CN109064511B - Method and device for measuring height of center of gravity of human body and related equipment - Google Patents
Method and device for measuring height of center of gravity of human body and related equipment Download PDFInfo
- Publication number
- CN109064511B CN109064511B CN201810961873.7A CN201810961873A CN109064511B CN 109064511 B CN109064511 B CN 109064511B CN 201810961873 A CN201810961873 A CN 201810961873A CN 109064511 B CN109064511 B CN 109064511B
- Authority
- CN
- China
- Prior art keywords
- human body
- image
- gravity
- center
- height
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000005484 gravity Effects 0.000 title claims abstract description 149
- 238000000034 method Methods 0.000 title claims abstract description 121
- 238000012545 processing Methods 0.000 claims abstract description 104
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 99
- 238000001514 detection method Methods 0.000 claims abstract description 59
- 238000013528 artificial neural network Methods 0.000 claims abstract description 31
- 230000006870 function Effects 0.000 claims description 63
- 230000008569 process Effects 0.000 claims description 54
- 238000013527 convolutional neural network Methods 0.000 claims description 39
- 238000004590 computer program Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 11
- 230000000306 recurrent effect Effects 0.000 claims description 10
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 238000005259 measurement Methods 0.000 abstract description 25
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 14
- 238000012937 correction Methods 0.000 description 6
- 238000011176 pooling Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000037396 body weight Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The application discloses a method for measuring the height of the gravity center of a human body, which comprises the steps of receiving an image to be tested; carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground; determining position information of the gravity center of the human body according to the person; calculating the image to be tested through a depth information generation algorithm to obtain the depth information of the image to be tested; under the depth information, calculating the pixel heights of the gravity center of the human body and the ground according to the position information; and carrying out regression processing on the pixel height through a regression neural network to obtain the height of the gravity center of the human body. The method can more conveniently realize the measurement of the height of the gravity center of the human body, effectively improve the measurement efficiency and reduce the measurement error; the application also discloses a device and equipment for measuring the height of the center of gravity of the human body and a computer readable storage medium, and the beneficial effects are also achieved.
Description
Technical Field
The application relates to the technical field of computer vision, in particular to a method for measuring the height of the center of gravity of a human body, and further relates to a device and equipment for measuring the height of the center of gravity of the human body and a computer readable storage medium.
Background
The gravity is the gravity of the earth to an object, the action point of the resultant force borne by the whole human body is the position of the gravity center of the human body, in the research of motion biomechanics, indexes such as the track, displacement, speed, acceleration and the like of the gravity center of the human body are important indexes for evaluating the motion condition of the human body, and the stability of the balance of the human body and the reasonability of the motion can be evaluated according to the position of the gravity center; when biomechanical analysis is performed on the motion technology, the motion rule is searched, and the center of gravity is not kept away. Therefore, it is important to correctly measure the position of the center of gravity of the human body to solve various problems of sports biomechanics.
The existing gravity center measuring method almost adopts a corresponding measuring platform device to realize the measurement of the center of a human body, and one method adopts a balancing plate, a weighing scale and other devices to carry out measurement, but the measuring method has the advantages of low measuring speed, large measuring error, low precision, single testing function, capability of obtaining only one piece of force measurement data and complicated testing process. The other is a gravity center measuring device which adopts a sensor and a singlechip for processing, data acquisition is carried out through the sensor, such as an angle sensor or a position sensor, and the like, and the data is sent to the singlechip for data processing, so that the gravity center position is obtained; however, this method can only measure the absolute height of the center of gravity of the human body, and cannot measure the relative height of the center of gravity of the human body standing high with respect to the ground. In addition, the general gravity center measuring device is large in size, inconvenient to carry and high in price.
Therefore, how to more conveniently realize the measurement of the height of the center of gravity of the human body, improve the measurement efficiency and reduce the measurement error is a problem to be solved urgently by the technical personnel in the field.
Disclosure of Invention
The method can more conveniently realize the measurement of the height of the center of gravity of the human body, effectively improve the measurement efficiency and reduce the measurement error; it is another object of the present application to provide a human body center of gravity height measuring device, apparatus, computer readable storage medium and system, also having the above beneficial effects.
In order to solve the technical problem, the application provides a method for measuring the height of the center of gravity of a human body, which comprises the following steps:
receiving an image to be tested;
carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground;
determining position information of the gravity center of the human body according to the person;
calculating the image to be tested through a depth information generation algorithm to obtain the depth information of the image to be tested;
under the depth information, calculating the pixel heights of the gravity center of the human body and the ground according to the position information;
and carrying out regression processing on the pixel height through a regression neural network to obtain the height of the gravity center of the human body.
Preferably, the performing target detection on the image to be tested through a target detection algorithm to obtain a target object includes:
processing the image to be tested through a convolutional neural network to obtain a convolution result;
processing the convolution result through an RPN algorithm to obtain the frame position of the target object;
extracting characteristic information at the position of the frame through a ROIAlign algorithm;
and determining the type of the target object according to the characteristic information, and outputting the target object.
Preferably, the determining the type of the target object according to the feature information and outputting the target object includes:
processing the characteristic information through a fully-connected neural network to obtain a first processing result;
and processing the first processing result based on a sofamax algorithm to obtain the target object and the position information of the target object.
Preferably, the determining the position information of the center of gravity of the human body according to the person includes:
processing the characteristic information corresponding to the target object through a convolutional neural network to obtain a second processing result;
processing the second processing result based on a sigmoid algorithm to obtain a position information set;
when the target object is the person, extracting a maximum element position in the position information set as position information of the center of gravity of the human body.
Preferably, the operating the image to be tested through a depth information generating algorithm to obtain the depth information of the image to be tested includes:
sequentially passing the image to be tested through convolutional neural networks corresponding to an encoder and a decoder to obtain the parallax corresponding to the image to be tested;
and processing the parallax through a bilinear sampler to obtain the depth information.
Preferably, the method for measuring the height of the center of gravity of the human body further comprises:
calculating a loss function of an image processing algorithm; wherein the image processing algorithm comprises the target detection algorithm, the depth information generation algorithm, and the recurrent neural network;
optimizing the loss function to obtain an optimal value of the loss function;
and acquiring the height of the center of gravity of the human body corresponding to the optimal value.
Preferably, the optimizing the loss function to obtain an optimal value of the loss function includes:
and optimizing the loss function by using an Adam optimization algorithm to obtain the optimal value.
In order to solve the above technical problem, the present application provides a human body center of gravity height measuring device, the device includes:
the data receiving module is used for receiving the image to be tested;
the target detection module is used for carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground;
the gravity center acquisition module is used for determining the position information of the gravity center of the human body according to the person;
the depth information generating module is used for calculating the image to be tested through a depth information generating algorithm to obtain the depth information of the image to be tested;
the pixel height calculation module is used for calculating the pixel heights of the gravity center of the human body and the ground according to the position information under the depth information;
and the gravity center height calculation module is used for performing regression processing on the pixel height through a regression neural network to obtain the gravity center height of the human body.
In order to solve the above technical problem, the present application provides a human body center of gravity height measuring apparatus, the apparatus includes:
a memory for storing a computer program;
and the processor is used for realizing the steps of any one of the human body gravity center height measuring methods when the computer program is executed.
In order to solve the above technical problem, the present application provides a computer-readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of any one of the above methods for measuring the height of the center of gravity of a human body.
The application provides a method for measuring the height of the gravity center of a human body, which comprises the steps of receiving an image to be tested; carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground; determining position information of the gravity center of the human body according to the person; calculating the image to be tested through a depth information generation algorithm to obtain the depth information of the image to be tested; under the depth information, calculating the pixel heights of the gravity center of the human body and the ground according to the position information; and carrying out regression processing on the pixel height through a regression neural network to obtain the height of the gravity center of the human body.
Therefore, the method for measuring the height of the gravity center of the human body provided by the application processes the image to be tested sent by a user through a target detection algorithm and a depth information generation algorithm, and performs regression processing on the result obtained based on the two algorithms through a regression neural network to obtain the height of the gravity center of the human body in the image to be tested; the measuring method can be realized only by a corresponding image processing algorithm without any other mechanical device, so that the cost is low, and meanwhile, the measuring speed is high and the precision is high; in addition, the measuring method can also obtain the current state of the human body and the relative height from the ground when the human body stands at a high position, can also detect the gravity center height of all the human bodies in the image to be tested, can replace the traditional gravity center measuring device at present, can also be applied to other occasions, such as video streams of construction sites, markets and roads, and has wide application prospect.
The human body gravity center height measuring device, the human body gravity center height measuring equipment and the computer readable storage medium have the beneficial effects, and are not repeated herein.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for measuring the height of the center of gravity of a human body according to the present application;
fig. 2 is a schematic flow chart of a target detection method provided in the present application;
fig. 3 is a schematic flowchart of a method for acquiring information about the position of the center of gravity of a human body according to the present application;
fig. 4 is a schematic flowchart of an image depth information obtaining method provided in the present application;
FIG. 5 is a block diagram of a process for measuring the height of the center of gravity of a human body according to the present application;
FIG. 6 is a block diagram of an object detection process provided herein;
FIG. 7 is a block diagram of a depth information generation process provided herein;
FIG. 8 is a block diagram of a recurrent neural network process provided herein;
FIG. 9 is a schematic view of a device for measuring the height of the center of gravity of a human body according to the present application;
fig. 10 is a schematic structural diagram of a human body gravity center height measuring device provided by the present application.
Detailed Description
The core of the application is to provide a method for measuring the height of the center of gravity of a human body, the method can more conveniently realize the measurement of the height of the center of gravity of the human body, effectively improve the measurement efficiency and reduce the measurement error; another core of the present application is to provide a device, an apparatus, a computer-readable storage medium, and a system for measuring the height of the center of gravity of a human body, which also have the above-mentioned advantages.
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Generally speaking, the traditional gravity center measurement method almost adopts a corresponding measurement platform device to realize the measurement of the center of the human body, the measurement speed is low, the measurement error is large, the precision is low, the test function is single, only one force measurement data can be obtained, and the test process is complicated. In addition, the traditional gravity center measuring method can only measure the absolute gravity center height of the human body generally, and cannot measure the relative gravity center height of the human body standing at a high position relative to the ground.
In order to solve the problems, the application provides a method for measuring the height of the center of gravity of the human body, and the measurement of the height of the center of gravity of the human body is realized through a related algorithm of deep learning. The rapid rise of deep learning in recent years solves many problems in real life, and is widely applied, which is mainly embodied in two aspects: the computer vision field and the natural language processing field. The application mainly relates to the field of computer vision, and various visual tasks are realized by learning feature extraction of images through a convolutional neural network and performing higher-level abstraction on features. The convolutional neural network has the advantages that the convolutional neural network is incomparable with a full-connection network, the number of parameters is small, and the convolutional neural network has good robustness on translation, scale transformation, rotation transformation and the like of an image.
Referring to fig. 1, fig. 1 is a schematic flow chart of a method for measuring a height of a center of gravity of a human body according to the present application, where the method includes:
s101: receiving an image to be tested;
specifically, when the height of the center of gravity of the human body needs to be measured for one or more persons in a certain image, the image can be sent to the processor, and further, the processor can process the image as the image to be tested. The source and type of the image to be tested are not limited in the application, and the image to be tested can be a picture shot by a user independently, or a certain video frame captured in a video by the user, which can be set by the user based on the actual requirements of the user.
S102: carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground;
specifically, since the measurement of the height of the human body heart needs to be realized, after an image to be tested sent by a user is obtained, a human body and the ground need to be determined in the image to be tested, at the moment, the target detection can be performed on the image to be tested through a target detection algorithm, so that a target object is determined from the image to be tested, and the target object is the human body and the ground. The implementation process of the target detection algorithm is a process of positioning a target in an input image and determining the type of the target, wherein the input image is the image to be tested, and the target is the target object.
As a preferred embodiment, please refer to fig. 2, fig. 2 is a schematic flowchart of a target detection method provided in the present application, where the target object is obtained by performing target detection on an image to be tested through a target detection algorithm, and the method may include:
s201: processing the image to be tested through a convolutional neural network to obtain a convolution result;
s202: processing the convolution result through an RPN algorithm to obtain the frame position of the target object;
s203: extracting characteristic information through a ROIAlign algorithm at the position of the frame;
s204: and determining the type of the target object according to the characteristic information, and outputting the target object.
Specifically, when the target detection is performed on the image to be detected, the target detection can be realized through a corresponding learning algorithm. Firstly, after receiving an image to be tested, a processor sends the image to be tested to a convolutional neural network for convolution processing so as to obtain a corresponding convolution result; the convolutional neural network is a deep learning structure inspired by the visual stimuli mechanism, and the number of stages is not limited in the present application. Further, the convolution result is processed through an RPN algorithm, the RPN algorithm also belongs to a convolution neural network, and the corresponding output is the frame position of the target object; it should be noted that the type of the target object obtained at this time is unknown, that is, only the frame position information of the target object can be determined, and it is not possible to determine whether the target object is a person or the ground. Therefore, at the position of the frame, the position type target object can be further subjected to feature extraction through a ROIAlign algorithm to obtain corresponding feature information, and finally the type of the target object can be determined according to the feature information and output.
Preferably, the determining the type of the target object based on the feature information and outputting the target object may include: processing the characteristic information through a full-connection neural network to obtain a first processing result; and processing the first processing result based on the sofamax algorithm to obtain the target object and the position information of the target object.
Specifically, when the type of the target object is determined and the target object is output, the target object may be processed through the fully-connected neural network to obtain a corresponding processing result, that is, the first processing result, where the number of layers of the fully-connected neural network is not limited in the present application; further, the first processing result can be processed through the sofamax algorithm, the corresponding output is the classification result of the target object, namely the output is human and/or ground, and correspondingly, the position information of the human and/or ground can be further obtained.
S103: determining position information of the gravity center of the human body according to the person;
on the basis of S102, the present step aims to determine the position information of the center of gravity of the target "person" after determining the target object in the category of "person". The position information may be coordinate information of a center point of a body weight in the image to be detected.
As a preferred embodiment, please refer to fig. 3, fig. 3 is a flowchart illustrating a method for acquiring information about the center of gravity of a human body according to the present application, where the determining information about the center of gravity of the human body according to a person may include:
s301: processing the characteristic information corresponding to the target object through a convolutional neural network to obtain a second processing result;
s302: processing the second processing result based on a sigmoid algorithm to obtain a position information set;
s303: when the target object is a human, the maximum element position is extracted as position information of the center of gravity of the human body in the position information set.
Specifically, the feature information of the target object may be convolved by the convolutional neural network to obtain a corresponding result, that is, the second processing result, it should be noted that the convolutional neural network is the same as the above-mentioned convolutional neural network, but the respective corresponding stages are not unique, and a user may set the result according to actual needs; further, the position information set can be obtained by processing the second processing result through the sigmoid algorithm, and since the position of the center of gravity of the human body in the image to be tested is higher than the ground position, when the target object is a human, the position information of the center of gravity of the human body can be extracted from the position information set as the maximum element position.
In addition, it should be noted that the process may be executed after S203 and at the same time with S204, or may be executed after S204, which is not limited in the present application.
S104: calculating the image to be tested through a depth information generation algorithm to obtain the depth information of the image to be tested;
specifically, the step aims to obtain the depth information of the image to be tested so as to realize the measurement of the height of the gravity center of the human body of the image to be tested. The depth information is the number of bits used for storing each pixel in the image, the image to be tested can be obtained by processing through a corresponding depth information generation algorithm, and monocular depth estimation or binocular depth estimation is adopted for the image to be tested, which is not limited in the application.
As a preferred embodiment, please refer to fig. 4, where fig. 4 is a schematic flow chart of an image depth information obtaining method provided in the present application, where the obtaining of the depth information of the image to be tested by performing the operation on the image to be tested through the depth information generating algorithm may include:
s401: sequentially passing the image to be tested through the convolutional neural networks corresponding to the encoder and the decoder to obtain the parallax corresponding to the image to be tested;
s402: and processing the parallax by a bilinear sampler to obtain depth information.
Specifically, for obtaining the depth information, the present application provides a more specific implementation method, that is, the implementation is realized by a convolutional neural network and a bilinear sampler corresponding to a decoder and the decoder. The method comprises the steps of firstly, obtaining the parallax of an image to be tested through the operation of the convolutional neural network on the image to be tested, and further, processing the parallax through a bilinear sampler to obtain the required depth information.
S105: under the depth information, calculating the pixel height between the gravity center of the human body and the ground according to the position information;
specifically, after the depth information of the image to be tested, the target object and the position of the center of gravity of the human body are obtained, the pixel height between the center of gravity of the human body and the ground of the target object can be obtained through calculation according to the position information of the center of gravity of the human body under the depth information, and the pixel height is the height of the center of gravity of the human body in the image to be tested.
S106: and (4) carrying out regression processing on the pixel height through a regression neural network to obtain the height of the gravity center of the human body.
Specifically, after the pixel height is obtained, regression processing can be performed on the pixel height through a regression neural network, so that the pixel height is mapped to the actual height of the gravity center of the human body through a corresponding mapping function, and therefore measurement of the height of the gravity center of the human body is achieved.
According to the method for measuring the height of the gravity center of the human body, the height of the gravity center of the human body can be measured only through a corresponding image processing algorithm, any other mechanical device is not needed, the cost is low, and meanwhile, the measuring speed is high and the precision is high; in addition, the measuring method can also obtain the current state of the human body and the relative height from the ground when the human body stands at a high position, can also detect the gravity center height of all the human bodies in the image to be tested, can replace the traditional gravity center measuring device at present, can also be applied to other occasions, such as video streams of construction sites, markets and roads, and has wide application prospect.
On the basis of the above embodiments, as a preferred embodiment, the method for measuring the height of the center of gravity of the human body may further include: calculating a loss function of an image processing algorithm; the image processing algorithm comprises a target detection algorithm, a depth information generation algorithm and a recurrent neural network; optimizing the loss function to obtain an optimal value of the loss function; and acquiring the height of the center of gravity of the human body corresponding to the optimal value.
Specifically, it can be seen from the above implementation method that the technical solution provided in the present application mainly relates to three parts, namely, a target detection process, a depth information acquisition process, and a process of performing regression processing based on a detection result obtained by the target detection process and depth information obtained by the depth information acquisition process, so that after the three processes, corresponding loss functions can be calculated respectively to further optimize each loss function to obtain respective corresponding optimal values, and when a loss function takes an optimal value, respective corresponding results are obtained to further obtain a final height of a center of gravity of a human body, which can effectively reduce loss and error degree, and improve measurement accuracy.
Preferably, the optimizing the loss function to obtain the optimal value of the loss function may include: and optimizing the loss function by using an Adam optimization algorithm to obtain an optimal value.
Specifically, for the optimization process of the loss function, the optimization process can be realized through an Adam optimization algorithm, Adam is a first-order optimization algorithm which can replace the traditional stochastic gradient descent process, the weights of the neural network can be updated iteratively based on training data, the specific realization process refers to the prior art, and details are not repeated herein.
On the basis of the above embodiments, the present application provides a more specific embodiment to introduce the technical solution in detail. Referring to fig. 5, fig. 5 is a frame diagram of a process for measuring the height of the center of gravity of a human body according to the present application. Firstly, inputting an image to be tested, respectively carrying out target detection processing and depth information generation processing on the image to be tested, and respectively obtaining respective processing results; and further, performing regression model correction processing based on the two processing results, and obtaining the height of the gravity center of the human body in the image to be tested according to the regression results. Hereinafter, the above three processes will be described in detail.
1. The target detection processing process can be divided into a forward propagation operation and a backward propagation operation, wherein the forward propagation operation corresponds to the target detection process of the image to be detected; the back propagation operation corresponds to the calculation of the target detection loss function.
(1) Forward propagation operation for target detection:
referring to fig. 6, fig. 6 is a block diagram of a target detection process provided in the present application, in which an input image is an image to be detected.
Firstly, after receiving an image to be tested I sent by a user, convolving the image to be tested by using a 3 x 3 convolution template:
wherein K represents a 3 x 3 convolution templateK (m, n) is the element of the mth row and the nth column in the convolution template; i (I, j) is an element of the ith row and the jth column in the image to be tested; the convolution result after convolution is stored in S1Performing the following steps;
further, each element in the convolution result is corrected by using a nonlinear function to obtain a processing result A1：
A1＝a(S1(i,j))；
Wherein a is an activation function in the neural network, and the selection of the value corresponds to different algorithms, such as a sigmiod algorithm, a ReLU algorithm and the like;
further, the corrected processing result A was subjected to the maximum pooling method (max pooling)1Pooling is performed, where pooling may be performed using local 2 x 2 regions, and the pooling results are stored in a matrix P1The method comprises the following steps:
P1＝max(x(2p,2q),x(2p+1,2q),x(2p,2q+1),x(2p+1,2q+1))；
wherein x (p, q) represents A1A certain adjacent 2 x 2 region;
thus, based on the three-layer structure, one stage (stage) of the convolutional neural network can be constructed, and in the present application, by stacking 5 stages, a main structure of the convolutional neural network in the left half of the frame diagram shown in fig. 6 can be constructed, and the final output of the main structure can be set as:
Cdet＝CNNdet(I)；
wherein, CNNdet() I.e. the main structure of the convolutional neural network, CdetRepresenting the output result of the subject architecture.
Next, as shown in FIG. 6, C may be paired firstdetProcessing by using an RPN algorithm (Region probable Network) to generate a frame candidate of the target object, and assuming that the output corresponding to the RPN algorithm is ROI (Region of interest), then:
ROI＝RPN(Cdet)；
further, the ith element ROI in the ROIiCorresponding convolutional neural network subject architecture output CdetRegion C ofiAs an input, the ROIAlign algorithm is used for processing, and the output corresponding to the ROIAlign algorithm is assumed to be RiWhere i corresponds to each target object in the image to be tested, then:
Ri＝ROIAlign(Ci)。
finally, for the result R of ROIAlign algorithmiTwo steps are respectively carried out, one is to RiUsing two fully-connected layers for treatment, the second is for RiProcessing using four convolutional layers, where the last layer of convolutional layers outputs a 1 x 56 matrix, and the corresponding results are FC's respectivelyiAnd CONVi；
For FCiFirstly, the classification result of the target Object is obtained by processing with the sofamax algorithm, and the classification result is stored in the ObjectiThe output layer of the sofamax algorithm comprises two output neurons, which correspond to two target objects, namely a human object and a ground object. Using one full connection layer pair FCiProcessing to obtain the position information of the target object, wherein the position information can comprise the width and the height of the target object and the pixel coordinates of the upper left corner of the target object in the image to be tested, and storing the position information in the BboxiThe method comprises the following steps:
Bboxi＝(x,y,w,h)；
wherein (x, y) is the pixel coordinate, w is the width, and h is the height.
For CONViThe sigmiod algorithm may be used for processing, assuming that the output corresponding to the sigmiod algorithm is G'iAnd then:
G′i＝sigmoid(CONVi)；
according to the classification result ObjectiIf ObjectiIf it is human, then it is right for Gi' all elements G ' of 'i(i, j) comparing the positions of the elements to obtain the position G of the center of gravity of the human body as the position of the maximum elementi：
Gi＝argmax(G′i(i,j))。
(2) Back propagation operation of target detection:
this section relates to the calculation of a loss function for the object detection process, which may include a loss function for object position information, a loss function for object classification, and a loss function for correcting the position of the center of gravity of the human body, corresponding to the above-mentioned object detection process.
Target object location information loss function:
wherein,
if the label information is the label information, taking a value of 1 when the label information is the target object, and otherwise, taking a value of 0; r () is smoothL1A function; t is tiThe subscript i corresponds to one of (x, y, w, h) as the output frame position of the target object;
subscripts i and (x) are the location labels of the target object*,y*,w*,h*) One of them corresponds to (x)*,y*,w*,h*) Pre-calibrating the position of the frame for a user;
are normalization constants of the respective algorithms.
Target object classification loss function:
wherein p isiA classification probability matrix of the target object output by the softmax algorithm;
is the tag information of the class to which the target object belongs, and, when the target object,
the value is 1, otherwise the value is 0; i is the ith target object in the image to be tested;
are normalization constants of the respective algorithms.
Correcting the position loss function of the center of gravity of the human body:
wherein, giThe probability matrix is the position of the gravity center point of the human body output by the sigmiod algorithm;
the label information of the gravity center position of the human body. i is the ith target object belonging to the category of 'people' in the image to be tested.
Are normalization constants of the respective algorithms.
In summary, the loss function L of the target detection processdetComprises the following steps:
Ldet＝Lbbox+Lcls+LG；
further, L can be optimized by AdamdetAnd optimizing to enable the optimal solution to be achieved, and acquiring the target object and the position information thereof corresponding to the optimal solution and the position information of the gravity center of the human body.
2. For the depth information generation process, the method can also be divided into a forward propagation operation and a backward propagation operation, wherein the forward propagation operation corresponds to the depth information generation process of the image to be detected; the back propagation operation corresponds to a computation process of the depth information generation loss function.
(1) Forward operation of depth information generation:
referring to fig. 7, fig. 7 is a frame diagram of a depth information generating process provided in the present application, in which an input image is an image to be detected.
Before obtaining the depth information, a depth image data set may be used to train a convolutional neural network corresponding to a depth information generation process, so that the convolutional neural network has a capability of generating the depth information, where the depth image data set may be a KITTI data set. In the application, a binocular depth estimation method is adopted, an image to be tested is input into a pair of calibrated binocular cameras, two identical scene images are obtained, wherein the two identical scene images are respectively Il、IrThe two are operated through a trained convolutional neural network to obtain the parallax dr、dlAnd further, according to the distance b and the focal length f between the binocular cameras, the depth information d of the image to be tested can be obtained:
therefore, the depth information generating process based on the convolutional neural network may include an operation process corresponding to the convolutional neural network corresponding to the encoder and the decoder, and a processing process of the bilinear sampler, which is specifically as follows:
first, a convolutional neural network structure may be formed using 14 stages as encoders in the depth information generation process, and the convolution result is stored in conv7b, where conv7b has a size IlIs/are as follows
Then:
wherein,
representing encoder convolutional nervesA network structure.
Further, transpose convolution may be performed on the encoder output conv7b using a 3 × 3 convolution template to obtain upconv7, the output upconv7 size of which is 2 times the input conv7b size; then, convolving upconv7 by using a 3-by-3 convolution template to obtain iconv7, wherein the convolution operation can keep the input size and the output size the same; thus, the two-layer structure can constitute a basic operation unit (unit) of the decoder, and in the present application, 7 units are used as a convolutional neural network structure of the decoder, and the final convolutional result is stored in disp:
wherein,
representing a decoder convolutional neural network structure; the disparities d are included in the disparitiesrAnd dl。
Finally, for drAnd an input left image IlReconstruction of the right image using a bilinear sampler s ()
To dlAnd an input right image IrReconstruction of left image using bilinear sampler s ()
Then:
further, to drGeneration of d using a bilinear samplerl(dr) To d is pairedlUsing bilinear samplingSample generation dr(dl) And then:
dl(dr)＝s(dr)；
dr(dl)＝s(dl)；
thereby can be based on
And obtaining the depth information.
(2) Reverse operation of depth information generation:
this section relates to the calculation of an object detection process loss function, which may include an image reconstruction loss function, a parallax smoothing loss function, and a left-right parallax symmetry loss function, corresponding to the object detection process described above.
Image reconstruction loss function:
wherein,
respectively correspond to the input images IlAnd IrThe reconstruction loss function of (1); n is the total pixel number of the image; α is a balance parameter, and in the present application, α is taken to be 0.85; SSIM () (structural similarity index) is a structural similarity function, a single-scale SSIM function is used in this application; | | | represents the L1 norm.
Parallax smoothing loss function:
wherein,
respectively correspond to the input images IlAnd IrThe disparity smoothing loss function of (1).
Left-right parallax symmetry loss function:
wherein,
respectively correspond to the input images IlAnd IrLeft and right disparity symmetry loss function.
In summary, the loss function L of the depth information generation processdetComprises the following steps:
further, L can be optimized by AdamdepAnd optimizing to enable the optimal solution to be achieved, and acquiring depth information corresponding to the optimal solution.
3. For the regression model correction process, the method can also be divided into a forward propagation operation and a backward propagation operation, wherein the forward propagation operation corresponds to the regression processing process; the back propagation operation corresponds to a calculation process of a regression processing loss function.
(1) Forward propagation operation of regression model correction:
firstly, obtaining a human body gravity center point G according to output results of a target detection process and a depth information generation processiPixel height h 'from ground'iI denotes a subscript index belonging to the "people" category in the image to be tested; secondly, establishing a regression neural network model, training a network, and learning a mapping function to realize the height h 'of the pixel'iMapping into actual human body gravity center height h'i：
hi＝h'iexp(dh(h'i))；
Wherein d ish(h'i) Namely the function mapping relation of the recurrent neural network learning.
Therefore, the recurrent neural network learning process may include a pixel height acquisition process and an actual height acquisition process, which are specifically as follows:
for the pixel height acquisition process, first, the output Object of the Object detection process is acquirediAnd judging the category, if the category belongs to the category of 'people', acquiring the output Bbox of the target detection processiAnd the ObjectiCorresponding human body gravity center position GiAcquiring an output d of the depth information generation process; further, from GiPosition in the image (i)max,jmax) And acquiring the depth of the position corresponding to the depth information d:
finally, can be at
Calculating human body gravity G under depthiDistance ground frame BboxiPixel height h 'in range'i。
Referring to fig. 8, fig. 8 is a block diagram of a recurrent neural network processing procedure provided in the present application for the actual altitude obtaining process.
Firstly, obtaining the output C of convolutional neural network main body architecture in the target detection processdetAnd determining the position of the frame of the person and the position of the frame of the ground at CdetOf corresponding region f'1、f'2As shown in FIG. 8, and using ROThe IAlign algorithm will become a fixed size W × H, where, in this application, taking W × H as 60 × 40, then:
f1＝ROIAlign(f'1)；
f2＝ROIAlign(f'2)；
further, f is1、f2Splicing the two layers together to form a single characteristic diagram f, inputting the single characteristic diagram f into two fully-connected layers for processing, and finally outputting dh(h'i) Comprises the following steps:
wherein i represents the number of the neurons of the last full-connection layer in the recurrent neural network, and the value of i is 1-128, omegaeConnection weight of the last fully-connected layer, aeIs the activation value of neuron i of the last fully connected layer.
(2) Back propagation operation of regression model correction:
this section involves the calculation of a loss function of the regression model correction process, which is the loss function of the recurrent neural network:
Lreg＝(y*-dh(h'i))2；
wherein, y*The regression neural network label is calculated in the following way:
wherein,
labeling information for the actual relative height of the gravity center of the ith individual human body in the image to be tested; h'iAnd obtaining the pixel height of the gravity center of the ith personal body obtained by the regression model correction process.
Further, L can be optimized by AdamregOptimizing to reach the optimal solution and obtaining the optimal solutionAnd solving the corresponding height of the gravity center of the human body.
According to the method for measuring the height of the center of gravity of the human body, the height of the center of gravity of the human body can be measured only through a corresponding image processing algorithm, any other mechanical device is not needed, the cost is low, and meanwhile the measuring speed is high and the precision is high; in addition, the measuring method can also obtain the current state of the human body and the relative height from the ground when the human body stands at a high position, can also detect the gravity center height of all the human bodies in the image to be tested, can replace the traditional gravity center measuring device at present, can also be applied to other occasions, such as video streams of construction sites, markets and roads, and has wide application prospect.
To solve the above problem, please refer to fig. 9, fig. 9 is a schematic diagram of a device for measuring height of center of gravity of a human body according to the present application, the device may include:
a data receiving module 10, configured to receive an image to be tested;
the target detection module 20 is used for performing target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground;
a center of gravity acquisition module 30 for determining position information of the center of gravity of the human body according to the person;
the depth information generating module 40 is used for calculating the image to be tested through a depth information generating algorithm to obtain the depth information of the image to be tested;
the pixel height calculating module 50 is used for calculating the pixel height between the gravity center of the human body and the ground according to the position information under the depth information;
and the gravity center height calculating module 60 is configured to perform regression processing on the pixel height through a regression neural network to obtain the height of the gravity center of the human body.
As a preferred embodiment, the object detection module 20 may include:
the convolution processing submodule is used for processing the image to be tested through a convolution neural network to obtain a convolution result;
the frame obtaining submodule is used for processing the convolution result through an RPN algorithm to obtain the frame position of the target object;
the characteristic extraction submodule is used for extracting characteristic information at the position of the frame through a ROIAlign algorithm;
and the category determining submodule is used for determining the category of the target object according to the characteristic information and outputting the target object.
As a preferred embodiment, the category determining submodule may be specifically configured to process the feature information through a fully-connected neural network to obtain a first processing result; and processing the first processing result based on the sofamax algorithm to obtain the target object and the position information of the target object.
As a preferred embodiment, the center of gravity acquiring module 30 may include:
the characteristic processing submodule is used for processing the characteristic information corresponding to the target object through the convolutional neural network to obtain a second processing result;
the position obtaining submodule is used for processing the second processing result based on a sigmoid algorithm to obtain a position information set;
and the position extraction submodule is used for extracting the maximum element position from the position information set as the position information of the gravity center of the human body when the target object is the human body.
As a preferred embodiment, the center of gravity height calculating module 60 may include:
the parallax obtaining submodule is used for sequentially passing the image to be tested through the convolutional neural networks corresponding to the encoder and the decoder to obtain the parallax corresponding to the image to be tested;
and the parallax processing sub-module is used for processing the parallax through the bilinear sampler to obtain the depth information.
As a preferred embodiment, the apparatus for measuring the height of the center of gravity of a human body may further include:
the loss function calculation module is used for calculating a loss function of the image processing algorithm; the image processing algorithm comprises a target detection algorithm, a depth information generation algorithm and a recurrent neural network;
the loss function optimization module is used for optimizing the loss function to obtain the optimal value of the loss function;
and the gravity center height acquisition module is used for acquiring the human body gravity center height corresponding to the optimal value.
As a preferred embodiment, the loss function optimization module may be specifically configured to perform optimization processing on the loss function through an Adam optimization algorithm to obtain an optimal value.
For the introduction of the apparatus provided in the present application, please refer to the above method embodiments, which are not described herein again.
To solve the above problem, please refer to fig. 10, fig. 10 is a schematic structural diagram of an apparatus for measuring a height of a center of gravity of a human body according to the present application, the apparatus may include:
a memory 1 for storing a computer program;
the processor 2 is configured to implement any of the above-mentioned steps of the method for measuring the height of the center of gravity of a human body when executing a computer program.
For the introduction of the device provided in the present application, please refer to the above method embodiment, which is not described herein again.
In order to solve the above problem, the present application further provides a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, can implement any of the above steps of the method for measuring the height of the center of gravity of a human body.
The computer-readable storage medium may include: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
For the introduction of the computer-readable storage medium provided in the present application, please refer to the above method embodiments, which are not described herein again.
The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The method, apparatus, device and computer readable storage medium for measuring the height of the center of gravity of a human body provided by the present application are described in detail above. The principles and embodiments of the present application are explained herein using specific examples, which are provided only to help understand the method and the core idea of the present application. It should be noted that, for those skilled in the art, it is possible to make several improvements and modifications to the present application without departing from the principle of the present application, and these improvements and modifications also fall into the elements of the protection scope of the claims of the present application.
Claims (8)
1. A method for measuring the height of the center of gravity of a human body is characterized by comprising the following steps:
receiving an image to be tested;
carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground;
determining position information of the gravity center of the human body according to the person;
calculating the image to be tested through a depth information generation algorithm to obtain the depth information of the image to be tested;
under the depth information, calculating the pixel heights of the gravity center of the human body and the ground according to the position information;
performing regression processing on the pixel height through a regression neural network to obtain the height of the gravity center of the human body;
the target detection of the image to be tested through a target detection algorithm to obtain a target object comprises the following steps:
processing the image to be tested through a convolutional neural network to obtain a convolution result;
processing the convolution result through an RPN algorithm to obtain the frame position of the target object;
extracting characteristic information at the position of the frame through a ROIAlign algorithm;
determining the type of the target object according to the characteristic information, and outputting the target object;
then, the determining the position information of the gravity center of the human body according to the person includes:
processing the characteristic information corresponding to the target object through a convolutional neural network to obtain a second processing result;
processing the second processing result based on a sigmoid algorithm to obtain a position information set;
when the target object is the person, extracting a maximum element position in the position information set as position information of the center of gravity of the human body.
2. The method for measuring the height of the center of gravity of a human body according to claim 1, wherein the determining the kind of the target object based on the feature information and outputting the target object comprises:
processing the characteristic information through a fully-connected neural network to obtain a first processing result;
and processing the first processing result based on a sofamax algorithm to obtain the target object and the position information of the target object.
3. The method for measuring the height of the center of gravity of a human body according to claim 1, wherein the obtaining of the depth information of the image to be tested by operating the image to be tested through a depth information generating algorithm comprises:
sequentially passing the image to be tested through convolutional neural networks corresponding to an encoder and a decoder to obtain the parallax corresponding to the image to be tested;
and processing the parallax through a bilinear sampler to obtain the depth information.
4. The method for measuring the height of the center of gravity of a human body according to any one of claims 1 to 3, further comprising:
calculating a loss function of an image processing algorithm; wherein the image processing algorithm comprises the target detection algorithm, the depth information generation algorithm, and the recurrent neural network;
optimizing the loss function to obtain an optimal value of the loss function;
and acquiring the height of the center of gravity of the human body corresponding to the optimal value.
5. The method for measuring the height of the center of gravity of the human body according to claim 4, wherein the optimizing the loss function to obtain the optimal value of the loss function comprises:
and optimizing the loss function by using an Adam optimization algorithm to obtain the optimal value.
6. A human body gravity center height measuring device is characterized by comprising:
the data receiving module is used for receiving the image to be tested;
the target detection module is used for carrying out target detection on the image to be detected through a target detection algorithm to obtain a target object; wherein the target object comprises a person and a ground;
the gravity center acquisition module is used for determining the position information of the gravity center of the human body according to the person;
the depth information generating module is used for calculating the image to be tested through a depth information generating algorithm to obtain the depth information of the image to be tested;
the pixel height calculation module is used for calculating the pixel heights of the gravity center of the human body and the ground according to the position information under the depth information;
the gravity center height calculation module is used for performing regression processing on the pixel height through a regression neural network to obtain the gravity center height of the human body;
the target detection module is specifically used for processing the image to be tested through a convolutional neural network to obtain a convolution result; processing the convolution result through an RPN algorithm to obtain the frame position of the target object; extracting characteristic information at the position of the frame through a ROIAlign algorithm; determining the type of the target object according to the characteristic information, and outputting the target object;
the gravity center obtaining module is specifically configured to process, through a convolutional neural network, feature information corresponding to the target object to obtain a second processing result; processing the second processing result based on a sigmoid algorithm to obtain a position information set; when the target object is the person, extracting a maximum element position in the position information set as position information of the center of gravity of the human body.
7. An apparatus for measuring the height of the center of gravity of a human body, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the method of measuring the height of the center of gravity of a human body as claimed in any one of claims 1 to 5 when said computer program is executed.
8. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, carries out the steps of the method for measuring the height of the center of gravity of a human body as claimed in any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810961873.7A CN109064511B (en) | 2018-08-22 | 2018-08-22 | Method and device for measuring height of center of gravity of human body and related equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810961873.7A CN109064511B (en) | 2018-08-22 | 2018-08-22 | Method and device for measuring height of center of gravity of human body and related equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109064511A CN109064511A (en) | 2018-12-21 |
| CN109064511B true CN109064511B (en) | 2022-02-15 |
Family
ID=64686846
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810961873.7A Active CN109064511B (en) | 2018-08-22 | 2018-08-22 | Method and device for measuring height of center of gravity of human body and related equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109064511B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110443843A (en) * | 2019-07-29 | 2019-11-12 | 东北大学 | A kind of unsupervised monocular depth estimation method based on generation confrontation network |
| CN110929756B (en) * | 2019-10-23 | 2022-09-06 | 广物智钢数据服务(广州)有限公司 | Steel size and quantity identification method based on deep learning, intelligent equipment and storage medium |
| CN111860652B (en) * | 2020-07-22 | 2022-03-29 | 中国平安财产保险股份有限公司 | Method, device, equipment and medium for measuring animal body weight based on image detection |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034247A (en) * | 2010-12-23 | 2011-04-27 | 中国科学院自动化研究所 | Motion capture method for binocular vision image based on background modeling |
| CN103679171A (en) * | 2013-09-24 | 2014-03-26 | 暨南大学 | A gait feature extraction method based on human body gravity center track analysis |
| CN105930871A (en) * | 2016-04-26 | 2016-09-07 | 浪潮电子信息产业股份有限公司 | Old people anti-falling technology based on improved intelligent learning algorithm |
| CN108209926A (en) * | 2018-01-08 | 2018-06-29 | 西安科技大学 | Human Height measuring system based on depth image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103927724B (en) * | 2014-04-30 | 2016-05-18 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of processing upper garment image |
-
2018
- 2018-08-22 CN CN201810961873.7A patent/CN109064511B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034247A (en) * | 2010-12-23 | 2011-04-27 | 中国科学院自动化研究所 | Motion capture method for binocular vision image based on background modeling |
| CN103679171A (en) * | 2013-09-24 | 2014-03-26 | 暨南大学 | A gait feature extraction method based on human body gravity center track analysis |
| CN105930871A (en) * | 2016-04-26 | 2016-09-07 | 浪潮电子信息产业股份有限公司 | Old people anti-falling technology based on improved intelligent learning algorithm |
| CN108209926A (en) * | 2018-01-08 | 2018-06-29 | 西安科技大学 | Human Height measuring system based on depth image |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度学习的机器人抓取位置检测方法;闫哲等;《高技术通讯》;20180115;第28卷(第1期);第58-65页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109064511A (en) | 2018-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110276316B (en) | Human body key point detection method based on deep learning | |
| CN111709409B (en) | Face living body detection method, device, equipment and medium | |
| CN111881705A (en) | Data processing, training and recognition method, device and storage medium | |
| JP7057959B2 (en) | Motion analysis device | |
| CA3042819A1 (en) | Method and system for postural analysis and measuring anatomical dimensions from a digital image using machine learning | |
| CN113362382A (en) | Three-dimensional reconstruction method and three-dimensional reconstruction device | |
| CN109064511B (en) | Method and device for measuring height of center of gravity of human body and related equipment | |
| JP6397379B2 (en) | CHANGE AREA DETECTION DEVICE, METHOD, AND PROGRAM | |
| CN111783748A (en) | Face recognition method and device, electronic equipment and storage medium | |
| CN113628261B (en) | Infrared and visible light image registration method in electric power inspection scene | |
| CN114359283A (en) | Defect detection method based on Transformer and electronic equipment | |
| KR20190061538A (en) | Method and apparatus of recognizing motion pattern base on combination of multi-model | |
| CN113610046B (en) | Behavior recognition method based on depth video linkage characteristics | |
| CN108898269A (en) | Electric power image-context impact evaluation method based on measurement | |
| CN115311186A (en) | Cross-scale attention confrontation fusion method for infrared and visible light images and terminal | |
| CN114648669A (en) | Motor train unit fault detection method and system based on domain-adaptive binocular parallax calculation | |
| CN113609999B (en) | Human body model building method based on gesture recognition | |
| CN114663880A (en) | Three-dimensional target detection method based on multi-level cross-modal self-attention mechanism | |
| CN116843725B (en) | River surface flow velocity measurement method and system based on deep learning optical flow method | |
| CN109448850A (en) | A kind of body-mass index prediction technique, device, electronic equipment and storage medium | |
| CN113111743A (en) | Personnel distance detection method and device | |
| US10607359B2 (en) | System, method, and apparatus to detect bio-mechanical geometry in a scene using machine vision for the application of a virtual goniometer | |
| CN114612979B (en) | Living body detection method and device, electronic equipment and storage medium | |
| CN116665293A (en) | Sitting posture early warning method and system based on monocular vision | |
| CN113808256B (en) | High-precision holographic human body reconstruction method combined with identity recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |