Method, System, Device and Medium for Determining Pedestrian Movement Trajectory

Abstract

The present invention relates to the technical field of trajectory tracking, and specifically provides a method, a system, a device and a medium for determining a pedestrian movement trajectory, aiming at the problem of how to accurately and effectively associate the pedestrian movement trajectories located in different image acquisition ranges and obtain the complete movement trajectory of the pedestrian without ensuring the continuous effectiveness of the human appearance features of the pedestrian. For this purpose, the present invention can obtain the human feature information corresponding to the pedestrian movement trajectories respectively through different image acquisition devices, and associate the pedestrian movement trajectories belonging to the same pedestrian according to the relationship features between the image acquisition devices and the human feature information. The present invention comprehensively considers the influence of the relationship features between the image acquisition devices and the human feature information on the association of the pedestrian movement trajectories, and can make the finally generated pedestrian movement trajectories corresponding to each pedestrian more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory tracking, and specifically provides a method, system, device and medium for determining a pedestrian movement trajectory. Background Art

[0002] Multi-target trajectory aggregation refers to obtaining the movement trajectories of multiple targets through multiple cameras to determine the overall movement trajectory of each target. Multi-target trajectory aggregation has a wide range of applications in scenarios such as shopping malls, parks, airports, and security. The multi-target trajectory aggregation methods in the prior art often use a single-lens tracking algorithm to obtain the movement trajectory of a target within a single camera, and then associate the movement trajectories between multiple cameras through the human appearance characteristics of the target, so as to obtain the complete movement trajectory of the target under multiple cameras. However, the human appearance characteristics are not always effective, such as the human appearance characteristics being blocked, blurred, or having similar clothing, which results in a poor effect of the overall movement trajectory of the target obtained by the existing multi-target trajectory aggregation methods.

[0003] There is a need in the art for a new solution for determining a pedestrian movement trajectory to solve the above problems. Summary of the Invention

[0004] The present invention aims to solve the above technical problems, that is, to solve or partially solve the problem of how to accurately and effectively associate the movement trajectories of pedestrians within the image acquisition range without ensuring the effectiveness of the human appearance characteristics of pedestrians all the time, and obtain the complete movement trajectory of pedestrians. The present invention provides a method, system, device and medium for determining a pedestrian movement trajectory, including:

[0005] In a first aspect, the present invention provides a method for determining a pedestrian movement trajectory, the method including:

[0006] Obtaining the movement trajectories of pedestrians within the image acquisition ranges of different image acquisition devices;

[0007] Obtaining the relationship characteristics between different image acquisition devices;

[0008] Obtaining the human feature information corresponding to each pedestrian movement trajectory;

[0009] Associating the pedestrian movement trajectories belonging to the same pedestrian according to the human feature information and the relationship characteristics, and generating the final pedestrian movement trajectories corresponding to each pedestrian.

[0010] In a technical solution of the above method for determining a pedestrian movement trajectory, the step of "associating the pedestrian movement trajectories belonging to the same pedestrian according to the human feature information and the relationship characteristics, and generating the final pedestrian movement trajectories corresponding to each pedestrian" specifically includes:

[0011] Determine the feature similarity between different pedestrian movement trajectories according to the human feature information;

[0012] Calculate the association score between different pedestrian movement trajectories according to the feature similarity and the relationship feature;

[0013] If the association score between two pedestrian movement trajectories is greater than a preset score threshold, determine that the two pedestrian movement trajectories belong to the same pedestrian's movement trajectory;

[0014] Associate the pedestrian movement trajectories belonging to the same pedestrian to generate the final pedestrian movement trajectory corresponding to each pedestrian; and / or,

[0015] The relationship feature includes the spatial relationship feature between different image acquisition devices and the temporal relationship feature between different image acquisition devices. Among them, the spatial relationship feature between two image acquisition devices is used to represent the association weight between the two image acquisition devices in the topological structure formed by the mutual association of all image acquisition devices, and the temporal relationship feature between two image acquisition devices is used to represent the probability that a pedestrian spends different durations moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device.

[0016] In a technical solution of the above pedestrian movement trajectory determination method, the step of "calculating the association score between different pedestrian movement trajectories according to the feature similarity and the relationship feature" specifically includes calculating the association score through the following formula:

[0017] z ij＝ e ij -decay + decay×s ij

[0018] Among them, the z ij represents the association score between two pedestrian movement trajectories, the e ij represents the feature similarity between two pedestrian movement trajectories, the s ij represents the relationship feature of the relationship features between the image acquisition devices corresponding to two pedestrian movement trajectories respectively, and the decay represents a preset decay factor;

[0019] When the relationship feature includes the spatial relationship feature between different image acquisition devices and the temporal relationship feature between different image acquisition devices, the s ij represents the product of the spatial relationship feature and the temporal relationship feature between the image acquisition devices corresponding to two pedestrian movement trajectories respectively.

[0020] In a technical solution of the above pedestrian movement trajectory determination method, the method further includes calculating the feature similarity e between two pedestrian movement trajectories through the following formulaij :

[0021] e ij = e i T ·e j

[0022] wherein, the e i represents a central vector of a pedestrian movement trajectory, the e j represents another central vector of a pedestrian movement trajectory, and the T represents the vector transpose of the central vector e i ;

[0023] The x i represents a feature vector determined by the i-th human body feature information, and c(x i ) represents the pedestrian movement trajectory to which the i-th human body feature information belongs. "{x|c(x) = c(x i )}" represents the set of feature vectors of all human body feature information corresponding to the pedestrian movement trajectory c(x i ), and the norm represents a normalization operation;

[0024] The x j represents a feature vector determined by the j-th human body feature information, and c(x j ) represents the pedestrian movement trajectory to which the j-th human body feature information belongs. "{x|c(x) = c(x j )}" represents the set of feature vectors of all human body feature information corresponding to the pedestrian movement trajectory c(x j ).

