训练得到可行驶区域分割模型,并利用该可行驶区域分割模型对目标图像进行区域分割,得到与目标图像对应的可行驶区域分割图像。通过本申请实施例,当机器人处于新的道路环境时,可以以在先图像作为输入,以同步采集的在先点云数据作为预期输出进行模型训练,而无需进行大量样本采集和人工标注工作,从而能够有效降低可行驶区域分割模型的训练成本,并提升可行驶区域分割模型的训练效率。
为了更清楚地说明本申请实施例中的技术方案,下面将对实施例或现有技术描述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为可行驶区域分割方法的使用过程示意图;
图2为本申请实施例中可行驶区域分割模型的训练过程的示意流程图;
图3为雷达与相机的外参标定过程的示意图;
图4为可行驶区域分割模型的示意图;
图5为本申请实施例中一种可行驶区域分割方法的一个实施例流程图;
图6为本申请实施例中一种可行驶区域分割装置的一个实施例结构图;
图7为本申请实施例中一种机器人的示意框图。
为使得本申请的发明目的、特征、优点能够更加的明显和易懂,下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,下面所描述的实施例仅仅是本申请一部分实施例,而非全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其它实施例,都属于本申请保护的范围。
应当理解,当在本说明书和所附权利要求书中使用时,术语“包括”指示所描述特征、整体、步骤、操作、元素和/或组件的存在,但并不排除一个或多个其它特征、整体、步骤、操作、元素、组件和/或其集合的存在或添加。
还应当理解,在此本申请说明书中所使用的术语仅仅是出于描述特定实施例的目的而并不意在限制本申请。如在本申请说明书和所附权利要求书中所使用的那样,除非上下文清楚地指明其它情况,否则单数形式的“一”、“一个”及“该”意在包括复数形式。
还应当进一步理解,在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合,并且包括这些组合。
如在本说明书和所附权利要求书中所使用的那样,术语“如果”可以依据上下文被解释为“当...时”或“一旦”或“响应于确定”或“响应于检测到”。类似地,短语“如果确定”或“如果检测到[所描述条件或事件]”可以依据上下文被解释为意指“一旦确定”或“响应于确定”或“一旦检测到[所描述条件或事件]”或“响应于检测到[所描述条件或事件]”。
另外,在本申请的描述中,术语“第一”、“第二”、“第三”等仅用于区分描述,而不能理解为指示或暗示相对重要性。
随着道路的建设,行驶环境也越来越复杂。在现有技术中,当机器人部署到新的道路
环境时,通常需要采集大量新的道路环境数据用于模型训练,以进行可行驶区域的分割。然而,现有方法的迭代时间较长,且数据标注过程依赖于人力,导致可行驶区域分割方法成本高而效率低。
有鉴于此,本申请实施例提供了一种可行驶区域分割方法、装置、计算机可读存储介质及机器人,以解决现有的可行驶区域分割方法成本高而效率低的问题。
需要说明的是,本申请方法的执行主体为机器人,可以包括但不限于扫地机、巡检机器人、送餐机器人、引导机器人等现有技术中任意一种常见的机器人。
请参阅图1,在本申请实施例中,可以使用预设的可行驶区域分割模型对图像进行区域分割,得到与该图像对应的可行驶区域图像,并可以利用该可行驶区域图像进行辅助导航,其中,可行驶区域分割模型为根据在先图像和在先点云数据训练得到的深度学习模型,具体地,图2示出了可行驶区域分割模型的训练过程,可以包括如下步骤:
步骤S201、获取在先图像和在先点云数据。
具体地,在先前某一时刻,可以通过预设的相机采集包含行驶环境的在先图像,在同一时刻,还可以通过预设的雷达采集同一行驶环境的在先点云数据。
可以理解的是,机器人上可以预先安装有雷达与相机,以进行导航、探索及运动控制等相关应用。
在本申请实施例中,为了保证在先图像和在先点云数据能够在同一时刻采集,此处,可以使用预设的同步触发器向相机和雷达发送同步的采集信号;之后,相机响应于采集信号进行在先图像的采集,雷达响应于采集信号进行在先点云数据的采集;机器人则可获取到在先图像和在先点云数据。
步骤S202、将在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据。
具体地,可以根据预设的数据映射关系将在先点云数据映射至在先图像,其中,数据映射关系可以为雷达坐标系至相机坐标系的平移矩阵和旋转矩阵。根据平移矩阵和旋转矩阵,可以将在先点云数据由雷达坐标系转换至相机坐标系中,即可得到转换点云数据。
