CN116069879A - A method, device, equipment and storage medium for predicting pedestrian trajectories - Google Patents
A method, device, equipment and storage medium for predicting pedestrian trajectories Download PDFInfo
- Publication number
- CN116069879A CN116069879A CN202211420061.4A CN202211420061A CN116069879A CN 116069879 A CN116069879 A CN 116069879A CN 202211420061 A CN202211420061 A CN 202211420061A CN 116069879 A CN116069879 A CN 116069879A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- current
- motion
- information
- trajectory
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000033001 locomotion Effects 0.000 claims abstract description 107
- 230000007246 mechanism Effects 0.000 claims abstract description 11
- 230000008569 process Effects 0.000 claims abstract description 11
- 238000004590 computer program Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000013459 approach Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000011273 social behavior Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明公开了一种预测行人轨迹的方法、装置、设备及存储介质，其方法包括：获取当前行人过去的历史轨迹信息，并通过对所述历史轨迹信息进行编码处理，得到当前行人的行人运动特征；利用社交注意力机制对所述行人运动特征进行处理，得到当前行人的权重信息，并利用所述权重信息和所述行人运动特征，得到当前行人的行人运动隐藏特征；利用所述当前行人的行人运动隐藏特征对当前行人的行人轨迹进行预测，得到当前行人的预测行人轨迹。
The invention discloses a method, device, equipment and storage medium for predicting pedestrian trajectory. The method includes: obtaining past historical trajectory information of the current pedestrian, and obtaining the current pedestrian movement of the pedestrian by encoding the historical trajectory information feature; use the social attention mechanism to process the pedestrian motion features to obtain the weight information of the current pedestrian, and use the weight information and the pedestrian motion features to obtain the pedestrian motion hidden features of the current pedestrian; use the current pedestrian The hidden features of the pedestrian movement are used to predict the trajectory of the current pedestrian, and the predicted trajectory of the current pedestrian is obtained.
Description
技术领域technical field
本发明涉及交通环境感知技术领域，特别涉及一种预测行人轨迹的方法、装置、设备及存储介质。The present invention relates to the technical field of traffic environment perception, in particular to a method, device, equipment and storage medium for predicting pedestrian trajectories.
背景技术Background technique
目前国内外研究者对人与人，人与车的交互特征的提取较为单一，无法充分利用其中的交互信息。人们开始加强对无人驾驶系统安全性的研究，预测行人的轨迹路线可让无人驾驶决策系统提前做出下一步的规划，避免交通事故的发生。但是行人的轨迹预测是很难的，首先行人的轨迹预测会受到很多因素的影响，例如和周边其他行人的运动轨迹，周边障碍物的分布，地面的情况等，会导致行人轨迹多种多样。其次是行人之间的运动轨迹是相互影响的，例如为了躲避周边其他人，行人会下意识的改变自己之前的运动轨迹去保护自己，防止被碰撞等。其中交通环境感知是无人驾驶技术中极为重要的一部分，而行人轨迹预测则是一个亟待解决和优化的难点。传统方法只能简单地进行线性序列预测，容易忽视人的运动信息特征，从而导致准确性差。随着神经网络在行人轨迹预测方面的使用，极大地增强了行人运动特征和意图收集与分析，虽然增加了预测的准确性，反而造成网络的负担过重、特征信息过多过杂，导致准确性达不到预期。At present, researchers at home and abroad only extract the interaction features of people and people, and people and vehicles, and cannot make full use of the interaction information. People have begun to strengthen research on the safety of unmanned driving systems. Predicting the trajectory of pedestrians can allow the unmanned driving decision-making system to make the next step in advance to avoid traffic accidents. But the trajectory prediction of pedestrians is very difficult. First, the trajectory prediction of pedestrians will be affected by many factors, such as the trajectory of other pedestrians around, the distribution of surrounding obstacles, the ground conditions, etc., which will lead to various pedestrian trajectories. Secondly, the movement trajectories of pedestrians are mutually influenced. For example, in order to avoid other people around, pedestrians will subconsciously change their previous movement trajectories to protect themselves and prevent collisions. Among them, traffic environment perception is an extremely important part of unmanned driving technology, and pedestrian trajectory prediction is a difficulty that needs to be solved and optimized urgently. Traditional methods can only simply predict linear sequences, which tend to ignore the characteristics of human motion information, resulting in poor accuracy. With the use of neural networks in pedestrian trajectory prediction, the collection and analysis of pedestrian motion characteristics and intentions has been greatly enhanced. Sex does not meet expectations.
发明内容Contents of the invention
根据本发明实施例提供的方案解决的技术问题是车辆在静止状态下人流湍急的行人过马路的交通场景预测行人轨迹的网络收集特征信息过杂导致准确性低的问题。The technical problem solved by the solution provided according to the embodiment of the present invention is the problem of low accuracy due to too much feature information collected by the network for predicting pedestrian trajectories in a traffic scene where vehicles are in a stationary state and pedestrians are crossing the road in a rush.
根据本发明实施例提供的一种预测行人轨迹的方法，包括：A method for predicting pedestrian trajectories according to an embodiment of the present invention includes:
获取当前行人过去的历史轨迹信息，并通过对所述历史轨迹信息进行编码处理，得到当前行人的行人运动特征；Obtain past historical trajectory information of the current pedestrian, and obtain the pedestrian motion characteristics of the current pedestrian by encoding the historical trajectory information;
利用社交注意力机制对所述行人运动特征进行处理，得到当前行人的权重信息，并利用所述权重信息和所述行人运动特征，得到当前行人的行人运动隐藏特征；Using a social attention mechanism to process the pedestrian motion features to obtain the weight information of the current pedestrian, and using the weight information and the pedestrian motion features to obtain the pedestrian motion hidden features of the current pedestrian;
利用所述当前行人的行人运动隐藏特征对当前行人的行人轨迹进行预测，得到当前行人的预测行人轨迹。The pedestrian trajectory of the current pedestrian is predicted by using the pedestrian motion hidden feature of the current pedestrian to obtain the predicted pedestrian trajectory of the current pedestrian.
根据本发明实施例提供的一种预测行人轨迹的装置，包括：A device for predicting pedestrian trajectories according to an embodiment of the present invention includes:
处理模块，用于获取当前行人过去的历史轨迹信息，并通过对所述历史轨迹信息进行编码处理，得到当前行人的行人运动特征；利用社交注意力机制对所述行人运动特征进行处理，得到当前行人的权重信息，并利用所述权重信息和所述行人运动特征，得到当前行人的行人运动隐藏特征；The processing module is used to obtain the past historical trajectory information of the current pedestrian, and obtain the pedestrian movement characteristics of the current pedestrian by encoding the historical trajectory information; use the social attention mechanism to process the pedestrian movement characteristics to obtain the current pedestrian movement characteristics. The weight information of the pedestrian, and using the weight information and the pedestrian motion feature to obtain the pedestrian motion hidden feature of the current pedestrian;
预测模块，用于利用所述当前行人的行人运动隐藏特征对当前行人的行人轨迹进行预测，得到当前行人的预测行人轨迹。The prediction module is used to predict the trajectory of the current pedestrian by using the hidden feature of the pedestrian movement of the current pedestrian to obtain the predicted trajectory of the current pedestrian.
本申请实施例提供的一种电子设备，包括：存储器；处理器；以及计算机程序；其中，所述计算机程序存储在所述存储器中，并被配置为由所述处理器执行以实现一种预测行人轨迹的方法。An electronic device provided in an embodiment of the present application includes: a memory; a processor; and a computer program; wherein the computer program is stored in the memory and is configured to be executed by the processor to realize a prediction approach to pedestrian trajectories.
本申请实施例提供的一种计算机可读存储介质，其上存储有计算机程序；所述计算机程序被处理器执行以实现一种预测行人轨迹的方法。A computer-readable storage medium provided by an embodiment of the present application stores a computer program thereon; the computer program is executed by a processor to implement a method for predicting pedestrian trajectories.
根据本发明实施例提供的方案，提升了行人运动隐藏特征提取准确率，并在符合社交规则的前提下提供了多条良好的预测轨迹可能，结果更具有多样性和可选择性。According to the solution provided by the embodiment of the present invention, the accuracy rate of pedestrian motion hidden feature extraction is improved, and multiple good prediction trajectory possibilities are provided under the premise of conforming to social rules, and the results are more diverse and selectable.
附图说明Description of drawings
此处所说明的附图用来提供对本发明的进一步理解，构成本发明的一部分，本发明的示意性实施例及其说明用于理解本发明，并不构成对本发明的不当限定。在附图中：The accompanying drawings described here are used to provide a further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to understand the present invention and do not constitute improper limitations to the present invention. In the attached picture:
图1是本发明实施例提供的一种预测行人轨迹的方法流程图；Fig. 1 is a flow chart of a method for predicting pedestrian trajectories provided by an embodiment of the present invention;
图2是本发明实施例提供的一种预测行人轨迹的装置示意图；Fig. 2 is a schematic diagram of a device for predicting pedestrian trajectories provided by an embodiment of the present invention;
图3是本发明实施例提供的基于社交行为的行人运动隐藏特征提取方法的流程图。Fig. 3 is a flowchart of a method for extracting hidden features of pedestrian movement based on social behavior provided by an embodiment of the present invention.
具体实施方式Detailed ways
以下结合附图对本发明的优选实施例进行详细说明，应当理解，以下所说明的优选实施例仅用于说明和解释本发明，并不用于限定本发明。The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described below are only used to illustrate and explain the present invention, and are not intended to limit the present invention.
本发明实施例适用于车辆在静止状态下人流湍急的行人过马路的交通场景预测行人轨迹的网络收集特征信息过杂环境下。The embodiment of the present invention is applicable to a traffic scene in which a vehicle is in a stationary state and pedestrians cross the road in a turbulent flow, and the feature information collected by the network for predicting pedestrian trajectories is too complex.
图1是本发明实施例提供的一种预测行人轨迹的方法流程图，如图1所示，包括：Fig. 1 is a flow chart of a method for predicting pedestrian trajectories provided by an embodiment of the present invention, as shown in Fig. 1 , including:
步骤S101：获取当前行人过去的历史轨迹信息，并通过对所述历史轨迹信息进行编码处理，得到当前行人的行人运动特征；Step S101: Obtain the past historical trajectory information of the current pedestrian, and obtain the pedestrian motion characteristics of the current pedestrian by encoding the historical trajectory information;
步骤S102：利用社交注意力机制对所述行人运动特征进行处理，得到当前行人的权重信息，并利用所述权重信息和所述行人运动特征，得到当前行人的行人运动隐藏特征；Step S102: Using the social attention mechanism to process the pedestrian motion features to obtain the weight information of the current pedestrian, and using the weight information and the pedestrian motion features to obtain the pedestrian motion hidden features of the current pedestrian;
步骤S103：利用所述当前行人的行人运动隐藏特征对当前行人的行人轨迹进行预测，得到当前行人的预测行人轨迹。Step S103: Predict the trajectory of the current pedestrian by using the hidden features of the pedestrian movement of the current pedestrian to obtain the predicted trajectory of the current pedestrian.
具体地说，所述通过对所述历史轨迹信息进行编码处理，得到当前行人的行人运动特征包括：通过全连接网络将所述当前行人过去的历史轨迹信息从坐标空间转化到特征空间；通过将所述特征空间和当前行人上一时刻的行人运动特征进行编码处理，得到当前行人的行人运动特征。Specifically, the process of encoding the historical trajectory information to obtain the pedestrian motion characteristics of the current pedestrian includes: converting the past historical trajectory information of the current pedestrian from the coordinate space to the feature space through a fully connected network; The feature space and the pedestrian motion features of the current pedestrian at the previous moment are encoded to obtain the pedestrian motion features of the current pedestrian.
其中，所述通过将所述特征空间和当前行人上一时刻的行人运动特征进行编码处理，得到当前行人的行人运动特征包括：Wherein, by encoding the feature space and the pedestrian motion characteristics of the current pedestrian at the last moment, obtaining the pedestrian motion characteristics of the current pedestrian includes:
其中，是指行人运动特征；是指函数；是指特征空间；是指函数的权重参数，是编码器的权重参数。in, Refers to the pedestrian movement characteristics; refers to the function; is the feature space; refers to the function The weight parameter of is the weight parameter of the encoder.
进一步地，所述利用社交注意力机制对所述行人运动特征进行处理，得到当前行人的权重信息包括：根据当前行人的行人运动特征，计算当前行人与其每个相邻行人的相对运动信息；利用所述相对运动信息，计算当前行人与其每个相邻行人的注意力权重；利用所述相对运动信息和所述注意力权重，得到当前行人的权重信息。Further, the process of using the social attention mechanism to process the pedestrian motion characteristics to obtain the weight information of the current pedestrian includes: calculating the relative motion information of the current pedestrian and each adjacent pedestrian according to the pedestrian motion characteristics of the current pedestrian; The relative motion information is used to calculate the attention weight of the current pedestrian and each adjacent pedestrian; and the weight information of the current pedestrian is obtained by using the relative motion information and the attention weight.
其中，所述利用所述权重信息和所述行人运动特征，得到当前行人的行人运动隐藏特征包括：Wherein, said using said weight information and said pedestrian motion features to obtain the pedestrian motion hidden features of current pedestrians includes:
其中，是行人运动隐藏特征；是相邻行人上一时刻的运动状态信息，是上一时刻周围行人的对行人未来轨迹的影响，权重信息是权重信息；是噪声。in, is the pedestrian motion hidden feature; is the motion state information of adjacent pedestrians at the last moment, It was the pedestrians around at the last moment for pedestrians Influence of future trajectories, weight information is the weight information; is noise.
具体地说，所述利用所述当前行人的行人运动隐藏特征对当前行人的行人轨迹进行预测，得到当前行人的预测行人轨迹包括：获取当前行人的初始运动状态信息，并利用所述当前行人的行人运动隐藏特征对所述初始运动状态信息进行更新处理，得到更新后的运动状态信息；通过将所述更新后的当前运动状态转化到坐标空间,得到当前行人的预测行人轨迹。Specifically, using the pedestrian motion hidden features of the current pedestrian to predict the pedestrian trajectory of the current pedestrian, and obtaining the predicted pedestrian trajectory of the current pedestrian includes: obtaining the initial motion state information of the current pedestrian, and using the current pedestrian's The pedestrian motion hidden features update the initial motion state information to obtain updated motion state information; by transforming the updated current motion state into a coordinate space, the predicted pedestrian trajectory of the current pedestrian is obtained.
其中，所述通过将所述更新后的当前运动状态转化到坐标空间,得到当前行人的预测行人轨迹包括：Wherein, said obtaining the predicted pedestrian trajectory of the current pedestrian by converting the updated current motion state into the coordinate space includes:
其中，是预测行人轨迹：是更新后的当前运动状态，是指函数的权重参数。in, is the predicted pedestrian trajectory: is the updated current motion state, refers to the function weight parameter.
图2是本发明实施例提供的一种预测行人轨迹的装置示意图，如图2所示，包括：处理模块，用于获取当前行人过去的历史轨迹信息，并通过对所述历史轨迹信息进行编码处理，得到当前行人的行人运动特征；利用社交注意力机制对所述行人运动特征进行处理，得到当前行人的权重信息，并利用所述权重信息和所述行人运动特征，得到当前行人的行人运动隐藏特征；预测模块，用于利用所述当前行人的行人运动隐藏特征对当前行人的行人轨迹进行预测，得到当前行人的预测行人轨迹。Fig. 2 is a schematic diagram of a device for predicting pedestrian trajectories provided by an embodiment of the present invention. As shown in Fig. 2 , it includes: a processing module for obtaining past historical trajectory information of the current pedestrian, and encoding the historical trajectory information processing to obtain the pedestrian motion characteristics of the current pedestrian; use the social attention mechanism to process the pedestrian motion characteristics to obtain the weight information of the current pedestrian, and use the weight information and the pedestrian motion characteristics to obtain the pedestrian motion of the current pedestrian Hidden feature; prediction module, used to predict the pedestrian trajectory of the current pedestrian by using the hidden feature of the pedestrian movement of the current pedestrian, and obtain the predicted pedestrian trajectory of the current pedestrian.
本申请实施例提供的一种电子设备，包括：存储器；处理器；以及计算机程序；其中，所述计算机程序存储在所述存储器中，并被配置为由所述处理器执行以实现一种预测行人轨迹的方法。An electronic device provided in an embodiment of the present application includes: a memory; a processor; and a computer program; wherein the computer program is stored in the memory and is configured to be executed by the processor to realize a prediction approach to pedestrian trajectories.
本申请实施例提供的一种计算机可读存储介质，其上存储有计算机程序；所述计算机程序被处理器执行以实现一种预测行人轨迹的方法。A computer-readable storage medium provided by an embodiment of the present application stores a computer program thereon; the computer program is executed by a processor to implement a method for predicting pedestrian trajectories.
图3是本发明实施例提供的基于社交行为的行人运动隐藏特征提取方法的流程图，包括以下步骤：Fig. 3 is a flowchart of a method for extracting hidden features of pedestrian movement based on social behavior provided by an embodiment of the present invention, including the following steps:
S1.在编码器中通过长短时序网络对过去的轨迹进行编码，得到行人的运动特征。S1. Past trajectories through long and short temporal networks in the encoder Encoded to get the pedestrian sports characteristics .
通过全连接网络将当前行人的轨迹序列从坐标空间转化到特征空间得到；through a fully connected network The trajectory sequence of the current pedestrian Transform from coordinate space to feature space to get ;
是编码器中全连接网络的权重参数。 is the fully connected network in the encoder weight parameter.
通过函数将线性嵌入后与行人之前的状态输入到编码器LSTM模块进行编码,直到观测序列结束，将所有信息都完成编码,并对行人u的运动特征进行更新；through function Will The state of the pedestrian after the linear embedding and before Input to the encoder LSTM module for encoding, until the end of the observation sequence, all information is encoded, and the motion characteristics of pedestrian u to update;
式中：是函数的权重参数，是编码器的权重参数，通过预训练微调初始化。In the formula: is a function The weight parameter of is the weight parameter of the encoder, initialized by pre-training fine-tuning.
长短时序网络内部结构循环单元，整个网络可以建立较长距离的时序依赖关系。首先读取上一时刻的序列，进行编码，得到隐藏状态和简单的序列传递规律，经由第一个门：遗忘门，输入一个在0到1之间数值给每个在记忆单元，1表示完全保留，0表示完全舍弃。经由第二个门：输入门，将确定什么样的信息内存放在记忆单元中。经由第三个门：更新门，去除掉不用需要的特征留下重要的特征，更新记忆元状态，得到下一时刻的序列传递规律。最后一个门：输出门，在这里将最后内部筛选的结果传递给外部的隐藏状态，从而得到下一时刻的隐藏状态，进一步得到下一时刻的序列。The internal structure of the long-short sequence network is a recurrent unit, and the entire network can establish a long-distance sequence dependency. First read the sequence at the previous moment, encode it, and get the hidden state and simple sequence transmission law. Through the first gate: the forget gate, input a value between 0 and 1 for each memory unit, and 1 means complete Reserved, 0 means completely discarded. Via the second gate: the input gate, it will be determined what information is stored in the memory cell. Through the third gate: the update gate, the unnecessary features are removed to leave important features, and the state of the memory element is updated to obtain the sequence transfer law at the next moment. The last gate: the output gate, where the final internal screening result is passed to the external hidden state, so as to obtain the hidden state at the next moment, and further obtain the sequence at the next moment.
S2. 将步骤S1中得到的行人的运动特征,通过社交注意力机制对待测行人产生权重信息来评估其他行人对待测行人的影响。S2. The motion characteristics of pedestrians obtained in step S1 , through the social attention mechanism to test pedestrians Generate weight information To evaluate the impact of other pedestrians on the pedestrian under test.
S21、计算行人与他周围相邻行人的相对运动信息；S21. Calculate pedestrians pedestrians around him The relative motion information of ;
行人和他周围交互密切的行人之间相对位置信息由进行计算，再将通过全连接网络映射至，得到行人和他周围交互密切的行人之间相对运动信息，由(1)行人和之间的欧氏距离相组合，(2)行人与行人的方位角(即的速度向量和和的连接向量之间的夹角)，以及(3)最近接近的距离(如果两个物体都保持当前速度，那么它们将来所能达到的最小距离)三个部分组成，式中为该全连接层的权重参数，计算公式如下所示：pedestrian Pedestrians who interact closely with him The relative position information between to calculate, and then through a fully connected network map to , get the pedestrian Pedestrians who interact closely with him Relative motion information , by (1) pedestrian and Combined with the Euclidean distance between, (2) pedestrian with pedestrians azimuth (ie The velocity vector and and The angle between the connection vectors), and (3) the nearest approach distance (if the two objects maintain the current speed, then the minimum distance they can reach in the future) consists of three parts, where is the weight parameter of the fully connected layer, and the calculation formula is as follows:
式中，是过去轨迹的序列数，本发明是将前八序列轨迹作为过去的轨迹。In the formula, is the sequence number of past trajectories, and the present invention regards the first eight sequence trajectories as past trajectories.
S22、计算每个相邻行人的注意力权重。S22. Calculate the attention weight of each adjacent pedestrian.
行人和之间的是通过全连接层将嵌入中的，是相邻行人的运动特征信息。pedestrian and between is through the fully connected layer Will embed middle, is an adjacent pedestrian motion feature information.
式中，是行人a与他周围相邻行人的交互运动信息，N是行人的总数，是和应用于运动特征信息的线性映射权重的公共行列, 是全连接层的权重参数。In the formula, is the interactive motion information between pedestrian a and the adjacent pedestrians around him, N is the total number of pedestrians, yes and the public ranks and columns of the linear mapping weights applied to the motion feature information, is a fully connected layer weight parameter.
S23、将和通过标量积和softmax运算得到每个相邻行人的注意力权重，是所有行人运动特征信息，表示行人的数量，、。S23. Will and The attention weight of each adjacent pedestrian is obtained by scalar product and softmax operation , is the motion feature information of all pedestrians, represents the number of pedestrians, , .
, ,
S3.在解码器中根据S2得到的权重信息，结合行人的运动状态和相邻行人的运动状态，得到有用的行人运动隐藏特征。S3. The weight information obtained in the decoder according to S2 , combined with pedestrian state of motion and adjacent pedestrians state of motion , to get useful hidden features of pedestrian motion .
式中，是相邻行人上一时刻的运动状态信息，是上一时刻周围行人的对行人未来轨迹的影响，是噪声。In the formula, is an adjacent pedestrian The exercise status information at the last moment, It was the pedestrians around at the last moment for pedestrians impact on future trajectories, is noise.
S4.根据S3中得到的运动隐藏特征和行人的当前运动状态预测行人轨迹。S4. Based on the motion hidden features obtained in S3 and the current motion state of the pedestrian Predicting Pedestrian Trajectories .
解码器中长短时序网络接收到的行人的初始的当前运动状态信息为,是编码器的状态级联高级噪声得到的。Pedestrians received by the long-short time series network in the decoder The initial current motion state information of is , is the encoder status Cascade Advanced Noise owned.
随后更新，就需要将上一时刻的运动状态信息和上一时刻注意力机制模块筛选到的有用的行人运动隐藏特征相结合到长短时序网络中得到的。update later , it is necessary to transfer the motion state information of the previous moment and last moment attention mechanism module The filtered useful hidden features of pedestrian motion are combined into long and short time series networks.
式中，是长短时序网络的解码单元函数，是解码器中长短时序网络的权重。In the formula, is the decoding unit function of the long and short time series network, are the weights of the long and short temporal networks in the decoder.
然后通过函数将更新后的当前运动状态转化到坐标空间,得到预测的未来轨迹：then pass The function will update the current motion state after Convert to the coordinate space to get the predicted future trajectory :
式中是函数的权重。In the formula is a function the weight of.
尽管上文对本发明进行了详细说明，但是本发明不限于此，本技术领域技术人员可以根据本发明的原理进行各种修改。因此，凡按照本发明原理所作的修改，都应当理解为落入本发明的保护范围。Although the present invention has been described in detail above, the present invention is not limited thereto, and various modifications can be made by those skilled in the art based on the principle of the present invention. Therefore, any modifications made according to the principles of the present invention should be understood as falling within the protection scope of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211420061.4A CN116069879B (en) | 2022-11-14 | 2022-11-14 | Method, device, equipment and storage medium for predicting pedestrian trajectory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211420061.4A CN116069879B (en) | 2022-11-14 | 2022-11-14 | Method, device, equipment and storage medium for predicting pedestrian trajectory |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116069879A true CN116069879A (en) | 2023-05-05 |
| CN116069879B CN116069879B (en) | 2023-06-20 |
Family
ID=86173865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211420061.4A Active CN116069879B (en) | 2022-11-14 | 2022-11-14 | Method, device, equipment and storage medium for predicting pedestrian trajectory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116069879B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111044045A (en) * | 2019-12-09 | 2020-04-21 | 中国科学院深圳先进技术研究院 | Navigation method, device and terminal device based on neural network |
| CN111311009A (en) * | 2020-02-24 | 2020-06-19 | 广东工业大学 | A Pedestrian Trajectory Prediction Method Based on Long Short-Term Memory |
| CN111459168A (en) * | 2020-04-23 | 2020-07-28 | 上海交通大学 | A fusion method and system for predicting pedestrian trajectories of autonomous vehicles crossing the street |
| CN112766561A (en) * | 2021-01-15 | 2021-05-07 | 东南大学 | Generating type confrontation track prediction method based on attention mechanism |
| WO2022222034A1 (en) * | 2021-04-20 | 2022-10-27 | 深圳市锐明技术股份有限公司 | Pedestrian trajectory prediction method and apparatus, electronic device, and storage medium |
-
2022
- 2022-11-14 CN CN202211420061.4A patent/CN116069879B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111044045A (en) * | 2019-12-09 | 2020-04-21 | 中国科学院深圳先进技术研究院 | Navigation method, device and terminal device based on neural network |
| CN111311009A (en) * | 2020-02-24 | 2020-06-19 | 广东工业大学 | A Pedestrian Trajectory Prediction Method Based on Long Short-Term Memory |
| CN111459168A (en) * | 2020-04-23 | 2020-07-28 | 上海交通大学 | A fusion method and system for predicting pedestrian trajectories of autonomous vehicles crossing the street |
| CN112766561A (en) * | 2021-01-15 | 2021-05-07 | 东南大学 | Generating type confrontation track prediction method based on attention mechanism |
| WO2022222034A1 (en) * | 2021-04-20 | 2022-10-27 | 深圳市锐明技术股份有限公司 | Pedestrian trajectory prediction method and apparatus, electronic device, and storage medium |
Non-Patent Citations (3)
| Title |
|---|
| ALEXANDRE ALAHI等: "Social LSTM:Human Trajectory Prediction in Crowded Spaces", 2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), pages 961 - 971 * |
| 李文礼等: "基于视野域机制的行人轨迹预测", 计算机应用研究, vol. 40, no. 1, pages 80 - 85 * |
| 欧阳俊等: "基于GAN和注意力机制的行人轨迹预测", 激光与光电子学进展, vol. 57, no. 14, pages 161 - 172 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116069879B (en) | 2023-06-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111942407B (en) | Trajectory prediction method, apparatus, device and storage medium | |
| CN109697852B (en) | Urban road congestion degree prediction method based on time sequence traffic events | |
| Tang et al. | Sparseocc: Rethinking sparse latent representation for vision-based semantic occupancy prediction | |
| CN115240425A (en) | Traffic prediction method based on multi-scale space-time fusion graph network | |
| CN114692762B (en) | A vehicle trajectory prediction method based on graph attention interaction mechanism | |
| CN114372116A (en) | A vehicle trajectory prediction method based on LSTM and spatiotemporal attention mechanism | |
| CN108399749A (en) | A kind of traffic trip needing forecasting method in short-term | |
| CN118823308A (en) | A target detection and tracking network model and method based on multi-source heterogeneous sensor information fusion | |
| CN115829171A (en) | Pedestrian trajectory prediction method combining space information and social interaction characteristics | |
| CN115293237A (en) | A vehicle trajectory prediction method based on deep learning | |
| CN116307152A (en) | A Spatiotemporal Interactive Dynamic Graph Attention Network for Traffic Prediction | |
| Chen et al. | Vulnerable road user trajectory prediction for autonomous driving using a data-driven integrated approach | |
| CN117133129B (en) | Traffic speed prediction method based on multi-component attention-seeking neural network | |
| CN114372503A (en) | Cluster vehicle motion trail prediction method | |
| Yang et al. | Multi-granularity scenarios understanding network for trajectory prediction | |
| Chen et al. | CAE‐GAN: A hybrid model for vehicle trajectory prediction | |
| CN114119615A (en) | A Radar Segmentation Method Fusing Spatial Attention and Self-Attention Transform Network | |
| CN114493034A (en) | Space-time global semantic representation learning method for regional flow prediction | |
| CN116052427A (en) | Inter-city inter-regional mobility prediction method and device based on private car travel track data | |
| CN119283896A (en) | A trajectory planning method for autonomous driving vehicles | |
| Jiang et al. | A Survey of Deep Learning-Based Pedestrian Trajectory Prediction: Challenges and Solutions | |
| CN116069879B (en) | Method, device, equipment and storage medium for predicting pedestrian trajectory | |
| CN117351712B (en) | Zhou Che track prediction method and system based on Cro-IntentFormer and fused with vehicle driving intention | |
| Li et al. | Trajectory prediction of cyclist based on spatial‐temporal multi‐graph network in crowded scenarios | |
| CN118172923A (en) | Deep learning modeling method for the interaction between left-turning vehicles and illegal bicycles at intersections |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |