Video motion tracking method, device, apparatus and storage medium
By extracting the identification features of moving targets from video stream sequences and establishing a prediction mechanism algorithm, the problem of tracking moving targets in complex environments or under occlusion in videos is solved, and continuous localization and trajectory tracking of moving targets are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-13
- Publication Date
- 2026-03-20
AI Technical Summary
In existing videos, moving targets are difficult to track effectively in complex environments or when they are occluded. In particular, when multiple objects are occluded and overlap, existing tracking algorithms need to re-recover the mechanism, which makes target localization difficult.
By extracting the recognition features of moving targets from the original images of the video stream sequence, analyzing motion parameters, and establishing a prediction mechanism algorithm, the moving target can be located when recognition fails, thereby achieving trajectory tracking.
When the moving target is occluded or the environment is complex, it can continuously track the trajectory of the moving target, improving the accuracy and stability of multi-target tracking in video.
Smart Images
Figure CN115620222B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of motion tracking, in particular to a video motion tracking method, device and equipment and storage medium. BACKGROUND
[0002] With the development of intelligent video monitoring, in the intelligent video monitoring system, motion target detection and tracking technology is the key and the fundamental, and is a popular topic in the field of computer vision.
[0003] At present, in the motion target recognition and tracking technology in the video, many tracking methods can well track a single object, but when multiple tracking objects appear in the scene, especially when there is mutual occlusion and overlap between the objects, part of the occlusion causes the motion target part features to be unable to be detected. Moreover, after the introduction of the occlusion interference, the existing tracking algorithm needs to recover the mechanism again, and the motion target can be repositioned when it appears again, thereby causing the motion target in the video to be difficult to effectively track in a complex environment or be occluded. SUMMARY
[0004] The main purpose of the present application is to provide a video motion tracking method, device and equipment and storage medium, which aims to solve the technical problem that the motion target in the existing video is difficult to effectively track in a complex environment or is occluded.
[0005] To achieve the above purpose, the present application provides a video motion tracking method, which comprises:
[0006] extracting a motion target from at least one original image of a video stream sequence, and obtaining the identification features of the motion target;
[0007] based on the identification features, analyzing and obtaining the motion parameters of the motion target in the original image of the current frame, and establishing a prediction mechanism algorithm according to at least one motion target and the motion parameters;
[0008] when the motion target recognition fails, positioning the motion target in the real-time image of the video stream sequence based on the prediction mechanism algorithm, so as to realize the trajectory tracking of the motion target.
[0009] Exemplarily, the extraction of the motion target from at least one original image of a video stream sequence and the obtaining of the identification features of the motion target comprise:
[0010] obtaining at least one original image of a video stream sequence, and detecting a motion region in the original image;
[0011] determining the motion target in the motion region;
[0012] acquiring at least one identification feature of the moving target, the identification feature being used for identification of the moving target in trajectory tracking.
[0013] For example, based on the identification feature, motion parameters of the moving target in the original image of the current frame are analyzed and obtained, and a prediction mechanism algorithm is established according to the moving target and the motion parameters of at least one frame, including:
[0014] Based on the identification feature, the moving target in the original image of the current frame is identified.
[0015] The position coordinates of the moving target in the original image of the current frame are calculated.
[0016] The position coordinates in the original image of at least one frame are associated to obtain at least one motion parameter of the moving target.
[0017] According to the parameter set formed by at least one motion parameter, the prediction mechanism algorithm is established.
[0018] For example, the moving target includes at least one identification feature, and before the moving target in the original image of the current frame is identified based on the identification feature, it includes:
[0019] At least one identification feature is combined to obtain a feature combination.
[0020] Then, the moving target in the original image of the current frame is identified based on the identification feature, including:
[0021] Based on the feature combination, the moving target corresponding to the feature combination in the original image of the current frame is identified.
[0022] For example, the moving target in the original image of the current frame is identified based on the identification feature, including:
[0023] The constraint condition of the identification feature is obtained, and the target feature of the moving object in the original image of the current frame is extracted based on the constraint condition.
[0024] The target feature is compared with the identification feature, and if the comparison result is consistent, the moving object is determined as the moving target.
[0025] For example, the target feature of the moving object in the original image of the current frame is extracted based on the constraint condition, including:
[0026] If the moving object extracted based on the constraint condition is multiple, at least one target feature is extracted in the original image of the current frame.
[0027] The target feature is compared with the identification feature, and if the comparison result is consistent, the moving object is determined as the moving target, comprising:
[0028] The at least one target feature is matched with the identification feature one by one, and the moving object corresponding to the target feature with a successful matching result is the moving target.
[0029] For example, when the moving target recognition fails, the moving target in the real-time image of the video stream sequence is located based on the prediction mechanism algorithm to realize the trajectory tracking of the moving target, comprising:
[0030] When the moving target recognition fails, the prediction mechanism algorithm is obtained, and the prediction mechanism algorithm includes image frame number and target parameter;
[0031] The current frame number corresponding to the real-time image of the video stream sequence is input into the image frame number of the prediction mechanism algorithm, the target parameter corresponding to the current frame number is calculated, and the target motion parameter is obtained;
[0032] The moving target is located based on the target motion parameter to obtain the current position of the moving target, so as to realize the trajectory tracking of the moving target.
[0033] For example, to achieve the above purpose, the application also provides a video motion tracking device, which comprises:
[0034] A feature extraction module is configured to extract a moving target from at least one original image of a video stream sequence and obtain identification features of the moving target;
[0035] A feature recognition module is configured to analyze and obtain motion parameters of the moving target in the original image of the current frame based on the identification features, and to establish a prediction mechanism algorithm based on at least one moving target and the motion parameters;
[0036] A trajectory tracking module is configured to locate the moving target in the real-time image of the video stream sequence based on the prediction mechanism algorithm when the moving target recognition fails, so as to realize the trajectory tracking of the moving target.
[0037] For example, to achieve the above purpose, the application also provides a video motion tracking device, which comprises a memory, a processor, and a video motion tracking program stored in the memory and executable on the processor, and the video motion tracking program is executed by the processor to realize the steps of the video motion tracking method as described above.
[0038] Exemplarily, to achieve the above object, the application further provides a computer storage medium, wherein a video motion tracking program is stored on the computer storage medium, and the video motion tracking program, when executed by a processor, implements the steps of the video motion tracking method as described above.
[0039] Compared with the prior art, in which a moving target in a video is difficult to be effectively tracked when the moving target is in a complex environment or is occluded, the application extracts a moving target from at least one original image of a video stream sequence, and obtains an identification feature of the moving target; based on the identification feature, the moving target is analyzed and a motion parameter of the moving target in the original image of a current frame is obtained, a prediction mechanism algorithm is established according to the moving target and the motion parameter of at least one frame, and when the moving target is not identified, the moving target in a real-time image of the video stream sequence is positioned based on the prediction mechanism algorithm, so as to realize trajectory tracking of the moving target. It can be understood that the motion parameter of the moving target in the video stream sequence is analyzed through the identification feature of the moving target, the prediction mechanism algorithm is created according to the motion parameter, and then when the moving target is occluded or is in a complex environment and is not identified, the moving target is continuously positioned according to the prediction mechanism algorithm, so as to realize trajectory tracking of the moving target. BRIEF DESCRIPTION OF DRAWINGS
[0040] Figure 1 is a flowchart of a first embodiment of the video motion tracking method of the application;
[0041] Figure 2 is a functional module schematic diagram of a preferred embodiment of the video motion tracking device of the application;
[0042] Figure 3 is a structural schematic diagram of a hardware running environment related to the embodiment scheme of the application.
[0043] The implementation of the object, the functional features and the advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION
[0044] It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.
[0045] The application provides a video motion tracking method, which will be described below with reference to Figure 1 , Figure 1 is a flowchart of the video motion tracking method of the application.
[0046] This application also provides embodiments of a video motion tracking method. It should be noted that although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here. The video motion tracking method can be applied to a computer. For ease of description, the execution entity description of each step of the video motion tracking method is omitted below. The video motion tracking method includes:
[0047] Step S110: Extract the moving target from at least one frame of the original image of the video stream sequence and obtain the recognition features of the moving target.
[0048] It should be noted that a video stream sequence refers to a video composed of original images arranged in chronological order. That is, a video stream sequence includes at least one original image frame, with each frame corresponding to a single scene. Video streams can be offline or streamed online in real-time.
[0049] A motion target refers to a moving object in a video that needs to be detected and tracked. This moving object can be a human (such as an athlete), an animal (such as a dog), or other subjects that need to be evaluated for motion. No specific limitations are made here. The following description will use an athlete as an example to illustrate the motion target.
[0050] Recognition features refer to the features of a moving target used to identify and detect that moving target. These include visual features, such as color features under various viewpoints, scales, and lighting conditions, as well as shape features, edge features, etc.
[0051] In this embodiment, at least one frame of original image is acquired from the video stream sequence, and the moving target is extracted from the original image. During this process, the contour of the moving target is obtained by performing a difference operation on two adjacent frames in the video stream sequence. When abnormal subject movement occurs in the original image, a significant difference will appear between adjacent frames. The data of adjacent frames are subtracted to obtain the absolute value of the brightness difference between the two original images. It is then determined whether this absolute value is greater than a difference threshold to assess the motion characteristics of the video stream sequence and determine whether there is motion in the original image. Essentially, by comparing the difference image with the difference threshold, it is determined whether each pixel in the original image is a moving target pixel or a background pixel, thereby extracting the moving target. For example, when the absolute value is greater than the difference threshold, it is determined that there is motion in the original image, and this object is the moving target. At this point, the recognition features of the moving target are extracted.
[0052] It should be noted that the method of extracting the moving target by difference operation between the original images of the adjacent two frames in the video stream sequence is an optional extraction scheme, and other schemes for extracting the moving target can also be used for determination of the moving target, such as edge detection, optical flow method, etc. The extraction schemes can be used independently or in combination, and are not limited here.
[0053] In the embodiment, the moving target extracted from the video stream sequence and the identification feature corresponding to the moving target are used for identification and detection of the moving target in other images of the video stream sequence, and can also be used for identification of the moving target in a real-time played video, thereby providing a basis for positioning and tracking of the moving target.
[0054] For example, the extraction of the moving target from at least one frame of the original image of the video stream sequence and the acquisition of the identification feature of the moving target include:
[0055] In step S111, at least one frame of the original image of the video stream sequence is acquired, and a moving region is detected in the original image;
[0056] In step S112, the moving target in the moving region is determined.
[0057] In step S113, at least one identification feature of the moving target is acquired, and the identification feature is used for identification of the moving target in trajectory tracking.
[0058] In the process of extracting the moving target from at least one frame of the original image of the video stream sequence, the detection of the moving region is faster than the identification and detection of the feature, which can improve the extraction efficiency of the moving target. Therefore, the moving region is detected from the original image by difference operation, and the moving target is further determined based on the moving region, that is, the moving target is extracted by refined difference operation.
[0059] After the moving target is determined in the original image of the video stream sequence, at least one identification feature of the moving target is extracted, and the identification feature is used for identification and detection of the moving target in other frames of the original image, thereby tracking the moving target.
[0060] The identification feature of the moving target is acquired, wherein the identification feature can be selected according to actual requirements, and can also be selected according to the feature weight when the moving target is extracted, so as to extract the feature convenient for detection or identification in the moving target.
[0061] For example, the identified features are color features, texture features, shape features, point, line, curve features corresponding to regions with obvious marks, perimeter, area, and centroid features of a moving target, and the like.
[0062] For example, the identified features are color features, texture features, shape features, point, line, curve features corresponding to regions with obvious marks, perimeter, area, and centroid features of a moving target, and the like.
[0063] In step S120, based on the identified features, motion parameters of the moving target in the original image of the current frame are analyzed and obtained, and a prediction mechanism algorithm is established according to the moving target and the motion parameters of at least one frame.
[0064] Based on the identified features, motion parameters of the moving target in the original image of the current frame are extracted, and the motion parameters of the moving target in the original image of the current frame are obtained. The motion parameters include position parameters, velocity parameters, acceleration parameters, and the like of the moving target, which are used to reflect the motion trajectory of the moving target.
[0065] The motion parameters of the moving target in at least one frame of the original image are extracted, and the motion parameters of the respective frames are obtained. The motion trajectory of the moving target can be analyzed by combining the at least one motion parameter. Therefore, the prediction mechanism algorithm is established by the motion parameters, and the moving target in the real-time image of the video stream sequence can be positioned by the prediction mechanism algorithm, so as to continue the trajectory tracking when the moving target is blocked or overlapped.
[0066] For example, the identified features are color features, texture features, shape features, point, line, curve features corresponding to regions with obvious marks, perimeter, area, and centroid features of a moving target, and the like.
[0067] In step S121, based on the identified features, the moving target in the original image of the current frame is identified.
[0068] In step S122, the position coordinates of the moving target in the original image of the current frame are calculated.
[0069] Step S123: Associatively process the position coordinates in at least one frame of the original image to obtain at least one motion parameter of the moving target;
[0070] Step S124: Establish the prediction mechanism algorithm based on the parameter set formed by at least one of the motion parameters.
[0071] Based on the recognition features of moving targets, moving targets are identified from the original image of the current frame of the video stream sequence. That is, by recognizing the moving object in the original image of the current frame, it is determined whether the features of the moving object correspond to the recognition features. If they correspond, the moving object is determined to be the moving target that needs to be tracked or detected.
[0072] After identifying the moving target in the original image of the current frame, the target's position coordinates in the original image are determined based on its centroid. The original image of the current frame corresponds to one moving target's position coordinates and image frame number, where the image frame number refers to the specific frame of the current original image in the video stream sequence. Since the original images of adjacent frames, or other frames, also correspond to specific position coordinates and image frame numbers, at least one position coordinate and image frame number from at least one original image are correlated to obtain the motion parameters of the moving target in the original images of different frames. Based on the motion parameters of at least one original image and its image frame number, a prediction mechanism algorithm is established.
[0073] For example, in establishing the prediction mechanism algorithm, motion parameters are calculated based on the frame time of adjacent original images. It can be understood that the frame time is determined by the temporal order of the original images in the video stream sequence. For instance, if 15 frames are taken per second, then the frame time between two adjacent frames is 1 / 15 of a second. That is, the time corresponding to the position coordinates increases incrementally in units of 1 / 15 seconds. It should be noted that the number of frames taken per second is determined based on the actual operation and is not specifically limited here. It can be understood that if the time of the original image in the current frame is n seconds, the time of the original image in the next frame is n+1 / 15 seconds.
[0074] In some scenarios, if the position coordinates of the moving target in the previous frame are (a1, b1, c1), and the position coordinates of the moving target in the current frame are (a2, b2, c2), and if direction 'a' represents the direction of the moving target's movement, and the frame time between adjacent original images is 1 / 15 second, then based on conventional mathematical algorithms, position coordinate information, and frame time information, the position parameters, velocity parameters, and acceleration parameters in the motion parameters are calculated. The motion parameters of the moving target in a frame of the original image and the number of frames in that original image constitute a set of historical data. A parameter set is established using at least one set of historical data.
[0075] The preset basic mechanism algorithm is acquired, the basic mechanism algorithm is trained through a parameter set, and an instantiated prediction mechanism algorithm is obtained. It should be noted that the basic mechanism algorithm can be a conventional mathematical algorithm (for example, y=kx+h, where k is the number of image frames, y is a certain parameter in the motion parameter, and the values of the algorithm coefficients k and h can be obtained through training to obtain the prediction mechanism algorithm), and the process of training the basic mechanism algorithm can be implemented through a convolutional neural network, which is not limited here.
[0076] For example, the moving target includes at least one identification feature, and before identifying the moving target in the original image of the current frame based on the identification feature, the method comprises:
[0077] Step A1, combining at least one identification feature to obtain a feature combination;
[0078] Then, identifying the moving target in the original image of the current frame based on the identification feature comprises:
[0079] Step A2, identifying the moving target corresponding to the feature combination in the original image of the current frame based on the feature combination.
[0080] Since the moving target has at least one identification feature, which can be used to identify the moving target in the original image, in order to increase the accuracy of identification, at least one identification feature can be combined to obtain a feature combination, and the moving target is detected through multiple identification features in the feature combination, and the moving target corresponding to the feature combination in the original image of the current frame is identified.
[0081] For example, identifying the moving target in the original image of the current frame based on the identification feature comprises:
[0082] Step S1211, obtaining a constraint condition of the identification feature, and extracting a target feature of a moving object in the original image of the current frame based on the constraint condition;
[0083] Step S1212, comparing the target feature with the identification feature, and if the comparison result is consistent, determining that the moving object is the moving target.
[0084] It should be noted that the constraint condition refers to the range in which the identification feature is located. For example, when the identification feature is a shape feature, the shape feature of the moving target can be expanded by 10%, to obtain a new shape with a larger area, which is the constraint condition of the shape feature. The moving object corresponding to the shape feature within the constraint condition will be detected. The further refined moving object is further identified, that is, the target feature of the moving object in the original image of the current frame is extracted. The target feature has the same attribute as the identification feature.
[0085] The target feature is compared with the recognition feature, and whether the moving object is the tracked moving target is determined according to the comparison result.
[0086] For example, if the comparison result of the target feature and the recognition feature is consistent, it is determined that the moving object is the moving target.
[0087] For example, if the comparison result of the target feature and the recognition feature is inconsistent, it is determined that the moving object is not the moving target, indicating that the moving target is not detected in the original image of the current frame.
[0088] When the moving target is not detected in the original image of the current frame, there are two cases:
[0089] Case one: the target feature recognition is not comprehensive, resulting in inconsistent comparison results. At this time, the step of extracting the target feature of the moving object in the original image of the current frame based on the constraint condition is performed again, and the moving target continues to be recognized.
[0090] Case two: there is no moving target in the original image of the current frame or the moving target is blocked and cannot be recognized. Then, the step of positioning the moving target in the real-time image of the video stream sequence based on the prediction mechanism algorithm when the moving target recognition fails is continued to be performed, the position of the moving target in the next frame is predicted, and the continuous tracking of the moving target can be realized.
[0091] For example, the extraction of the target feature of the moving object in the original image of the current frame based on the constraint condition comprises:
[0092] Step B1: If the moving object extracted based on the constraint condition is multiple, at least one target feature is extracted in the original image of the current frame;
[0093] The target feature is compared with the recognition feature, and if the comparison result is consistent, it is determined that the moving object is the moving target, comprising:
[0094] Step B2: The at least one target feature is matched with the recognition feature one by one, and the moving object corresponding to the target feature with a successful matching result is the moving target.
[0095] If there are multiple moving objects extracted from the original image based on the constraint condition, the extraction operation of the target feature is performed on each moving object. Therefore, in the process of determining whether the moving object is the moving target, multiple target features and recognition features are matched one by one, and the moving object corresponding to the target feature with a successful matching result is the moving target.
[0096] Step S130, when the moving target recognition fails, the moving target in the real-time image of the video stream sequence is located based on the prediction mechanism algorithm to realize the trajectory tracking of the moving target.
[0097] When the moving target recognition fails, it means that there is no moving target in the image of the current frame or the moving target is blocked and overlapped and cannot be recognized. Then, the position of the moving target in the real-time image of the next frame or the following frames can be predicted based on the prediction mechanism algorithm, i.e. the moving target is located to realize the trajectory tracking of the moving target.
[0098] For example, when the moving target recognition fails, the moving target in the real-time image of the video stream sequence is located based on the prediction mechanism algorithm to realize the trajectory tracking of the moving target, including:
[0099] Step S131, when the moving target recognition fails, the prediction mechanism algorithm is obtained, and the prediction mechanism algorithm includes image frame number and target parameter.
[0100] Step S132, the current frame number corresponding to the real-time image of the video stream sequence is input into the image frame number of the prediction mechanism algorithm, the target parameter corresponding to the current frame number is calculated, and the target motion parameter is obtained.
[0101] Step S133, the moving target is located based on the target motion parameter to obtain the current position of the moving target to realize the trajectory tracking of the moving target.
[0102] When the moving target recognition fails, the prediction mechanism algorithm is obtained, and the prediction mechanism algorithm includes image frame number and target parameter. The image frame number refers to the frame number of the real-time image of the current frame in the video stream sequence, and the target parameter refers to the motion parameter of the moving target in the real-time image corresponding to the image frame number.
[0103] Therefore, the current frame number corresponding to the real-time image in the video stream sequence is input into the image frame number of the prediction mechanism algorithm, and the target parameter corresponding to the current frame number is calculated by the prediction mechanism algorithm, i.e. the target motion parameter. Different parameters in the target motion parameter, such as position parameter, speed parameter and acceleration parameter, are calculated by different prediction mechanism algorithms. The moving target in the real-time image can be located by the target motion parameter, the problem of repositioning after the moving target disappears is avoided, and the continuous trajectory tracking of the moving target is realized, which is of great benefit to the motion evaluation or tracking technology.
[0104] Compared with the prior art, the motion target in a video is difficult to track effectively when the motion target is in a complex environment or is occluded. The motion target is extracted from at least one original image of a video stream sequence, and identification features of the motion target are obtained. Motion parameters of the motion target in the original image of a current frame are analyzed and obtained based on the identification features. A prediction mechanism algorithm is established according to the motion target and the motion parameters of at least one frame. When the motion target identification fails, the motion target in a real-time image of the video stream sequence is positioned based on the prediction mechanism algorithm, so as to realize trajectory tracking of the motion target. It can be understood that the motion parameters of the motion target in the video stream sequence are analyzed based on the identification features of the motion target, the prediction mechanism algorithm is created according to the motion parameters, and then when the motion target is occluded or is in a complex environment and identification fails, the motion target is continuously positioned based on the prediction mechanism algorithm, so as to realize trajectory tracking of the motion target.
[0105] As shown in the figure, Figure 2 The video motion tracking device provided by the application comprises:
[0106] The feature extraction module 401 is configured to extract a motion target from at least one original image of a video stream sequence, and obtain identification features of the motion target.
[0107] The feature identification module 402 is configured to analyze and obtain motion parameters of the motion target in the original image of a current frame based on the identification features, and establish a prediction mechanism algorithm according to the motion target and the motion parameters of at least one frame.
[0108] The trajectory tracking module 403 is configured to position the motion target in a real-time image of the video stream sequence based on the prediction mechanism algorithm when the motion target identification fails, so as to realize trajectory tracking of the motion target.
[0109] In a possible design, the feature extraction module 401 is specifically configured to:
[0110] Obtain at least one original image of the video stream sequence, and detect a motion region in the original image.
[0111] Determine the motion target in the motion region.
[0112] Obtain at least one identification feature of the motion target, and the identification feature is used for identification of the motion target in trajectory tracking.
[0113] In a possible design, the feature identification module 402 is specifically configured to:
[0114] identify the moving object in the original image of the current frame based on the identified features;
[0115] calculate the position coordinates of the moving object in the original image of the current frame;
[0116] perform association processing on the position coordinates in the original image of at least one frame to obtain at least one motion parameter of the moving object;
[0117] establish the prediction mechanism algorithm according to a parameter set formed based on at least one motion parameter.
[0118] In a possible design, the feature identification module 402 is further configured to:
[0119] combine at least one identified feature to obtain a feature combination;
[0120] In a possible design, the feature identification module 402 is further configured to:
[0121] identify the moving object corresponding to the feature combination in the original image of the current frame based on the feature combination.
[0122] In a possible design, the feature identification module 402 is further configured to:
[0123] obtain a constraint condition of the identified feature, and extract a target feature of a moving object in the original image of the current frame based on the constraint condition;
[0124] compare the target feature with the identified feature, and determine that the moving object is the moving object if the comparison result is consistent.
[0125] In a possible design, the feature identification module 402 is further configured to:
[0126] if the moving object extracted based on the constraint condition is multiple, at least one target feature is extracted in the original image of the current frame;
[0127] In a possible design, the feature identification module 402 is further configured to:
[0128] match the at least one target feature with the identified feature one by one, and the moving object corresponding to a target feature that passes the matching is the moving object.
[0129] In a possible design, the trajectory tracking module 403 is specifically configured to:
[0130] When the moving target recognition fails, a prediction mechanism algorithm is acquired, the prediction mechanism algorithm including image frame number and target parameter;
[0131] A current frame number corresponding to a real-time image of the video stream sequence is input into the image frame number of the prediction mechanism algorithm, the target parameter corresponding to the current frame number is calculated, and a target motion parameter is obtained;
[0132] The moving target is positioned based on the target motion parameter, and a current position of the moving target is obtained, so as to realize trajectory tracking of the moving target.
[0133] The video motion tracking device specific embodiment of the present application is basically the same as each embodiment of the above-mentioned video motion tracking method, and will not be repeated here.
[0134] In addition, the present application also provides a video motion tracking device. As shown in Figure 3 Figure 3 is a structural diagram of a hardware running environment involved in the embodiment scheme of the present application.
[0135] In one possible implementation, Figure 3 that is, a structural diagram of a hardware running environment of the video motion tracking device.
[0136] As shown in Figure 3 , the video motion tracking device can include a processor 701, a communication interface 702, a memory 703, and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 complete mutual communication through the communication bus 704, the memory 703 is used to store a computer program, and the processor 701 is used to execute the program stored on the memory 703 to realize the steps of the video motion tracking method.
[0137] The communication bus 704 mentioned in the above-mentioned video motion tracking device can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 704 can be divided into an address bus, a data bus, and a control bus, etc. For the convenience of representation, only one thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.
[0138] The communication interface 702 is used for communication between the above-mentioned video motion tracking device and other devices.
[0139] The memory 703 may include random access memory (RMD) or non-volatile memory (NM), such as at least one disk storage device. Optionally, the memory 703 may also be at least one storage device located remotely from the aforementioned processor 701.
[0140] The processor 701 mentioned above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0141] The specific implementation of the video motion tracking device in this application is basically the same as the embodiments of the video motion tracking method described above, and will not be repeated here.
[0142] Furthermore, this application also proposes a computer storage medium storing a video motion tracking program, which, when executed by a processor, implements the steps of the video motion tracking method described above.
[0143] The specific implementation of the computer storage medium in this application is basically the same as the embodiments of the video motion tracking method described above, and will not be repeated here.
[0144] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0145] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0146] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, can also be through hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of the prior art contribution can be embodied in the form of software products, the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), including a number of instructions to make a terminal device (may be a mobile phone, computer, server, device, or network equipment, etc.) executes the method described in various embodiments of the present application.
[0147] The above is only the preferred embodiment of the present application, not therefore limit the patent scope of the present application, all use the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, equivalent structure or equivalent process transformation, or, are also included in the patent protection scope of the present application.
Claims
1. A video motion tracking method, characterized in that, The method includes: Extract the moving target from at least one frame of the original image in the video stream sequence and obtain the identification features of the moving target; Based on the identified features, the motion parameters of the moving target in the original image of the current frame are analyzed and obtained. Based on the moving target and the motion parameters of at least one frame, a prediction mechanism algorithm is established. When the moving target recognition fails, the moving target in the real-time image of the video stream sequence is located based on the prediction mechanism algorithm to achieve trajectory tracking of the moving target; The step involves analyzing and obtaining the motion parameters of the moving target in the original image of the current frame based on the identified features, and establishing a prediction mechanism algorithm based on the moving target and the motion parameters of at least one frame, including: Based on the recognition features, the moving target in the original image of the current frame is identified; Calculate the position coordinates of the moving target in the original image of the current frame; The position coordinates in at least one frame of the original image are correlated to obtain at least one motion parameter of the moving target; The prediction mechanism algorithm is established based on a parameter set formed by at least one of the motion parameters.
2. The method as described in claim 1, characterized in that, Extracting a moving target from at least one frame of the original image in the video stream sequence and obtaining the recognition features of the moving target includes: Acquire at least one frame of the original image from the video stream sequence, and detect the motion region in the original image; Determine the moving target within the motion region; At least one of the identification features of the moving target is obtained, and the identification feature is used to identify the moving target during trajectory tracking.
3. The method as described in claim 1, characterized in that, The moving target includes at least one of the recognition features, and before recognizing the moving target in the original image of the current frame based on the recognition features, the process includes: At least one of the identification features is combined to obtain the feature combination; The step of identifying the moving target in the original image of the current frame based on the identified features includes: Based on the feature combination, the moving target corresponding to the feature combination in the original image of the current frame is identified.
4. The method as described in claim 1, characterized in that, The step of identifying the moving target in the original image of the current frame based on the identified features includes: Obtain the constraints of the recognition features, and extract the target features of the moving objects in the original image of the current frame based on the constraints; The target features are compared with the identification features. If the comparison results are consistent, the moving object is determined to be the moving target.
5. The method as described in claim 4, characterized in that, The step of extracting target features of moving objects in the original image of the current frame based on the constraints includes: If there are multiple moving objects extracted based on the constraints, then at least one target feature is extracted from the original image of the current frame. The target feature is compared with the recognition feature. If the comparison result is consistent, the moving object is determined to be the moving target, including: The at least one target feature is matched one by one with the recognition feature, and the moving object corresponding to the successfully matched target feature is the moving target.
6. The method as described in claim 1, characterized in that, When the moving target recognition fails, the moving target in the real-time image of the video stream sequence is located based on the prediction mechanism algorithm to achieve trajectory tracking of the moving target, including: When the moving target recognition fails, the prediction mechanism algorithm is obtained, which includes the number of image frames and target parameters; The current frame number corresponding to the real-time image of the video stream sequence is input into the image frame number of the prediction mechanism algorithm to calculate the target parameter corresponding to the current frame number and obtain the target motion parameter; The moving target is located based on the target motion parameters to obtain the current position of the moving target, so as to realize the trajectory tracking of the moving target.
7. A video motion tracking device, characterized in that, The device includes: The feature extraction module is used to extract a moving target from at least one frame of the original image of the video stream sequence and obtain the recognition features of the moving target; The feature recognition module is used to analyze and obtain the motion parameters of the moving target in the original image of the current frame based on the recognition features, and to establish a prediction mechanism algorithm based on the moving target and the motion parameters of at least one frame. The trajectory tracking module is used to locate the moving target in the real-time image of the video stream sequence based on the prediction mechanism algorithm when the moving target recognition fails, so as to achieve trajectory tracking of the moving target; The video motion tracking device is used to achieve: Based on the recognition features, the moving target in the original image of the current frame is identified; Calculate the position coordinates of the moving target in the original image of the current frame; The position coordinates in at least one frame of the original image are correlated to obtain at least one motion parameter of the moving target; The prediction mechanism algorithm is established based on a parameter set formed by at least one of the motion parameters.
8. A video motion tracking device, characterized in that, The video motion tracking device includes a memory, a processor, and a video motion tracking program stored in the memory and executable on the processor, wherein the video motion tracking program, when executed by the processor, implements the steps of the video motion tracking method as described in any one of claims 1 to 6.
9. A computer storage medium, characterized in that, The computer storage medium stores a video motion tracking program, which, when executed by a processor, implements the steps of the video motion tracking method as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Multi-target tracking method, device and equipment and storage medium
CN111292352A