可以理解的是,在进行数据映射之前,可以预先进行外参标定,得到平移矩阵和旋转矩阵。图3所示即为外参标定的示意图,相机可以通过标定平面的二维码或棋盘格来计算标定平面在相机坐标系下的表示,同时,雷达发出的光束落在标定平面上(图中黑点),利用雷达点在雷达坐标系下的坐标和平面方程在相机坐标系下的坐标,构建点在平面上的约束从而求解得到平移矩阵和旋转矩阵。
在得到平移矩阵和旋转矩阵之后,可以根据下式将在先点云数据转换至相机坐标系中:
其中,(x1,y1)为转换后得到的转换点云数据,(x2,y2,z2)为在先点云数据,I为单位矩阵,t为平移矩阵,R为旋转矩阵。
具体地,旋转矩阵R可以表示为:
R=Rx·Ry·Rz
R=Rx·Ry·Rz
其中,Rx为绕x轴旋转的旋转矩阵,Ry为绕y轴旋转的旋转矩阵,Rz为绕z轴旋转的旋转矩阵;θx为绕x轴旋转的欧拉角,θy为绕y轴旋转的欧拉角,θz为绕z轴旋转的欧拉角。
步骤S203、根据转换点云数据确定在先分割图像。
在本申请实施例中,具体可以根据转换点云数据的高度信息确定在先分割图像。其中,在先分割图像可以为由转换点云数据分割出的可行驶区域图像。
在本申请实施例的一种具体实现方式中,在先点云数据可以为三维数据(x2,y2,z2),对应的转换点云数据可以为二维数据(x1,y1),而z2值即为转换点云数据的高度信息。根据转换点数据的高度信息,可以确定可行驶区域图像。例如,可以将预设的高度范围内的点云数据确定为障碍物数据,并排除该高度范围包含的区域,得到可行驶区域,进而得到在先分割图像。
可以理解的是,高度范围可以根据实际情况进行具体化、情景化设置,本申请对此不作限定。
步骤S204、根据在先图像和在先分割图像对深度学习模型进行训练,得到可行驶区域分割模型。
在本申请实施例中,可以将在先图像作为深度学习模型的输入,将在先分割图像作为深度学习模型的预期输出,对深度学习模型进行训练,得到可行驶区域分割模型。具体地,可以将在先图像输入至深度学习模型,并获取深度学习模型的输出图像;根据输出图像和在先分割图像计算深度学习模型的训练损失值;根据训练损失值对深度学习模型进行模型参数调整,可得到如图4所示的可行驶区域分割模型,其中,可行驶区域分割模型可以包括骨干网络、颈部网络和可行驶区域分割头。
在本申请实施例的一种具体实现方式中,可以计算输出图像和在先分割图像之间的可行驶区域交并比,即可以计算输出图像与在先分割图像之间可行驶区域的交集和并集的比值,之后,可以根据该比值计算训练损失值。例如,可将该比值作为训练损失值。
在得到训练损失值后,可以根据该训练损失值对深度学习模型进行参数调整。例如,可以通过反向更新的方式对深度学习模型进行参数调整。
可以理解的是,当训练损失值小于预设的损失值阈值时,或者,当模型训练的迭代次数大于预设的次数阈值时,可以认为得到的可行驶区域分割模型较为准确,此时,可使用该可行驶区域分割模型对图像进行区域分割,请参阅图5,本申请实施例中一种可行驶区域分割方法的一个实施例可以包括:
步骤S501、获取待分割的目标图像。
在本申请实施例中,可以通过相机采集待分割的目标图像,其中,待分割的目标图像可以为有待进行可行驶区域分割的目标图像。
可以理解的是,当机器人位于新的道路环境时,或者,当机器人开始导航时,相机可以实时采集到一张或者多张待分割的图像,为了便于叙述,下面将以采集到的图像中的任意一张图像作为待分割的目标图像为例,对本申请实施例中的可行驶区域分割方法进行说明。
在本申请实施例的一种具体实现方式中,为了提升可行驶区域分割模型的分割效率,可以在利用可行驶区域分割模型对目标图像进行区域分割之前,对目标图像进行预处理,具体可以根据实际需要对目标图像进行现有技术中常见的图像预处理操作。此处,可以优选地对目标分割图像进行图像标准化处理,再进行图像归一化处理。具体地,可使用下式对目标图像进行标准化处理:
其中,xi为目标图像的图像像素值,μ为目标图像的图像像素值的均值,σ为目标图像的图像像素值的标准方差,N为目标图像的图像像素数量,X'为标准化的目标图像。
之后,可对标准化的目标图像进行归一化处理,得到归一化的目标图像,具体地,可使用下式对标准化的目标图像进行归一化处理:
其中,x'i为标准化的目标图像的图像像素值,max(x')、min(x')分别为标准化的目标图像的图像像素值的最大值与最小值,X”为归一化的目标图像。
在得到归一化的目标图像后,可利用可行驶区域分割模型对归一化的目标图像进行区域分割。
步骤S502、使用预设的可行驶区域分割模型对目标图像进行区域分割,得到与目标图像对应的可行驶区域分割图像。
在本申请实施例中,可以将标准化的目标图像作为可行驶区域分割模型的输入,并将可行驶区域分割模型的输出作为对应的可行驶区域分割图像。
