WO2019144263A1 - 可移动平台的控制方法、设备、计算机可读存储介质 - Google Patents

可移动平台的控制方法、设备、计算机可读存储介质 Download PDF

Info

Publication number
WO2019144263A1
WO2019144263A1 PCT/CN2018/073769 CN2018073769W WO2019144263A1 WO 2019144263 A1 WO2019144263 A1 WO 2019144263A1 CN 2018073769 W CN2018073769 W CN 2018073769W WO 2019144263 A1 WO2019144263 A1 WO 2019144263A1
Authority
WO
WIPO (PCT)
Prior art keywords
target object
feature
tracking
feature part
preset
Prior art date
Application number
PCT/CN2018/073769
Other languages
English (en)
French (fr)
Inventor
庞磊
赵丛
张李亮
朱高
李思晋
刘尧
Original Assignee
深圳市大疆创新科技有限公司
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 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to PCT/CN2018/073769 priority Critical patent/WO2019144263A1/zh
Priority to CN201880001860.3A priority patent/CN109155067A/zh
Publication of WO2019144263A1 publication Critical patent/WO2019144263A1/zh
Priority to US16/935,709 priority patent/US11227388B2/en
Priority to US17/648,179 priority patent/US12002221B2/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/73—Deblurring; Sharpening
    • 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
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/757—Matching configurations of points or features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/17—Terrestrial scenes taken from planes or by drones
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
    • 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
    • G06T2207/20024—Filtering details
    • G06T2207/20028—Bilateral filtering
    • 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
    • G06T2207/20172—Image enhancement details
    • G06T2207/20192—Edge enhancement; Edge preservation
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30196—Human being; Person
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30196—Human being; Person
    • G06T2207/30201—Face

Definitions

  • the present invention relates to the field of electronic technologies, and in particular, to a control method, device, and computer readable storage medium for a mobile platform.
  • the existing tracking strategy is to track a feature part of the target object with a distinct feature as a target.
  • the size ratio of the tracking frame of the feature part of the target object also changes in the captured image. This will affect the effect of tracking.
  • the size of the tracking frame of the feature part of the target object is relatively large in the captured image, which may cause the tracking speed to be slow, thereby easily causing the target object to be tracked and lost, and tracking
  • the reliability of the target is worse.
  • the tracking frame of the feature part of the target object occupies a small size in the captured image, which may cause the feature of the tracked target object to be blurred and tracked. The reliability is worse.
  • the invention provides a control method, a device and a computer readable storage medium of a movable platform, which can prevent the target object of the tracking from being lost, and can improve the reliability of the tracking.
  • a first aspect of the embodiments of the present invention provides a method for controlling a mobile platform, including:
  • the second feature part of the target object is tracked and switched to track the first feature part of the target object.
  • a second aspect of the embodiments of the present invention provides a mobile platform, including: a memory and a processor;
  • the memory is configured to store program code
  • the processor is configured to invoke the program code, when the program code is executed, to perform the following operations:
  • the second feature part of the target object is tracked and switched to track the first feature part of the target object.
  • a third aspect of the embodiments of the present invention provides a computer readable storage medium, where the computer readable storage medium stores computer instructions, and when the computer instructions are executed, implementing the first aspect of the embodiment of the present invention.
  • the control method of the mobile platform is not limited to, but not limited to, but not limited to
  • the matching parameter corresponding to the first feature part may be determined according to the second feature part of the target object and the first feature part, and the matching parameter is matched according to the first feature part. Determining a first feature portion of the target object in the first feature portion identified, and switching the second feature portion of the target object to the first target object when the tracking parameter of the target object meets the preset tracking condition
  • the feature part is tracked, in this way, the movable platform can be tracked to the feature parts that are compatible with the tracking parameters of the current target object, thereby preventing the loss of the tracking target object and improving the reliability of the tracking.
  • 1 is a schematic flow chart of a control method of a mobile platform
  • FIG. 2 is a schematic diagram of switching the tracking of the human body of the target object to tracking the head and shoulders of the target object when the far field tracking switches the near field tracking;
  • 3 is a schematic diagram of switching the head and shoulder of the target object to track the human body of the target object when the near field tracking switches the far field tracking;
  • FIG. 4 is a schematic diagram of a head-side prediction of a human body according to a preset proportional relationship and a target object when far-field tracking switches near-field tracking;
  • FIG. 5 is a schematic diagram of predicting the head and shoulder according to the joint point of the target object when the far field tracking switches the near field tracking;
  • FIG. 6 is a schematic diagram of predicting a human body according to a preset proportional relationship and a head and shoulder of a target object when the near field tracking is switched to the far field tracking;
  • FIG. 7 is a schematic diagram of predicting a human body according to joint points of a target object when near-field tracking switches far-field tracking;
  • Figure 8 is a schematic structural view of a movable platform.
  • first, second, third, etc. may be used to describe various information in the present invention, such information should not be limited to these terms. These terms are used to distinguish the same type of information from each other.
  • first information may also be referred to as the second information without departing from the scope of the invention.
  • second information may also be referred to as the first information.
  • word “if” may be interpreted as "at time”, or "when", or "in response to determination.”
  • a control method of a mobile platform may include, but is not limited to, a drone, a ground robot (such as an unmanned vehicle, etc.).
  • the mobile platform can be configured with a shooting device and can capture captured images through a shooting device.
  • the mobile platform can be configured with a pan/tilt, which can carry a shooting device (such as a camera, a camera, etc.) to stabilize and/or adjust the shooting device.
  • the mobile platform is taken as an example for the unmanned aerial vehicle as an example. It can be understood that the drone of the following part of the present invention can be replaced by the mobile platform.
  • the traditional tracking technology is to track a single feature part of the target object, such as using the human body of the target object as a tracking target, or using a preset part of the human body of the target object (for example, the head of the human body) as a tracking target.
  • a preset part of the human body of the target object for example, the head of the human body
  • the size ratio of the tracking frame of the feature part of the target object in the captured image also changes, so that Will affect the effect of tracking.
  • the size of the tracking frame of the feature part of the target object is relatively large in the captured image, which may cause the tracking speed to be slow, thereby easily causing the target object to be tracked and lost, and tracking
  • the reliability of the UAV is worse.
  • the tracking frame of the feature part of the target object occupies a small size in the captured image, which may cause the feature of the tracked target object to be blurred and tracked. The reliability is worse.
  • the preset part of the human body of the target object when the distance between the UAV and the target object is relatively close, the preset part of the human body of the target object may be used as the tracking target, that is, the preset part of the human body of the target object to the target object. Tracking; when the distance between the drone and the target object is long, the human body of the target object can be used as the tracking target, that is, the drone can track the human body of the target object, and the above manner can achieve better tracking effect.
  • the following four situations may be included:
  • the first case near field tracking.
  • the near-field tracking method may be adopted, that is, the preset part of the human body of the target object (such as the head, or the head and the shoulder (which may be simply referred to as the head and shoulders), etc.) Track the target and track the preset part of the target's human body.
  • the second case far field tracking.
  • the far-field tracking method can be adopted, that is, the human body of the target object is used as the tracking target, and the human body of the target object is tracked.
  • the third case switching from far-field tracking to near-field tracking.
  • the distance between the drone and the target object is far, and the drone uses the far-field tracking method to track the human body of the target object. Then, the distance between the drone and the target object is getting closer and closer.
  • the far-field tracking mode is switched to the near-field tracking mode, that is, The drone no longer tracks the human body of the target object, but uses the preset part of the human body of the target object as a tracking target, and tracks the preset part of the human body of the target object.
  • the fourth case switching from near-field tracking to far-field tracking.
  • the distance between the drone and the target object is relatively close, and the drone uses the near-field tracking method to track the preset part of the human body of the target object. Then, the distance between the drone and the target object is further and farther.
  • the near field tracking mode is switched to the far field tracking mode, that is, The drone no longer tracks the preset part of the target object, but uses the human body of the target object as the tracking target and tracks the human body of the target object.
  • FIG. 1 is a flowchart of a control method of a mobile platform, the method includes:
  • Step 101 Acquire a captured image.
  • the execution body of the method may be a mobile platform, and may further be a processor of the mobile platform, where the processor may be one or more, and the processor may be a general purpose processor or a dedicated processor. processor.
  • the mobile platform may be configured with a photographing device, and during the tracking of the target object by the movable platform, the photographing device may photograph the target object to obtain a captured image, and the processor of the mobile platform may acquire the photographed The image is taken, wherein the target object is an object tracked by the movable platform.
  • Step 102 identifying a first feature portion from the captured image.
  • the processor of the movable platform may identify the first feature from the captured image, and further, a neural network (eg, a convolutional neural network) may be utilized to identify the first feature in the image.
  • a neural network eg, a convolutional neural network
  • the neural network may return the position of the first feature part in the captured image and the corresponding image area.
  • the position in the captured image and the corresponding image area may be represented by a detection frame of the first feature part, that is, the first feature part may be represented by a detection frame.
  • Step 103 Determine a second feature portion of the target object in the captured image.
  • the processor may determine a second feature portion of the target object from the captured image, wherein determining the second feature portion of the target object from the captured image may adopt a tracking algorithm: acquiring the current one After the frame is captured, the target area is determined according to the position of the second feature part of the target object in the captured image of the previous frame, and then, in the target area of the current frame of the captured image, the second object of the target object in the captured image of the previous frame is searched. The image region that the feature portion most matches, and the most matching image region is determined as the second feature portion of the target object in the current frame of the captured image.
  • the second feature part of the target object in the current frame of the captured image may be represented by a detection frame.
  • Step 104 Determine a matching parameter corresponding to the first feature part according to the second feature part of the target object and the first feature part.
  • the first feature portion identified from the captured image may include a plurality of features, wherein the first feature portion includes a feature portion of the target object.
  • the second feature portion of the target object needs to be tracked to track the first feature portion of the target object. Therefore, it is necessary to determine from the first feature portion identified in the captured image which of the first feature portions is the first feature portion of the target object.
  • the matching parameter corresponding to the first feature part may be determined according to the second feature part of the target object and the first feature part, wherein each first feature part may correspond to a matching parameter
  • the matching parameter can be used to indicate the likelihood that a certain first feature location is the first feature portion of the target object.
  • the matching parameter includes one or more of a coincidence degree matching parameter, an image similarity matching parameter, and a geometric similarity matching parameter. It can be understood that the matching parameter may be a coincidence degree matching parameter and an image similarity matching parameter.
  • One of the geometric similarity matching parameters may also be determined according to a plurality of matching degree matching parameters, image similarity matching parameters, and geometric similarity matching parameters.
  • Step 105 Determine a first feature part of the target object from the first feature part according to the matching parameter corresponding to the first feature part.
  • a plurality of first feature portions may be identified from the captured image, wherein the first feature portion of the plurality of first feature portions includes a first feature portion of the target object, and may correspond to each of the first feature portions.
  • the matching parameter determines which first feature location is the first feature part of the target object.
  • determining the first feature portion of the target object from the first feature portion according to the matching parameter corresponding to the first feature portion comprises: determining, from the matching parameters corresponding to the first feature portion The largest matching parameter determines the first feature portion of the largest matching parameter as the first feature portion of the target object. Specifically, when the matching parameter is larger, it may be considered that the first feature part corresponding to the matching parameter is more likely to be the first feature part of the target object, and a maximum matching parameter may be determined from the determined matching parameter, and The first feature portion of the largest matching parameter is determined as the first feature portion of the target object.
  • Step 106 If the tracking parameter of the target object meets the preset tracking condition, the second feature part of the target object is tracked and switched to track the first feature part of the target object.
  • the tracking parameter of the target object may include: a size ratio of the tracking frame of the target object in the captured image, and/or a distance between the target object and the movable platform.
  • the movable platform performs tracking on the target object by tracking the second feature part of the target object, and when the tracking parameter of the target object satisfies the preset tracking condition, the target is Tracking the second feature portion of the object may result in poor reliability of the tracking. At this time, it may be switched to track the first feature portion of the target object that is compatible with the tracking parameter of the current target object.
  • the first feature portion is a preset portion of the human body
  • the second feature portion is a human body
  • the movable platform uses the far field tracking strategy to track the second feature part of the target object, and once the tracking parameter of the target object is determined to meet the second preset
  • the movable platform switches to the near field tracking strategy, and the second feature portion of the target object is tracked to track the first feature portion of the target object.
  • the tracking parameter of the target object satisfies the second preset tracking condition, and may include, but is not limited to, the tracking frame of the target object has a size ratio of the captured image greater than or equal to a preset second ratio threshold, and/or the target object.
  • the distance from the movable platform is less than or equal to the preset second distance.
  • the movable platform and the target object are illustrated.
  • the distance is small, and the distance tracking can be switched from the far field tracking to the near field tracking, that is, the tracking of the human body of the target object is switched to the preset part of the human body of the target object.
  • the preset second ratio threshold and the preset second distance may be configured according to experience, and no limitation is imposed thereon.
  • the movable platform tracks the human body of the target object.
  • the matching parameters corresponding to each head and shoulder are determined according to the human body 211 and the head and shoulders 201, the head and shoulders 202 and the head and shoulders 203 of the target object, and the head and shoulders 201 are determined as the head and shoulders of the target object according to the matching parameters.
  • the tracking parameter of the target object satisfies the second preset tracking condition at a certain moment, the tracking of the human body of the target object is switched to track the head and shoulders 201 of the target object.
  • the first feature portion is a human body
  • the second feature portion is a preset portion of the human body, if the tracking parameter of the target object satisfies the first When the tracking condition is preset, the second feature portion of the target object is switched to track the first feature portion of the target object.
  • the movable platform adopts a near field tracking strategy to track the second feature part of the target object, and once the tracking parameter of the target object is determined to satisfy the first preset
  • the movable platform switches to the far field tracking strategy, and the second feature portion of the target object is tracked to track the first feature portion of the target object.
  • the tracking parameter of the target object satisfies the first preset tracking condition, and may include, but is not limited to, the tracking frame of the target object has a size ratio of the captured image that is less than or equal to a preset first percentage threshold, and/or the target object.
  • the distance from the movable platform is greater than or equal to the preset first distance.
  • the movable platform and the target object are illustrated.
  • the distance is large, and it can be switched from near-field tracking to far-field tracking, that is, the target part of the target object is tracked and switched to the body of the target object.
  • the preset first ratio threshold and the preset first distance may be configured according to experience, and no limitation is imposed thereon.
  • the movable platform tracks the head and shoulders of the human body of the target object.
  • the matching parameters of each human body are determined according to the head and shoulders 231 of the target object, the human body 221, the human body 222, and the human body 223, and the human body 221 is determined as the target object according to the matching parameters, at a certain moment.
  • the tracking parameter of the target object satisfies the first preset tracking condition
  • the head and shoulder of the target object are tracked and switched to track the human body 221 of the target object.
  • determining the matching parameter corresponding to the first feature part according to the second feature part of the target object and the first feature part may be implemented in the following feasible manners:
  • the matching parameter of the second feature part and the first feature part may be determined according to the second feature part of the target object and the first feature part.
  • the matching parameter of the second feature part of the target object and the first feature part may represent a degree of matching between the second feature part of the target object and the first feature part, and the higher the matching degree, the first feature is represented. The more likely the part is the first feature of the target object.
  • the matching parameters of the human body 211 and the head and shoulder 201 of the target object may be determined according to the human body 211 and the head and shoulder 201 of the target object, according to the human body 211 and the head of the target object.
  • the shoulder 202 determines the matching parameters of the human body 211 of the target object and the head and shoulders 202, and determines the matching parameters of the human body 211 and the head and shoulders 203 of the target object according to the human body 211 and the head and shoulders 203 of the target object.
  • the determination process of the matching parameters will be explained in detail below.
  • the matching parameters are one of a coincidence degree matching parameter, an image similarity matching parameter, and a geometric similarity matching parameter. Therefore, the coincidence degree matching parameter between the human body 211 and the head and shoulder 201 of the target object can be determined according to the human body 211 and the head and shoulder 201 of the target object, wherein the coincidence degree matching parameter is used to indicate the degree of coincidence of the two image regions, and the degree of coincidence
  • the matching parameter can be represented by the ratio of the intersection of the two image regions to the union of the two image regions.
  • the ratio between the human body 211 of the target object and the head and shoulders 201 and the union of the human body 211 and the head and shoulders 201 of the target object can be used as the coincidence between the human body 211 and the head and shoulder 201 as the target object.
  • Degree matching parameters By determining the coincidence degree matching coefficient between the human body 211 and the head and shoulder 201 of the target object, the degree of coincidence between the human body 211 of the target object and the head and shoulders 201 can be known, wherein when the degree of coincidence is higher, the head and shoulders 201 can be represented. The more likely it is the head and shoulders of the target object.
  • the image similarity matching parameter between the human body 211 and the head and shoulder 201 of the target object may be determined according to the human body 211 and the head and shoulder 201 of the target object, wherein the image similarity matching parameter is used to represent the images in the two image regions. similarity.
  • the image similarity matching parameter can be determined by using a histogram of the images in the two image regions.
  • the histograms of the human body 211 and the head and shoulders 201 of the target object are respectively determined, and the normalized correlation between the two histograms is calculated. Coefficients (such as Pap singer distance, histogram intersection distance, etc.), the normalized correlation coefficient is used as an image similarity matching parameter.
  • the image similarity matching parameter can be determined by determining the image similarity matching parameter between the human body 211 and the head and shoulder 201 of the target object, wherein when the degree of similarity is higher, the head can be represented. The more likely the shoulder 201 is to be the head and shoulders of the target object.
  • the geometric degree matching parameter between the human body 211 and the head and shoulder 201 of the target object may be determined according to the human body 211 and the head and shoulder 201 of the target object, wherein the geometric degree matching parameter is used to indicate the degree of matching of the size of the two image regions.
  • the degree of change of the target object's own motion is not large, and the distance between the target and the movable platform is relatively small. Therefore, the proportional relationship between the feature parts of the target object in the image should be Preset ratio.
  • the ratio of the area of the human body 211 and the head and shoulder 201 of the target object can be determined by the area of the human body 211 of the target object and the area of the head and shoulder 201, and then the ratio of the area to the preset ratio can be further determined.
  • the preset ratio may be a size ratio between the human body and the head and shoulder, and the geometric degree matching parameter is determined according to the difference.
  • the matching parameter between the human body 211 and the head and shoulders 201 can be determined according to one or more of the coincidence degree matching parameter, the image similarity matching parameter, and the geometric similarity matching parameter, that is, those skilled in the art can
  • One of the coincidence degree matching parameter, the image similarity matching parameter, and the geometric similarity matching parameter is used as a matching parameter between the human body 211 and the head and shoulder 201, and the coincidence degree matching parameter, the image similarity matching parameter, and the geometric similarity may also be used.
  • the plurality of degree matching parameters are fused to determine matching parameters between the human body 211 and the head and shoulders 201, and are not specifically limited herein. With the method described above, the matching parameters between the human body 211 and the head and shoulders 202 can also be determined, and the matching parameters between the human body 211 and the head and shoulders 201 can also be determined.
  • the matching parameters of the second feature part 231 and the first feature part 221 may be determined according to the second feature part 231 and the first feature part 221 of the target object;
  • the matching parameter of the second feature part 231 and the first feature part 222 may be determined according to the second feature part 231 and the first feature part 222 of the target object;
  • the second feature part 231 and the first feature part 223 of the target object may be determined according to the second feature part 231 and the first feature part 223 Matching parameters of the second feature portion 231 and the first feature portion 223.
  • the second feature portion 231 and the first feature portion 222, and the second feature portion 231 and the first feature portion 223, refer to the foregoing section. I won't go into details here.
  • the third feature portion may be predicted, wherein the third feature portion is the first feature portion of the target object predicted according to the second feature portion of the target object. Then, matching parameters of the third feature portion and the first feature portion are determined according to the third feature portion and the first feature portion.
  • the predicting the third feature part of the target object may include: predicting the third feature part according to the preset proportional relationship and the second feature part of the target object.
  • a joint point of the target object is determined according to the second feature portion of the target object, and the third feature portion is predicted according to the joint point.
  • the preset proportional relationship may be a proportional relationship between the first characteristic part of the person and the second characteristic part of the person, and may be an empirical value.
  • the third feature portion 212 that is, the head and shoulder 212 of the predicted target object, may be predicted according to the preset proportional relationship and the human body 211 of the target object.
  • the pre-set proportional relationship is the proportional relationship between the human body and the head and shoulders of the human body.
  • the joint point of the target object may be determined according to the human body 211 of the target object, and then the head and shoulder may be predicted according to the joint point of the target object, for example, from the target object.
  • the joint point of the shoulder and the joint point of the eye are determined in the off node, and the head and shoulder 213 of the target object are predicted based on the joint point of the shoulder and the joint point of the eye.
  • the predicted head and shoulders 213 and head and shoulders 201 can be determined based on the predicted head and shoulders 213 and the head and shoulders 201.
  • the predicted head and shoulders 213 and head and shoulders 202 define the predicted head and shoulders 213.
  • the predicted head and shoulders 213 and head and shoulders 203 determine the matching parameters of the predicted head and shoulders 213 and head and shoulders 203. For details on how to determine matching parameters, refer to the previous section, and details are not described here.
  • the human body 232 that is, the predicted human body 232, may be predicted according to a preset proportional relationship and the head and shoulders 231 of the target object, wherein the preset proportional relationship It is the proportional relationship between the human body and the head and shoulders of the human body.
  • the joint point of the target object may be determined according to the head and shoulder 231 of the target object, and then the human body 233 is predicted according to the joint point of the target object, for example, from the target object.
  • the joint points the joint point of the foot, the joint point of the shoulder, and the joint point of the eye are determined, and the human body 233 of the target object is predicted based on the joint point of the foot, the joint point of the shoulder, and the joint point of the eye.
  • the predicted matching parameters of the human body 233 and the human body 221 can be determined according to the predicted human body 233 and the human body 221, and the predicted human body 233 and the human body 222 determine the matching parameters of the predicted human body 233 and the human body 222, and the prediction is performed.
  • the human body 233 and the human body 223 determine the predicted human body 233 and the human body 223. For details on how to determine matching parameters, refer to the previous section, and details are not described here.
  • the distance between the upper edge of the tracking frame of the first feature portion of the largest matching parameter and the upper edge of the tracking frame of the second feature portion may also be determined in consideration of the case of multi-person coincidence.
  • determining the first feature part of the maximum matching parameter as the first feature part of the target object may include: when the distance is less than or equal to a preset distance threshold (the preset distance threshold may be configured according to experience, When no limitation is made, the first feature portion of the largest matching parameter may be determined as the first feature portion of the target object.
  • the maximum matching parameter is re-determined on the basis of excluding the first feature portion of the largest matching parameter, and so on.
  • the first The feature portion 201 is determined as the first feature portion of the target object.
  • the distance is greater than the preset distance threshold, the first feature portion 201 is excluded, and the first feature portion corresponding to the largest matching parameter is re-determined from the first feature portion 202 and the first feature portion 203, and so on.
  • the human body of the target object is tracked, and when the movable platform approaches the target object, the size ratio of the tracking frame of the target object in the captured image becomes larger, thereby causing The tracking speed is slow, so the far-field tracking can be switched to near-field tracking, that is, the head and shoulder of the target object are tracked to improve the tracking efficiency.
  • the head and shoulder of the target object are tracked.
  • the tracking frame of the head and shoulder of the target object will have a smaller proportion in the captured image, and the accuracy of the tracking will be The difference is made, therefore, the near field tracking can be switched to the far field tracking, that is, the human body of the target object is tracked to improve the accuracy of the tracking.
  • the size of the tracking frame may change to some extent, the size of the tracking frame may not match the size of the tracked target object. Therefore, other auxiliary methods may be used to determine the switching condition to improve the switching accuracy. For example, using the depth sensor technology and the image projection relationship, the depth corresponding to the target object in the captured image is found, thereby obtaining the distance between the movable platform and the target object, and if the distance is too close (for example, less than or equal to 3 meters), the operation is switched to Near field tracking, if the distance is too far (such as greater than 4 meters), cut to far field tracking. In addition, it is also possible to directly measure the distance between the target object and the movable platform by means of ranging (such as binocular ranging, ultrasonic ranging or lidar ranging).
  • ranging such as binocular ranging, ultrasonic ranging or lidar ranging
  • the application scenarios applicable to the foregoing embodiments may include, but are not limited to:
  • the target object is a human
  • the tracking frame size is detected. If the size of the tracking frame in the captured image is found to be greater than or equal to 30%, the head and shoulder of the target object are automatically detected. , that is, near field tracking.
  • the target object is far away from the drone, if the size of the tracking frame in the captured image is less than 10%, the near field tracking is switched to the far field tracking, and the human body of the target object is automatically detected.
  • the drone has the face-lifting function, after the face is successfully scanned, the near-field tracking (such as tracking the head and shoulders) is started.
  • the target object is far away from the drone, it can automatically switch to far-field tracking (for example, tracking).
  • the human body can achieve the effect of automatically focusing on the user.
  • the drone can take off directly from the user's hand, fly to the user's oblique upper back, and start tracking the head and shoulders of the target object. After the drone flies out, if the target object is far away from the drone, it can automatically switch to far-field tracking, that is, tracking the human body of the target object.
  • the drone Surround video from near and far. After the drone is focused on the human body, the drone spirally surrounds the video, and the drone can take off directly from the user's hand and start the spiral flight shooting, that is, tracking the head and shoulder of the target object. After the drone flies out, if the target object is far away from the drone, it can automatically switch to far-field tracking, that is, tracking the human body of the target object.
  • an embodiment of the present invention further provides a removable platform 30, including a memory 31 and a processor 32 (such as one or more processors).
  • the memory is configured to store program code
  • the processor is configured to invoke the program code, when the program code is executed, to perform the following operations: acquiring a captured image;
  • the second feature part of the target object is tracked and switched to track the first feature part of the target object.
  • the processor determines the matching parameter corresponding to the first feature part according to the second feature part of the target object and the first feature part
  • the processor is specifically configured to: according to the second feature part of the target object and the first feature part A matching parameter of the second feature portion and the first feature portion is determined.
  • the determining, by the processor, the matching parameter corresponding to the first feature part according to the second feature part of the target object and the first feature part is specifically: predicting a third feature part of the target object;
  • the third feature portion is a first feature portion of the target object predicted according to the second feature portion;
  • the third feature portion is determined according to the third feature portion of the target object and the first feature portion Matching parameters of the first feature.
  • the method is specifically configured to: according to a proportional relationship between the second feature part of the target object and the third feature part of the target object, and the second feature part prediction center The third feature part of the target object.
  • the method is specifically configured to: determine joint point information of the target object according to the second feature part of the target object; and predict the location according to the joint point information The third feature part of the target object.
  • the processor is configured to: when the first feature part of the target object is determined from the first feature part, according to the matching parameter corresponding to the first feature part, the matching parameter corresponding to the first feature part The largest matching parameter is determined; the first feature portion of the largest matching parameter is determined as the first feature portion of the target object.
  • the first feature part is a human body
  • the second feature part is a preset part of a human body
  • the processor will target the target object when a tracking parameter of the target object satisfies a preset tracking condition
  • the method is specifically configured to: if the tracking parameter of the target object meets the first preset tracking condition, The second feature portion is tracked and switched to track the first feature portion of the target object.
  • the tracking parameter of the target object satisfies the first preset tracking condition, and the tracking frame of the target object has a size ratio of the captured image that is less than or equal to a preset first percentage threshold, and/or The distance between the target object and the movable platform is greater than or equal to a preset first distance.
  • the first feature part is a preset part of a human body
  • the second feature part is a human body
  • the processor will target the target object when a tracking parameter of the target object satisfies a preset tracking condition
  • the method is specifically configured to: if the tracking parameter of the target object meets the second preset tracking condition, the target object is to be The second feature portion is tracked and switched to track the first feature portion of the target object.
  • the tracking parameter of the target object satisfies the second preset tracking condition, including: the tracking frame of the target object has a size ratio of the captured image greater than or equal to a preset second ratio threshold, and/or The distance between the target object and the movable platform is less than or equal to a preset second distance.
  • the preset part comprises: a head, or a head and a shoulder.
  • the processor is further configured to: determine a distance between an upper edge of the tracking frame of the first feature part of the maximum matching parameter and an upper edge of the tracking frame of the second feature part;
  • the specific feature is: when the distance is less than or equal to the preset distance threshold, the first feature of the largest matching parameter is The part is determined as the first feature part of the target object.
  • the matching parameter includes one or more of a coincidence degree matching parameter, an image similarity matching parameter, and a geometric similarity matching parameter.
  • the embodiment of the present invention further provides a computer readable storage medium storing computer instructions, where the computer instructions are executed (if the computer instructions are executed by the processor), The control method of the movable platform described in FIG.
  • the system, apparatus, module or unit set forth in the above embodiments may be implemented by a computer chip or an entity, or by a product having a certain function.
  • a typical implementation device is a computer, and the specific form of the computer may be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver, and a game control.
  • embodiments of the invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or a combination of software and hardware. Moreover, embodiments of the invention may take the form of a computer program product embodied on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) including computer usable program code.
  • computer usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • these computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device.
  • the instruction means implements the functions specified in one or more blocks of the flowchart or in a flow or block diagram of the flowchart.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Remote Sensing (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Human Computer Interaction (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Astronomy & Astrophysics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Image Analysis (AREA)

Abstract

本发明实施例一种可移动平台的控制方法、设备、计算机可读存储介质，所述方法包括：获取拍摄图像；从所述拍摄图像中识别出第一特征部位；确定所述拍摄图像中目标对象的第二特征部位；根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数；根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位；若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。应用本发明实施例，可以防止跟踪目标对象的丢失，并提高跟踪的可靠性。

Description

可移动平台的控制方法、设备、计算机可读存储介质 技术领域
本发明涉及电子技术领域，尤其是涉及一种可移动平台的控制方法、设备、计算机可读存储介质。
背景技术
现有的跟踪策略是把目标对象的一个具有明显特征的特征部位作为目标来跟踪。通常在对目标对象的某个固定的特征部位进行跟踪的过程中，由于可移动平台与目标对象的距离在变化，目标对象的特征部位的跟踪框在拍摄图像中的尺寸占比也随之变化，这样会影响跟踪的效果。
例如，当可移动平台与目标对象的距离很近时，则目标对象的特征部位的跟踪框在拍摄图像中的尺寸占比较大，会造成跟踪速度变慢，进而容易造成目标对象跟踪丢失，跟踪的可靠性变差；当可移动平台与目标对象的距离较远时，则目标对象的特征部位的跟踪框在拍摄图像中的尺寸占比较小，会造成跟踪到的目标对象的特征模糊，跟踪的可靠性变差。
发明内容
本发明提供一种可移动平台的控制方法、设备、计算机可读存储介质，可以防止跟踪的目标对象丢失，并可以提高跟踪的可靠性。
本发明实施例第一方面，提供一种可移动平台的控制方法，包括：
获取拍摄图像；
从所述拍摄图像中识别出第一特征部位；
确定所述拍摄图像中目标对象的第二特征部位；
根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部 位对应的匹配参数；
根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位；
若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
本发明实施例第二方面，提供一种可移动平台，包括：存储器和处理器；
所述存储器，用于存储程序代码；所述处理器，用于调用所述程序代码，当程序代码被执行时，用于执行以下操作：
获取拍摄图像；
从所述拍摄图像中识别出第一特征部位；
确定所述拍摄图像中目标对象的第二特征部位；
根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数；
根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位；
若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
本发明实施例第三方面，提供一种计算机可读存储介质，所述计算机可读存储介质上存储有计算机指令，所述计算机指令被执行时，实现如本发明实施例第一方面所述的可移动平台的控制方法。
基于上述技术方案，本发明实施例中，可以根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数，根据第一特征部位对应的匹配参数从图像中识别的第一特征部位中确定目标对象的第一特征部位，并在目标对象的跟踪参数满足预设跟踪条件时，将对目标对象的 第二特征部位进行跟踪切换到对目标对象的第一特征部位进行跟踪，通过这种方式，可以使得可移动平台对与当前目标对象的跟踪参数相适应的特征部位来进行跟踪，从而防止跟踪目标对象的丢失，并提高跟踪的可靠性。
附图说明
为了更加清楚地说明本发明实施例中的技术方案，下面将对本发明实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明中记载的一些实施例，对于本领域普通技术人员来讲，还可以根据本发明实施例的这些附图获得其它的附图。
图1是一个可移动平台的控制方法的流程示意图；
图2是远场跟踪切换近场跟踪时，将对目标对象的人体进行跟踪切换到对目标对象的头肩进行跟踪的示意图；
图3是近场跟踪切换远场跟踪时，将对目标对象的头肩进行跟踪切换到对目标对象的人体进行跟踪的示意图；
图4是远场跟踪切换近场跟踪时，根据预设的比例关系和目标对象的人体预测头肩的示意图；
图5是远场跟踪切换近场跟踪时，根据目标对象的关节点来预测头肩的示意图；
图6是近场跟踪切换远场跟踪时，根据预设的比例关系和目标对象的头肩预测人体的示意图；
图7是近场跟踪切换远场跟踪时，根据目标对象的关节点来预测人体的示意图；
图8是一个可移动平台的结构示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。另外，在不冲突的情况下，下述的实施例及实施例中的特征可以相互组合。
本发明使用的术语仅仅是出于描述特定实施例的目的，而非限制本发明。本发明和权利要求书所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式，除非上下文清楚地表示其它含义。应当理解的是，本文中使用的术语“和/或”是指包含一个或多个相关联的列出项目的任何或所有可能组合。
尽管在本发明可能采用术语第一、第二、第三等来描述各种信息，但这些信息不应限于这些术语。这些术语用来将同一类型的信息彼此区分开。例如，在不脱离本发明范围的情况下，第一信息也可以被称为第二信息，类似地，第二信息也可以被称为第一信息。取决于语境，此外，所使用的词语“如果”可以被解释成为“在……时”，或者，“当……时”，或者，“响应于确定”。
本发明实施例中提出一种可移动平台的控制方法，所述可移动平台可以包括但不限于无人机、地面机器人(例如无人车等)。可移动平台可以配置拍摄设备，并可以通过拍摄设备采集拍摄图像，另外，可移动平台可以配置有云台，云台可以承载拍摄设备(如相机、摄像机等)以为拍摄设备增稳和/或调整。本发明实施例中以可移动平台为无人机为例来进行示意性说明，可以理解的是本文下述部分的无人机可以可移动平台来进行替换。
传统跟踪技术是对目标对象的单一特征部位进行跟踪，如将目标对象的人体作为跟踪目标，或者，将目标对象的人体的预设部位(例如人体的头部)作为跟踪目标。然而，在对目标对象的单一特征部位进行跟踪的过程中，由于可移动平台与目标对象的距离在变化，目标对象的特征部位的跟踪框在拍 摄图像中的尺寸占比也随之变化，这样会影响跟踪的效果。
例如，当可移动平台与目标对象的距离很近时，则目标对象的特征部位的跟踪框在拍摄图像中的尺寸占比较大，会造成跟踪速度变慢，进而容易造成目标对象跟踪丢失，跟踪的可靠性变差；当无人机与目标对象的距离较远时，则目标对象的特征部位的跟踪框在拍摄图像中的尺寸占比较小，会造成跟踪到的目标对象的特征模糊，跟踪的可靠性变差。
针对上述发现，本发明实施例中，当无人机与目标对象的距离较近时，可以将目标对象的人体的预设部位作为跟踪目标，即无人机对目标对象的人体的预设部位进行跟踪；当无人机与目标对象的距离较远时，可以将目标对象的人体作为跟踪目标，即无人机对目标对象的人体进行跟踪，上述方式可以达到更好的跟踪效果。其中，在无人机对目标对象进行跟踪的过程中，可以包括如下四种情况：
第一种情况：近场跟踪。当无人机与目标对象的距离较近时，可以采用近场跟踪方式，即将目标对象的人体的预设部位(如头部，或者头部和肩部(可以简称为头肩)等)作为跟踪目标，并对目标对象的人体的预设部位进行跟踪。
第二种情况：远场跟踪。当无人机与目标对象的距离较远时，可以采用远场跟踪方式，即将目标对象的人体作为跟踪目标，并对目标对象的人体进行跟踪。
第三种情况：由远场跟踪切换到近场跟踪。首先，无人机与目标对象的距离较远，无人机采用远场跟踪方式，对目标对象的人体进行跟踪。然后，无人机与目标对象的距离越来越近，当无人机与目标对象之间的距离小于或等于某个距离阈值时，将远场跟踪方式切换为近场跟踪方式，也就是说，无人机不再对目标对象的人体进行跟踪，而是将目标对象的人体的预设部位作为跟踪目标，并对目标对象的人体的预设部位进行跟踪。
第四种情况：由近场跟踪切换到远场跟踪。首先，无人机与目标对象的距离较近，无人机采用近场跟踪方式，对目标对象的人体的预设部位进行跟踪。然后，无人机与目标对象的距离越来越远，当无人机与目标对象之间的距离大于或等于某个距离阈值时，将近场跟踪方式切换为远场跟踪方式，也就是说，无人机不再对目标对象的人体的预设部位进行跟踪，而是将目标对象的人体作为跟踪目标，并对目标对象的人体进行跟踪。
以下结合具体实施例，对远场跟踪切换到近场跟踪、近场跟踪切换远场跟踪的过程进行说明。参见图1所示，为可移动平台的控制方法的流程图，该方法包括：
步骤101，获取拍摄图像。
具体地，所述方法的执行主体可以为可移动平台，进一步地可以为可移动平台的处理器，其中，所述处理器可以为一个或多个，所述处理器可以为通用处理器或者专用处理器。
如前所述，可移动平台可以配置有拍摄设备，在可移动平台对目标对象进行跟踪的过程中，拍摄设备可以对目标对象进行拍摄以获取拍摄图像，可移动平台的处理器可以获取所述拍摄图像，其中，目标对象为可移动平台跟踪的对象。
步骤102，从拍摄图像中识别出第一特征部位。
具体地，可移动平台的处理器可以从拍摄图像中识别出第一特征部位，进一步地，可以利用神经网络(例如卷积神经网络)来识别图像中的第一特征部位。其中，神经网络在检测到拍摄图像中的第一特征部位后，可以返回所述第一特征部位在拍摄图像中的位置和对应的图像区域。其中，所述在拍摄图像中的位置和对应的图像区域可以以第一特征部位的检测框来表示，即第一特征部位可以以检测框表示。
步骤103，确定拍摄图像中目标对象的第二特征部位。
具体地，处理器在获取到拍摄图像时，可以从拍摄图像中确定目标对象的第二特征部位，其中，从拍摄图像中确定目标对象的第二特征部位可以采用追踪算法：在获取到当前一帧拍摄图像后，根据上一帧拍摄图像中目标对象的第二特征部位的位置确定目标区域，然后，在当前一帧拍摄图像的目标区域中查找与上一帧拍摄图像中目标对象的第二特征部位最匹配的图像区域，并将所述最匹配的图像区域确定为当前一帧拍摄图像中目标对象的第二特征部位。其中，当前一帧拍摄图像中目标对象的第二特征部位可以以检测框表示。
步骤104，根据目标对象的第二特征部位和所述第一特征部位确定第一特征部位对应的匹配参数。
具体地，从拍摄图像中识别的第一特征部位可能包括多个，其中，第一特征部位中包括目标对象的特征部位。在远场跟踪切换到近场跟踪，或者近场跟踪切换到远场跟踪时，需要将对目标对象的第二特征部位进行跟踪切换到对目标对象的第一特征部位进行跟踪。因此，需要从拍摄图像中识别的第一特征部位中确定到底哪一个第一特征部位是目标对象的第一特征部位。
为了确定目标对象的第一特征部位，可以根据目标对象的第二特征部位和所述第一特征部位确定第一特征部位对应的匹配参数，其中，每一个第一特征部位可以对应一个匹配参数，匹配参数可以用于表示某个第一特征部位是目标对象的第一特征部位的可能性。其中，所述匹配参数包括重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种或多种，可以理解的是，匹配参数可以是重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种，也可以是根据重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的多种来确定的。
步骤105，根据第一特征部位对应的匹配参数从第一特征部位中确定目标对象的第一特征部位。
具体的，如前所述，可能从拍摄图像中识别多个第一特征部位，其中， 所述多个第一特征部位中包括目标对象的第一特征部位，可以根据每一个第一特征部位对应的匹配参数来确定哪一个第一特征部位为目标对象的第一特征部位。
在某些实施例中，根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位，包括：从所述第一特征部位对应的匹配参数中确定最大的匹配参数，将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。具体地，当匹配参数越大时，可以认为与所述匹配参数对应的第一特征部位越有可能是目标对象的第一特征部位，可以从确定的匹配参数中确定一个最大的匹配参数，将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
步骤106，若目标对象的跟踪参数满足预设跟踪条件时，将对目标对象的第二特征部位进行跟踪切换到对目标对象的第一特征部位进行跟踪。
具体地，目标对象的跟踪参数可以包括：该目标对象的跟踪框在拍摄图像中的尺寸占比，和/或，该目标对象与可移动平台之间的距离。在目标对象的跟踪参数不满足预设跟踪条件时，可移动平台通过对目标对象的第二特征部位进行跟踪实现对目标对象的跟踪，当目标对象的跟踪参数满足预设跟踪条件时，对目标对象的第二特征部位进行跟踪可能会导致跟踪的可靠性变差，此时，可以切换为对与当前目标对象的跟踪参数相适应的目标对象的第一特征部位进行跟踪。
在某些实施例中，针对远场跟踪切换到近场跟踪的情况，所述第一特征部位为人体的预设部位，所述第二特征部位为人体，若目标对象的跟踪参数满足第二预设跟踪条件时，则将对目标对象的第二特征部位进行跟踪切换到对目标对象的第一特征部位进行跟踪。具体地，在目标对象的跟踪参数满足第二预设跟踪条件之前，可移动平台采用远场跟踪策略，对目标对象的第二特征部位进行跟踪，一旦确定目标对象的跟踪参数满足第二预设跟踪条件时，可移动平台切换到近场跟踪策略，将对目标对象的第二特征部位进行跟踪切 换到对目标对象的第一特征部位进行跟踪。其中，目标对象的跟踪参数满足第二预设跟踪条件，可以包括但不限于：目标对象的跟踪框在拍摄图像的尺寸占比大于或等于预设第二占比阈值，和/或，目标对象与可移动平台之间的距离小于或等于预设第二距离。
其中，在该尺寸占比大于或等于预设第二占比阈值，和/或，该目标对象与可移动平台之间的距离小于或等于预设第二距离时，说明可移动平台与目标对象的距离较小，可以从远场跟踪切换到近场跟踪，即将对目标对象的人体进行跟踪切换到对目标对象的人体的预设部位进行跟踪。此外，预设第二占比阈值、预设第二距离可以根据经验配置，对此不做限制。
例如，如图2所示，在目标对象的跟踪参数不满足第二预设跟踪条件时，可移动平台对目标对象的人体进行跟踪。采用如前所述的方法，根据目标对象的人体211和头肩201、头肩202和头肩203确定每一个头肩对应的匹配参数，根据匹配参数确定头肩201为目标对象的头肩，当在某一个时刻，目标对象的跟踪参数满足第二预设跟踪条件时，将对目标对象的人体进行跟踪切换到对目标对象的头肩201进行跟踪。
在某些实施例中，针对近场跟踪切换到远场跟踪的情况，所述第一特征部位为人体，所述第二特征部位为人体的预设部位，若目标对象的跟踪参数满足第一预设跟踪条件时，则将对目标对象的第二特征部位进行跟踪切换到对目标对象的第一特征部位进行跟踪。具体地，在目标对象的跟踪参数满足第一预设跟踪条件之前，可移动平台采用近场跟踪策略，对目标对象的第二特征部位进行跟踪，一旦确定目标对象的跟踪参数满足第一预设跟踪条件时，可移动平台切换到远场跟踪策略，将对目标对象的第二特征部位进行跟踪切换到对目标对象的第一特征部位进行跟踪。其中，目标对象的跟踪参数满足第一预设跟踪条件，可以包括但不限于：目标对象的跟踪框在拍摄图像的尺寸占比小于或等于预设第一占比阈值，和/或，目标对象与可移动平台之间的距离大于或等于预设第一距离。
其中，在该尺寸占比小于或等于预设第一占比阈值，和/或，该目标对象 与可移动平台之间的距离大于或等于预设第一距离时，说明可移动平台与目标对象的距离较大，可以从近场跟踪切换到远场跟踪，即将对目标对象的人体的预设部位进行跟踪切换到对目标对象的人体进行跟踪。此外，预设第一占比阈值、预设第一距离可以根据经验配置，对此不做限制。
例如，如图3所示，在目标对象的跟踪参数不满足第一预设跟踪条件时，可移动平台对目标对象的人体的头肩进行跟踪。采用如前所述的方法，根据目标对象的头肩231和人体221、人体222和人体223确定每一个人体对应的匹配参数，根据匹配参数确定人体221为目标对象的人体，当在某一个时刻，目标对象的跟踪参数满足第一预设跟踪条件时，将对目标对象的头肩进行跟踪切换到对目标对象的人体221进行跟踪。
在某些实施例中，根据目标对象的第二特征部位和第一特征部位确定第一特征部位对应的匹配参数，可以通过如下几种可行的方式实现：
在一种可行的实施方式中，可以根据目标对象的第二特征部位和第一特征部位确定该第二特征部位与该第一特征部位的匹配参数。具体地，目标对象的第二特征部位与第一特征部位的匹配参数，可以表示目标对象的第二特征部位与第一特征部位之间的匹配程度，匹配程度越高，表示某个第一特征部位越有可能为目标对象的第一特征部位。
参见图2所示，针对远场跟踪切换近场跟踪的情况，可以根据目标对象的人体211和头肩201确定目标对象的人体211与头肩201的匹配参数，根据目标对象的人体211和头肩202确定目标对象的人体211与头肩202的匹配参数，根据目标对象的人体211和头肩203确定目标对象的人体211与头肩203的匹配参数。下面将详细解释匹配参数的确定过程。
如前所述，匹配参数是重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种。因此，可以根据目标对象的人体211和头肩201确定目标对象的人体211和头肩201之间的重合程度匹配参数，其中，重合程度匹配参数用于表示两个图像区域的重合程度，重合程度匹配参数可以 用两个图像区域的交集与两个图像区域的并集之比来表示。在本实施方式中，即可以通过目标对象的人体211与头肩201的交集与目标对象的人体211与头肩201的并集之比，作为目标对象的人体211和头肩201之间的重合程度匹配参数。通过确定目标对象的人体211和头肩201之间的重合程度匹配系数，可以知道目标对象的人体211与头肩201之间的重合程度，其中，当重合程度越高时，可以表示头肩201越有可能是目标对象的头肩。
再例如，可以根据目标对象的人体211和头肩201确定目标对象的人体211和头肩201之间的图像相似程度匹配参数，其中，图像相似程度匹配参数用于表示两个图像区域内图像的相似程度。图像相似程度匹配参数可以用两个图像区域内图像的直方图来确定，在本实施方式中，分别确定目标对象的人体211和头肩201的直方图，计算两个直方图的归一化相关系数(如巴氏距离，直方图相交距离等)，将所述归一化相关系数作为图像相似程度匹配参数，当然，本领域技术人员可以采用其他方式来确定图像相似程度匹配参数，在这里不作具体的限定。通过确定目标对象的人体211和头肩201之间的图像相似程度匹配参数，可以知道目标对象的人体211与头肩201之间的图像相似程度，其中，当相似程度越高时，可以表示头肩201越有可能是目标对象的头肩。
再例如，可以根据目标对象的人体211和头肩201确定目标对象的人体211和头肩201之间的几何程度匹配参数，其中，几何程度匹配参数用于表示两个图像区域的大小匹配程度。通常，在连续图像帧中，目标对象的自身运动变化程度不会很大，与可移动平台之间的距离变化也比较小，因此，图像中目标对象的特征部位之间的比例关系应该是呈预设比例。在本实施方式中，即可以通过目标对象的人体211的面积和头肩201的面积，确定目标对象的人体211与头肩201的面积之比，然后进一步地确定面积之比与预设比例之间的差异。其中，在本实施方式中，预设比例可以是人体与头肩之间的大小比例，根据所述差异确定几何程度匹配参数。通过确定目标对象的人体211 和头肩201之间的几何程度匹配参数，可以知道目标对象的人体211与头肩201之间的大小匹配程度，其中，当大小匹配程度时，可以表示头肩201越有可能是目标对象的头肩。
可以理解的是，人体211和头肩201之间的匹配参数可以根据重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种或多种来确定，即本领域技术人员可以将重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种作为人体211和头肩201之间的匹配参数，也可以将重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的多种进行融合以确定人体211和头肩201之间的匹配参数，在这里，不作具体的限定。采用如上所述的方法，还可以确定人体211和头肩202之间的匹配参数，还可以确定人体211和头肩201之间的匹配参数。
参见图3所示，针对近场跟踪切换到远场跟踪的情况，可以根据目标对象的第二特征部位231和第一特征部位221确定第二特征部位231与第一特征部位221的匹配参数；可以根据目标对象的第二特征部位231和第一特征部位222确定第二特征部位231与第一特征部位222的匹配参数；可以根据目标对象的第二特征部位231和第一特征部位223确定第二特征部位231与第一特征部位223的匹配参数。其中，具体确定第二特征部位231与第一特征部位221、第二特征部位231与第一特征部位222和第二特征部位231与第一特征部位223的匹配参数的方法请参见前述部分，此处不再赘述。
在另一种可行的实施方式中，可以预测第三特征部位，其中，第三特征部位为根据目标对象的第二特征部位预测的目标对象的第一特征部位。然后，根据第三特征部位和第一特征部位确定该第三特征部位与第一特征部位的匹配参数。
其中，预测目标对象的第三特征部位，可以包括：根据预设的比例关系和目标对象的第二特征部位预测第三特征部位。或者，根据目标对象的第二特征部位确定目标对象的关节点，根据所述关节点预测第三特征部位。其中， 预设的比例关系可以是通常情况下人的第一特征部位与人的第二特征部位之间的比例关系，可以是一个经验值。
参见图4所示，针对远场跟踪切换近场跟踪的情况，可以根据预设的比例关系和目标对象的人体211预测第三特征部位212，即预测的目标对象的头肩212，其中，所述预设的比例关系是人体和人体的头肩之间的比例关系。
参见图5所示，针对远场跟踪切换近场跟踪的情况，可以根据目标对象的人体211来确定目标对象的关节点，然后根据目标对象的关节点来预测头肩，例如，从目标对象的关节点中确定肩膀的关节点和眼睛的关节点，根据肩膀的关节点和眼睛的关节点来预测得到目标对象的头肩213。
在得到预测的头肩213之后，就可以根据预测的头肩213和头肩201确定预测的头肩213和头肩201的匹配参数，预测的头肩213和头肩202确定预测的头肩213和头肩202的匹配参数，预测的头肩213和头肩203确定预测的头肩213和头肩203的匹配参数。具体确定匹配参数的方法请参见前述部分，此处不再赘述。
参见图6所示，针对近场跟踪切换远场跟踪的情况，可以根据预设的比例关系和目标对象的头肩231预测人体232，即预测的人体232，其中，所述预设的比例关系是人体和人体的头肩之间的比例关系。
参见图7所示，针对近场跟踪切换远场跟踪的情况，可以根据目标对象的头肩231来确定目标对象的关节点，然后根据目标对象的关节点来预测人体233，例如，从目标对象的关节点中确定脚的关节点、肩膀的关节点和眼睛的关节点，根据脚的关节点、肩膀的关节点和眼睛的关节点来预测得到目标对象的人体233。
在得到预测的人体233之后，就可以根据预测的人体233和人体221确定预测的人体233和人体221的匹配参数，预测的人体233和人体222确定预测的人体233和人体222的匹配参数，预测的人体233和人体223确定预 测的人体233和人体223。具体确定匹配参数的方法请参见前述部分，此处不再赘述。
在上述实施例中，考虑到多人重合的情况，还可以确定最大的匹配参数的第一特征部位的跟踪框的上边沿与第二特征部位的跟踪框的上边沿之间的距离。进一步的，将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位，可以包括：当所述距离小于或等于预设距离阈值(该预设距离阈值可以根据经验配置，对此不做限制)时，可以将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。此外，当所述距离大于预设距离阈值时，则在排除最大的匹配参数的第一特征部位的基础上，重新确定最大的匹配参数，以此类推。
例如，在图2中，若最大的匹配参数的第一特征部位201的跟踪框的上边沿与第二特征部位211的跟踪框的上边沿之间的距离小于预设距离阈值，则将第一特征部位201确定为目标对象的第一特征部位。当该距离大于预设距离阈值时，则排除第一特征部位201，并从第一特征部位202和第一特征部位203中重新确定最大的匹配参数对应的第一特征部位，以此类推。
在上述实施例中，在远场跟踪时，跟踪的是目标对象的人体，在可移动平台接近目标对象时，目标对象的人体的跟踪框在拍摄图像中的尺寸占比会变大，进而导致跟踪速度变慢，因此，可以将远场跟踪切换到近场跟踪，即跟踪的是目标对象的头肩，以提高跟踪的效率。在近场跟踪时，跟踪的是目标对象的头肩，在可移动平台远离目标对象时，目标对象的头肩的跟踪框在拍摄图像中的尺寸占比会变小，进而跟踪的准确度会变差，因此，可以将近场跟踪切换到远场跟踪，即跟踪的是目标对象的人体，以提高跟踪的准确度。
由于跟踪框的大小可能会发生一定的变化，跟踪框的大小可能与跟踪的目标对象的大小不一致，因此，可以使用其它辅助方式来确定切换条件，以提高切换准确度。例如，采用深度传感器技术及图像投影关系，找到拍摄图像中目标对象对应处的深度，从而得到可移动平台与目标对象之间的距离，如果距离过近(如小于或等于3米)则切换为近场跟踪，如果距离过远(如 大于4米)则切为远场跟踪。此外，也可以直接采用测距的方式(如双目测距、超声波测距或激光雷达测距等)测出目标对象与可移动平台之间的距离。
在一个例子中，上述实施例所适用的应用场景可以包括但不限于：
自拍模式。在目标对象是人时，无人机检测到目标对象时，进行跟踪框大小检测，如果发现跟踪框在拍摄图像中的尺寸占比大于或等于30％，则自动对目标对象的头肩进行检测，即进行近场跟踪。当目标对象离无人机较远时，如发现跟踪框在拍摄图像中的尺寸占比小于10％，则将近场跟踪切换到远场跟踪，自动对目标对象的人体进行检测。
扫脸起飞。在无人机具有扫脸起飞功能时，在扫脸成功后，开始进行近场跟踪(例如跟踪头肩)，当目标对象离无人机较远时，可以自动切换为远场跟踪(例如跟踪人体)，从而可以达到自动以开机用户为焦点的效果。
由近及远的录像。无人机可以直接从用户手上起飞，向用户的斜后上方飞行，并开始跟踪目标对象的头肩。在无人机飞出后，若目标对象离无人机较远，则可以自动切换为远场跟踪，即跟踪目标对象的人体。
由近及远的环绕录像。在无人机对焦到人体后，无人机向外螺旋环绕录像，无人机可以直接从用户手上起飞，并开始进行螺旋线飞行拍摄，即跟踪目标对象的头肩。在无人机飞出后，若目标对象离无人机较远，则可以自动切换为远场跟踪，即跟踪目标对象的人体。
基于与上述方法同样的构思，参见图8所示，本发明实施例中还提供一种可移动平台30，包括存储器31和处理器32(如一个或多个处理器)。
在一个例子中，所述存储器，用于存储程序代码；所述处理器，用于调用所述程序代码，当程序代码被执行时，用于执行以下操作：获取拍摄图像；
从所述拍摄图像中识别出第一特征部位；
确定所述拍摄图像中目标对象的第二特征部位；
根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数；
根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位；
若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
所述处理器根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数时具体用于：根据目标对象的第二特征部位和所述第一特征部位确定所述第二特征部位与所述第一特征部位的匹配参数。
优选的，所述处理器根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数时具体用于：预测所述目标对象的第三特征部位；其中，所述第三特征部位为根据所述第二特征部位预测的目标对象的第一特征部位；根据目标对象的第三特征部位和所述第一特征部位确定所述第三特征部位与所述第一特征部位的匹配参数。
优选的，所述处理器预测所述目标对象的第三特征部位时具体用于：根据目标对象的第二特征部位与目标对象的第三特征部位的比例关系和所述第二特征部位预测所述目标对象的第三特征部位。
优选的，所述处理器预测所述目标对象的第三特征部位时具体用于：根据所述目标对象的第二特征部位确定所述目标对象的关节点信息；根据所述关节点信息预测所述目标对象的第三特征部位。
优选的，所述处理器根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位时具体用于：从所述第一特征部位对应的匹配参数中确定最大的匹配参数；将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
优选的，所述第一特征部位为人体，所述第二特征部位为人体的预设部位，所述处理器在所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪时具体用于：若所述目标对象的跟踪参数满足第一预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特 征部位进行跟踪。
所述目标对象的跟踪参数满足第一预设跟踪条件，包括：所述目标对象的跟踪框在所述拍摄图像的尺寸占比小于或等于预设第一占比阈值，和/或，所述目标对象与所述可移动平台之间的距离大于或等于预设第一距离。
优选的，所述第一特征部位为人体的预设部位，所述第二特征部位为人体，所述处理器在所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪时具体用于：若所述目标对象的跟踪参数满足第二预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
所述目标对象的跟踪参数满足第二预设跟踪条件，包括：所述目标对象的跟踪框在所述拍摄图像的尺寸占比大于或等于预设第二占比阈值，和/或，所述目标对象与所述可移动平台之间的距离小于或等于预设第二距离。
优选的，所述预设部位包括：头部，或者头部和肩部。
优选的，所述处理器还用于：确定所述最大的匹配参数的第一特征部位的跟踪框的上边沿与所述第二特征部位的跟踪框的上边沿之间的距离；
所述处理器将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位时具体用于：当所述距离小于或等于预设距离阈值时，将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
优选的，所述匹配参数包括重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种或多种。
基于与上述方法同样的发明构思，本发明实施例中还提供一种计算机可读存储介质上存储有计算机指令，所述计算机指令被执行(如所述计算机指令被处理器执行)时，可以实现图1所述的可移动平台的控制方法。
上述实施例阐明的系统、装置、模块或单元，可以由计算机芯片或实体实现，或者由具有某种功能的产品来实现。一种典型的实现设备为计算机，计算机的具体形式可以是个人计算机、膝上型计算机、蜂窝电话、相机电话、 智能电话、个人数字助理、媒体播放器、导航设备、电子邮件收发设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任意几种设备的组合。
为了描述的方便，描述以上装置时以功能分为各种单元分别描述。当然，在实施本发明时可以把各单元的功能在同一个或多个软件和/或硬件中实现。
本领域内的技术人员应明白，本发明实施例可提供为方法、系统、或计算机程序产品。因此，本发明可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本发明实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本发明是参照根据本发明实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可以由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其它可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其它可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
而且，这些计算机程序指令也可以存储在能引导计算机或其它可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或者多个流程和/或方框图一个方框或者多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其它可编程数据处理设备，使得在计算机或者其它可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其它可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
以上所述仅为本发明实施例而已，并不用于限制本发明。对于本领域技术人员来说，本发明可以有各种更改和变化。凡在本发明的精神和原理之内 所作的任何修改、等同替换、改进，均应包含在本发明的权利要求范围之内。

Claims (27)

  1. 一种可移动平台的控制方法，其特征在于，所述方法包括：
    获取拍摄图像；
    从所述拍摄图像中识别出第一特征部位；
    确定所述拍摄图像中目标对象的第二特征部位；
    根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数；
    根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位；
    若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
  2. 根据权利要求1所述的方法，其特征在于，根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数包括：
    根据目标对象的第二特征部位和所述第一特征部位确定所述第二特征部位与所述第一特征部位的匹配参数。
  3. 根据权利要求1所述的方法，其特征在于，根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数包括：
    预测所述目标对象的第三特征部位；其中，所述第三特征部位为根据所述第二特征部位预测的目标对象的第一特征部位；
    根据目标对象的第三特征部位和所述第一特征部位确定所述第三特征部位与所述第一特征部位的匹配参数。
  4. 根据权利要求3所述的方法，其特征在于，
    预测所述目标对象的第三特征部位，包括：
    根据目标对象的第二特征部位与目标对象的第三特征部位的比例关系和所述第二特征部位预测所述目标对象的第三特征部位。
  5. 根据权利要求3所述的方法，其特征在于，
    预测所述目标对象的第三特征部位，包括：
    根据所述目标对象的第二特征部位确定所述目标对象的关节点信息；
    根据所述关节点信息预测所述目标对象的第三特征部位。
  6. 根据权利要求1所述的方法，其特征在于，
    根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位，包括：
    从所述第一特征部位对应的匹配参数中确定最大的匹配参数；将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
  7. 根据权利要求1所述的方法，其特征在于，
    所述第一特征部位为人体，所述第二特征部位为人体的预设部位，
    所述若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪包括：
    若所述目标对象的跟踪参数满足第一预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
  8. 根据权利要求7所述的方法，其特征在于，
    所述目标对象的跟踪参数满足第一预设跟踪条件，包括：所述目标对象的跟踪框在所述拍摄图像的尺寸占比小于或等于预设第一占比阈值，和/或，所述目标对象与所述可移动平台之间的距离大于或等于预设第一距离。
  9. 根据权利要求1所述的方法，其特征在于，
    所述第一特征部位为人体的预设部位，所述第二特征部位为人体，
    所述若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪包括：
    若所述目标对象的跟踪参数满足第二预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
  10. 根据权利要求9所述的方法，其特征在于，
    所述目标对象的跟踪参数满足第二预设跟踪条件，包括：所述目标对象的跟踪框在所述拍摄图像的尺寸占比大于或等于预设第二占比阈值，和/或，所述目标对象与所述可移动平台之间的距离小于或等于预设第二距离。
  11. 根据权利要求7或9所述的方法，其特征在于，
    所述预设部位包括：头部，或者头部和肩部。
  12. 根据权利要求6所述的方法，其特征在于，
    所述方法还包括：确定所述最大的匹配参数的第一特征部位的跟踪框的上边沿与所述第二特征部位的跟踪框的上边沿之间的距离；
    所述将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位包括：当所述距离小于或等于预设距离阈值时，将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
  13. 根据权利要求1-12任一项所述的方法，其特征在于，
    所述匹配参数包括重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种或多种。
  14. 一种可移动平台，其特征在于，包括：存储器和处理器；
    所述存储器，用于存储程序代码；
    所述处理器，用于调用所述程序代码，当程序代码被执行时，用于执行以下操作：获取拍摄图像；
    从所述拍摄图像中识别出第一特征部位；
    确定所述拍摄图像中目标对象的第二特征部位；
    根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数；
    根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位；
    若所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
  15. 根据权利要求14所述的可移动平台，其特征在于，所述处理器根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数时具体用于：根据目标对象的第二特征部位和所述第一特征部位确定所述第二特征部位与所述第一特征部位的匹配参数。
  16. 根据权利要求14所述的可移动平台，其特征在于，
    所述处理器根据目标对象的第二特征部位和所述第一特征部位确定所述第一特征部位对应的匹配参数时具体用于：预测所述目标对象的第三特征部位；其中，所述第三特征部位为根据所述第二特征部位预测的目标对象的第一特征部位；根据目标对象的第三特征部位和所述第一特征部位确定所述第三特征部位与所述第一特征部位的匹配参数。
  17. 根据权利要求16所述的可移动平台，其特征在于，
    所述处理器预测所述目标对象的第三特征部位时具体用于：根据目标对象的第二特征部位与目标对象的第三特征部位的比例关系和所述第二特征部位预测所述目标对象的第三特征部位。
  18. 根据权利要求16所述的可移动平台，其特征在于，
    所述处理器预测所述目标对象的第三特征部位时具体用于：根据所述目标对象的第二特征部位确定所述目标对象的关节点信息；根据所述关节点信息预测所述目标对象的第三特征部位。
  19. 根据权利要求14所述的可移动平台，其特征在于，
    所述处理器根据所述第一特征部位对应的匹配参数从所述第一特征部位中确定目标对象的第一特征部位时具体用于：从所述第一特征部位对应的匹配参数中确定最大的匹配参数；将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
  20. 根据权利要求14所述的可移动平台，其特征在于，
    所述第一特征部位为人体，所述第二特征部位为人体的预设部位，
    所述处理器在所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进 行跟踪时具体用于：若所述目标对象的跟踪参数满足第一预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
  21. 根据权利要求20所述的可移动平台，其特征在于，
    所述目标对象的跟踪参数满足第一预设跟踪条件，包括：所述目标对象的跟踪框在所述拍摄图像的尺寸占比小于或等于预设第一占比阈值，和/或，所述目标对象与所述可移动平台之间的距离大于或等于预设第一距离。
  22. 根据权利要求14所述的可移动平台，其特征在于，
    所述第一特征部位为人体的预设部位，所述第二特征部位为人体，
    所述处理器在所述目标对象的跟踪参数满足预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪时具体用于：若所述目标对象的跟踪参数满足第二预设跟踪条件时，将对所述目标对象的第二特征部位进行跟踪切换到对所述目标对象的第一特征部位进行跟踪。
  23. 根据权利要求22所述的可移动平台，其特征在于，
    所述目标对象的跟踪参数满足第二预设跟踪条件，包括：所述目标对象的跟踪框在所述拍摄图像的尺寸占比大于或等于预设第二占比阈值，和/或，所述目标对象与所述可移动平台之间的距离小于或等于预设第二距离。
  24. 根据权利要求20或22所述的可移动平台，其特征在于，
    所述预设部位包括：头部，或者头部和肩部。
  25. 根据权利要求19所述的可移动平台，其特征在于，
    所述处理器还用于：确定所述最大的匹配参数的第一特征部位的跟踪框的上边沿与所述第二特征部位的跟踪框的上边沿之间的距离；
    所述处理器将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位时具体用于：当所述距离小于或等于预设距离阈值时，将最大的匹配参数的第一特征部位确定为目标对象的第一特征部位。
  26. 根据权利要求14-25任一项所述的可移动平台，其特征在于，
    所述匹配参数包括重合程度匹配参数、图像相似程度匹配参数、几何相似程度匹配参数中的一种或多种。
  27. 一种计算机可读存储介质，其特征在于，
    所述计算机可读存储介质上存储有计算机指令，所述计算机指令被执行时，实现权利要求1-13任一项所述的可移动平台的控制方法。
PCT/CN2018/073769 2018-01-23 2018-01-23 可移动平台的控制方法、设备、计算机可读存储介质 WO2019144263A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
PCT/CN2018/073769 WO2019144263A1 (zh) 2018-01-23 2018-01-23 可移动平台的控制方法、设备、计算机可读存储介质
CN201880001860.3A CN109155067A (zh) 2018-01-23 2018-01-23 可移动平台的控制方法、设备、计算机可读存储介质
US16/935,709 US11227388B2 (en) 2018-01-23 2020-07-22 Control method and device for mobile platform, and computer readable storage medium
US17/648,179 US12002221B2 (en) 2018-01-23 2022-01-17 Control method and device for mobile platform, and computer readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/073769 WO2019144263A1 (zh) 2018-01-23 2018-01-23 可移动平台的控制方法、设备、计算机可读存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/935,709 Continuation US11227388B2 (en) 2018-01-23 2020-07-22 Control method and device for mobile platform, and computer readable storage medium

Publications (1)

Publication Number Publication Date
WO2019144263A1 true WO2019144263A1 (zh) 2019-08-01

Family

ID=64806196

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/073769 WO2019144263A1 (zh) 2018-01-23 2018-01-23 可移动平台的控制方法、设备、计算机可读存储介质

Country Status (3)

Country Link
US (2) US11227388B2 (zh)
CN (1) CN109155067A (zh)
WO (1) WO2019144263A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2021144260A (ja) * 2018-06-15 2021-09-24 ソニーグループ株式会社 情報処理装置、情報処理方法、プログラム、および情報処理システム
CN111145212B (zh) * 2019-12-03 2023-10-03 浙江大华技术股份有限公司 一种目标追踪处理方法及装置
CN112101288B (zh) * 2020-09-25 2024-02-13 北京百度网讯科技有限公司 安全帽佩戴检测方法、装置、设备及存储介质
CN112884804A (zh) * 2020-12-09 2021-06-01 深圳云天励飞技术股份有限公司 行动对象追踪方法及相关设备
US11394865B1 (en) 2021-06-28 2022-07-19 Western Digital Technologies, Inc. Low-power, fast-response machine learning autofocus enhancements
CN113469041A (zh) * 2021-06-30 2021-10-01 北京市商汤科技开发有限公司 一种图像处理方法、装置、计算机设备和存储介质
WO2023065165A1 (zh) * 2021-10-20 2023-04-27 深圳市大疆创新科技有限公司 可移动平台的控制方法、装置、可移动平台和存储介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101251895A (zh) * 2008-03-13 2008-08-27 上海交通大学 基于高斯混合模型的人体跟踪方法
CN105654512A (zh) * 2015-12-29 2016-06-08 深圳羚羊微服机器人科技有限公司 一种目标跟踪方法和装置
CN104081757B (zh) * 2012-02-06 2017-10-17 索尼公司 图像处理设备、图像处理方法、图像处理系统

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004107266A1 (en) * 2003-05-29 2004-12-09 Honda Motor Co., Ltd. Visual tracking using depth data
TWI351001B (en) * 2007-11-21 2011-10-21 Ind Tech Res Inst Method and apparatus for adaptive object detection
JP4784678B2 (ja) * 2009-04-22 2011-10-05 カシオ計算機株式会社 撮像装置、撮像方法及びプログラム
JPWO2013001941A1 (ja) * 2011-06-27 2015-02-23 日本電気株式会社 物体検知装置、物体検知方法および物体検知プログラム
US20130100140A1 (en) * 2011-10-25 2013-04-25 Cywee Group Limited Human body and facial animation systems with 3d camera and method thereof
CN106485735A (zh) * 2015-09-01 2017-03-08 南京理工大学 基于立体视觉技术的人体目标识别与跟踪方法
JP6755713B2 (ja) * 2016-05-25 2020-09-16 キヤノン株式会社 追尾装置、追尾方法及びプログラム
CN106598226B (zh) * 2016-11-16 2019-05-21 天津大学 一种基于双目视觉和深度学习的无人机人机交互方法
WO2018195979A1 (zh) * 2017-04-28 2018-11-01 深圳市大疆创新科技有限公司 一种跟踪控制方法、装置及飞行器
US11048919B1 (en) * 2018-05-30 2021-06-29 Amazon Technologies, Inc. Person tracking across video instances

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101251895A (zh) * 2008-03-13 2008-08-27 上海交通大学 基于高斯混合模型的人体跟踪方法
CN104081757B (zh) * 2012-02-06 2017-10-17 索尼公司 图像处理设备、图像处理方法、图像处理系统
CN105654512A (zh) * 2015-12-29 2016-06-08 深圳羚羊微服机器人科技有限公司 一种目标跟踪方法和装置

Also Published As

Publication number Publication date
US20220138946A1 (en) 2022-05-05
CN109155067A (zh) 2019-01-04
US11227388B2 (en) 2022-01-18
US20200349704A1 (en) 2020-11-05
US12002221B2 (en) 2024-06-04

Similar Documents

Publication Publication Date Title
WO2019144263A1 (zh) 可移动平台的控制方法、设备、计算机可读存储介质
US11812132B2 (en) Imaging device, control method therefor, and recording medium
US9321173B2 (en) Tracking and following people with a mobile robotic device
CN109891189B (zh) 策划的摄影测量
WO2019179441A1 (zh) 智能设备的焦点跟随方法、装置、智能设备及存储介质
BR112020006277A2 (pt) aparelho de captura de imagem e método de controle para o mesmo
US10547774B2 (en) Image processing device, image processing method, and program
WO2019179443A1 (zh) 智能设备的持续唤醒方法、装置、智能设备和存储介质
WO2019061079A1 (zh) 一种对焦处理方法及设备
WO2018120033A1 (zh) 一种辅助用户寻物的方法及装置
WO2021026804A1 (zh) 基于云台的目标跟随方法、装置、云台和计算机存储介质
US11184550B2 (en) Image capturing apparatus capable of automatically searching for an object and control method thereof, and storage medium
CN112602319B (zh) 一种对焦装置、方法及相关设备
RU2762998C2 (ru) Оборудование фиксации изображений и способ управления для него
CN108475072A (zh) 一种跟踪控制方法、装置及飞行器
CN114267041B (zh) 场景中对象的识别方法及装置
WO2019104569A1 (zh) 一种对焦方法、设备及可读存储介质
CN110490131B (zh) 一种拍摄设备的定位方法、装置、电子设备及存储介质
JP2023057157A (ja) 撮像装置及びその制御方法、プログラム
WO2020014864A1 (zh) 位姿确定方法、设备、计算机可读存储介质
WO2019144296A1 (zh) 可移动平台的控制方法、装置和可移动平台
WO2020019353A1 (zh) 跟踪控制方法、设备、计算机可读存储介质
WO2022021093A1 (zh) 拍摄方法、拍摄装置及存储介质
US11314968B2 (en) Information processing apparatus, control method, and program
JP2017016592A (ja) 主被写体検出装置、主被写体検出方法及びプログラム

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: 18902061

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18902061

Country of ref document: EP

Kind code of ref document: A1