[0025] In a technical solution of the above pedestrian movement trajectory determination method, the relationship features include spatial relationship features between different image acquisition devices and temporal relationship features between different image acquisition devices. Among them, the spatial relationship features between two image acquisition devices are used to represent the association weight between the two image acquisition devices in the topological structure formed by the mutual association of all image acquisition devices, and the temporal relationship features between two image acquisition devices are used to represent the probability that a pedestrian spends different lengths of time moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device. The step of "acquiring the spatial relationship features and temporal relationship features between different image acquisition devices" specifically includes:

[0026] Acquiring human feature samples obtained from pedestrian images collected by each image acquisition device, wherein the human feature samples include human features extracted from the pedestrian images and the device numbers of the image acquisition devices that collect the pedestrian images;

[0027] For each human feature sample, obtain similar human feature samples that are similar to the human features of the current human feature sample from all human feature samples, and form a matching sample pair by combining the current human feature sample with each similar human feature sample respectively;

[0028] For each matching sample pair, obtain the two human feature samples x i and x j in the current matching sample pair, and their respective device numbers c i and c j of the image acquisition devices. Based on the device numbers c i and c j , establish an image acquisition device pair (c i , c j ); obtain the acquisition times ts i and ts j of the pedestrian images corresponding to the two human feature samples x i and x j in the current matching sample pair (c i , c j ). Determine the time interval |ts i - ts j | between the two acquisition times ts i and ts j . Add the time interval |ts i - ts j | to the time interval list of the current image acquisition device pair (c i , c j );

[0029] For each image acquisition device pair, perform binning processing on the time interval list of the current image acquisition device pair (c i , c j ) according to the preset number of bins. Establish a histogram of the current image acquisition device pair (c i , c j ) based on the number of time intervals falling into each bin after binning processing. Among them, the abscissa of the histogram represents the time it takes for a pedestrian to move from the image acquisition range of the image acquisition device with device number c i to the image acquisition range of the image acquisition device with device number c j , and the ordinate represents the probability of taking that time;

[0030] Based on the histogram of each image acquisition device pair, determine the spatial relationship feature and time relationship feature between the two image acquisition devices in each image acquisition device pair.

[0031] In a technical solution of the above method for determining the pedestrian movement trajectory, the step of "determining the spatial relationship feature and the temporal relationship feature between the two image acquisition devices in each pair of image acquisition devices according to the histogram of each pair of image acquisition devices" specifically includes:

[0032] Determine the temporal relationship feature between the two image acquisition devices in the pair of image acquisition devices according to the ordinate of the histogram of the pair of image acquisition devices;

[0033] Determine the spatial relationship feature between the two image acquisition devices according to the following formula:

[0034]

[0035] where k ij refers to the spatial relationship feature between the two image acquisition devices c i and c j corresponding to the two human feature samples x i and x j in the matching sample pair, n ij refers to the total number of time intervals in the time interval list of the pair of image acquisition devices (c i , c j ), and refers to the sum of the total number of time intervals of all pairs of image acquisition devices.

[0036] In a technical solution of the above method for determining the pedestrian movement trajectory, in the step of "establishing the histogram of the current pair of image acquisition devices (c i , c j ) according to the number of time intervals falling into each bin after binning processing", the specific steps include:

[0037] Perform Gaussian smoothing on the number of time intervals falling into each bin after binning processing to obtain the number of time intervals in each bin after Gaussian smoothing;

[0038] Normalize the number of time intervals in each bin after Gaussian smoothing to obtain the normalization result corresponding to each bin;

[0039] Use the value corresponding to each bin as the abscissa and the normalization result corresponding to each bin as the ordinate to establish the histogram of the current pair of image acquisition devices (c i , c j ).

[0040] In a technical solution of the above method for determining a pedestrian's movement trajectory, the human body feature information includes human body features extracted from a pedestrian image collected by an image acquisition device, the acquisition time of the pedestrian image, the device number of the image acquisition device, and the trajectory number of the pedestrian movement trajectory to which the human body feature belongs; the step of "associating the pedestrian movement trajectories belonging to the same pedestrian according to the human body feature information and the relationship feature to generate the final pedestrian movement trajectory corresponding to each pedestrian" specifically includes:

[0041] Step S1: For each piece of human body feature information, obtain similar human body feature information that is similar to the human body feature of the current human body feature information from all human body feature information, and form a matching feature pair by combining the current human body feature information with each piece of similar human body feature information respectively;

[0042] Step S2: For each matching feature pair, determine whether the trajectory numbers corresponding to the two pieces of human body feature information in the current matching feature pair are the same;

[0043] Step S21: If they are the same, determine that the pedestrian movement trajectories to which the two pieces of human body feature information in the current matching feature pair belong are the pedestrian movement trajectories of the same pedestrian;

[0044] Step S22: If they are different, obtain the histogram of the pair of image acquisition devices corresponding to the device numbers of the two pieces of human body feature information in the current matching feature pair, and obtain the spatial relationship feature between the two image acquisition devices corresponding to the current matching feature pair according to the histogram;

[0045] Obtain the time interval determined by the acquisition times corresponding to the two pieces of human body feature information in the current matching feature pair, query the histogram with the time interval as the abscissa to obtain the corresponding ordinate, and use the ordinate as the time relationship feature;

[0046] According to the human body feature information, the spatial relationship feature, and the time relationship feature, determine whether the pedestrian movement trajectories to which the two pieces of human body feature information in the current matching feature pair belong are the pedestrian movement trajectories of the same pedestrian; if so, modify the trajectory numbers corresponding to the two pieces of human body feature information to the same trajectory number;

[0047] Step S3: Arrange the human body feature information with the same trajectory number in the order of the acquisition time corresponding to the human body feature information from the earliest to the latest, and generate the final pedestrian movement trajectory corresponding to each pedestrian according to the arranged human body feature information.

[0048] In a technical solution of the above method for determining a pedestrian's movement trajectory, the method further includes obtaining similar human body features to the current target human body feature from all target human body features through the following steps, where the target human body feature is the human body feature sample or the human body feature information:

[0049] Obtain the feature similarity between the current target human feature and other target human features respectively;

[0050] Obtain the start time beg of the pedestrian movement trajectory to which the current target human feature belongs i and the end time end i , and obtain the start time beg of the pedestrian movement trajectory to which each of the other target human features belongs ij and the end time end ij ;

[0051] For each of the other target human features, if the feature similarity corresponding to the other target human feature is greater than the similarity threshold and the start time beg ij and the end time end ij satisfy beg ij < end i +tw and end ij < beg i -tw, then regard the other target human feature as the similar human feature of the current target human feature, where tw is a preset time window.

[0052] In a second aspect, the present invention provides a pedestrian movement trajectory determination system, and the system includes:

[0053] A pedestrian movement trajectory acquisition module, which is configured to acquire the pedestrian movement trajectories within the image acquisition ranges of different image acquisition devices;

[0054] A relationship feature acquisition module, which is configured to acquire the relationship features between different image acquisition devices;

[0055] A human feature information acquisition module, which is configured to acquire the human feature information corresponding to each pedestrian movement trajectory;

[0056] A final pedestrian movement trajectory generation module, which is configured to associate the pedestrian movement trajectories belonging to the same pedestrian according to the human feature information and the relationship features, and generate the final pedestrian movement trajectories corresponding to each pedestrian.

[0057] In a third aspect, a control device is provided. The control device includes a processor and a storage device. The storage device is adapted to store multiple program codes, and the program codes are adapted to be loaded and run by the processor to execute the pedestrian movement trajectory determination method described in any one of the technical solutions of the above-mentioned pedestrian movement trajectory determination method.

[0058] In a fourth aspect, a computer-readable storage medium is provided, in which multiple program codes are stored, and the program codes are adapted to be loaded and run by a processor to execute the pedestrian motion trajectory determination method described in any one of the technical solutions of the above-mentioned pedestrian motion trajectory determination method.

[0059] In the case of adopting the above technical solutions, the present invention can obtain the human feature information corresponding to the pedestrian motion trajectories located within the image acquisition ranges of different image acquisition devices, and associate the pedestrian motion trajectories belonging to the same pedestrian according to the relationship features between different image acquisition devices and the human feature information corresponding to the pedestrian motion trajectories, so as to obtain the final pedestrian motion trajectories corresponding to each pedestrian. Through the above configuration method, the present invention comprehensively considers the influence of the relationship features between image acquisition devices and the human feature information corresponding to pedestrian motion trajectories on the association of pedestrian motion trajectories, prevents the association of pedestrian motion trajectories of different pedestrians with similar human feature information, enhances the association of pedestrian motion trajectories of the same pedestrian with different human feature information, and makes the generated final pedestrian motion trajectories corresponding to each pedestrian more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Referring to the accompanying drawings, the disclosure of the present invention will become easier to understand. It is easy for those skilled in the art to understand that these drawings are only for the purpose of illustration and are not intended to limit the protection scope of the present invention. Among them:

[0061] Figure 1 is a schematic flowchart of the main steps of a pedestrian motion trajectory determination method according to an embodiment of the present invention;

[0062] Figure 2 is a schematic flowchart of the main steps of a pedestrian motion trajectory determination method according to an embodiment of the present invention;

[0063] Figure 3 is a main structural block diagram of a pedestrian motion trajectory determination system according to an embodiment of the present invention;

[0064] Figure 4 is a main structural block diagram of a pedestrian motion trajectory determination system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] The following describes some embodiments of the present invention with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principle of the present invention and are not intended to limit the protection scope of the present invention.

[0066] In the description of the present invention, a "module" and a "processor" may include hardware, software, or a combination of both. A module may include a hardware circuit, various appropriate sensors, communication ports, a memory, and may also include a software part, such as program code, or may be a combination of software and hardware. A processor may be a central processing unit, a microprocessor, an image processor, a digital signal processor, or any other appropriate processor. The processor has data and / or signal processing functions. The processor may be implemented in software, in hardware, or in a combination of both. A non-transitory computer-readable storage medium includes any appropriate medium that can store program code, such as a magnetic disk, a hard disk, an optical disk, a flash memory, a read-only memory, a random access memory, and so on. The term "A and / or B" represents all possible combinations of A and B, such as only A, only B, or A and B. The term "at least one of A or B" or "at least one of A and B" has a meaning similar to "A and / or B" and may include only A, only B, or A and B. The singular terms "a" and "the" may also include the plural form.

[0067] Refer to the appendix Figure 1 , Figure 1 is a schematic diagram of the main steps of a pedestrian motion trajectory determination method according to an embodiment of the present invention. As Figure 1 shown, the pedestrian motion trajectory determination method in the embodiment of the present invention mainly includes the following steps S101-step S104.

[0068] Step S101: Obtain the pedestrian motion trajectories within the image acquisition ranges of different image acquisition devices.

[0069] In this embodiment, the pedestrian motion trajectories can be obtained from within the image acquisition ranges of different image acquisition devices.

[0070] In one implementation, a multi-object tracking algorithm can be used to obtain the pedestrian motion trajectories of multiple pedestrians within the image acquisition range of the image acquisition device. The multi-object tracking algorithm includes, but is not limited to, DeepSort (Simple Online and Realtime Tracking with a Deep Association Metric).

[0071] Step S102: Obtain the relationship features between different image acquisition devices.

[0072] In this embodiment, the relationship features between different image acquisition devices can be obtained. The relationship features include, but are not limited to, time relationship features, space relationship features, and relationship features obtained by combining time relationship features and space relationship features, etc.

[0073] In one embodiment, the relationship features may include spatial relationship features between different image acquisition devices and temporal relationship features between different image acquisition devices. Among them, the spatial relationship features between two image acquisition devices are used to represent the association weight between the two image acquisition devices in the topological structure formed by the mutual association of all image acquisition devices, and the temporal relationship features between two image acquisition devices are used to represent the probability that a pedestrian spends different lengths of time moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device. The spatial relationship features between two image acquisition devices can be used to characterize the association weight between the two image acquisition devices in the topological structure formed by the mutual association of all image acquisition devices, and the association weight can represent the degree of association between the two image acquisition devices. For example, if a pedestrian frequently appears between two image acquisition devices, it can be considered that the association weight between the two image acquisition devices is relatively high. The temporal relationship features between two image acquisition devices can be used to characterize the probability that a pedestrian spends different lengths of time moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device. According to the topological structure between the image acquisition devices, the distance relationship between two image acquisition devices can be confirmed. Then, when a pedestrian spends a short time moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device that is farther away, the probability of this situation is relatively low.

[0074] In one embodiment, a large amount of pedestrian movement trajectory data located within the image acquisition ranges of different image acquisition devices without pedestrian movement trajectory annotation can be read, and these pedestrian movement trajectory data can be used to train a preset relationship feature model. The trained relationship feature model is used to obtain the relationship features between different image acquisition devices.

[0075] Step S103: Obtain the human feature information corresponding to each pedestrian movement trajectory.

[0076] In this embodiment, the human feature information corresponding to the pedestrian movement trajectory can be obtained according to the captured images acquired by the image acquisition device corresponding to each pedestrian movement trajectory.

[0077] In one embodiment, the human feature information may be a human feature vector, and feature extraction can be performed on the captured images acquired by the image acquisition device to obtain the human feature vector.

[0078] In one embodiment, the human feature information may be a human feature sequence, that is, according to multiple captured images corresponding to the pedestrian movement trajectory, the human feature vector of each captured image is extracted, and the set of human feature vectors of multiple captured images is used as the human feature sequence. The human feature information may also be a human feature vector in the human feature sequence. For example, any one of the human feature vectors in the human feature sequence is obtained as the human feature information, or the average vector including multiple human feature vectors in the human feature sequence is taken as the human feature information.

[0079] Step S104: Correlate the pedestrian movement trajectories belonging to the same pedestrian according to the human feature information and the relationship feature, and generate the final pedestrian movement trajectory corresponding to each pedestrian.

[0080] In this embodiment, it is possible to comprehensively determine whether different pedestrian movement trajectories belong to the same pedestrian according to the human feature information corresponding to the pedestrian movement trajectory and the relationship feature of the image acquisition device, and correlate the pedestrian movement trajectories belonging to the same pedestrian to generate the final pedestrian movement trajectory corresponding to each pedestrian.

[0081] Based on the above steps S101 - S104, the embodiment of the present invention can obtain the human feature information corresponding to the pedestrian movement trajectories respectively within the image acquisition ranges of different image acquisition devices, and correlate the pedestrian movement trajectories belonging to the same pedestrian according to the relationship feature between different image acquisition devices and the human feature information corresponding to the pedestrian movement trajectory, so as to obtain the final pedestrian movement trajectory corresponding to each pedestrian. Through the above configuration method, the embodiment of the present invention comprehensively considers the influence of the relationship feature between image acquisition devices and the human feature information corresponding to the pedestrian movement trajectory on the correlation of pedestrian movement trajectories, prevents the correlation of pedestrian movement trajectories of different pedestrians with similar human feature information, enhances the correlation of pedestrian movement trajectories of the same pedestrian with different human feature information, and makes the generated final pedestrian movement trajectory corresponding to each pedestrian more accurate.

[0082] The following further describes steps S102 and S104.

[0083] In one embodiment of the embodiment of the present invention, step S102 may further include the following steps S1021 to S1025:

[0084] Step S1021: Obtain the human feature samples obtained from the pedestrian images captured by each image acquisition device, where the human feature samples include the human features extracted from the pedestrian images and the device numbers of the image acquisition devices that captured the pedestrian images.

[0085] In this embodiment, feature extraction can be performed on the pedestrian images collected by each image acquisition device to obtain multiple human feature samples. The set of human feature samples can be expressed as X = {x 1 , x 2 , …, x n}, and the set of device numbers of the corresponding image acquisition devices can be expressed as C = {c 1 , c 2 , …, c n}, where n is the number of human feature samples.

[0086] Step S1022: For each human feature sample, obtain similar human feature samples that are similar to the human feature of the current human feature sample from all human feature samples, and form a matching sample pair by combining the current human feature sample with each similar human feature sample respectively.

[0087] In one embodiment, the top-k nearest neighbor human feature samples of the human feature samples can be retrieved. The top-k nearest neighbors corresponding to each human feature sample x i can be expressed as n(x i ) = {X i1 , x i2 , …, X ik}. The algorithms used in the retrieval process include but are not limited to brute-force retrieval, IVFPQ algorithm (Inverted File System Product Quantization), HNSW algorithm (Hierarchcal Navigable Small World graphs), etc.

[0088] In one embodiment, according to the following steps S10221 to S10223, obtain human feature samples that are similar to the human feature of the current human feature sample, where the target human feature in steps S10221 to S10223 is the human feature sample:

[0089] Step S10221: Obtain the feature similarity between the current target human feature and other target human features respectively;

[0090] Step S10222: Obtain the start time beg i and end time end i of the pedestrian movement trajectory to which the current target human feature belongs, and obtain the start time beg ij and end time end ij of the pedestrian movement trajectory to which each other target human feature belongs;

[0091] Step S10223: For each other target human feature, if the feature similarity corresponding to the other target human feature is greater than the similarity threshold and the start time beg ij and the end time end ij satisfy beg ij <end i +tw and end ij <beg i -tw, then use the other target human feature as the similar human feature of the current target human feature, where tw is a preset time window.

[0092] In this embodiment, the feature similarity corresponding to the target human feature can be the vector inner product between the two human feature vectors corresponding to the current target human feature and the other target human feature. Those skilled in the art can set the value of the similarity threshold according to the actual application needs.

[0093] In one embodiment, for each other target human feature, when the other target human feature meets the following conditions, the other target human feature can be used as the similar human feature of the current target human feature:

[0094] (1) The feature similarity corresponding to the other target human feature is greater than the similarity threshold;

[0095] (2) The maximum time interval between the acquisition time of the pedestrian movement trajectory corresponding to the current target human feature and the acquisition time of the pedestrian movement trajectory corresponding to the other human feature is less than a preset first trajectory time interval threshold.

[0096] Those skilled in the art can set the value of the first trajectory time interval threshold according to the actual application needs.

[0097] In one embodiment, for each other target human feature, when the other target human feature meets the following conditions, the other target human feature can be used as the similar human feature of the current target human feature:

[0098] (1) The feature similarity corresponding to the other target human feature is greater than the similarity threshold;

[0099] (2) The median difference between the median of the acquisition time of the pedestrian movement trajectory corresponding to the current target human feature and the median of the acquisition time of the pedestrian movement trajectory corresponding to the other human feature is less than a preset second trajectory time interval threshold.

[0100] Those skilled in the art can set the value of the second trajectory time interval threshold according to the actual application needs.

[0101] Step S1023: For each pair of matching samples, obtain the two human feature samples x in the current pair of matching samples i and x j respectively corresponding device numbers c of the image acquisition devices i and c j , and establish an image acquisition device pair (c i and c j ) according to the device numbers c i and c j ; obtain the acquisition times ts i and ts j of the pedestrian images corresponding to the two human feature samples x i and x j in the current pair of matching samples (x i and ts j ), determine the time interval |ts i -ts j | between the two acquisition times ts i and ts j , and add the time interval |ts i -ts j | to the time interval list of the current image acquisition device pair (c i and c j ).

[0102] In this embodiment, the device numbers c i and c j of the image acquisition devices corresponding to the pair of matching samples (x i and c j ) can be used to establish an image acquisition device pair (c i and c j ), and the time interval of the acquisition times of the pedestrian images corresponding to x i and x j is added to the time interval list of the image acquisition device pair (c i and c j ).

[0103] Step S1024: For each image acquisition device pair, perform binning processing on the time interval list of the current image acquisition device pair (c i and c j ) according to the preset number of bins, and establish a histogram of the current image acquisition device pair (c i and c j ) based on the number of time intervals falling into each bin after the binning processing. Among them, the abscissa of the histogram represents the movement of the pedestrian from the image acquisition range of the image acquisition device with the device number c i to the image acquisition device with the device number c jThe time taken for the image acquisition range of the image acquisition device, and the ordinate represents the probability of the time taken.

[0104] In this embodiment, step S1024 may further include the following steps S10241 to S10243:

[0105] Step S10241: Perform Gaussian smoothing on the number of time intervals falling into each bin after binning processing to obtain the number of time intervals in each bin after Gaussian smoothing;

[0106] Step S10242: Normalize the number of time intervals in each bin after Gaussian smoothing to obtain the normalization result corresponding to each bin;

[0107] Step S10243: Use the value corresponding to each bin as the abscissa and the normalization result corresponding to each bin as the ordinate to establish a histogram of the current image acquisition device pair (c i , c j ).

[0108] In this embodiment, the purpose of Gaussian smoothing is to avoid excessive discreteness of the number of time intervals in each bin. The purpose of normalization is to make the sum of the normalization results corresponding to all bins equal to 1. The maximum-minimum normalization method can be used to normalize the number of time intervals in each bin after Gaussian smoothing. Those skilled in the art can set the value of the preset number of bins according to the actual application needs.

[0109] In one embodiment, the process of normalizing the number of time intervals in each bin after Gaussian smoothing can use the zero-mean normalization (Z-score standardization) method or the probability normalization method.

[0110] Step S1025: Determine the spatial relationship feature and temporal relationship feature between the two image acquisition devices in each image acquisition device pair according to the histogram of each image acquisition device pair.

[0111] In this embodiment, step S1025 may further include the following steps S10251 to S10252:

[0112] Step S10251: Determine the temporal relationship feature between the two image acquisition devices in the image acquisition device pair according to the ordinate of the histogram of the image acquisition device pair.

[0113] Step S10252: Determine the spatial relationship feature between the two image acquisition devices according to the following formula (1):

[0114]

[0115] Among them, k ij refers to the spatial relationship feature between the two human feature samples x i and x j in the matching sample pair, corresponding to the image acquisition devices c i and c j respectively, and n ij refers to the total number of time intervals in the time interval list of the image acquisition device pair (c i , c j ), and refers to the sum of the total number of time intervals of all image acquisition device pairs.

[0116] In an implementation manner of the embodiment of the present invention, step S104 may further include the following steps S1041 to step S1044:

[0117] Step S1041: Determine the feature similarity between different pedestrian movement trajectories according to the human feature information.

[0118] In an implementation manner, a human feature vector may be confirmed according to the human feature information, and the inner product between the human feature vectors corresponding to different pedestrian trajectories may be obtained to determine the feature similarity between different pedestrian trajectories.

[0119] In an implementation manner, the feature similarity e between two pedestrian movement trajectories may be calculated according to the following formula (2) ij :

[0120] e ij = e i T ·e j (2)

[0121] Among them, e i represents the central vector of a pedestrian movement trajectory, e j represents the central vector of another pedestrian movement trajectory, T represents the vector transpose of the central vector e i ;

[0122] x i represents the feature vector determined by the i-th human feature information, c(x i ) represents the pedestrian movement trajectory to which the i-th human feature information belongs, "{x|c(x) = c(x i )}" represents the set of feature vectors of all human feature information corresponding to the pedestrian movement trajectory c(x i ), and norm represents the normalization operation;

[0123] x jdenotes the feature vector determined by the j-th human body feature information, and c(x j ) denotes the pedestrian motion trajectory to which the j-th human body feature information belongs. "{x|c(x) = c(x j )}" denotes the set of feature vectors of all human body feature information corresponding to the pedestrian motion trajectory c(x j ).

[0124] Step S1042: Calculate the association score between different pedestrian motion trajectories according to the feature similarity and the relationship feature.

[0125] In this embodiment, the association score between different pedestrian trajectories can be calculated according to the following formula (3):

[0126] z ij = e ij -decay + decay × s ij (3)

[0127] where z ij denotes the association score between two pedestrian motion trajectories, e ij denotes the feature similarity between two pedestrian motion trajectories, and s ij denotes the relationship feature between the image acquisition devices corresponding to two pedestrian motion trajectories respectively.

[0128] In one embodiment, s ij can represent the product of the spatial relationship feature and the temporal relationship feature between the image acquisition devices corresponding to two pedestrian motion trajectories respectively. The time interval between the moments when the human body feature samples corresponding to two pedestrian motion trajectories are collected can be queried, and the vertical coordinate corresponding to the bin where the time interval is located can be obtained from the histogram established according to the above steps S1021 to S1024. The vertical coordinate is used as the temporal relationship feature between the image acquisition devices corresponding to two pedestrian motion trajectories respectively, and the spatial relationship feature between the image acquisition devices corresponding to two pedestrian motion trajectories is calculated according to formula (1). The temporal relationship feature and the spatial relationship feature are multiplied to obtain s ij . decay represents a preset decay factor.

[0129] Step S1043: If the association score between two pedestrian motion trajectories is greater than a preset score threshold, it is determined that the two pedestrian motion trajectories belong to the pedestrian motion trajectories of the same pedestrian.

[0130] Step S1044: Associate the pedestrian motion trajectories belonging to the same pedestrian to generate the pedestrian motion trajectories corresponding to each pedestrian finally.

[0131] Those skilled in the art can set the value of the preset score threshold according to the specific situation of the actual application.

[0132] For the pedestrian movement trajectories belonging to the same pedestrian, the pedestrian movement trajectories can be sorted according to the acquisition time of the pedestrian movement trajectories, and the final pedestrian movement trajectories corresponding to each pedestrian can be generated based on the sorted pedestrian movement trajectories.

[0133] In an implementation manner of the embodiment of the present invention, the human body feature information includes the human body features extracted from the pedestrian images collected by the image acquisition device, the acquisition moment of the pedestrian images, the device number of the image acquisition device, and the trajectory number of the pedestrian movement trajectory to which the human body features belong. Step S104 may further include the following steps S1045 to step S1047:

[0134] Step S1045: For each piece of human body feature information, obtain similar human body feature information that is similar to the human body features of the current human body feature information from all the human body feature information, and form a matching feature pair by combining the current human body feature information with each piece of similar human body feature information respectively.

[0135] In this implementation manner, the method described in the above steps S10221 to S10223 can be used to obtain similar human body feature information that is similar to the human body features of the current human body feature information, where the target human body features in steps S10221 to S10223 are the human body feature information.

[0136] Step S1046: For each matching feature pair, determine whether the trajectory numbers corresponding to the two pieces of human body feature information in the current matching feature pair are the same.

[0137] In this implementation manner, step S1046 may further include the following steps S10461 and step S10462:

[0138] Step S10461: If they are the same, determine that the pedestrian movement trajectories to which the two pieces of human body feature information in the current matching feature pair belong are the pedestrian movement trajectories of the same pedestrian;

[0139] Step S10462: If they are different, obtain the histogram of the pair of image acquisition devices corresponding to the device numbers of the two pieces of human body feature information in the current matching feature pair, and obtain the spatial relationship feature between the two image acquisition devices corresponding to the current matching feature pair according to the histogram;

[0140] Obtain the time interval determined by the acquisition moments corresponding to the two pieces of human body feature information in the current matching feature pair, query the histogram with the time interval as the abscissa to obtain the corresponding ordinate, and use the ordinate as the time relationship feature;

[0141] Based on the human body feature information, spatial relationship features, and temporal relationship features, determine whether the pedestrian movement trajectories to which the two human body feature information in the current matching feature pair belong are the pedestrian movement trajectories of the same pedestrian; if so, modify the trajectory numbers corresponding to the two human body feature information to the same trajectory number.

[0142] Step S1047: Arrange the human body feature information with the same trajectory number in the order of the acquisition time corresponding to the human body feature information from the earliest to the latest, and generate the final pedestrian movement trajectory corresponding to each pedestrian according to the arranged human body feature information.

[0143] In one embodiment, refer to the appendix Figure 2 , Figure 2 It is a schematic diagram of the main steps of the pedestrian movement trajectory determination method according to an embodiment of the present invention. The pedestrian movement trajectory determination method may include the following steps S201 to step S206:

[0144] Step S201: Form a matching sample pair.

[0145] In this embodiment, step S201 is similar to the methods described in the foregoing steps S1021 and step S1022. For simplicity of description, it will not be elaborated here.

[0146] Step S202: Establish a histogram of the image acquisition device pair.

[0147] In this embodiment, step S202 is similar to the methods described in the foregoing steps S1023 and step S1024. For simplicity of description, it will not be elaborated here.

[0148] Step S203: Obtain the pedestrian movement trajectory to be processed.

[0149] In this embodiment, the pedestrian movement trajectory that needs to be confirmed can be obtained.

[0150] Step S204: Obtain the human body feature information similar to the human body feature information corresponding to the pedestrian trajectory to be processed, and form a matching feature pair.

[0151] In this embodiment, step S204 is similar to the method described in the foregoing step S1045. For simplicity of description, it will not be elaborated here.

[0152] Step S205: Generate the final pedestrian movement trajectory according to the matching feature pair and the histogram.

[0153] In this embodiment, step S205 is similar to the methods described in the foregoing steps S1046 and step S1047. For simplicity of description, it will not be elaborated here.

[0154] It should be noted that although the above embodiments describe the various steps in a specific order, those skilled in the art can understand that for the purpose of achieving the effects of the present invention, it is not necessary to execute the different steps in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the protection scope of the present invention.

[0155] Furthermore, the present invention also provides a pedestrian motion trajectory determination system.

[0156] Refer to the appendix Figure 3 , Figure 3 is the main structural block diagram of a pedestrian motion trajectory determination system according to an embodiment of the present invention. As Figure 3 shown, the pedestrian motion trajectory determination system in the embodiment of the present invention may include a pedestrian motion trajectory acquisition module, a relationship feature acquisition module, a human feature information acquisition module, and a final pedestrian motion trajectory generation module. In this embodiment, the pedestrian motion trajectory acquisition module may be configured to acquire the pedestrian motion trajectories within the image acquisition ranges of different image acquisition devices. The relationship feature acquisition module may be configured to acquire the relationship features between different image acquisition devices. The human feature information acquisition module may be configured to acquire the human feature information corresponding to each pedestrian motion trajectory. The final pedestrian motion trajectory generation module may be configured to associate the pedestrian motion trajectories belonging to the same pedestrian according to the human feature information and the relationship features, and generate the final pedestrian motion trajectories corresponding to each pedestrian.

[0157] In one implementation manner, refer to the appendix Figure 4 , Figure 4 is the main structural block diagram of a pedestrian motion trajectory determination system according to an implementation manner of an embodiment of the present invention. As Figure 4 shown, the pedestrian trajectory determination system may include a spatio-temporal relationship learning module and a trajectory association module. The spatio-temporal relationship learning module may include a matching sample pair sub-module and a histogram statistics sub-module. The trajectory association module may include a matching feature pair sub-module and a final pedestrian trajectory generation sub-module. In this implementation manner, the matching sample pair sub-module may be configured to acquire the human feature samples obtained from the pedestrian images acquired by each image acquisition device. For each human feature sample, obtain the similar human feature samples with similar human features to the current human feature sample from all human feature samples, and form a matching sample pair by combining the current human feature sample with each similar human feature sample respectively. The histogram statistics sub-module may be configured to, for each matching sample pair, obtain the device numbers c i and x j of the image acquisition devices corresponding to the two human feature samples x i and c j in the current matching sample pair respectively, and according to the device number ci and c j Establish an image acquisition device pair (c i , c j ); Obtain the current matching sample pair (c i , c j ) and the two human feature samples x i and x j in the corresponding pedestrian images of each, and the acquisition times ts i of ts j , determine the time interval |ts i -ts j | between the two acquisition times ts i -ts j |, add the time interval |ts i -ts j | to the time interval list of the current image acquisition device pair (c i , c j ). For each image acquisition device pair, perform binning processing on the time interval list of the current image acquisition device pair (c i , c j ), and establish the current image acquisition device pair (c i , c j) histogram. The matching feature pair sub-module can be configured to obtain, for each piece of human feature information, similar human feature information that is similar to the human feature of the current human feature information from all human feature information, and form a matching feature pair by combining the current human feature information with each piece of similar human feature information respectively. The final pedestrian trajectory generation sub-module can be configured to, for each matching feature pair, determine whether the trajectory numbers corresponding to the two pieces of human feature information in the current matching feature pair are the same; if they are the same, it is determined that the pedestrian movement trajectories to which the two pieces of human feature information in the current matching feature pair belong are the pedestrian movement trajectories of the same pedestrian; if they are different, obtain the histogram of the image acquisition device pair corresponding to the device numbers corresponding to the two pieces of human feature information in the current matching feature pair, and obtain the spatial relationship feature between the two image acquisition devices corresponding to the current matching feature pair according to the histogram; obtain the time interval determined by the acquisition times corresponding to the two pieces of human feature information in the current matching feature pair, query the histogram with the time interval as the abscissa to obtain the corresponding ordinate, and use the ordinate as the time relationship feature; according to the human feature information, spatial relationship feature and time relationship feature, determine whether the pedestrian movement trajectories to which the two pieces of human feature information in the current matching feature pair belong are the pedestrian movement trajectories of the same pedestrian; if so, modify the trajectory numbers corresponding to the two pieces of human feature information to the same trajectory number; arrange the human feature information with the same trajectory number in the order of the acquisition time corresponding to the human feature information from the earliest to the latest, and generate the final pedestrian movement trajectory according to the arranged human feature information.

[0158] The above pedestrian movement trajectory determination system is used to execute Figure 1 and Figure 2 the embodiments of the pedestrian movement trajectory determination method shown. The technical principles, the technical problems solved and the technical effects produced by the two are similar. Those skilled in the art of this technology can clearly understand that, for the convenience and conciseness of description, the specific working process and related descriptions of the pedestrian movement trajectory determination system can refer to the content described in the embodiments of the pedestrian movement trajectory determination method, and will not be repeated here.

[0159] Those skilled in the art can understand that all or part of the processes in the methods of the above-mentioned embodiments of the present invention can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device, medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium that can carry the computer program code. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0160] Furthermore, the present invention also provides a control device. In an embodiment of the control device according to the present invention, the control device includes a processor and a storage device. The storage device can be configured to store a program for executing the pedestrian movement trajectory determination method of the above-mentioned method embodiment. The processor can be configured to execute the program in the storage device, and the program includes, but is not limited to, the program for executing the pedestrian movement trajectory determination method of the above-mentioned method embodiment. For the sake of convenience of description, only the parts related to the embodiments of the present invention are shown. For the specific technical details not disclosed, please refer to the method part of the embodiments of the present invention. The control device can be a control device formed by various electronic devices.

[0161] Furthermore, the present invention also provides a computer-readable storage medium. In an embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program for executing the pedestrian movement trajectory determination method of the above-mentioned method embodiment. The program can be loaded and run by a processor to implement the above-mentioned pedestrian movement trajectory determination method. For the sake of convenience of description, only the parts related to the embodiments of the present invention are shown. For the specific technical details not disclosed, please refer to the method part of the embodiments of the present invention. The computer-readable storage medium can be a storage device formed by various electronic devices. Optionally, the computer-readable storage medium in the embodiments of the present invention is a non-transitory computer-readable storage medium.

[0162] Furthermore, it should be understood that since the setting of each module is only to illustrate the functional units of the device of the present invention, the corresponding physical devices of these modules can be the processor itself, or a part of the software in the processor, a part of the hardware, or a part of the combination of software and hardware. Therefore, the number of each module in the figure is only illustrative.

[0163] Those skilled in the art can understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principle of the present invention. Therefore, the technical solutions after splitting or combining will all fall within the protection scope of the present invention.

[0164] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims

1. A method for determining a pedestrian movement trajectory, characterized in that, the method includes: Obtaining the pedestrian movement trajectories within the image acquisition ranges of different image acquisition devices; Obtaining the relationship features between different image acquisition devices; the relationship features include the spatial relationship features and the temporal relationship features between different image acquisition devices. Among them, the spatial relationship features between two image acquisition devices are used to represent the association weight between the two image acquisition devices in the topological structure formed by the mutual association of all image acquisition devices, and the temporal relationship features between two image acquisition devices are used to represent the probability that a pedestrian spends different durations moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device; Obtaining the human feature information corresponding to each pedestrian movement trajectory; According to the human feature information and the relationship features, associating the pedestrian movement trajectories belonging to the same pedestrian to generate the final pedestrian movement trajectories corresponding to each pedestrian; The step of "obtaining the relationship features between different image acquisition devices" specifically includes: Obtaining the human feature samples obtained from the pedestrian images collected by each image acquisition device, where the human feature samples include the human features extracted from the pedestrian images and the device numbers of the image acquisition devices that collected the pedestrian images; For each human feature sample, obtaining the similar human feature samples that are similar to the human features of the current human feature sample from all human feature samples, and forming a matching sample pair by combining the current human feature sample with each similar human feature sample; For each pair of matching samples, obtain two human feature samples x in the current pair of matching samples i and x j respectively corresponding device numbers c of the image acquisition devices i and c j . According to the device numbers c i and c j , establish a pair of image acquisition devices (c i , c j ); Obtain the acquisition times ts i and ts j of the pedestrian images respectively corresponding to the two human feature samples x i and x j in the current pair of matching samples (c i and ts j ). Determine the time interval |ts i -ts j | between the two acquisition times ts i and ts j ; Add the time interval |ts i -ts j | to the time interval list of the current pair of image acquisition devices (c i , c j ). For each pair of image acquisition devices, perform binning on the time interval list of the current pair of image acquisition devices (c i , c j ) according to the preset number of bins, and establish a histogram of the current pair of image acquisition devices (c i , c j ) based on the number of time intervals falling into each bin after binning. Wherein, the abscissa of the histogram represents the time taken for a pedestrian to move from the image acquisition range of the image acquisition device with device number c i to the image acquisition range of the image acquisition device with device number c j , and the ordinate represents the probability of taking the time; According to the histograms of each pair of image acquisition devices, determining the spatial relationship features and the temporal relationship features between the two image acquisition devices in each pair of image acquisition devices; The step of "according to the human feature information and the relationship features, associating the pedestrian movement trajectories belonging to the same pedestrian to generate the final pedestrian movement trajectories corresponding to each pedestrian" specifically includes: Determining the feature similarity between different pedestrian movement trajectories according to the human feature information; Calculating the association score between different pedestrian movement trajectories according to the feature similarity and the relationship features; If the association score between two pedestrian movement trajectories is greater than a preset score threshold, it is determined that the two pedestrian movement trajectories belong to the pedestrian movement trajectories of the same pedestrian; Associating the pedestrian movement trajectories belonging to the same pedestrian to generate the final pedestrian movement trajectories corresponding to each pedestrian.

2. The method for determining a pedestrian movement trajectory according to claim 1, characterized in that, the step of "calculating the association score between different pedestrian movement trajectories according to the feature similarity and the relationship features" specifically includes calculating the association score through the following formula: z ij = e ij -decay + decay × s ij Among them, the z ij represents the association score between the motion trajectories of two pedestrians, the e ij represents the feature similarity between the motion trajectories of two pedestrians, the s ij represents the relationship feature between the image acquisition devices corresponding to the motion trajectories of two pedestrians respectively, and the decay represents a preset decay factor; When the relationship features include the spatial relationship features and the temporal relationship features between different image acquisition devices, the s ij represents the product of the spatial relationship features and the temporal relationship features between the image acquisition devices corresponding to the motion trajectories of two pedestrians respectively.

3. The method for determining a pedestrian movement trajectory according to claim 2, characterized in that, The method further includes calculating the feature similarity e between the movement trajectories of two pedestrians by the following formula ij :[[]]END]] e ij = e i T ·e j Among them, the e i represents the central vector of a pedestrian's movement trajectory, and the e j represents the central vector of another pedestrian's movement trajectory. The T represents the vector transpose of the central vector e i ; The said x i represents the feature vector determined by the i-th human body feature information, and the c(x i ) represents the pedestrian motion trajectory to which the i-th human body feature information belongs. "{x|c(x) = c(x i )}" represents the set of feature vectors of all human body feature information corresponding to the pedestrian motion trajectory c(x i ), and the norm represents the normalization operation; The said x j represents the feature vector determined by the j-th human body feature information, and the c(x j ) represents the pedestrian motion trajectory to which the j-th human body feature information belongs, "{x|c(x) = c(x j )}" represents the set of feature vectors of all human body feature information corresponding to the pedestrian motion trajectory c(x j ).

4. The method for determining a pedestrian movement trajectory according to claim 1, characterized in that, the step of "according to the histograms of each pair of image acquisition devices, determining the spatial relationship features and the temporal relationship features between the two image acquisition devices in each pair of image acquisition devices" specifically includes: Determine the time relationship feature between two image acquisition devices in the image acquisition device pair according to the ordinate of the histogram of the image acquisition device pair; Determine the spatial relationship feature between the two image acquisition devices according to the following formula: where k ij refers to the spatial relationship feature between the two human feature samples x i and x j in the matching sample pair, corresponding to the image acquisition devices c i and c j respectively, and n ij refers to the total number of time intervals in the time interval list of the image acquisition device pair (c i , c j ), and refers to the sum of the total number of time intervals of all image acquisition device pairs.

5. The method for determining a pedestrian motion trajectory according to claim 1, characterized in that Establishing a histogram of the current image acquisition device pair (c i , c j ) according to the number of time intervals falling into each bin after binning processing specifically includes: Perform Gaussian smoothing on the number of time intervals falling into each bin after binning processing to obtain the number of time intervals in each bin after Gaussian smoothing; Normalize the number of time intervals in each bin after Gaussian smoothing to obtain the normalization result corresponding to each bin; Taking the value corresponding to each bin as the abscissa and the normalization result corresponding to each bin as the ordinate, a histogram of the current image acquisition device pair (c i , c j ) is established.

6. The method for determining a pedestrian motion trajectory according to claim 1 or 4 or 5, characterized in that The human body feature information includes human body features extracted from pedestrian images collected by an image acquisition device, the acquisition time of the pedestrian image, the device number of the image acquisition device, and the trajectory number of the pedestrian motion trajectory to which the human body feature belongs; the step of "associating the pedestrian motion trajectories belonging to the same pedestrian according to the human body feature information and the relationship feature to generate the final pedestrian motion trajectory corresponding to each pedestrian" further includes: Step S1: For each human body feature information, obtain similar human body feature information that is similar to the human body feature of the current human body feature information from all human body feature information, and form a matching feature pair by combining the current human body feature information with each similar human body feature information respectively; Step S2: For each matching feature pair, determine whether the trajectory numbers corresponding to the two human body feature information in the current matching feature pair are the same; Step S21: If they are the same, determine that the pedestrian motion trajectories to which the two human body feature information in the current matching feature pair belong are the pedestrian motion trajectories of the same pedestrian; Step S22: If they are different, obtain the histogram of the image acquisition device pair corresponding to the device numbers of the two image acquisition devices corresponding to the two human body feature information in the current matching feature pair, and obtain the spatial relationship feature between the two image acquisition devices corresponding to the current matching feature pair according to the histogram; Obtain the time interval determined by the acquisition times corresponding to the two human body feature information in the current matching feature pair, query the histogram with the time interval as the abscissa to obtain the corresponding ordinate, and use the ordinate as the time relationship feature; According to the human body feature information, the spatial relationship feature and the time relationship feature, determine whether the pedestrian motion trajectories to which the two human body feature information in the current matching feature pair belong are the pedestrian motion trajectories of the same pedestrian; if so, modify the trajectory numbers corresponding to the two human body feature information to the same trajectory number; Step S3: Arrange the human body feature information with the same trajectory number in the order of the acquisition time corresponding to the human body feature information from the earliest to the latest, and generate the final pedestrian motion trajectory corresponding to each pedestrian according to the arranged human body feature information.

7. The method for determining a pedestrian motion trajectory according to claim 6, characterized in that The method further includes obtaining similar human body features to the current target human body feature from all target human body features through the following steps, where the target human body feature is the human body feature sample or the human body feature information: Obtain the feature similarity between the current target human body feature and other target human body features respectively; Obtain the start time beg of the pedestrian movement trajectory to which the current target human feature belongs i and the end time end i , obtain the start time beg of the pedestrian movement trajectory to which each of the other target human features belongs ij and the end time end ij ; For each of the other target human body features, if the feature similarity corresponding to the other target human body feature is greater than the similarity threshold and the start time beg ij and the end time end ij satisfy beg ij <end i +tw and end ij <beg i -tw, then the other target human body feature is used as a similar human body feature of the current target human body feature, where tw is a preset time window.

8. A pedestrian motion trajectory determination system Characterized in that The system includes: A pedestrian motion trajectory acquisition module configured to acquire the pedestrian motion trajectories within the image acquisition ranges of different image acquisition devices; A relationship feature acquisition module configured to acquire the relationship features between different image acquisition devices; the relationship features include the spatial relationship features and the temporal relationship features between different image acquisition devices, where the spatial relationship features between two image acquisition devices are used to represent the association weight between the two image acquisition devices in the topological structure formed by the mutual association of all image acquisition devices, and the temporal relationship features between two image acquisition devices are used to represent the probability that a pedestrian spends different durations moving from the image acquisition range of one image acquisition device to the image acquisition range of another image acquisition device; A human body feature information acquisition module configured to acquire the human body feature information corresponding to each pedestrian motion trajectory; A final pedestrian motion trajectory generation module configured to associate the pedestrian motion trajectories belonging to the same pedestrian according to the human body feature information and the relationship features, and generate the final pedestrian motion trajectories corresponding to each pedestrian; The relationship feature acquisition module is further configured to: obtain the human body feature samples obtained from the pedestrian images acquired by each image acquisition device, where the human body feature samples include the human body features extracted from the pedestrian images and the device numbers of the image acquisition devices that acquired the pedestrian images; For each human body feature sample, obtain similar human body feature samples similar to the human body features of the current human body feature from all human body feature samples, and form a matching sample pair by combining the current human body feature sample with each similar human body feature sample respectively; For each pair of matching samples, obtain two human feature samples x in the current pair of matching samples i and x j corresponding device numbers c of their respective image acquisition devices i and c j , and based on the device numbers c i and c j establish an image acquisition device pair (c i , c j ); obtain the acquisition times ts i and ts j of the two pedestrian images corresponding to the two human feature samples x i and x j in the current pair of matching samples (c i and ts j ), determine the time interval |ts i -ts j | between the two acquisition times ts i and ts j , and add the time interval |ts i -ts j | to the time interval list of the current image acquisition device pair (c i , c j ); For each pair of image acquisition devices, bin the time interval list of the current pair of image acquisition devices (c i , c j ) according to the preset number of bins, and establish a histogram of the current pair of image acquisition devices (c i , c j ) based on the number of time intervals falling into each bin after binning. The abscissa of the histogram represents the time taken for a pedestrian to move from the image acquisition range of the image acquisition device numbered c i to the image acquisition range of the image acquisition device numbered c j , and the ordinate represents the probability of taking the time; Determine the spatial relationship features and the temporal relationship features between the two image acquisition devices in each pair of image acquisition devices according to the histograms of each pair of image acquisition devices; The final pedestrian motion trajectory generation module is further configured to: determine the feature similarity between different pedestrian motion trajectories according to the human body feature information; Calculate the association scores between different pedestrian motion trajectories according to the feature similarity and the relationship features; If the association score between two pedestrian motion trajectories is greater than a preset score threshold, determine that the two pedestrian motion trajectories belong to the pedestrian motion trajectories of the same pedestrian; Associate the pedestrian motion trajectories belonging to the same pedestrian to generate the final pedestrian motion trajectories corresponding to each pedestrian.

9. A control device, including a processor and a storage device, the storage device being adapted to store multiple program codes Characterized in that The program codes are adapted to be loaded and run by the processor to execute the pedestrian motion trajectory determination method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing multiple program codes, characterized in that, the program codes are adapted to be loaded and run by a processor to execute the pedestrian motion trajectory determination method according to any one of claims 1 to 7.

Images