可以理解的是,可行驶区域分割图像可用于指示道路环境中可行驶的区域,而不可行驶的区域可以为存在障碍物的区域,或者与机器人之间存在障碍物的区域。请参阅图1,在得到可行驶区域分割图像后,可以利用该可行区域分割图像进行辅助导航,例如,可以根据可行区域分割图像进行导航路径规划和避障规划等具体应用。
在实际使用过程中,当部署到新的道路环境时,环境中可能会出现原训练集中不存在的未知物体,此时可使用上述方法对可行驶区域分割模型进行实时训练,并使用可行驶区域模型进行可行驶区域分割,实现在多种场景、多种环境的使用。
综上所述,本申请实施例可以根据同步采集的在先图像和在先点云数据训练得到可行驶区域分割模型,并利用该可行驶区域分割模型对目标图像进行区域分割,得到与目标图像对应的可行驶区域分割图像。通过本申请实施例,当机器人处于新的道路环境时,可以以在先图像作为输入,以同步采集的在先点云数据作为预期输出进行模型训练,而无需进行大量样本采集和人工标注工作,从而能够有效降低可行驶区域分割模型的训练成本,并提升可行驶区域分割模型的训练效率。
应理解,上述实施例中各步骤的序号的大小并不意味着执行顺序的先后,各过程的执行顺序应以其功能和内在逻辑确定,而不应对本申请实施例的实施过程构成任何限定。
对应于上文实施例所述的一种可行驶区域分割方法,图6示出了本申请实施例提供的一种可行驶区域分割装置的一个实施例结构图。
本实施例中,一种可行驶区域分割装置可以包括:
图像获取模块601,用于获取待分割的目标图像;
区域分割模块602,用于使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像;
其中,所述可行驶区域分割模型为根据在先图像和在先点云数据训练得到的深度学习模型;所述在先图像为在所述目标图像之前采集的图像;所述在先点云数据为与所述在先图像同步采集的点云数据;以所述在先图像作为所述深度学习模型的训练输入,以在先分割图像作为所述深度学习模型的预期训练输出;所述在先分割图像为根据所述在先点云数据确定的可行驶区域分割图像。
在本申请实施例的一种具体实现方式中,所述可行驶区域分割装置还可以包括:
数据获取模块,用于获取所述在先图像和所述在先点云数据;
数据转化模块,用于将所述在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据;
图像确定模块,用于根据所述转换点云数据确定所述在先分割图像;
模型训练模块,用于根据所述在先图像和所述在先分割图像对所述深度学习模型进行训练,得到所述可行驶区域分割模型。
在本申请实施例的一种具体实现方式中,所述数据获取模块可以包括:
信号发送单元,用于通过预设的同步触发器向相机和雷达发送同步的采集信号,以使所述相机和所述雷达进行同步数据采集;
图像采集单元,用于获取所述相机响应于所述采集信号所采集的所述在先图像;
数据采集单元,用于获取所述雷达响应于所述采集信号所采集的所述在先点云数据。
在本申请实施例的一种具体实现方式中,所述数据转化模块可以包括:
矩阵获取单元,用于获取由所述雷达坐标系至所述相机坐标系的平移矩阵和旋转矩阵;
数据转换单元,用于按照所述平移矩阵和所述旋转矩阵将所述在先点云数据由所述雷达坐标系转换至所述相机坐标系中,得到所述转换点云数据。
在本申请实施例的一种具体实现方式中,所述模型训练模块可以包括:
图像获取单元,用于将所述在先图像输入至所述深度学习模型,并获取所述深度学习模型的输出图像;
损失计算单元,用于根据所述输出图像和所述在先分割图像计算所述深度学习模型的训练损失值;
参数调整单元,用于根据所述训练损失值对所述深度学习模型进行模型参数调整,得到所述可行驶区域分割模型。
在本申请实施例的一种具体实现方式中,所述损失计算单元可以包括:
交并比计算子单元,用于计算所述输出图像和所述在先分割图像之间的可行驶区域交并比;
损失值计算子单元,用于根据所述可行驶区域交并比计算所述训练损失值。
在本申请实施例的一种具体实现方式中,所述可行驶区域分割装置还可以包括:
标准化处理模块,用于对所述目标图像进行标准化处理,得到标准化的所述目标图像;
归一化处理模块,用于对标准化的所述目标图像进行归一化处理,得到归一化的所述目标图像;
所述区域分割模块还可以包括:
区域分割单元,用于使用所述可行驶区域分割模型对归一化的所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像。
所属领域的技术人员可以清楚地了解到,为描述的方便和简洁,上述描述的装置,模块和单元的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在上述实施例中,对各个实施例的描述都各有侧重,某个实施例中没有详述或记载的部分,可以参见其它实施例的相关描述。
图7示出了本申请实施例提供的一种机器人的示意框图,为了便于说明,仅示出了与本申请实施例相关的部分。
如图7所示,该实施例的机器人7包括:处理器70、存储器71以及存储在所述存储器71中并可在所述处理器70上运行的计算机程序72。所述处理器70执行所述计算机程序72时实现上述各个可行驶区域分割方法实施例中的步骤,例如图5所示的步骤S501至步骤S502。或者,所述处理器70执行所述计算机程序72时实现上述各装置实施例中各模块/单元的功能,例如图6所示模块601至模块602的功能。
示例性的,所述计算机程序72可以被分割成一个或多个模块/单元,所述一个或者多个模块/单元被存储在所述存储器71中,并由所述处理器70执行,以完成本申请。所述一
个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段,该指令段用于描述所述计算机程序72在所述机器人7中的执行过程。
本领域技术人员可以理解,图7仅仅是机器人7的示例,并不构成对机器人7的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件,例如所述机器人7还可以包括输入输出设备、网络接入设备、总线等。
所述处理器70可以是中央处理单元(Central Processing Unit,CPU),还可以是其它通用处理器、数字信号处理器(Digital Signal Processor,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或者其它可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器71可以是所述机器人7的内部存储单元,例如机器人7的硬盘或内存。所述存储器71也可以是所述机器人7的外部存储设备,例如所述机器人7上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。进一步地,所述存储器71还可以既包括所述机器人7的内部存储单元也包括外部存储设备。所述存储器71用于存储所述计算机程序以及所述机器人7所需的其它程序和数据。所述存储器71还可以用于暂时地存储已经输出或者将要输出的数据。
所属领域的技术人员可以清楚地了解到,为了描述的方便和简洁,仅以上述各功能单元、模块的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能单元、模块完成,即将所述装置的内部结构划分成不同的功能单元或模块,以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中,上述集成的单元既可以采用硬件的形式实现,也可以采用软件功能单元的形式实现。另外,各功能单元、模块的具体名称也只是为了便于相互区分,并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在上述实施例中,对各个实施例的描述都各有侧重,某个实施例中没有详述或记载的部分,可以参见其它实施例的相关描述。
本领域普通技术人员可以意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本
申请的范围。
在本申请所提供的实施例中,应该理解到,所揭露的装置/机器人和方法,可以通过其它的方式实现。例如,以上所描述的装置/机器人实施例仅仅是示意性的,例如,所述模块或单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口,装置或单元的间接耦合或通讯连接,可以是电性,机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能单元可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现,也可以采用软件功能单元的形式实现。
所述集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读存储介质中。基于这样的理解,本申请实现上述实施例方法中的全部或部分流程,也可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储于一计算机可读存储介质中,该计算机程序在被处理器执行时,可实现上述各个方法实施例的步骤。其中,所述计算机程序包括计算机程序代码,所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读存储介质可以包括:能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是,所述计算机可读存储介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减,例如在某些司法管辖区,根据立法和专利实践,计算机可读存储介质不包括电载波信号和电信信号。
以上所述实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围,均应包含在本申请的保护范围之内。
Claims (10)
- 一种可行驶区域分割方法,其特征在于,包括:获取待分割的目标图像;使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像;其中,所述可行驶区域分割模型为根据在先图像和在先点云数据训练得到的深度学习模型;所述在先图像为在所述目标图像之前采集的图像;所述在先点云数据为与所述在先图像同步采集的点云数据;以所述在先图像作为所述深度学习模型的训练输入,以在先分割图像作为所述深度学习模型的预期训练输出;所述在先分割图像为根据所述在先点云数据确定的可行驶区域分割图像。
- 根据权利要求1所述的可行驶区域分割方法,其特征在于,在使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像之前,还包括:获取所述在先图像和所述在先点云数据;将所述在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据;根据所述转换点云数据确定所述在先分割图像;根据所述在先图像和所述在先分割图像对所述深度学习模型进行训练,得到所述可行驶区域分割模型。
- 根据权利要求2所述的可行驶区域分割方法,其特征在于,所述获取所述在先图像和所述在先点云数据,包括:通过预设的同步触发器向相机和雷达发送同步的采集信号,以使所述相机和所述雷达进行同步数据采集;获取所述相机响应于所述采集信号所采集的所述在先图像;获取所述雷达响应于所述采集信号所采集的所述在先点云数据。
- 根据权利要求2所述的可行驶区域分割方法,其特征在于,所述将所述在先点云数据由预设的雷达坐标系转换至预设的相机坐标系中,得到转换点云数据,包括:获取由所述雷达坐标系至所述相机坐标系的平移矩阵和旋转矩阵;按照所述平移矩阵和所述旋转矩阵将所述在先点云数据由所述雷达坐标系转换至所述相机坐标系中,得到所述转换点云数据。
- 根据权利要求2所述的可行驶区域分割方法,其特征在于,所述根据所述在先图像 和所述在先分割图像对所述深度学习模型进行训练,得到所述可行驶区域分割模型,包括:将所述在先图像输入至所述深度学习模型,并获取所述深度学习模型的输出图像;根据所述输出图像和所述在先分割图像计算所述深度学习模型的训练损失值;根据所述训练损失值对所述深度学习模型进行模型参数调整,得到所述可行驶区域分割模型。
- 根据权利要求5所述的可行驶区域分割方法,其特征在于,所述根据所述输出图像和所述在先分割图像计算所述深度学习模型的训练损失值,包括:计算所述输出图像和所述在先分割图像之间的可行驶区域交并比;根据所述可行驶区域交并比计算所述训练损失值。
- 根据权利要求1至6中任一项所述的可行驶区域分割方法,其特征在于,在使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像之前,还包括:对所述目标图像进行标准化处理,得到标准化的所述目标图像;对标准化的所述目标图像进行归一化处理,得到归一化的所述目标图像;所述使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像,包括:使用所述可行驶区域分割模型对归一化的所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像。
- 一种可行驶区域分割装置,其特征在于,包括:图像获取模块,用于获取待分割的目标图像;区域分割模块,用于使用预设的可行驶区域分割模型对所述目标图像进行区域分割,得到与所述目标图像对应的可行驶区域分割图像;其中,所述可行驶区域分割模型为根据在先图像和在先点云数据训练得到的深度学习模型;所述在先图像为在所述目标图像之前采集的图像;所述在先点云数据为与所述在先图像同步采集的点云数据;以所述在先图像作为所述深度学习模型的训练输入,以在先分割图像作为所述深度学习模型的预期训练输出;所述在先分割图像为根据所述在先点云数据确定的可行驶区域分割图像。
- 一种计算机可读存储介质,所述计算机可读存储介质存储有计算机程序,其特征在于,所述计算机程序被处理器执行时实现如权利要求1至7中任一项所述的可行驶区域分割方法的步骤。
- 一种机器人,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序,其特征在于,所述处理器执行所述计算机程序时实现如权利要求1至7中任一项所述的可行驶区域分割方法的步骤。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310733916.7 | 2023-06-19 | ||
| CN202310733916.7A CN116883652A (zh) | 2023-06-19 | 2023-06-19 | 一种可行驶区域分割方法、装置、可读存储介质及机器人 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2024259938A1 true WO2024259938A1 (zh) | 2024-12-26 |
Family
ID=88267163
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2023/141525 Ceased WO2024259938A1 (zh) | 2023-06-19 | 2023-12-25 | 一种可行驶区域分割方法、装置、可读存储介质及机器人 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN116883652A (zh) |
| WO (1) | WO2024259938A1 (zh) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116883652A (zh) * | 2023-06-19 | 2023-10-13 | 深圳市优必选科技股份有限公司 | 一种可行驶区域分割方法、装置、可读存储介质及机器人 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109117691A (zh) * | 2017-06-23 | 2019-01-01 | 百度在线网络技术(北京)有限公司 | 可行驶区域检测方法、装置、设备及存储介质 |
| CN109543600A (zh) * | 2018-11-21 | 2019-03-29 | 成都信息工程大学 | 一种实现可行驶区域检测方法及系统和应用 |
| WO2021128777A1 (en) * | 2019-12-23 | 2021-07-01 | Suzhou Zhijia Science & Technologies Co., Ltd. | Method, apparatus, device, and storage medium for detecting travelable region |
| CN113343875A (zh) * | 2021-06-18 | 2021-09-03 | 深圳亿嘉和科技研发有限公司 | 一种用于机器人的可行驶区域感知方法 |
| CN113421217A (zh) * | 2020-03-02 | 2021-09-21 | 北京京东乾石科技有限公司 | 可行驶区域检测方法和装置 |
| CN116883652A (zh) * | 2023-06-19 | 2023-10-13 | 深圳市优必选科技股份有限公司 | 一种可行驶区域分割方法、装置、可读存储介质及机器人 |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109102543B (zh) * | 2018-08-17 | 2021-04-02 | 深圳蓝胖子机器智能有限公司 | 基于图像分割的物体定位方法、设备和存储介质 |
| CN110189341B (zh) * | 2019-06-05 | 2021-08-10 | 北京青燕祥云科技有限公司 | 一种图像分割模型训练的方法、图像分割的方法及装置 |
| CN114981840B (zh) * | 2020-11-11 | 2025-06-13 | 深圳元戎启行科技有限公司 | 基于点云数据的地面分割方法、装置和计算机设备 |
| US12123734B2 (en) * | 2021-12-01 | 2024-10-22 | Motional Ad Llc | Automatic annotation of drivable road segments |
-
2023
- 2023-06-19 CN CN202310733916.7A patent/CN116883652A/zh active Pending
- 2023-12-25 WO PCT/CN2023/141525 patent/WO2024259938A1/zh not_active Ceased
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109117691A (zh) * | 2017-06-23 | 2019-01-01 | 百度在线网络技术(北京)有限公司 | 可行驶区域检测方法、装置、设备及存储介质 |
| CN109543600A (zh) * | 2018-11-21 | 2019-03-29 | 成都信息工程大学 | 一种实现可行驶区域检测方法及系统和应用 |
| WO2021128777A1 (en) * | 2019-12-23 | 2021-07-01 | Suzhou Zhijia Science & Technologies Co., Ltd. | Method, apparatus, device, and storage medium for detecting travelable region |
| CN113421217A (zh) * | 2020-03-02 | 2021-09-21 | 北京京东乾石科技有限公司 | 可行驶区域检测方法和装置 |
| CN113343875A (zh) * | 2021-06-18 | 2021-09-03 | 深圳亿嘉和科技研发有限公司 | 一种用于机器人的可行驶区域感知方法 |
| CN116883652A (zh) * | 2023-06-19 | 2023-10-13 | 深圳市优必选科技股份有限公司 | 一种可行驶区域分割方法、装置、可读存储介质及机器人 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116883652A (zh) | 2023-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111340797B (zh) | 一种激光雷达与双目相机数据融合检测方法及系统 | |
| CN114140527B (zh) | 一种基于语义分割的动态环境双目视觉slam方法 | |
| CN115496923B (zh) | 一种基于不确定性感知的多模态融合目标检测方法及装置 | |
| CN116912417B (zh) | 基于人脸三维重建的纹理贴图方法、装置、设备和存储介质 | |
| CN113743391A (zh) | 应用于低速自主驾驶机器人的三维障碍物检测系统与方法 | |
| WO2019161813A1 (zh) | 动态场景的三维重建方法以及装置和系统、服务器、介质 | |
| WO2024197815A1 (zh) | 一种工程机械的建图、方法、设备、及可读存储介质 | |
| CN114119780B (zh) | 一种图像标注方法、装置及电子设备 | |
| CN108648194B (zh) | 基于cad模型三维目标识别分割和位姿测量方法及装置 | |
| CN115410167A (zh) | 目标检测与语义分割方法、装置、设备及存储介质 | |
| CN110246147A (zh) | 视觉惯性里程计方法、视觉惯性里程计装置及移动设备 | |
| WO2018120027A1 (zh) | 一种障碍物检测方法及装置 | |
| CN115620094B (zh) | 关键点的标注方法、装置、电子设备及存储介质 | |
| CN110774319A (zh) | 一种机器人及其定位方法和装置 | |
| CN116385994A (zh) | 一种三维道路线提取方法及相关设备 | |
| CN116129112A (zh) | 一种核酸检测机器人的口腔三维点云分割方法及机器人 | |
| WO2024083006A1 (zh) | 一种三维成像方法、装置、设备和存储介质 | |
| CN115761164A (zh) | 逆透视ipm图像生成的方法和装置 | |
| CN109949414A (zh) | 室内地图的构建方法及装置 | |
| WO2024259938A1 (zh) | 一种可行驶区域分割方法、装置、可读存储介质及机器人 | |
| CN113724311B (zh) | 深度图获取方法、设备及存储介质 | |
| CN118736513A (zh) | 数据处理方法、装置、电子设备、存储介质及车辆 | |
| KR20220144456A (ko) | Svm 원 영상을 기반으로 근접 주행 환경을 인식하는 방법 및 시스템 | |
| CN116642490A (zh) | 基于混合地图的视觉定位导航方法、机器人及存储介质 | |
| CN117930169A (zh) | 目标检测方法、终端设备及计算机可读存储介质 |
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: 23942203 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |