WO2023273498A1 - 深度检测方法及装置、电子设备和存储介质 - Google Patents

深度检测方法及装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2023273498A1
WO2023273498A1 PCT/CN2022/085913 CN2022085913W WO2023273498A1 WO 2023273498 A1 WO2023273498 A1 WO 2023273498A1 CN 2022085913 W CN2022085913 W CN 2022085913W WO 2023273498 A1 WO2023273498 A1 WO 2023273498A1
Authority
WO
WIPO (PCT)
Prior art keywords
target object
key point
frame
point detection
detected
Prior art date
Application number
PCT/CN2022/085913
Other languages
English (en)
French (fr)
Inventor
赵佳
谢符宝
刘文韬
钱晨
Original Assignee
上海商汤智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海商汤智能科技有限公司 filed Critical 上海商汤智能科技有限公司
Publication of WO2023273498A1 publication Critical patent/WO2023273498A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20112—Image segmentation details
    • G06T2207/20132—Image cropping

Definitions

  • the present disclosure relates to the technical field of computers, and in particular to a depth detection method and device, electronic equipment and a storage medium.
  • the depth information can reflect the distance of the human body in the image relative to the image acquisition device, and based on the depth information, the human body object in the image can be spatially positioned.
  • the monocular camera is a relatively common and widely used image acquisition device.
  • the monocular camera can only provide two-dimensional information. How to accurately determine the depth information of the human body in the image based on the image collected by the monocular camera has become the current A burning problem.
  • the present disclosure proposes a technical solution for depth detection.
  • a deep detection method including:
  • the frame to be detected includes a target object; perform key point detection on the target object according to the frame to be detected, and obtain a key point detection result; determine the target object based on the key point detection result
  • the characteristic length of the target area in the target area wherein the target area includes the head area and/or the shoulder area, and the characteristic length is used to characterize the size information of the target area in the target object; according to the characteristic length of the target area , determining the depth information of the target object in the frame to be detected.
  • the performing key point detection on the target object according to the frame to be detected to obtain the key point detection result includes: according to the position information of the target object in the reference frame, The target object in the frame to be detected performs key point detection to obtain a key point detection result, wherein the reference frame is a video frame before the frame to be detected in the target video to which the frame to be detected belongs .
  • the performing key point detection on the target object in the frame to be detected according to the position information of the target object in the reference frame to obtain a key point detection result includes: according to The first position of the target object in the reference frame is clipped to the frame to be detected to obtain a clipping result; the key point detection is performed on the target object in the clipping result to obtain the key point detection result.
  • the performing key point detection on the target object in the frame to be detected according to the position information of the target object in the reference frame to obtain a key point detection result includes: obtaining A second position of the target area of the target object in the reference frame; according to the second position, the frame to be detected is cropped to obtain a cropping result; key points are performed on the target object in the cropping result Detect to obtain the key point detection result.
  • the obtaining the second position of the target area of the target object in the reference frame includes: identifying the target area in the reference frame by using a first neural network to obtain The second position output by the first neural network; and/or, according to the key point detection result corresponding to the reference frame, the second position of the target area in the reference frame is obtained.
  • the key point detection results include head key points, left shoulder key points, and right shoulder key points; based on the key point detection results, determining the features of the target area in the target object
  • the length includes: obtaining the first characteristic length of the target area according to the distance between the left shoulder key point and the right shoulder key point; according to the distance between the head key point and the shoulder center point, Acquire the second characteristic length of the target area, wherein the shoulder center point is the middle point between the left shoulder key point and the right shoulder key point; according to the first characteristic length and/or the second A characteristic length is used to determine the characteristic length of the target area.
  • the depth information includes a depth distance
  • the depth distance includes a distance between the target object and an optical center of a collection device
  • the collection device includes image collection of the target object
  • the device; according to the characteristic length of the target area, determining the depth information of the target object in the frame to be detected includes: obtaining the preset characteristic length of the target area, and the preset of the acquisition device Equipment parameters: determining the depth distance according to the proportional relationship between the preset characteristic length and the characteristic length of the target area, and the preset equipment parameters.
  • the depth information includes an offset angle
  • the offset angle includes a spatial angle of the target object relative to an optical axis of a collection device
  • the collection device includes An image acquisition device
  • the method further includes: acquiring preset device parameters of the acquisition device; and determining the offset angle according to the preset device parameters and the key point detection result.
  • the method further includes: determining a position of the target object in a three-dimensional space according to depth information of the target object.
  • a depth detection device including:
  • An acquisition module configured to acquire a frame to be detected, the frame to be detected includes a target object; a key point detection module, configured to perform key point detection on the target object according to the frame to be detected, to obtain a key point detection result; A length determination module, configured to determine a characteristic length of a target region in the target object based on the key point detection result, wherein the target region includes a head region and/or a shoulder region, and the characteristic length is used to characterize Size information of the target area in the target object; a depth detection module configured to determine the depth information of the target object in the frame to be detected according to the characteristic length of the target area.
  • the key point detection module is configured to: perform key point detection on the target object in the frame to be detected according to the position information of the target object in the reference frame, and obtain key points A point detection result, wherein the reference frame is a video frame before the frame to be detected in the target video to which the frame to be detected belongs.
  • the key point detection module is further configured to: clip the frame to be detected according to the first position of the target object in the reference frame to obtain a clipping result; The target object in the clipping result is subjected to key point detection to obtain the key point detection result.
  • the key point detection module is further configured to: acquire a second position of the target area of the target object in the reference frame; Clipping the frame to obtain a clipping result; performing key point detection on the target object in the clipping result to obtain the key point detection result.
  • the key point detection module is further configured to: use a first neural network to identify the target area in the reference frame to obtain a second position output by the first neural network; and /or, obtain the second position of the target area in the reference frame according to the key point detection result corresponding to the reference frame.
  • the key point detection results include head key points, left shoulder key points, and right shoulder key points;
  • the characteristic length determination module is configured to: The distance between the key points is to obtain the first characteristic length of the target area; according to the distance between the head key point and the shoulder center point, the second characteristic length of the target area is obtained, wherein the The shoulder center point is an intermediate point between the left shoulder key point and the right shoulder key point; the characteristic length of the target area is determined according to the first characteristic length and/or the second characteristic length.
  • the depth information includes a depth distance
  • the depth distance includes a distance between the target object and an optical center of a collection device
  • the collection device includes image collection of the target object
  • the depth detection module is used to: obtain the preset characteristic length of the target area, and the preset device parameters of the acquisition device; according to the difference between the preset characteristic length and the characteristic length of the target area The proportional relationship of and the preset device parameters determine the depth distance.
  • the depth information includes an offset angle
  • the offset angle includes a spatial angle of the target object relative to an optical axis of a collection device
  • the collection device includes An image acquisition device
  • the device is further configured to: obtain preset device parameters of the acquisition device; and determine the offset angle according to the preset device parameters and the key point detection results.
  • the apparatus is further configured to: determine the position of the target object in a three-dimensional space according to the depth information of the target object.
  • an electronic device including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to call the instructions stored in the memory to execute the above-mentioned method.
  • a computer-readable storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the above method is implemented.
  • a computer program product including computer readable codes, and when the computer readable codes are run in an electronic device, a processor in the electronic device executes the above method.
  • the key point detection is performed on the target object according to the frame to be detected containing the target object to obtain the key point detection result, and the characteristic length of the target area in the target object is determined based on the key point detection result, so that according to the The characteristic length determines the depth information of the target object.
  • the depth estimation of the target object in the frame to be detected can be performed based on the characteristic length of the target area in the target object, because the characteristic length is used to characterize the head area of the target object And/or the size information of the shoulder area, when the orientation or posture of the target object changes, the characteristic length is not easily disturbed, and the value is relatively stable, so the depth information obtained based on the characteristic length can be more accurate, improving the Accuracy and robustness of depth detection.
  • Fig. 1 shows a flowchart of a depth detection method according to an embodiment of the present disclosure.
  • FIG. 2 shows a schematic diagram of a target area according to an embodiment of the present disclosure.
  • Fig. 3 shows a flowchart of a depth detection method according to an embodiment of the present disclosure.
  • FIG. 4 shows a block diagram of a depth detection device according to an embodiment of the present disclosure.
  • Fig. 5 shows a schematic diagram of an application example according to the present disclosure.
  • FIG. 6 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • FIG. 7 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • Fig. 1 shows a flowchart of a depth detection method according to an embodiment of the present disclosure.
  • the method can be performed by a depth detection device, and the depth detection device can be an electronic device such as a terminal device or a server, and the terminal device can be user equipment (User Equipment, UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal Digital assistants (Personal Digital Assistant, PDA), handheld devices, computing devices, in-vehicle devices, wearable devices, etc.
  • the method may be implemented by a processor invoking computer-readable instructions stored in a memory.
  • the method can be performed by a server.
  • the method may include:
  • Step S11 acquiring a frame to be detected, where the frame to be detected contains a target object.
  • the frame to be detected may be any image frame that requires depth detection, for example, it may be an image frame extracted from a captured video, or an image frame obtained by capturing an image.
  • the number of frames to be detected is not limited in the embodiments of the present disclosure, and may include one frame or multiple frames.
  • the depth detection is performed on the frames at the same time, or the depth detection can be performed on multiple frames to be detected sequentially in a certain order.
  • the frame to be detected includes the target object to be subjected to depth detection.
  • the type of the target object is not limited in the embodiments of the present disclosure, and may include various human objects, animal objects, or some mechanical objects, such as robots. Subsequent disclosed embodiments are described by taking the target object as a person object as an example. Implementations in which the target object is other types can be flexibly expanded by referring to the subsequent disclosed embodiments, and will not be elaborated one by one.
  • the number of target objects contained in the frame to be detected is also not limited in the embodiments of the present disclosure, and may contain one or more target objects, which can be flexibly determined according to actual conditions.
  • frame extraction may be performed from a video to obtain one or more frames to be detected, wherein the frame extraction may include One or more methods such as frame-by-frame extraction, frame sampling at a certain interval, or random frame sampling.
  • the image acquisition of the target object may also be performed to obtain the frame to be detected; in some possible implementation manners, the frame to be detected may also be obtained by reading from a database.
  • Step S12 performing key point detection on the target object according to the frame to be detected, and obtaining a key point detection result.
  • the key point detection result may include the position of the detected key point in the frame to be detected.
  • the number and types of detected key points can be flexibly determined according to the actual situation.
  • the number of detected key points can include 2 to 150, etc.
  • the detected key points can be Contains 14 limb key points of the human body (such as head key points, shoulder key points, neck key points, elbow key points, wrist key points, crotch key points, leg key points and foot key points, etc.) , or include 59 outline key points on the outline of the human body (such as some key points on the periphery of the head or the periphery of the shoulders) and the like.
  • the detected key points may also only include three key points including the key point of the head, the key point of the left shoulder and the key point of the right shoulder.
  • the frame to be detected can be input into any neural network with key point detection function to realize key point detection; in some possible implementations, It is also possible to identify the key points of the frame to be detected by using the relevant key point recognition algorithm to obtain the key point detection result; Part of the image area is subjected to key point detection to obtain key point detection results, etc.
  • step S12 reference may be made to the following disclosed embodiments in detail, which will not be expanded here.
  • Step S13 based on the key point detection result, determine the characteristic length of the target area in the target object.
  • the target area may include the head area and/or the shoulder area
  • the head area of the target object may be the area where the head of the target object is located, such as the area formed between the key points of the head and the key points of the neck
  • the head area may be the area where the shoulder and neck of the target object are located, such as the area formed between the key points of the neck and the key points of the shoulder.
  • Fig. 2 shows a schematic diagram of a target area according to an embodiment of the present disclosure.
  • the head key can be point
  • the key point of the left shoulder and the key point of the right shoulder are connected by the head and shoulders box, which is used as the target area.
  • the head-shoulders frame can be a rectangle as shown in Figure 2. It can be seen from Figure 2 that the head-shoulders frame can be connected to the head key point at the head vertex of the target object and the left shoulder key point at the left shoulder joint. and the right shoulder key point at the right shoulder joint.
  • the head-shoulders frame may also be in other shapes, such as polygons, circles, or other irregular shapes.
  • the characteristic length of the target area can be used to represent the size information of the target area in the target object, and the size information can reflect relatively stable features in the target area, so the characteristic length representing the size information is affected by the orientation or direction of the target object in the frame to be detected.
  • the influence of attitude is small, and the value is more stable.
  • the size information can be flexibly determined according to the actual situation, and is not limited to the following disclosed embodiments.
  • the size information may include the size of some parts of the target object with stable features, such as the size between the head and shoulders and/or the size between the shoulders; in some possible implementations , the size information can also be the larger or smaller value among the sizes of multiple feature-stabilized parts; in some possible implementations, the size information can also include the size between multiple feature-stabilized parts Ratio, etc., such as the size ratio between the head and shoulders and the size between the shoulders, or the size ratio between the shoulders and the head and shoulders.
  • the manner of determining the feature length according to the key point detection result may also change accordingly.
  • Step S14 according to the characteristic length of the target area, determine the depth information of the target object in the frame to be detected.
  • the information content contained in the depth information can be flexibly determined according to the actual situation, and any information that can reflect the depth of the target object in the three-dimensional space can be used as a realization method of the depth information.
  • the depth information may include a depth distance and/or an offset angle.
  • the depth distance can be the distance between the target object and the collection device, and the collection device can be any device that collects images of the target object.
  • the collection device can be a static image collection device, such as a camera, etc. ;
  • the collection device may also be a device for collecting dynamic images, such as a video camera or a camera.
  • the depth distance can be the distance between the target object and the collection device, the distance can be the distance between the target object and the collection device as a whole, or the distance between the target object and a certain equipment part of the collection device, in some possible
  • the distance between the target object and the optical center of the acquisition device may be used as the depth distance.
  • the offset angle may be an offset angle of the target object relative to the collection device, and in a possible implementation manner, the offset angle may be a spatial angle of the target object relative to the optical axis of the collection device.
  • the actual length corresponding to the characteristic length can be relatively easily determined as an a priori estimated value of the characteristic length.
  • the prior estimation value can be used as reference information, combined with the feature length, to realize the monocular distance measurement of the frame to be detected, and obtain the depth information of the frame to be detected.
  • the method of determining the depth information based on the characteristic length can be flexibly determined according to the actual situation. Any implementation method of monocular distance measurement can be used in the implementation process of step S14. For example, according to the characteristic length and the prior estimate value of the characteristic length, Combined with some relevant parameters in the acquisition process of the frame to be detected to jointly calculate the depth distance and so on.
  • the specific implementation process of step S14 can be referred to the following disclosed embodiments in detail, and will not be expanded here.
  • the key point detection is performed on the target object according to the frame to be detected containing the target object to obtain the key point detection result, and the characteristic length of the target area in the target object is determined based on the key point detection result, so that according to the The characteristic length determines the depth information of the target object.
  • the depth estimation of the target object in the frame to be detected can be performed based on the characteristic length of the target area in the target object, because the characteristic length is used to characterize the head area of the target object And/or the size information of the shoulder area, when the orientation or posture of the target object changes, the characteristic length is not easily disturbed, and the value is relatively stable, so the depth information obtained based on the characteristic length can be more accurate, improving the Accuracy and robustness of depth detection.
  • step S12 may include:
  • the key point detection is performed on the target object in the frame to be detected, and the key point detection result is obtained.
  • the reference frame may be a video frame located before the frame to be detected in the target video, and the target video may be a video including the frame to be detected.
  • the reference frame can be the previous frame of the frame to be detected in the target video, and in some possible implementations, the reference frame can also be the frame in the target video, located before the frame to be detected and connected to the frame to be detected.
  • the distance between the video frames does not exceed the preset distance, the number of preset distances can be flexibly determined according to the actual situation, and can be one or more frames apart, which is not limited in this embodiment of the present disclosure.
  • the position of the target object in the reference frame may be relatively close to the position of the target object in the frame to be detected.
  • the position information of the target object in the reference frame According to the position information of the target object in the reference frame, the position information of the target object in the frame to be detected can be roughly determined. In this case, more targeted key point detection can be performed on the target object in the frame to be detected, and The amount of detected data will be smaller, so that more accurate key point detection results can be obtained, and the efficiency of key point detection can also be improved.
  • the key point detection method of the target object in the frame to be detected can be flexibly determined according to the actual situation, for example, according to the position of the target object in the reference frame The information is cropped to detect the frame and then the key point detection is carried out, or according to the position information of the target object in the reference frame, the key point detection is directly performed on the image area corresponding to the position in the frame to be detected, etc.
  • Various possible implementation methods can be detailed See the following disclosed embodiments, which will not be expanded here.
  • the key point detection is performed on the target object in the frame to be detected, and the key point detection result is obtained, including:
  • the first position may be the overall position coordinates of the target object in the reference frame.
  • the first position may be the position coordinates of the body frame of the target object in the reference frame.
  • the manner of clipping the frame to be detected according to the first position is also not limited in the embodiments of the present disclosure, and is not limited to the following disclosed embodiments.
  • the first coordinates of the human body frame in the reference frame can be determined according to the first position, and combined with the corresponding relationship between the position coordinates between the reference frame and the frame to be detected, it can be determined that the human body frame of the target object is in the frame to be detected.
  • the second coordinates in the frame are detected, and the frame to be detected is cropped based on the second coordinates to obtain a cropping result.
  • the first coordinates of the body frame in the reference frame and the border length of the body frame can also be determined according to the first position, and combined with the position coordinate correspondence between the reference frame and the frame to be detected, determine The second coordinates of the human body frame of the target object in the frame to be detected, and the frame to be detected is cropped based on the second coordinates and the frame length to obtain a clipping result, wherein, the clipping based on the second coordinates and the frame length can be based on the first
  • the two coordinates determine the position of the clipping endpoint, and the frame length determines the length of the clipping result.
  • the length of the clipping result can be consistent with the frame length.
  • the length of the clipping result can also be proportional to the frame length, such as N times the frame length, etc., N can be any value not less than 1, etc.
  • the target object in the frame to be detected can be preliminarily positioned according to the first position of the target object in the reference frame, and the clipping result can be obtained.
  • the key point detection based on the clipping result can reduce the detection cost on the one hand.
  • the amount of data can improve the detection efficiency.
  • the accuracy of key point detection can be improved.
  • the key point detection is performed on the target object in the frame to be detected, and the key point detection result is obtained, including:
  • the second position may be the position coordinates of the target area of the target object in the reference frame.
  • the target area may include the head area and/or the shoulder area, so in a possible implementation
  • the second position may be the position coordinates of the head and shoulders frame of the target object in the reference frame.
  • the implementation form can be flexibly determined according to the actual situation, for example, it can be realized by performing head and shoulder frame and/or key point recognition on the reference frame, see the following publications for details Embodiment, do not expand here.
  • the key point detection method for the target object in the clipping result can be the same as the key point detection method based on the clipping result obtained at the first position, or it can be different. Do unfold.
  • step S14 Since in step S14, the depth information of the target object is determined according to the characteristic length of the target area, the key point detection result can be obtained according to the second position of the target area of the target object in the reference frame in the embodiment of the present disclosure.
  • the method can focus on the target area more specifically, thereby further reducing the amount of data processing, and obtaining the characteristic length of the target area more accurately, thereby further improving the accuracy and efficiency of depth detection.
  • obtaining the second position of the target area of the target object in the reference frame may include:
  • the second position of the target area in the reference frame is obtained.
  • the first neural network may be any network used to determine the second position, and its implementation form is not limited in the embodiments of the present disclosure.
  • the first neural network may be an object area detection network for identifying the second location of the object area directly from the reference frame.
  • the object area detection network may be faster based on Regional Convolutional Neural Networks (Faster Regions with Convolutional Neural Networks, Faster RCNN); in some possible implementations, the first neural network can also be a key point detection network, which is used to detect one or more key points in the reference frame Points are identified, and then the second position of the target area in the reference frame is determined according to the positions of the identified key points.
  • the reference frame may also be used as the frame to be detected for depth detection.
  • the reference frame may have undergone key point detection and a corresponding key point detection result has been obtained. Therefore, in some possible implementation manners, the second position of the target area in the reference frame may be obtained according to the key point detection result corresponding to the reference frame.
  • the key point detection may also be directly performed on the reference frame to obtain the key point detection result.
  • the key point detection method reference may be made to other disclosed embodiments, which will not be repeated here.
  • the second position of the target area in the reference frame can be flexibly determined in multiple ways according to the actual situation of the reference frame, which improves the flexibility and versatility of depth detection; and in some possible implementations
  • the second position can be determined directly based on the intermediate result of the reference frame in the depth detection, thereby reducing the repeated calculation of data and improving the depth detection. efficiency and precision.
  • the key point detection is performed on the target object in the clipping result to obtain the key point detection result, which may include:
  • the second neural network is used to perform key point detection on the target object in the clipping result to obtain a key point detection result.
  • the second neural network may be any neural network used to realize key point detection, and its implementation mode is not limited in the embodiments of the present disclosure, wherein, when the first neural network may be a key point detection network, the second The second neural network may be implemented in the same or different manner as the first neural network.
  • key point detection may also be performed on the target object in the clipping result through a related key point recognition algorithm, and the key point recognition algorithm to be used is also not limited in the embodiments of the present disclosure.
  • the key point detection results may include head key points, left shoulder key points, and right shoulder key points.
  • FIG. 3 shows a flowchart of a depth detection method according to an embodiment of the present disclosure. As shown in FIG. 3 , in a possible implementation, step S13 may include:
  • Step S131 according to the distance between the key point of the left shoulder and the key point of the right shoulder, the first characteristic length of the target area is obtained.
  • Step S132 obtain the second characteristic length of the target area according to the distance between the key point of the head and the center point of the shoulder, wherein the center point of the shoulder is the middle point between the key point of the left shoulder and the key point of the right shoulder.
  • Step S133 determine the characteristic length of the target area according to the first characteristic length and/or the second characteristic length.
  • the first characteristic length may be a characteristic length reflecting the distance between the shoulders of the target object, and in a possible implementation manner, the first characteristic length may be determined according to the distance between the left shoulder key point and the right shoulder key point .
  • the second characteristic length may be a characteristic length reflecting the distance between the head and shoulders of the target object.
  • the second characteristic length may be determined according to the distance between the head key point and the shoulder center point.
  • the shoulder center point can reflect the center position of the shoulder of the target object.
  • the position of the shoulder center point can be determined according to the positions of the left shoulder key point and the right shoulder key point; in a possible implementation
  • the center point of the shoulder can also be directly used as the key point to be detected, which can be directly obtained from the key point detection result.
  • step S133 the method of determining the characteristic length of the target area according to the first characteristic length and/or the second characteristic length can be flexibly determined according to the actual situation.
  • the first characteristic length and the second characteristic length can be combined
  • the larger value of the length is used as the characteristic length of the target area; in some possible implementations, the smaller value of the first characteristic length and the second characteristic length, or the average value of the two, or Is the ratio of the two, etc., as the characteristic length of the target area.
  • the characteristic length of the target area can be obtained based on the first characteristic length and the second characteristic length that are less interfered by the orientation or posture of the target object, so that the frames to be detected collected at any angle
  • more accurate depth detection results can be obtained, and the stability, robustness and precision of depth detection can be improved.
  • step S14 may include:
  • the depth distance is determined according to the proportional relationship between the preset characteristic length and the characteristic length of the target area, as well as preset equipment parameters.
  • the preset characteristic length may be the actual characteristic length of the target area under normal circumstances, that is, the a priori estimated value of the characteristic length in the above disclosed embodiments.
  • the value of the preset characteristic length can be flexibly changed according to different definitions of the characteristic length, and is not limited to the following disclosed embodiments.
  • the preset characteristic length when the characteristic length is a larger value between the first characteristic length and the second characteristic length, can be set to 25-40cm. In an example, the The preset feature length can be set to 32cm.
  • the preset device parameters may be some calibration parameters of the collection device itself, and the types and types of parameters contained therein may be flexibly determined according to the actual situation of the collection device.
  • the preset device parameters may include an internal reference matrix of the acquisition device, and the internal reference matrix may include one or more focal length parameters of the camera, and principal point positions of one or more cameras.
  • the way to obtain the preset device parameters is not limited in the embodiments of the present disclosure.
  • the preset device parameters can be directly obtained according to the actual situation of the acquisition device.
  • you can also The preset device parameters are obtained by calibrating the acquisition device.
  • the proportional relationship between the preset characteristic length and the characteristic length Based on the proportional relationship between the preset characteristic length and the characteristic length, the proportional relationship between the target object and the target object in the actual scene under normal circumstances can be determined, combined with the preset device parameters, the depth distance of the target object in the actual scene can be determined .
  • the process of calculating the depth distance can be flexibly selected according to actual conditions, and is not limited to the following disclosed embodiments.
  • the process of determining the depth distance according to the preset characteristic length, characteristic length and preset device parameters can be expressed by the following formulas (1) and (2):
  • d is the depth distance
  • C is the preset feature length
  • L is the feature length of the target area
  • f x and fy are the camera internal reference matrix
  • the focal length parameter in , f is the parameter value determined according to the focal length parameter.
  • the proportional relationship between the characteristic length and the relatively stable preset characteristic length can be used to determine the depth distance simply and conveniently in combination with the preset device parameters of the acquisition device. It is more accurate and can improve the accuracy and efficiency of depth detection.
  • the method proposed in the embodiment of the present disclosure may further include:
  • the manner of determining the offset angle can also be flexibly selected, and is not limited to the following disclosed embodiments.
  • the offset angle may be determined according to preset device parameters and the position coordinates of the center point of the head and shoulders in the key point detection result.
  • the center point of the head and shoulders may be the center point of the head and shoulders frame mentioned in the above disclosed embodiments.
  • the position coordinates of the key points of the head, the key points of the left shoulder and the key points of the right shoulder may be Determine the overall position coordinates of the head and shoulders frame, and determine the position coordinates of the center point of the head and shoulders based on the overall position coordinates of the head and shoulders frame; in some possible implementations, the center point of the head and shoulders can also be directly used as the key to be detected point, so that the position coordinates of the center point of the head and shoulders can be directly obtained in the key point detection results.
  • the process of determining the offset angle according to the preset equipment parameters and the position coordinates of the center point of the head and shoulders can be expressed by the following formulas (3) and (4):
  • ⁇ x is the offset angle in the x-axis direction
  • ⁇ y is the offset angle in the y-axis direction
  • (x, y) is the position coordinate of the center point of the head and shoulders
  • f x and f y are the camera internal parameters matrix
  • the focal length parameter in , u 0 and v 0 are the principal point positions in the camera internal reference matrix K.
  • the offset angle can be determined simply and conveniently by using the preset equipment parameters and the key point detection results obtained in the depth detection process. This determination method does not need to obtain additional data, and is easy to calculate, which can improve the depth detection. efficiency and convenience.
  • the method proposed in the embodiment of the present disclosure may further include:
  • the position of the target object in the three-dimensional space is determined.
  • the position of the target object in the three-dimensional space may be the three-dimensional coordinates of the target object in the three-dimensional space.
  • the way to determine the position in the three-dimensional space based on the depth information can be flexibly selected according to the actual situation.
  • the two-dimensional coordinates of the target object in the frame to be detected can be determined according to the key point detection results of the target object.
  • the two-dimensional coordinates are combined with the depth distance and/or offset angle in the depth information, so as to determine the three-dimensional coordinates of the target object in the three-dimensional space.
  • the depth information can be used to perform three-dimensional positioning of the target object, so as to realize various operations such as interaction with the target object.
  • the distance and angle between the target object and the smart air conditioner can be determined according to the position of the target object in three-dimensional space, so as to dynamically adjust the wind direction and/or wind speed of the smart air conditioner; in some possible
  • the target object can also be positioned in the game scene based on the position of the target object in the three-dimensional space in the AR game platform, so that the human-computer interaction in the AR scene can be realized more realistically and naturally.
  • the present disclosure also provides depth detection devices, electronic equipment, computer-readable storage media, and programs, all of which can be used to implement any depth detection method provided by the present disclosure.
  • depth detection devices electronic equipment, computer-readable storage media, and programs, all of which can be used to implement any depth detection method provided by the present disclosure.
  • FIG. 4 shows a block diagram of a depth detection device according to an embodiment of the present disclosure.
  • device 20 includes:
  • the obtaining module 21 is configured to obtain a frame to be detected, where the frame to be detected contains a target object.
  • the key point detection module 22 is configured to perform key point detection on the target object according to the frame to be detected, and obtain a key point detection result.
  • the characteristic length determination module 23 is used to determine the characteristic length of the target region in the target object based on the key point detection result, wherein the target region includes the head region and/or the shoulder region, and the characteristic length is used to characterize the target region in the target object. Size Information.
  • the depth detection module 24 is configured to determine the depth information of the target object in the frame to be detected according to the characteristic length of the target area.
  • the key point detection module is used to: perform key point detection on the target object in the frame to be detected according to the position information of the target object in the reference frame, and obtain the key point detection result, wherein the reference frame is The video frame before the frame to be detected in the target video to which the frame to be detected belongs.
  • the key point detection module is further configured to: clip the frame to be detected according to the first position of the target object in the reference frame to obtain a clipping result; perform key point detection on the target object in the clipping result, Get key point detection results.
  • the key point detection module is further used to: obtain the second position of the target area of the target object in the reference frame; according to the second position, the frame to be detected is cropped to obtain the cropping result; the cropping result The key point detection is performed on the target object in , and the key point detection result is obtained.
  • the key point detection module is further configured to: use the first neural network to identify the target area in the reference frame to obtain the second position output by the first neural network; and/or, according to the reference frame The corresponding key point detection result obtains the second position of the target area in the reference frame.
  • the key point detection results include head key points, left shoulder key points, and right shoulder key points; the feature length determination module is used to: obtain The first characteristic length of the target area; the second characteristic length of the target area is obtained according to the distance between the head key point and the shoulder center point, where the shoulder center point is the middle point between the left shoulder key point and the right shoulder key point ; Determine the characteristic length of the target area according to the first characteristic length and/or the second characteristic length.
  • the depth information includes a depth distance
  • the depth distance includes a distance between the target object and the optical center of the acquisition device
  • the acquisition device includes a device for image acquisition of the target object
  • the depth detection module is used to: acquire The preset characteristic length of the target area, and the preset device parameters of the acquisition device; according to the proportional relationship between the preset characteristic length and the characteristic length of the target area, and the preset device parameters, the depth distance is determined.
  • the depth information includes an offset angle
  • the offset angle includes a spatial angle of the target object relative to the optical axis of the acquisition device
  • the acquisition device includes an image acquisition device for the target object
  • the device is also used for: Obtain the preset device parameters of the acquisition device; determine the offset angle according to the preset device parameters and key point detection results.
  • the device is further configured to: determine the position of the target object in the three-dimensional space according to the depth information of the target object.
  • the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the methods described in the method embodiments above, and its specific implementation can refer to the description of the method embodiments above. For brevity, here No longer.
  • Fig. 5 shows a schematic diagram of an application example according to the present disclosure.
  • the application example of the present disclosure proposes a depth detection method, which may include the following process:
  • Step S31 using the Faster RCNN neural network to detect the head and shoulders frame of the human body on the first frame of the target video, and obtain the position of the head and shoulders frame in the first frame.
  • Step S32 starting from the second frame of the target video, using the video frame as the frame to be detected, using the previous frame of the frame to be detected as the reference frame, and passing the key point detection network according to the second position of the head and shoulders frame in the reference frame Carry out key point detection on the frame to be detected, obtain the position coordinates of the three key points of the head key point, left shoulder key point, and right shoulder key point, and use the circumscribed rectangle of the three key points as the head and shoulder frame in the frame to be detected .
  • Step S33 determining a feature length L in the frame to be detected, wherein the feature length L may be a larger value among the first feature length and the second feature length.
  • the first characteristic length may be the length of the line segment between the left shoulder key point and the right shoulder key point
  • the second characteristic length may be the length of the line segment between the shoulder center point and the head key point.
  • Step S34 according to one or more of the feature length L, the preset feature length C, the center point of the head and shoulders frame in the frame to be detected, and the camera intrinsic parameter matrix K, determine the depth information of the target object:
  • the depth distance can be calculated according to the formulas (1) and (2) in the above disclosed embodiments according to the characteristic length L obtained in step S33, the preset characteristic length C, and the camera internal reference matrix K d;
  • the depth information can also be calculated according to the position (x, y) of the center point of the head and shoulders frame in the frame to be detected and the camera internal reference matrix K, through the formulas (3) and (4) in the above disclosed embodiments The offset angle in .
  • the frame next to the frame to be detected in the target video may be used as the frame to be detected, and the depth detection is performed again in step S32.
  • the characteristic length defined based on the three key points of the top of the head, left shoulder, and right shoulder can be used as the basis for depth estimation.
  • the characteristic length is less disturbed by the orientation and posture of the human body. In complex scenes such as facing away from the camera or partially occluded, it can achieve more accurate and robust depth detection, more applicable scenes, and more stable ranging results.
  • the writing order of each step does not mean a strict execution order and constitutes any limitation on the implementation process.
  • the specific execution order of each step should be based on its function and possible
  • the inner logic is OK.
  • Embodiments of the present disclosure also provide a computer-readable storage medium, on which computer program instructions are stored, and the above-mentioned method is implemented when the computer program instructions are executed by a processor.
  • the computer readable storage medium may be a non-volatile computer readable storage medium or a volatile computer readable storage medium.
  • An embodiment of the present disclosure also proposes an electronic device, including: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to invoke the instructions stored in the memory to execute the above method.
  • An embodiment of the present disclosure also provides a computer program product, including computer readable codes.
  • the processor in the device executes the method for implementing the depth detection method provided in any of the above embodiments. instruction.
  • the embodiments of the present disclosure also provide another computer program product, which is used for storing computer-readable instructions, and when the instructions are executed, the computer executes the operation of the depth detection method provided by any of the above-mentioned embodiments.
  • Electronic devices may be provided as terminals, servers, or other forms of devices.
  • FIG. 6 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • the electronic device 800 may be a terminal such as a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, or a personal digital assistant.
  • electronic device 800 may include one or more of the following components: processing component 802, memory 804, power supply component 806, multimedia component 808, audio component 810, input/output (I/O) interface 812, sensor component 814 , and the communication component 816.
  • the processing component 802 generally controls the overall operations of the electronic device 800, such as those associated with display, telephone calls, data communications, camera operations, and recording operations.
  • the processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above method. Additionally, processing component 802 may include one or more modules that facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802 .
  • the memory 804 is configured to store various types of data to support operations at the electronic device 800 . Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phonebook data, messages, pictures, videos, and the like.
  • the memory 804 can be implemented by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic or Optical Disk Magnetic Disk
  • the power supply component 806 provides power to various components of the electronic device 800 .
  • Power components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for electronic device 800 .
  • the multimedia component 808 includes a screen providing an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user.
  • the touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensor may not only sense a boundary of a touch or swipe action, but also detect duration and pressure associated with the touch or swipe action.
  • the multimedia component 808 includes a front camera and/or a rear camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capability.
  • the audio component 810 is configured to output and/or input audio signals.
  • the audio component 810 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device 800 is in operation modes, such as call mode, recording mode and voice recognition mode. Received audio signals may be further stored in memory 804 or sent via communication component 816 .
  • the audio component 810 also includes a speaker for outputting audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, and the like. These buttons may include, but are not limited to: a home button, volume buttons, start button, and lock button.
  • Sensor assembly 814 includes one or more sensors for providing status assessments of various aspects of electronic device 800 .
  • the sensor component 814 can detect the open/closed state of the electronic device 800, the relative positioning of components, such as the display and the keypad of the electronic device 800, the sensor component 814 can also detect the electronic device 800 or a Changes in position of components, presence or absence of user contact with electronic device 800 , electronic device 800 orientation or acceleration/deceleration and temperature changes in electronic device 800 .
  • Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact.
  • Sensor assembly 814 may also include an optical sensor, such as a complementary metal-oxide-semiconductor (CMOS) or charge-coupled device (CCD) image sensor, for use in imaging applications.
  • CMOS complementary metal-oxide-semiconductor
  • CCD charge-coupled device
  • the sensor component 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor or a temperature sensor.
  • the communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices.
  • the electronic device 800 can access a wireless network based on a communication standard, such as a wireless network (WiFi), a second generation mobile communication technology (2G) or a third generation mobile communication technology (3G), or a combination thereof.
  • the communication component 816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication.
  • the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID Radio Frequency Identification
  • IrDA Infrared Data Association
  • UWB Ultra Wide Band
  • Bluetooth Bluetooth
  • electronic device 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A programmable gate array (FPGA), controller, microcontroller, microprocessor or other electronic component implementation for performing the methods described above.
  • ASICs application specific integrated circuits
  • DSPs digital signal processors
  • DSPDs digital signal processing devices
  • PLDs programmable logic devices
  • FPGA field programmable A programmable gate array
  • controller microcontroller, microprocessor or other electronic component implementation for performing the methods described above.
  • a non-volatile computer-readable storage medium such as the memory 804 including computer program instructions, which can be executed by the processor 820 of the electronic device 800 to implement the above method.
  • FIG. 7 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • the electronic device 1900 may be provided as a server.
  • electronic device 1900 includes processing component 1922 , which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions executable by processing component 1922 , such as application programs.
  • the application programs stored in memory 1932 may include one or more modules each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above method.
  • Electronic device 1900 may also include a power supply component 1926 configured to perform power management of electronic device 1900, a wired or wireless network interface 1950 configured to connect electronic device 1900 to a network, and an input-output (I/O) interface 1958 .
  • the electronic device 1900 can operate based on the operating system stored in the memory 1932, such as the Microsoft server operating system (Windows Server TM ), the graphical user interface-based operating system (Mac OS X TM ) introduced by Apple Inc., and the multi-user and multi-process computer operating system (Unix TM ), a free and open source Unix-like operating system (Linux TM ), an open source Unix-like operating system (FreeBSD TM ), or the like.
  • Microsoft server operating system Windows Server TM
  • Mac OS X TM graphical user interface-based operating system
  • Unix TM multi-user and multi-process computer operating system
  • Linux TM free and open source Unix-like operating system
  • FreeBSD TM open source Unix-like operating system
  • a non-transitory computer-readable storage medium such as a memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to implement the above method.
  • the present disclosure can be a system, method and/or computer program product.
  • a computer program product may include a computer readable storage medium having computer readable program instructions thereon for causing a processor to implement various aspects of the present disclosure.
  • a computer readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device.
  • a computer readable storage medium may be, for example, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Computer-readable storage media include: portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory), static random access memory (SRAM), compact disc read only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanically encoded device, such as a printer with instructions stored thereon A hole card or a raised structure in a groove, and any suitable combination of the above.
  • RAM random access memory
  • ROM read-only memory
  • EPROM erasable programmable read-only memory
  • flash memory static random access memory
  • SRAM static random access memory
  • CD-ROM compact disc read only memory
  • DVD digital versatile disc
  • memory stick floppy disk
  • mechanically encoded device such as a printer with instructions stored thereon
  • a hole card or a raised structure in a groove and any suitable combination of the above.
  • computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., pulses of light through fiber optic cables), or transmitted electrical signals.
  • Computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or downloaded to an external computer or external storage device over a network, such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • a network adapter card or a network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
  • Computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or Source or object code written in any combination, including object-oriented programming languages—such as Smalltalk, C++, etc., and conventional procedural programming languages—such as the “C” language or similar programming languages.
  • Computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement.
  • the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as via the Internet using an Internet service provider). connect).
  • LAN local area network
  • WAN wide area network
  • an electronic circuit such as a programmable logic circuit, field programmable gate array (FPGA), or programmable logic array (PLA)
  • FPGA field programmable gate array
  • PDA programmable logic array
  • These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine such that when executed by the processor of the computer or other programmable data processing apparatus , producing an apparatus for realizing the functions/actions specified in one or more blocks in the flowchart and/or block diagram.
  • These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause computers, programmable data processing devices and/or other devices to work in a specific way, so that the computer-readable medium storing instructions includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks in flowcharts and/or block diagrams.
  • each block in a flowchart or block diagram may represent a module, a portion of a program segment, or an instruction that includes one or more Executable instructions.
  • the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions.
  • the computer program product can be specifically realized by means of hardware, software or a combination thereof.
  • the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK) etc. Wait.
  • a software development kit Software Development Kit, SDK

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)

Abstract

本公开涉及一种深度检测方法及装置、电子设备和存储介质，所述方法包括：获取待检测帧，所述待检测帧包含目标对象；根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果；基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，其中，所述目标区域包括头部区域和/或肩部区域，所述特征长度用于表征所述目标对象中目标区域的尺寸信息；根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息。

Description

深度检测方法及装置、电子设备和存储介质
本申请要求在2021年6月28日提交中国专利局、申请号为202110719313.2、发明名称为“深度检测方法及装置、电子设备和存储介质”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。
技术领域
本公开涉及计算机技术领域，尤其涉及一种深度检测方法及装置、电子设备和存储介质。
背景技术
深度信息可以反映图像中的人体相对于图像采集设备的距离，基于深度信息，可以对图像中的人体对象进行空间定位。
单目摄像头是一种较为常见和广泛应用的图像采集设备，然而单目摄像头往往仅能提供二维信息，如何基于单目摄像头采集的图像，较为准确地确定图像中人体的深度信息，成为目前一个亟待解决的问题。
发明内容
本公开提出了一种深度检测的技术方案。
根据本公开的一方面，提供了一种深度检测方法，包括：
获取待检测帧，所述待检测帧包含目标对象；根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果；基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，其中，所述目标区域包括头部区域和/或肩部区域，所述特征长度用于表征所述目标对象中目标区域的尺寸信息；根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息。
在一种可能的实现方式中，所述根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果，包括：根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，其中，所述参考帧为所述待检测帧所属的目标视频中，位于所述待检测帧之前的视频帧。
在一种可能的实现方式中，所述根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，包括：根据所述参考帧中所述目标对象的第一位置，对所述待检测帧进行裁剪，得到裁剪结果；对所述裁剪结果中的目标对象进行关键点检测，得到所述关键点检测结果。
在一种可能的实现方式中，所述根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，包括：获取所述目标对象的目标区域在所述参考帧中的第二位置；根据所述第二位置，对所述待检测帧进行裁剪，得到裁剪结果；对所述裁剪结果中的目标对象进行关键点检测，得到所述关键点检测结果。
在一种可能的实现方式中，所述获取所述目标对象的目标区域在所述参考帧中的第二位置，包括：通过第一神经网络对所述参考帧中的目标区域进行识别，得到所述第一神经网络输出的第二位置；和/或，根据所述参考帧对应的关键点检测结果，得到所述目标区域在所述参考帧中的第二位置。
在一种可能的实现方式中，所述关键点检测结果包括头部关键点、左肩关键点以及右肩关键点；所述基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，包括：根据所述左肩关键点与所述右肩关键点之间的距离，获取所述目标区域的第一特 征长度；根据所述头部关键点与肩部中心点之间的距离，获取所述目标区域的第二特征长度，其中，所述肩部中心点为所述左肩关键点与所述右肩关键点的中间点；根据所述第一特征长度和/或所述第二特征长度，确定所述目标区域的特征长度。
在一种可能的实现方式中，所述深度信息包括深度距离，所述深度距离包括所述目标对象与采集设备的光心之间的距离，所述采集设备包括对所述目标对象进行图像采集的设备；所述根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息，包括：获取所述目标区域的预设特征长度，以及所述采集设备的预设设备参数；根据所述预设特征长度与所述目标区域的特征长度之间的比例关系，以及所述预设设备参数，确定所述深度距离。
在一种可能的实现方式中，所述深度信息包括偏移角度，所述偏移角度包括所述目标对象相对于采集设备的光轴的空间角度，所述采集设备包括对所述目标对象进行图像采集的设备；所述方法还包括：获取所述采集设备的预设设备参数；根据所述预设设备参数以及所述关键点检测结果，确定所述偏移角度。
在一种可能的实现方式中，所述方法还包括：根据所述目标对象的深度信息，确定所述目标对象在三维空间中的位置。
根据本公开的一方面，提供了一种深度检测装置，包括：
获取模块，用于获取待检测帧，所述待检测帧包含目标对象；关键点检测模块，用于根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果；特征长度确定模块，用于基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，其中，所述目标区域包括头部区域和/或肩部区域，所述特征长度用于表征所述目标对象中目标区域的尺寸信息；深度检测模块，用于根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息。
在一种可能的实现方式中，所述关键点检测模块用于：根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，其中，所述参考帧为所述待检测帧所属的目标视频中，位于所述待检测帧之前的视频帧。
在一种可能的实现方式中，所述关键点检测模块进一步用于：根据所述参考帧中所述目标对象的第一位置，对所述待检测帧进行裁剪，得到裁剪结果；对所述裁剪结果中的目标对象进行关键点检测，得到所述关键点检测结果。
在一种可能的实现方式中，所述关键点检测模块进一步用于：获取所述目标对象的目标区域在所述参考帧中的第二位置；根据所述第二位置，对所述待检测帧进行裁剪，得到裁剪结果；对所述裁剪结果中的目标对象进行关键点检测，得到所述关键点检测结果。
在一种可能的实现方式中，所述关键点检测模块进一步用于：通过第一神经网络对所述参考帧中的目标区域进行识别，得到所述第一神经网络输出的第二位置；和/或，根据所述参考帧对应的关键点检测结果，得到所述目标区域在所述参考帧中的第二位置。
在一种可能的实现方式中，所述关键点检测结果包括头部关键点、左肩关键点以及右肩关键点；所述特征长度确定模块用于：根据所述左肩关键点与所述右肩关键点之间的距离，获取所述目标区域的第一特征长度；根据所述头部关键点与肩部中心点之间的距离，获取所述目标区域的第二特征长度，其中，所述肩部中心点为所述左肩关键点与所述右肩关键点的中间点；根据所述第一特征长度和/或所述第二特征长度，确定所述目标区域的特征长度。
在一种可能的实现方式中，所述深度信息包括深度距离，所述深度距离包括所述目标对象与采集设备的光心之间的距离，所述采集设备包括对所述目标对象进行图像采集的设备；所述深度检测模块用于：获取所述目标区域的预设特征长度，以及所述采集设 备的预设设备参数；根据所述预设特征长度与所述目标区域的特征长度之间的比例关系，以及所述预设设备参数，确定所述深度距离。
在一种可能的实现方式中，所述深度信息包括偏移角度，所述偏移角度包括所述目标对象相对于采集设备的光轴的空间角度，所述采集设备包括对所述目标对象进行图像采集的设备；所述装置还用于：获取所述采集设备的预设设备参数；根据所述预设设备参数以及所述关键点检测结果，确定所述偏移角度。
在一种可能的实现方式中，所述装置还用于：根据所述目标对象的深度信息，确定所述目标对象在三维空间中的位置。
根据本公开的一方面，提供了一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为调用所述存储器存储的指令，以执行上述方法。
根据本公开的一方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。
根据本公开的一方面，提供了一种计算机程序产品，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行上述方法。
在本公开实施例中，通过根据包含目标对象的待检测帧，对目标对象进行关键点检测以得到关键点检测结果，并基于关键点检测结果确定目标对象中目标区域的特征长度，从而根据该特征长度确定目标对象的深度信息，通过本公开实施例，可以基于目标对象中目标区域的特征长度，来对待检测帧中的目标对象进行深度估计，由于该特征长度用于表征目标对象头部区域和/或肩部区域的尺寸信息，在目标对象的朝向或姿势改变的情况下，该特征长度也不易受到干扰，数值较为稳定，因此基于该特征长度所得到的深度信息可以更加准确，提升了深度检测的精度和鲁棒性。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。
图1示出根据本公开实施例的深度检测方法的流程图。
图2示出根据本公开实施例的目标区域的示意图。
图3示出根据本公开实施例的深度检测方法的流程图。
图4示出根据本公开实施例的深度检测装置的框图。
图5示出根据本公开一应用示例的示意图。
图6示出根据本公开实施例的电子设备的框图。
图7示出根据本公开实施例的电子设备的框图。
具体实施方式
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。 另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
另外，为了更好地说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。
图1示出根据本公开实施例的深度检测方法的流程图。该方法可以由深度检测装置执行，深度检测装置可以是终端设备或服务器等电子设备，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字助理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。在一些可能的实现方式中，该方法可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。或者，可以通过服务器执行该方法。如图1所示，该方法可以包括：
步骤S11，获取待检测帧，待检测帧包含目标对象。
其中，待检测帧可以是具有深度检测需求的任意图像帧，比如可以是从拍摄的视频中提取的图像帧，或是拍摄图像得到的图像帧等。待检测帧的数量在本公开实施例中不做限制，可以包含一帧或多帧，在待检测帧包含多帧的情况下，可以根据本公开实施例提出的深度检测方法对多帧待检测帧同时进行深度检测，也可以按照一定的顺序，对多帧待检测帧依次进行深度检测等。
待检测帧中包含待进行深度检测的目标对象，目标对象的类型在本公开实施例中不做限制，可以包括各类人物对象、动物对象或是部分机械对象，比如机器人等。后续各公开实施例均以目标对象为人物对象为例进行说明，目标对象为其他类型的实现方式可以参考后续各公开实施例进行灵活扩展，不再一一阐述。
待检测帧中包含的目标对象数量在本公开实施例中同样不做限制，可以包含一个或多个目标对象，根据实际情况灵活决定。
获取待检测帧的方式在本公开实施例中也不做限制，在一种可能的实现方式中，可以从视频中进行帧提取以得到一帧或多帧待检测帧，其中，帧提取可以包括逐帧提取、按照一定的间隔进行帧采样或是随机帧采样等一种或多种方式。在一种可能的实现方式中，也可以对目标对象进行图像采集来得到待检测帧；在一些可能的实现方式中，还可以从数据库中读取以得到待检测帧等。
步骤S12，根据待检测帧，对目标对象进行关键点检测，得到关键点检测结果。
其中，关键点检测结果可以包括检测到的关键点在待检测帧中的位置。其中，检测到的关键点数量和类型可以根据实际情况灵活决定，在一些可能的实现方式中，检测到的关键点数量可以包括2～150个等，在一个示例中，检测到的关键点可以包含人体的14个肢体关键点(如头部关键点、肩部关键点、颈部关键点、手肘关键点、手腕关键点、胯部关键点、腿部关键点以及足部关键点等)，或是包含人体外围轮廓上的59个轮廓关键点(如头部外围或是肩部外围上的一些关键点)等。在一种可能的实现方式中，为了减小计算量，检测到的关键点也可以仅包含头部关键点、左肩关键点以及右肩关键点共三个关键点。
关键点检测的方式可以根据实际情况灵活决定，在一种可能的实现方式中，可以将待检测帧输入具有关键点检测功能的任意神经网络以实现关键点检测；在一些可能的实现方式中，也可以通过相关的关键点识别算法，对待检测帧进行关键点识别以得到关键点检测结果；在一些可能的实现方式中，还可以根据目标对象在待检测帧中的位置，对待检测帧中的部分图像区域进行关键点检测，以得到关键点检测结果等。步骤S12的一些可能的具体实现方式可以详见下述各公开实施例，在此先不做展开。
步骤S13，基于关键点检测结果，确定目标对象中目标区域的特征长度。
其中，目标区域可以包括头部区域和/或肩部区域，目标对象的头部区域可以是目标对象头部所在的区域，比如头部关键点和颈部关键点之间所构成的区域；肩部区域则可以是目标对象肩颈部所在的区域，比如颈部关键点和肩部关键点之间所构成的区域。
图2示出根据本公开实施例的目标区域的示意图，如图2所示，在一种可能的实现方式中，在目标区域包括头部区域和肩部区域的情况下，可以将头部关键点、左肩关键点和右肩关键点连接而成的头肩框，作为目标区域。在一个示例中，头肩框可以是如图2所示的矩形，从图2中可以看出，头肩框可以通过连接目标对象头部顶点的头部关键点、左肩关节处的左肩关键点和右肩关节处的右肩关键点所得到。在一个示例中，头肩框也可以为其他性状，比如多边形、圆形或是其他不规则的形状等。
目标区域的特征长度可以用于表征目标对象中目标区域的尺寸信息，该尺寸信息可以反映目标区域内较为稳定的特征，故表征该尺寸信息的特征长度受到待检测帧中目标对象的朝向或是姿态的影响较小，数值较为稳定。
尺寸信息包含的内容可以根据实际情况灵活决定，不局限于下述各公开实施例。在一些可能的实现方式中，该尺寸信息可以包含目标对象某些特征稳定的部位的尺寸，比如头肩部位之间的尺寸和/或肩部之间的尺寸等；在一些可能的实现方式中，该尺寸信息也可以是多个特征稳定的部位的尺寸中的较大值或较小值等；在一些可能的实现方式中，该尺寸信息还可以包含多个特征稳定的部位之间的尺寸比值等，比如头肩部位之间的尺寸与肩部之间的尺寸比值，或是肩部之间的尺寸与头肩部位之间的尺寸比值等。
随着表征的尺寸信息的不同，根据关键点检测结果确定特征长度的方式也可以随之发生变化，详见下述各公开实施例，在此先不做展开。
步骤S14，根据目标区域的特征长度，确定待检测帧中目标对象的深度信息。
其中，深度信息包含的信息内容可以根据实际情况灵活决定，任何可以反映目标对象在三维空间中的深度情况的信息，均可以作为深度信息的实现方式。在一种可能的实现方式中，深度信息可以包括深度距离和/或偏移角度。
深度距离可以是目标对象与采集设备之间的距离，采集设备可以是对目标对象进行图像采集的任意设备，在一些可能的实现方式中，该采集设备可以是静态图像的采集设备，如照相机等；在一些可能的实现方式中，该采集设备也可以是采集动态图像的设备，比如摄像机或是摄像头等。
深度距离可以是目标对象与采集设备之间的距离，该距离可以是目标对象与采集设备整体之间的距离，也可以是目标对象与采集设备的某个设备部件之间的距离，在一些可能的实现方式中，可以将目标对象与采集设备的光心之间的距离，作为深度距离。
偏移角度可以是目标对象相对于采集设备的偏移角度，在一种可能的实现方式中，该偏移角度可以是目标对象相对于采集设备的光轴的空间角度。
由于目标区域的特征长度的数值较为稳定，因此可以较为容易地确定该特征长度对应的实际长度，作为该特征长度的先验估计值。该先验估计值可以作为参考信息，结合特征长度，实现待检测帧的单目测距，得到待检测帧的深度信息。其中，基于特征长度确定深度信息的方式可以根据实际情况灵活决定，任何单目测距的实现方式均可以用于步骤S14的实现过程中，比如可以根据特征长度以及特征长度的先验估计值，结合待检测帧在采集过程中的一些相关参数来共同计算深度距离等。步骤S14的具体实现过程可以详见下述各公开实施例，在此先不做展开。
在本公开实施例中，通过根据包含目标对象的待检测帧，对目标对象进行关键点检测以得到关键点检测结果，并基于关键点检测结果确定目标对象中目标区域的特征长度，从而根据该特征长度确定目标对象的深度信息，通过本公开实施例，可以基于目标对象中目标区域的特征长度，来对待检测帧中的目标对象进行深度估计，由于该特征长度用 于表征目标对象头部区域和/或肩部区域的尺寸信息，在目标对象的朝向或姿势改变的情况下，该特征长度也不易受到干扰，数值较为稳定，因此基于该特征长度所得到的深度信息可以更加准确，提升了深度检测的精度和鲁棒性。
在一种可能的实现方式中，步骤S12可以包括：
根据目标对象在参考帧中的位置信息，对待检测帧中的目标对象进行关键点检测，得到关键点检测结果。
其中，参考帧可以是目标视频中位于待检测帧之前的视频帧，目标视频可以是包含待检测帧的视频。在一些可能的实现方式中，参考帧可以是目标视频中待检测帧的前一帧，在一些可能的实现方式中，参考帧也可以是目标视频中，位于待检测帧以前且与待检测帧之间的距离不超过预设距离的视频帧，预设距离的数量可以根据实际情况灵活决定，可以是间隔一帧或多帧等，在本公开实施例中不做限定。
由于参考帧位于待检测帧之前，且与待检测帧的距离不超过预设距离，因此参考帧中目标对象的位置，和待检测帧中目标对象的位置可能较为接近，在这种情况下，根据目标对象在参考帧中的位置信息，可以大致确定出待检测帧中目标对象的位置信息，在这种情况下，可以对待检测帧中的目标对象进行更有针对性的关键点检测，且检测的数据量也会较小，从而可以得到更为准确的关键点检测结果，也可以提升关键点检测的效率。
在一些可能的实现方式中，根据目标对象在参考帧中的位置信息，对待检测帧中的目标对象进行关键点检测的方式可以根据实际情况灵活决定，比如可以根据目标对象在参考帧中的位置信息对待检测帧进行裁剪后再进行关键点检测，或是根据目标对象在参考帧中的位置信息，直接对待检测帧中对应位置的图像区域进行关键点检测等，各种可能的实现方式可以详见下述各公开实施例，在此先不做展开。
通过本公开实施例，可以根据目标对象在参考帧中的位置信息，对待检测帧实现更有针对性的关键点检测，提升关键点检测的效率和精度，从而提升深度检测方法的效率和精度。
在一种可能的实现方式中，根据目标对象在参考帧中的位置信息，对待检测帧中的目标对象进行关键点检测，得到关键点检测结果，包括：
根据所述参考帧中所述目标对象的第一位置，对待检测帧进行裁剪，得到裁剪结果；
对裁剪结果中的目标对象进行关键点检测，得到关键点检测结果。
其中，第一位置可以是参考帧中目标对象整体的位置坐标，比如在目标对象为人物对象的情况下，该第一位置可以是目标对象的人体框在参考帧中的位置坐标。
根据第一位置对待检测帧进行裁剪的方式在本公开实施例中同样不做限制，不局限于下述各公开实施例。在一种可能的实现方式中，可以根据第一位置，确定参考帧中人体框的第一坐标，并结合参考帧和待检测帧之间的位置坐标对应关系，确定目标对象的人体框在待检测帧中的第二坐标，基于该第二坐标对待检测帧进行裁剪以得到裁剪结果。
在一些可能的实现方式中，也可以根据第一位置，确定参考帧中人体框的第一坐标，以及人体框的边框长度，并结合参考帧和待检测帧之间的位置坐标对应关系，确定目标对象的人体框在待检测帧中的第二坐标，基于该第二坐标和边框长度来对待检测帧进行裁剪以得到裁剪结果，其中，基于第二坐标和边框长度的裁剪，可以是根据第二坐标确定裁剪端点的位置，并边框长度确定裁剪结果的长度，在一个示例中，裁剪结果的长度可以与边框长度一致，在一个示例中，裁剪结果的长度也可以与边框长度成比例，比如为边框长度的N倍等，N可以为不小于1的任意数值等。
对裁剪结果中的目标对象进行关键点检测的方式可以根据实际情况灵活决定，详见下述各公开实施例，在此先不做展开。
通过本公开实施例，可以根据参考帧中目标对象的第一位置，对待检测帧中的目标 对象进行初步定位，得到裁剪结果，基于该裁剪结果进行的关键点检测，一方面可以减小检测的数据量，提高检测效率，另一方面由于裁剪后目标对象在裁剪结果中所占的比例较大，因此可以提升关键点检测的精度。
在一种可能的实现方式中，根据目标对象在参考帧中的位置信息，对待检测帧中的目标对象进行关键点检测，得到关键点检测结果，包括：
获取目标对象的目标区域在参考帧中的第二位置；
根据第二位置，对待检测帧进行裁剪，得到裁剪结果；
对裁剪结果中的目标对象进行关键点检测，得到关键点检测结果。
其中，第二位置可以是目标对象的目标区域在参考帧中的位置坐标，如上述各公开实施例所述，目标区域可以包括头部区域和/或肩部区域，故在一种可能的实现方式中，该第二位置可以是目标对象的头肩框在参考帧中的位置坐标。
如何确定目标区域在参考帧中的第二位置，其实现形式可以根据实际情况灵活决定，比如可以通过对参考帧进行头肩框和/或关键点识别等方式进行实现，详见下述各公开实施例，在此先不做展开。
根据第二位置对待检测帧进行裁剪的方式，可以参考根据第一位置对待检测帧进行裁剪的方式，在此不再赘述。
对裁剪结果中的目标对象进行关键点检测的方式，可以与根据第一位置所得到的裁剪结果进行关键点检测的方式相同，也可以不同，详见下述各公开实施例，在此先不做展开。
由于步骤S14中，是根据目标区域的特征长度来确定目标对象的深度信息，因此本公开实施例中可以根据参考帧中目标对象的目标区域所在的第二位置来得到关键点检测结果，这种方式可以更为针对性地关注目标区域，从而进一步减小数据的处理量，更为精确地得到目标区域的特征长度，从而更进一步地提升了深度检测的精度和效率。
在一种可能的实现方式中，获取目标对象的目标区域在参考帧中的第二位置，可以包括：
通过第一神经网络对参考帧中的目标区域进行识别，得到第一神经网络输出的第二位置；和/或，
根据参考帧对应的关键点检测结果，得到目标区域在参考帧中的第二位置。
其中，第一神经网络可以是用于确定第二位置的任意网络，其实现形式在本公开实施例中不做限制。在一些可能的实现方式中，第一神经网络可以是目标区域检测网络，用于直接从参考帧中识别目标区域的第二位置，在一个示例中，该目标区域检测网络可以是更快的基于区域的卷积神经网络(Faster Regions with Convolutional Neural Networks，Faster RCNN)；在一些可能的实现方式中，第一神经网络也可以是关键点检测网络，用于对参考帧中的一个或多个关键点进行识别，继而根据识别到的关键点位置，确定参考帧中目标区域的第二位置。
在一些可能的实现方式中，参考帧也可能作为待检测帧进行深度检测，在这种情况下，参考帧可能已经经历过关键点检测并得到对应的关键点检测结果。因此，在一些可能的实现方式中，可以根据参考帧对应的关键点检测结果，来得到目标区域在参考帧中的第二位置。
在一些可能的实现方式中，也可以直接对参考帧进行关键点检测以得到关键点检测结果，关键点检测的方式可以参考其他各公开实施例，在此不再赘述。
通过本公开实施例，可以根据参考帧的实际情况，灵活地采用多种方式确定目标区域在参考帧中的第二位置，提升了深度检测的灵活性和通用性；而且在一些可能的实现方式中，在位于待检测帧以前的参考帧参与过深度检测的情况下，可以直接基于参考帧在深度检测中得到的中间结果来确定第二位置，从而减小数据的重复计算，提升深度检 测的效率和精度。
在一种可能的实现方式中，对裁剪结果中的目标对象进行关键点检测，得到关键点检测结果，可以包括：
通过第二神经网络对裁剪结果中的目标对象进行关键点检测，得到关键点检测结果。
其中，第二神经网络可以是用于实现关键点检测的任意神经网络，其实现方式在本公开实施例中不做限制，其中，在第一神经网络可以是关键点检测网络的情况下，第二神经网络可以与第一神经网络的实现方式相同或不同。
在一些可能的实现方式中，也可以通过相关的关键点识别算法对裁剪结果中的目标对象进行关键点检测，采用何种关键点识别算法在本公开实施例中同样不做限制。
在一些可能的实现方式中，关键点检测结果可以包括头部关键点、左肩关键点以及右肩关键点。图3示出根据本公开实施例的深度检测方法的流程图，如图3所示，在一种可能的实现方式中，步骤S13可以包括：
步骤S131，根据左肩关键点与右肩关键点之间的距离，获取目标区域的第一特征长度。
步骤S132，根据头部关键点与肩部中心点之间的距离，获取目标区域的第二特征长度，其中，肩部中心点为左肩关键点与右肩关键点的中间点。
步骤S133，根据第一特征长度和/或第二特征长度，确定目标区域的特征长度。
其中，第一特征长度可以是反映目标对象肩部之间距离的特征长度，在一种可能的实现方式中，该第一特征长度可以根据左肩关键点和右肩关键点之间的距离所确定。
第二特征长度可以是反映目标对象头肩之间距离的特征长度，在一种可能的实现方式中，该第二特征长度可以根据头部关键点与肩部中心点之间的距离所确定。
肩部中心点可以反映目标对象肩部的中心位置，在一种可能的实现方式中，可以根据左肩关键点和右肩关键点的位置，确定肩部中心点的位置；在一种可能的实现方式中，肩部中心点也可以直接作为被检测的关键点，从关键点检测结果中直接得到。
步骤S133中，根据第一特征长度和/或第二特征长度确定目标区域的特征长度的方式可以根据实际情况灵活决定，在一种可能的实现方式中，可以将第一特征长度和第二特征长度中的较大值，作为目标区域的特征长度；在一些可能的实现方式中，也可以将第一特征长度和第二特征长度中的较小值，或是二者的平均值，亦或是二者的比值等，作为目标区域的特征长度。
通过本公开实施例，可以基于受到目标对象的朝向或是姿态干扰较小的第一特征长度与第二特征长度，来得到目标区域的特征长度，使得在对任意角度下采集到的待检测帧进行深度检测的情况下，均可以得到较为准确的深度检测结果，提高深度检测的稳定性、鲁棒性和精度。
在一种可能的实现方式中，步骤S14可以包括：
获取目标区域的预设特征长度，以及采集设备的预设设备参数；
根据预设特征长度与目标区域的特征长度之间的比例关系，以及预设设备参数，确定深度距离。
其中，预设特征长度可以是通常情况下目标区域的实际特征长度，即上述公开实施例中的特征长度的先验估计值。预设特征长度的数值可以根据特征长度定义的不同而灵活发生变化，不局限于下述各公开实施例。在一种可能的实现方式中，在特征长度为第一特征长度和第二特征长度之间的较大值的情况下，可以将预设特征长度设置为25-40cm，在一个示例中，该预设特征长度可以设置为32cm。
预设设备参数可以是采集设备本身的一些标定参数，其包含的参数类型和种类可以根据采集设备的实际情况灵活决定。在一些可能的实现方式中，预设设备参数可以包括采集设备的内参矩阵，该内参矩阵中可以包含相机的一个或多个焦距参数，以及一个或 多个相机的主点位置等。
获取预设设备参数的方式在本公开实施例中不做限定，在一些可能的实现方式中，可以根据采集设备的实际情况直接获取该预设设备参数，在一些可能的实现方式中，也可以通过对采集设备进行标定来获得该预设设备参数。
基于预设特征长度与特征长度之间的比例关系，可以确定目标对象与通常情况下实际场景中目标对象之间的比例关系，结合预设设备参数，可以确定目标对象在实际场景中的深度距离。该计算深度距离的过程可以根据实际情况灵活选择，不局限于下述公开实施例。在一个示例中，根据预设特征长度、特征长度以及预设设备参数确定深度距离的过程，可以通过下述公式(1)和(2)进行表示：
Figure PCTCN2022085913-appb-000001
Figure PCTCN2022085913-appb-000002
其中，d为深度距离，C为预设特征长度，L为目标区域的特征长度，f x和 fy为相机内参矩阵
Figure PCTCN2022085913-appb-000003
中的焦距参数，f为根据焦距参数所确定的参数值。
通过本公开实施例，可以利用特征长度与较为稳定的预设特征长度之间的比例关系，结合采集设备的预设设备参数，简单便捷地确定深度距离，这种确定方式计算量较小且结果较为精确，可以提升深度检测的精度和效率。
在一种可能的实现方式中，本公开实施例提出的方法还可以包括：
获取采集设备的预设设备参数；
根据预设设备参数以及关键点检测结果，确定偏移角度。
其中，预设设备参数的实现形式与获取方式可以参考上述各公开实施例，在此不再赘述。
根据预设设备参数以及关键点检测结果，确定偏移角度的方式也可以灵活选择，不局限于下述各公开实施例。在一些可能的实现方式中，可以根据预设设备参数以及关键点检测结果中头肩中心点的位置坐标，来确定偏移角度。
其中，头肩中心点可以是上述公开实施例中提到的头肩框的中心点，在一些可能的实现方式中，可以根据头部关键点、左肩关键点和右肩关键点的位置坐标，确定头肩框整体的位置坐标，并基于该头肩框整体的位置坐标，确定头肩中心点的位置坐标；在一些可能的实现方式中，也可以直接将头肩中心点作为待检测的关键点，从而在关键点检测结果中直接获取到头肩中心点的位置坐标。
在一个示例中，根据预设设备参数以及头肩中心点的位置坐标确定偏移角度的过程，可以通过下述公式(3)和(4)进行表示：
Figure PCTCN2022085913-appb-000004
Figure PCTCN2022085913-appb-000005
其中，θ x为在x轴方向上的偏移角度，θ y为在y轴方向上的偏移角度，(x,y)为头肩中心点的位置坐标，f x和f y为相机内参矩阵
Figure PCTCN2022085913-appb-000006
中的焦距参数，u 0和v 0为相机内参矩阵K中的主点位置。
通过本公开实施例，可以利用预设设备参数和深度检测过程中得到的关键点检测结果，简单便捷地确定偏移角度，这种确定方式无需获取额外的数据，且便于计算，可以提升深度检测的效率和便捷程度。
在一种可能的实现方式中，本公开实施例提出的方法还可以包括：
根据目标对象的深度信息，确定目标对象在三维空间中的位置。
其中，目标对象在三维空间中的位置，可以是目标对象在三维空间中的三维坐标。基于深度信息确定三维空间中的位置的方式可以根据实际情况灵活选择，在一种可能的实现方式中，可以根据目标对象的关键点检测结果，确定目标对象在待检测帧中的二维坐标，并将该二维坐标与深度信息中的深度距离和/或偏移角度等进行结合，从而确定目标对象在三维空间中的三维坐标。
在确定目标对象在三维空间中的位置以后，可以基于该三维的位置信息，对目标对象进行人脸识别、活体识别、路线跟踪或是应用到虚拟现实(Virtual Reality，VR)或增强现实(Augmented Reality，AR)等场景中。通过本公开实施例，可以利用深度信息对目标对象进行三维定位，从而与目标对象实现各种方式的交互等操作。比如，在一些可能的实现方式中，可以根据目标对象在三维空间中的位置，确定目标对象与智能空调之间的距离和角度，从而动态调整智能空调的风向和/或风速；在一些可能的实现方式中，也可以在AR游戏平台中，基于目标对象在三维空间中的位置，对目标对象在游戏场景中进行定位，从而可以更加真实自然地实现AR场景中的人机互动。
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。
此外，本公开还提供了深度检测装置、电子设备、计算机可读存储介质、程序，上述均可用来实现本公开提供的任一种深度检测方法，相应技术方案和描述和参见方法部分的相应记载，不再赘述。
图4示出根据本公开实施例的深度检测装置的框图。如图4所示，装置20包括：
获取模块21，用于获取待检测帧，待检测帧包含目标对象。
关键点检测模块22，用于根据待检测帧，对目标对象进行关键点检测，得到关键点检测结果。
特征长度确定模块23，用于基于关键点检测结果，确定目标对象中目标区域的特征长度，其中，目标区域包括头部区域和/或肩部区域，特征长度用于表征目标对象中目标区域的尺寸信息。
深度检测模块24，用于根据目标区域的特征长度，确定待检测帧中目标对象的深度信息。
在一种可能的实现方式中，关键点检测模块用于：根据目标对象在参考帧中的位置信息，对待检测帧中的目标对象进行关键点检测，得到关键点检测结果，其中，参考帧为待检测帧所属的目标视频中，位于待检测帧之前的视频帧。
在一种可能的实现方式中，关键点检测模块进一步用于：根据参考帧中目标对象的 第一位置，对待检测帧进行裁剪，得到裁剪结果；对裁剪结果中的目标对象进行关键点检测，得到关键点检测结果。
在一种可能的实现方式中，关键点检测模块进一步用于：获取目标对象的目标区域在参考帧中的第二位置；根据第二位置，对待检测帧进行裁剪，得到裁剪结果；对裁剪结果中的目标对象进行关键点检测，得到关键点检测结果。
在一种可能的实现方式中，关键点检测模块进一步用于：通过第一神经网络对参考帧中的目标区域进行识别，得到第一神经网络输出的第二位置；和/或，根据参考帧对应的关键点检测结果，得到目标区域在参考帧中的第二位置。
在一种可能的实现方式中，关键点检测结果包括头部关键点、左肩关键点以及右肩关键点；特征长度确定模块用于：根据左肩关键点与右肩关键点之间的距离，获取目标区域的第一特征长度；根据头部关键点与肩部中心点之间的距离，获取目标区域的第二特征长度，其中，肩部中心点为左肩关键点与右肩关键点的中间点；根据第一特征长度和/或第二特征长度，确定目标区域的特征长度。
在一种可能的实现方式中，深度信息包括深度距离，深度距离包括目标对象与采集设备的光心之间的距离，采集设备包括对目标对象进行图像采集的设备；深度检测模块用于：获取目标区域的预设特征长度，以及采集设备的预设设备参数；根据预设特征长度与目标区域的特征长度之间的比例关系，以及预设设备参数，确定深度距离。
在一种可能的实现方式中，深度信息包括偏移角度，偏移角度包括目标对象相对于采集设备的光轴的空间角度，采集设备包括对目标对象进行图像采集的设备；装置还用于：获取采集设备的预设设备参数；根据预设设备参数以及关键点检测结果，确定偏移角度。
在一种可能的实现方式中，装置还用于：根据目标对象的深度信息，确定目标对象在三维空间中的位置。
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其具体实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述。
应用场景示例
图5示出根据本公开一应用示例的示意图，如图5所示，本公开应用示例提出一种深度检测方法，可以包括如下过程：
步骤S31，使用Faster RCNN神经网络，对目标视频的第一帧进行人体的头肩框检测，得到第一帧中头肩框的位置。
步骤S32，从目标视频的第二帧开始，将该视频帧作为待检测帧，将待检测帧的上一帧作为参考帧，根据参考帧中头肩框的第二位置，通过关键点检测网络对待检测帧进行关键点检测，得到头部关键点、左肩关键点以及右肩关键点这三个关键点的位置坐标，并将该三个关键点的外接矩形作为待检测帧中的头肩框。
步骤S33，确定待检测帧中的特征长度L，其中，特征长度L可以为：第一特征长度与第二特征长度中的较大值。第一特征长度可以为左肩关键点与右肩关键点之间线段的长度，第二特征长度可以为肩部中心点到头部关键点之间线段的长度。
步骤S34，根据特征长度L、预设特征长度C、待检测帧中的头肩框中心点和相机内参矩阵K中的一种或多种，确定目标对象的深度信息：
其中，步骤S33中定义的特征长度，对应的实际距离是可以先验估计的，对于成年人而言，该特征长度一般约为32cm，因此可以将预设特征长度设置为C＝32cm，因此，本公开应用示例中，可以根据步骤S33中得到的特征长度L、预设特征长度C，以及相机内参矩阵K，可以通过上述公开实施例中的公式(1)和(2)，计算出深度距离d；
在一个示例中，还可以根据待检测帧中头肩框中心点的位置(x，y)和相机内参矩 阵K，通过上述公开实施例中的公式(3)和(4)，计算出深度信息中的偏移角度。
在一个示例中，在通过步骤S34确定待检测帧中目标对象的深度信息以后，还可以将目标视频中待检测帧的下一帧作为待检测帧，并回到步骤S32重新进行深度检测。
通过本公开应用示例，可以利用基于头顶、左肩、右肩3个关键点定义的特征长度作为深度估计的依据，该特征长度受人体朝向和姿势的干扰较小，在目标对象侧身面对镜头、背对镜头或是部分遮挡等复杂场景下，均可以较为准确和鲁棒地实现深度检测，适用场景更多，测距结果更稳定。
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是非易失性计算机可读存储介质或易失性计算机可读存储介质。
本公开实施例还提出一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为调用所述存储器存储的指令，以执行上述方法。
本公开实施例还提供了一种计算机程序产品，包括计算机可读代码，当计算机可读代码在设备上运行时，设备中的处理器执行用于实现如上任一实施例提供的深度检测方法的指令。
本公开实施例还提供了另一种计算机程序产品，用于存储计算机可读指令，指令被执行时使得计算机执行上述任一实施例提供的深度检测方法的操作。
电子设备可以被提供为终端、服务器或其它形态的设备。
图6示出根据本公开实施例的电子设备的框图。如图6所示，电子设备800可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。
参照图6，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多 个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如互补金属氧化物半导体(CMOS)或电荷耦合装置(CCD)图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如无线网络(WiFi)，第二代移动通信技术(2G)或第三代移动通信技术(3G)，或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。
图7示出根据本公开实施例的电子设备的框图。如图7所示，电子设备1900可以被提供为一服务器。参照图7，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如微 软服务器操作系统(Windows Server TM)，苹果公司推出的基于图形用户界面操作系统(Mac OS X TM)，多用户多进程的计算机操作系统(Unix TM),自由和开放原代码的类Unix操作系统(Linux TM)，开放原代码的类Unix操作系统(FreeBSD TM)或类似。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是(但不限于)电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方 框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
该计算机程序产品可以具体通过硬件、软件或其结合的方式实现。在一个可选实施例中，所述计算机程序产品具体体现为计算机存储介质，在另一个可选实施例中，计算机程序产品具体体现为软件产品，例如软件开发包(Software Development Kit，SDK)等等。
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。

Claims (13)

  1. 一种深度检测方法，其特征在于，包括：
    获取待检测帧，所述待检测帧包含目标对象；
    根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果；
    基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，其中，所述目标区域包括头部区域和/或肩部区域，所述特征长度用于表征所述目标对象中目标区域的尺寸信息；
    根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果，包括：
    根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，其中，所述参考帧为所述待检测帧所属的目标视频中，位于所述待检测帧之前的视频帧。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，包括：
    根据所述参考帧中所述目标对象的第一位置，对所述待检测帧进行裁剪，得到裁剪结果；
    对所述裁剪结果中的目标对象进行关键点检测，得到所述关键点检测结果。
  4. 根据权利要求2或3所述的方法，其特征在于，所述根据所述目标对象在参考帧中的位置信息，对所述待检测帧中的所述目标对象进行关键点检测，得到关键点检测结果，包括：
    获取所述目标对象的目标区域在所述参考帧中的第二位置；
    根据所述第二位置，对所述待检测帧进行裁剪，得到裁剪结果；
    对所述裁剪结果中的目标对象进行关键点检测，得到所述关键点检测结果。
  5. 根据权利要求4所述的方法，其特征在于，所述获取所述目标对象的目标区域在所述参考帧中的第二位置，包括：
    通过第一神经网络对所述参考帧中的目标区域进行识别，得到所述第一神经网络输出的第二位置；和/或，
    根据所述参考帧对应的关键点检测结果，得到所述目标区域在所述参考帧中的第二位置。
  6. 根据权利要求1至5中任意一项所述的方法，其特征在于，所述关键点检测结果包括头部关键点、左肩关键点以及右肩关键点；
    所述基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，包括：
    根据所述左肩关键点与所述右肩关键点之间的距离，获取所述目标区域的第一特征长度；
    根据所述头部关键点与肩部中心点之间的距离，获取所述目标区域的第二特征长度，其中，所述肩部中心点为所述左肩关键点与所述右肩关键点的中间点；
    根据所述第一特征长度和/或所述第二特征长度，确定所述目标区域的特征长度。
  7. 根据权利要求1至6中任意一项所述的方法，其特征在于，所述深度信息包括深度 距离，所述深度距离包括所述目标对象与采集设备的光心之间的距离，所述采集设备包括对所述目标对象进行图像采集的设备；
    所述根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息，包括：
    获取所述目标区域的预设特征长度，以及所述采集设备的预设设备参数；
    根据所述预设特征长度与所述目标区域的特征长度之间的比例关系，以及所述预设设备参数，确定所述深度距离。
  8. 根据权利要求1至7中任意一项所述的方法，其特征在于，所述深度信息包括偏移角度，所述偏移角度包括所述目标对象相对于采集设备的光轴的空间角度，所述采集设备包括对所述目标对象进行图像采集的设备；
    所述方法还包括：
    获取所述采集设备的预设设备参数；
    根据所述预设设备参数以及所述关键点检测结果，确定所述偏移角度。
  9. 根据权利要求1至8中任意一项所述的方法，其特征在于，所述方法还包括：
    根据所述目标对象的深度信息，确定所述目标对象在三维空间中的位置。
  10. 一种深度检测装置，其特征在于，包括：
    获取模块，用于获取待检测帧，所述待检测帧包含目标对象；
    关键点检测模块，用于根据所述待检测帧，对所述目标对象进行关键点检测，得到关键点检测结果；
    特征长度确定模块，用于基于所述关键点检测结果，确定所述目标对象中目标区域的特征长度，其中，所述目标区域包括头部区域和/或肩部区域，所述特征长度用于表征所述目标对象中目标区域的尺寸信息；
    深度检测模块，用于根据所述目标区域的特征长度，确定所述待检测帧中所述目标对象的深度信息。
  11. 一种电子设备，其特征在于，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器被配置为调用所述存储器存储的指令，以执行权利要求1至9中任意一项所述的方法。
  12. 一种计算机可读存储介质，其上存储有计算机程序指令，其特征在于，所述计算机程序指令被处理器执行时实现权利要求1至9中任意一项所述的方法。
  13. 一种计算机程序产品，其特征在于，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现权利要求1至9中任意一项所述的方法。
PCT/CN2022/085913 2021-06-28 2022-04-08 深度检测方法及装置、电子设备和存储介质 WO2023273498A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110719313.2A CN113344999A (zh) 2021-06-28 2021-06-28 深度检测方法及装置、电子设备和存储介质
CN202110719313.2 2021-06-28

Publications (1)

Publication Number Publication Date
WO2023273498A1 true WO2023273498A1 (zh) 2023-01-05

Family

ID=77479127

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/085913 WO2023273498A1 (zh) 2021-06-28 2022-04-08 深度检测方法及装置、电子设备和存储介质

Country Status (3)

Country Link
CN (1) CN113344999A (zh)
TW (1) TW202301275A (zh)
WO (1) WO2023273498A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113344999A (zh) * 2021-06-28 2021-09-03 北京市商汤科技开发有限公司 深度检测方法及装置、电子设备和存储介质
CN113345000A (zh) * 2021-06-28 2021-09-03 北京市商汤科技开发有限公司 深度检测方法及装置、电子设备和存储介质

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150213702A1 (en) * 2014-01-27 2015-07-30 Atlas5D, Inc. Method and system for behavior detection
CN105007396A (zh) * 2015-08-14 2015-10-28 山东诚海电子科技有限公司 一种用于课堂教学的定位方法及装置
CN110225400A (zh) * 2019-07-08 2019-09-10 北京字节跳动网络技术有限公司 一种动作捕捉方法、装置、移动终端及存储介质
CN111084606A (zh) * 2019-10-12 2020-05-01 深圳壹账通智能科技有限公司 基于图像识别的视力检测方法、装置、及计算机设备
CN112525352A (zh) * 2020-11-24 2021-03-19 深圳市高巨创新科技开发有限公司 一种基于人脸识别的红外测温补偿方法及终端
CN113345000A (zh) * 2021-06-28 2021-09-03 北京市商汤科技开发有限公司 深度检测方法及装置、电子设备和存储介质
CN113344999A (zh) * 2021-06-28 2021-09-03 北京市商汤科技开发有限公司 深度检测方法及装置、电子设备和存储介质
CN113888619A (zh) * 2021-10-29 2022-01-04 北京市商汤科技开发有限公司 距离确定方法、装置、电子设备及存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102890785A (zh) * 2011-07-19 2013-01-23 上海上大海润信息系统有限公司 一种服务机器人目标识别与定位的方法
CN109492608B (zh) * 2018-11-27 2019-11-05 腾讯科技（深圳）有限公司 图像分割方法、装置、计算机设备及存储介质
CN111561906B (zh) * 2020-05-25 2022-03-11 北京洛必德科技有限公司 机器人单目测距方法、系统、电子设备和计算机存储介质
CN112365586B (zh) * 2020-11-25 2023-07-18 厦门瑞为信息技术有限公司 3d人脸建模与立体判断方法及嵌入式平台的双目3d人脸建模与立体判断方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150213702A1 (en) * 2014-01-27 2015-07-30 Atlas5D, Inc. Method and system for behavior detection
CN105007396A (zh) * 2015-08-14 2015-10-28 山东诚海电子科技有限公司 一种用于课堂教学的定位方法及装置
CN110225400A (zh) * 2019-07-08 2019-09-10 北京字节跳动网络技术有限公司 一种动作捕捉方法、装置、移动终端及存储介质
CN111084606A (zh) * 2019-10-12 2020-05-01 深圳壹账通智能科技有限公司 基于图像识别的视力检测方法、装置、及计算机设备
CN112525352A (zh) * 2020-11-24 2021-03-19 深圳市高巨创新科技开发有限公司 一种基于人脸识别的红外测温补偿方法及终端
CN113345000A (zh) * 2021-06-28 2021-09-03 北京市商汤科技开发有限公司 深度检测方法及装置、电子设备和存储介质
CN113344999A (zh) * 2021-06-28 2021-09-03 北京市商汤科技开发有限公司 深度检测方法及装置、电子设备和存储介质
CN113888619A (zh) * 2021-10-29 2022-01-04 北京市商汤科技开发有限公司 距离确定方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN113344999A (zh) 2021-09-03
TW202301275A (zh) 2023-01-01

Similar Documents

Publication Publication Date Title
EP3511864A1 (en) Method and apparatus for synthesizing virtual and real objects
WO2023273498A1 (zh) 深度检测方法及装置、电子设备和存储介质
US20210158560A1 (en) Method and device for obtaining localization information and storage medium
EP2927787A1 (en) Method and device for displaying picture
WO2022188305A1 (zh) 信息展示方法及装置、电子设备、存储介质及计算机程序
TWI718631B (zh) 人臉圖像的處理方法及裝置、電子設備和儲存介質
WO2023273499A1 (zh) 深度检测方法及装置、电子设备和存储介质
US20200402321A1 (en) Method, electronic device and storage medium for image generation
WO2022134475A1 (zh) 点云地图构建方法及装置、电子设备、存储介质和程序
TWI752473B (zh) 圖像處理方法及裝置、電子設備和電腦可讀儲存媒體
WO2022017140A1 (zh) 目标检测方法及装置、电子设备和存储介质
WO2022193466A1 (zh) 图像处理方法及装置、电子设备和存储介质
WO2023051356A1 (zh) 一种虚拟对象的显示方法及装置、电子设备和存储介质
WO2022121577A1 (zh) 图像处理方法及装置
WO2023155532A1 (zh) 位姿检测方法及装置、电子设备和存储介质
CN111860388A (zh) 图像处理方法及装置、电子设备和存储介质
CN112184787A (zh) 图像配准方法及装置、电子设备和存储介质
CN110930351A (zh) 一种光斑检测方法、装置及电子设备
WO2022179080A1 (zh) 定位方法、装置、电子设备、存储介质、程序及产品
WO2022183656A1 (zh) 数据生成方法、装置、设备、存储介质及程序
CN114581525A (zh) 姿态确定方法及装置、电子设备和存储介质
TWI770531B (zh) 人臉識別方法、電子設備和儲存介質
CN112613447A (zh) 关键点检测方法及装置、电子设备和存储介质
CN112767541A (zh) 三维重建方法及装置、电子设备和存储介质
CN112270737A (zh) 一种纹理映射方法及装置、电子设备和存储介质

Legal Events

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

Ref document number: 22831329

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE