CN114387549A - A visual detection system and method for pedestrians at zebra crossing based on deep learning - Google Patents
A visual detection system and method for pedestrians at zebra crossing based on deep learningDownload PDFInfo
- Publication number
- CN114387549A CN114387549ACN202210027599.2ACN202210027599ACN114387549ACN 114387549 ACN114387549 ACN 114387549ACN 202210027599 ACN202210027599 ACN 202210027599ACN 114387549 ACN114387549 ACN 114387549A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- pedestrians
- pedestrian
- zebra crossing
- deep learning
- 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.)
- Pending
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
Description
Translated fromChinese技术领域technical field
本发明涉及视觉检测技术领域，特别是涉及一种基于深度学习的斑马线礼让行人视觉检测系统及方法。The invention relates to the technical field of visual detection, in particular to a system and method for visual detection of zebra crossing polite pedestrians based on deep learning.
背景技术Background technique
在当前混合交通环境条件下，行人在道路交通系统中由于缺乏相应的保护设备而成为道路交通参与者中的弱势群体。由于行人在出行过程中需要穿越道路，为帮助行人安全通行，在道路上设置了斑马线区域。斑马线区域是在车行道上用斑马线等标线标示的规定车辆需减速礼让行人过街的地方，目的是防止车辆快速行驶时伤及行人。因此，斑马线区域常常成为行人与车辆交互最频繁的区域。In the current mixed traffic environment, pedestrians have become a vulnerable group among road traffic participants due to the lack of corresponding protective equipment in the road traffic system. Since pedestrians need to cross the road during travel, zebra crossing areas are set up on the road to help pedestrians pass safely. The zebra crossing area is a place marked with a zebra crossing and other markings on the roadway where vehicles are required to slow down and give way to pedestrians to cross the street, in order to prevent pedestrians from being injured when vehicles are driving fast. Therefore, the zebra crossing area often becomes the area where pedestrians and vehicles interact most frequently.
目前，由于前端违法监控摄像头在技术层面存在检测的局限性，面对周围环境复杂、语义信息差等问题导致礼让行人的检测效果不理想，因此，研发出能适应复杂环境的斑马线礼让行人的检测方法是本领域技术人员亟待解决的问题。At present, due to the technical limitations of the front-end illegal surveillance cameras, the detection effect of polite pedestrians is not ideal due to the complex surrounding environment and poor semantic information. The method is an urgent problem to be solved by those skilled in the art.
发明内容SUMMARY OF THE INVENTION
本发明的目的是为了提供一种基于深度学习的斑马线礼让行人视觉检测系统及方法，通过深度学习技术，对在复杂交通场景下的机动车不礼让行人的行为进行检测和识别，具有高效、准确、便捷、智能化程度高的特点。The purpose of the present invention is to provide a visual detection system and method for zebra crossing pedestrians based on deep learning, which can detect and recognize the behavior of vehicles in a complex traffic scene, which is highly efficient and accurate. , convenience and high degree of intelligence.
为实现上述目的，本发明提供了如下方案：For achieving the above object, the present invention provides the following scheme:
第一方面，本发明提供一种基于深度学习的斑马线礼让行人视觉检测系统，包括：In a first aspect, the present invention provides a deep learning-based visual detection system for zebra crossing polite pedestrians, including:
参数配置模块，用于设置所述视觉检测系统的检测参数；a parameter configuration module for setting detection parameters of the visual detection system;
通信模块，用于将设置的所述视觉检测系统的检测参数发送给信息采集及处理模块；a communication module for sending the set detection parameters of the visual detection system to the information collection and processing module;
信息采集及处理模块，用于检测并识别复杂交通场景下的行人、机动车目标，捕捉目标的行为并对捕捉到的行为进行判别。The information collection and processing module is used to detect and identify pedestrians and motor vehicles in complex traffic scenes, capture the behavior of the target and discriminate the captured behavior.
可选的，所述视觉检测系统的检测参数包括：车辆检测区域、行人斑马线区域、当前车道的斜率以及接入相机的参数。Optionally, the detection parameters of the visual detection system include: a vehicle detection area, a pedestrian zebra crossing area, a slope of the current lane, and parameters for accessing a camera.
可选的，所述通信模块具体采用Http通信协议。Optionally, the communication module specifically adopts the Http communication protocol.
可选的，所述信息采集及处理模块包括：Optionally, the information collection and processing module includes:
行人车辆检测模块，用于实时检测当前视频帧中的人、车实例，通过NMS算法对满足IOU及中心点距离阈值要求的检测框进行二次筛选；The pedestrian and vehicle detection module is used to detect people and vehicles in the current video frame in real time, and the NMS algorithm is used to secondary screen the detection frames that meet the IOU and center point distance threshold requirements;
多目标追踪模块，用于对检测到的人车实例进行特征提取和匹配，同时通过L-K光流算法及CIOU算法追踪人车实例在连续多帧中的目标轨迹；The multi-target tracking module is used for feature extraction and matching of detected instances of people and vehicles, and at the same time, the target trajectories of instances of people and vehicles in consecutive frames are tracked through L-K optical flow algorithm and CIOU algorithm;
行为判定模块，用于对进入检测区域的车辆及行人进行移动侦测，判断车辆是否礼让行人；The behavior determination module is used to detect the movement of vehicles and pedestrians entering the detection area, and determine whether the vehicle is courteous to pedestrians;
细粒度识别模块，用于采用深度学习技术对检测到的未礼让行人的机动车及移动行人进行二次精细识别。The fine-grained recognition module is used for secondary fine-grained recognition of motor vehicles and moving pedestrians detected by using deep learning technology.
可选的，所述行为判定模块判断车辆未礼让行人的条件为：Optionally, the condition for the behavior determination module to determine that the vehicle does not yield to pedestrians is:
行人在预设时间内沿斑马线方向横向运动，此时在斑马线区域内检测到车辆，若该车辆与移动行人的距离小于设置的阈值且车辆在前t秒时间内沿车道方向移动，判断该车辆未礼让行人；Pedestrians move laterally in the direction of the zebra crossing within a preset time. At this time, a vehicle is detected in the zebra crossing area. If the distance between the vehicle and the moving pedestrian is less than the set threshold and the vehicle moves in the direction of the lane within the first t seconds, the vehicle is judged. not yielding to pedestrians;
行人穿过斑马线移动，此时在斑马线区域内检测到车辆，且该车辆在前t秒内沿车道方向以超过阈值速度行驶，判断该车辆未礼让行人。Pedestrians move across the zebra crossing. At this time, a vehicle is detected in the zebra crossing area, and the vehicle travels in the direction of the lane at a speed exceeding the threshold within the first t seconds, and it is judged that the vehicle has not yielded to the pedestrian.
可选的，还包括：Optionally, also include:
车牌识别模块，用于采用深度学习技术对检测到的不礼让行人机动车的车牌号码进行识别。The license plate recognition module is used to recognize the license plate number of the detected vehicle that does not yield to pedestrians by using deep learning technology.
可选的，所述车牌识别模块包括：Optionally, the license plate recognition module includes:
车牌检测单元，用于提取机动车的感兴趣区域，通过深度学习技术检测机动车的车牌位置，利用位置、大小信息对检测到的车牌进行过滤；The license plate detection unit is used to extract the area of interest of the motor vehicle, detect the license plate position of the motor vehicle through deep learning technology, and filter the detected license plate by using the position and size information;
车牌矫正单元，用于通过矫正网络使检测到的车牌倾斜对齐，保证字符识别的准确性；The license plate correction unit is used to align the detected license plate obliquely through the correction network to ensure the accuracy of character recognition;
车牌识别单元，用于将矫正后的车牌送入字符识别网络，通过CTC算法实现车牌字符的不定长识别。The license plate recognition unit is used to send the corrected license plate into the character recognition network, and realize the indefinite length recognition of the license plate characters through the CTC algorithm.
第二方面，本发明还提供一种基于深度学习的斑马线礼让行人视觉检测方法，包括：In a second aspect, the present invention also provides a deep learning-based visual detection method for zebra crossing polite pedestrians, including:
S10、设置所述视觉检测系统的检测参数；S10, setting the detection parameters of the visual detection system;
S20、将设置的所述视觉检测系统的检测参数发送给信息采集及处理模块；S20, sending the set detection parameters of the visual detection system to the information collection and processing module;
S30、检测并识别复杂交通场景下的行人、机动车目标，捕捉目标的行为并对捕捉到的行为进行判别。S30. Detect and identify pedestrian and motor vehicle targets in a complex traffic scene, capture the behavior of the target and discriminate the captured behavior.
可选的，步骤S10中，视觉检测系统的检测参数包括：车辆检测区域、行人斑马线区域、当前车道的斜率以及接入相机的参数。Optionally, in step S10, the detection parameters of the visual detection system include: a vehicle detection area, a pedestrian zebra crossing area, a slope of the current lane, and parameters for accessing the camera.
可选的，步骤S30包括：Optionally, step S30 includes:
S301、实时检测当前视频帧中的人、车实例，通过NMS算法对满足IOU及中心点距离阈值要求的检测框进行二次筛选；S301. Detect the instances of people and vehicles in the current video frame in real time, and perform secondary screening on the detection frames that meet the IOU and center point distance threshold requirements through the NMS algorithm;
S302、对检测到的人车实例进行特征提取和匹配，同时通过L-K光流算法及CIOU算法追踪人车实例在连续多帧中的目标轨迹；S302. Perform feature extraction and matching on the detected instances of people and vehicles, and simultaneously track the target trajectory of the instances of people and vehicles in multiple consecutive frames through the L-K optical flow algorithm and the CIOU algorithm;
S303、对进入检测区域的车辆及行人进行移动侦测，判断车辆是否礼让行人；S303. Perform motion detection on vehicles and pedestrians entering the detection area, and determine whether the vehicle yields to pedestrians;
S304、采用深度学习技术对检测到的未礼让行人的机动车及移动行人进行二次精细识别。S304 , using the deep learning technology to perform secondary fine-grained identification on the detected motor vehicles and moving pedestrians who have not yielded to the pedestrian.
根据本发明提供的具体实施例，本发明公开了以下技术效果：According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:
本发明提供的基于深度学习的斑马线礼让行人视觉检测系统及方法，该方法通过深度学习技术，对在复杂交通场景下的行人、机动车等进行实时的目标识别和追踪，对于机动车不礼让行人、机动车超速行驶等违反道路交通规范章程的行为进行实时的捕捉和判别，并且能够根据实际需要对相关车辆的车牌进行准确识别，协助交通监管者进行高效、便捷、智能的交通监管，提高了识别的效率和准确率，减少了漏检、误检的发生。The deep learning-based visual detection system and method for pedestrians at zebra crossings provided by the present invention, through the deep learning technology, can realize real-time target recognition and tracking of pedestrians, motor vehicles, etc. , motor vehicle speeding and other violations of road traffic regulations are captured and discriminated in real time, and the license plates of relevant vehicles can be accurately identified according to actual needs, assisting traffic supervisors in efficient, convenient and intelligent traffic supervision. The efficiency and accuracy of identification reduce the occurrence of missed detection and false detection.
附图说明Description of drawings
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例中所需要使用的附图作简单的介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings used in the embodiments. Obviously, the drawings in the following description are only some of the present invention. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without any creative effort.
图1为本发明基于深度学习的斑马线礼让行人视觉检测方法的流程图；Fig. 1 is the flow chart of the pedestrian visual detection method of zebra crossing courtesy based on deep learning of the present invention;
图2为本发明基于深度学习的斑马线礼让行人视觉检测系统的工作流程图。FIG. 2 is a working flow chart of the visual detection system for pedestrians at zebra crossing based on deep learning of the present invention.
具体实施方式Detailed ways
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
本发明的目的是为了提供一种基于深度学习的斑马线礼让行人视觉检测系统及方法，通过深度学习技术，对在复杂交通场景下的机动车不礼让行人的行为进行检测和识别，具有高效、准确、便捷、智能化程度高的特点。The purpose of the present invention is to provide a visual detection system and method for zebra crossing pedestrians based on deep learning, which can detect and recognize the behavior of vehicles in a complex traffic scene, which is highly efficient and accurate. , convenience and high degree of intelligence.
为使本发明的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本发明作进一步详细的说明。In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.
实施例1：Example 1:
本发明提供的一种基于深度学习的斑马线礼让行人视觉检测系统，包括：The present invention provides a zebra crossing courtesy pedestrian visual detection system based on deep learning, including:
参数配置模块，用于设置视觉检测系统的检测参数；The parameter configuration module is used to set the detection parameters of the visual inspection system;
通信模块，用于将设置的视觉检测系统的检测参数发送给信息采集及处理模块；The communication module is used to send the set detection parameters of the visual detection system to the information collection and processing module;
信息采集及处理模块，用于检测并识别复杂交通场景下的行人、机动车目标，捕捉目标的行为并对捕捉到的行为进行判别。The information collection and processing module is used to detect and identify pedestrians and motor vehicles in complex traffic scenes, capture the behavior of the target and discriminate the captured behavior.
参数配置模块设置在系统管理平台(服务器)上，信息采集及处理模块部署在AI边缘计算平台上，两者之间通过通信模块进行信息交互，通信模块具体采用Http通信协议，通信模块包括在信息采集及处理模块上。The parameter configuration module is set on the system management platform (server), and the information collection and processing module is deployed on the AI edge computing platform. The two exchange information through the communication module. The communication module uses the Http communication protocol, and the communication module is included in the information. acquisition and processing module.
其中，视觉检测系统的检测参数包括：车辆检测区域、行人斑马线区域、当前车道的斜率以及接入相机的参数。Among them, the detection parameters of the visual detection system include: vehicle detection area, pedestrian zebra crossing area, the slope of the current lane, and the parameters of the access camera.
由信息采集及处理模块获取监控场景图像后利用参数配置模块提供的界面绘制功能进行区域绘制，生成相应参数后通过HTTP协议由参数配置模块发送给信息采集及处理模块。After the information acquisition and processing module obtains the monitoring scene image, the interface drawing function provided by the parameter configuration module is used to draw the area, and the corresponding parameters are generated and sent to the information acquisition and processing module by the parameter configuration module through the HTTP protocol.
进一步地，信息采集及处理模块包括：Further, the information collection and processing module includes:
行人车辆检测模块，用于实时检测当前视频帧中的人、车实例，通过NMS算法对满足IOU及中心点距离阈值要求的检测框进行二次筛选。The pedestrian and vehicle detection module is used to detect people and vehicles in the current video frame in real time. The NMS algorithm is used to secondary screen the detection frames that meet the IOU and center point distance threshold requirements.
行人车辆检测模块采用深度学习技术实时检测当前视频，辅以NMS算法对检测结果进行二次筛选，该筛选方法通过对检测框的IOU和中心点距离综合判断，过滤低于阈值并且合并超过一定阈值的相交检测框，避免出现同一目标被多个检测框覆盖的现象，保证检出准确率。The pedestrian and vehicle detection module uses deep learning technology to detect the current video in real time, supplemented by the NMS algorithm for secondary screening of the detection results. This screening method comprehensively judges the IOU and the center point distance of the detection frame, and filters below the threshold and merges more than a certain threshold. It avoids the phenomenon that the same target is covered by multiple detection frames and ensures the detection accuracy.
多目标追踪模块，用于对检测到的人车实例进行特征提取和匹配，同时通过L-K光流算法及CIOU算法追踪人车实例在连续多帧中的目标轨迹。The multi-target tracking module is used to perform feature extraction and matching on the detected instances of people and vehicles, and at the same time track the target trajectories of instances of people and vehicles in multiple consecutive frames through the L-K optical flow algorithm and the CIOU algorithm.
通过对图像特征点的光流追踪及匹配算法，保证检测到的同一目标在连续多帧图像中能被正确追踪。Through the optical flow tracking and matching algorithm of image feature points, it is ensured that the same detected target can be correctly tracked in consecutive multi-frame images.
行为判定模块，用于对进入检测区域的车辆及行人进行移动侦测，判断车辆是否礼让行人。The behavior determination module is used to perform motion detection on vehicles and pedestrians entering the detection area, and determine whether the vehicle is courteous to pedestrians.
进一步地，行为判定模块判断车辆未礼让行人的条件为：Further, the condition for the behavior determination module to determine that the vehicle does not yield to the pedestrian is:
行为判定模块对进入检测区域的车辆及行人进行移动侦测，当行人在预设时间内沿斑马线方向横向运动时，对此时所检测到的车辆位置及时空行为进行判断，在当前车辆进入斑马线区域后，该车辆与移动行人的距离小于设置的阈值且车辆在前t秒时间内沿车道方向移动，判断该车辆未礼让行人；The behavior determination module performs motion detection on vehicles and pedestrians entering the detection area. When pedestrians move laterally in the direction of the zebra crossing within a preset time, the detected vehicle position and space-time behavior are determined at this time. When the current vehicle enters the zebra crossing After entering the area, the distance between the vehicle and the moving pedestrian is less than the set threshold and the vehicle moves in the direction of the lane within the first t seconds, and it is judged that the vehicle has not yielded to the pedestrian;
当行人穿过斑马线移动时，此时在斑马线区域内检测到车辆，且该车辆在前t秒内沿车道方向以超过阈值速度行驶，基于车辆无法在1秒内停车的先验条件，判断该车辆未礼让行人。When a pedestrian moves across the zebra crossing, a vehicle is detected in the zebra crossing area, and the vehicle travels along the lane at a speed exceeding the threshold within the first t seconds. Based on the prior condition that the vehicle cannot stop within 1 second, determine the Vehicles fail to yield to pedestrians.
细粒度识别模块，用于采用深度学习技术对检测到的未礼让行人的机动车及移动行人进行二次精细识别，提升检测准确率。该模块是为了弥补多任务检测模型在分类识别任务上存在的不足，可能存在行人、车辆误检的情况。The fine-grained recognition module is used to use deep learning technology to perform secondary fine-grained recognition of detected motor vehicles and mobile pedestrians that have not yielded to pedestrians, so as to improve the detection accuracy. This module is to make up for the shortcomings of the multi-task detection model in the classification and recognition tasks, and there may be false detections of pedestrians and vehicles.
进一步地，还包括：Further, it also includes:
车牌识别模块，用于采用深度学习技术对检测到的不礼让行人机动车的车牌号码进行识别。The license plate recognition module is used to recognize the license plate number of the detected vehicle that does not yield to pedestrians by using deep learning technology.
进一步地，车牌识别模块包括：Further, the license plate recognition module includes:
车牌检测单元，用于提取机动车的感兴趣区域，通过深度学习技术检测机动车的车牌位置，利用位置、大小信息对检测到的车牌进行过滤；The license plate detection unit is used to extract the area of interest of the motor vehicle, detect the license plate position of the motor vehicle through deep learning technology, and filter the detected license plate by using the position and size information;
车牌矫正单元，用于通过矫正网络使检测到的车牌倾斜对齐，保证字符识别的准确性；The license plate correction unit is used to align the detected license plate obliquely through the correction network to ensure the accuracy of character recognition;
车牌识别单元，用于将矫正后的车牌送入字符识别网络，通过CTC算法实现车牌字符的不定长识别。The license plate recognition unit is used to send the corrected license plate into the character recognition network, and realize the indefinite length recognition of the license plate characters through the CTC algorithm.
本发明提供的基于深度学习的斑马线礼让行人视觉检测系统，通过深度学习技术及服务器与AI边缘计算平台的通信，能够在设定的区域内对复杂交通场景下的行人、机动车等目标进行实时检测、追踪和识别，并根据行人及机动车的位置及时空行为，对机动车不礼让行人、机动车超速行驶等违反道路交通规范章程的行为进行实时的捕捉和判别，并且对相关车辆的车牌进行识别，提高了交通监管的效率和准确性。The deep learning-based visual detection system for pedestrians at zebra crossing politeness provided by the present invention, through the deep learning technology and the communication between the server and the AI edge computing platform, can conduct real-time detection of pedestrians, motor vehicles and other targets in complex traffic scenes within a set area. Detect, track and identify, and based on the location and space-time behavior of pedestrians and motor vehicles, real-time capture and identification of behaviors that violate road traffic regulations, such as motor vehicles disrespecting pedestrians and motor vehicles speeding, and the license plates of related vehicles. The identification improves the efficiency and accuracy of traffic supervision.
实施例2：Example 2:
本发明还提供一种基于深度学习的斑马线礼让行人视觉检测方法，如图1所示，该方法包括：The present invention also provides a deep learning-based visual detection method for zebra crossing polite pedestrians, as shown in FIG. 1 , the method includes:
S10、设置视觉检测系统的检测参数；S10. Set detection parameters of the visual detection system;
S20、将设置的视觉检测系统的检测参数发送给信息采集及处理模块；S20, sending the set detection parameters of the visual detection system to the information collection and processing module;
S30、检测并识别复杂交通场景下的行人、机动车目标，捕捉目标的行为并对捕捉到的行为进行判别。S30. Detect and identify pedestrian and motor vehicle targets in a complex traffic scene, capture the behavior of the target and discriminate the captured behavior.
上述步骤S10中，视觉检测系统的检测参数包括：车辆检测区域、行人斑马线区域、当前车道的斜率以及接入相机的参数。In the above step S10, the detection parameters of the visual detection system include: a vehicle detection area, a pedestrian zebra crossing area, a slope of the current lane, and parameters for accessing the camera.
上述步骤S30包括：The above step S30 includes:
S301、实时检测当前视频帧中的人、车实例，通过NMS算法对满足IOU及中心点距离阈值要求的检测框进行二次筛选；S301. Detect the instances of people and vehicles in the current video frame in real time, and perform secondary screening on the detection frames that meet the IOU and center point distance threshold requirements through the NMS algorithm;
S302、对检测到的人车实例进行特征提取和匹配，同时通过L-K光流算法及CIOU算法追踪人车实例在连续多帧中的目标轨迹；S302. Perform feature extraction and matching on the detected instances of people and vehicles, and simultaneously track the target trajectory of the instances of people and vehicles in multiple consecutive frames through the L-K optical flow algorithm and the CIOU algorithm;
S303、对进入检测区域的车辆及行人进行移动侦测，判断车辆是否礼让行人；S303. Perform motion detection on vehicles and pedestrians entering the detection area, and determine whether the vehicle yields to pedestrians;
S304、采用深度学习技术对检测到的未礼让行人的机动车及移动行人进行二次精细识别；S304, using deep learning technology to perform secondary fine identification on detected motor vehicles and moving pedestrians that do not yield to pedestrians;
车辆未礼让行人的判定条件为：The judgment conditions for vehicles failing to yield to pedestrians are:
行为判定模块对进入检测区域的车辆及行人进行移动侦测，当行人在预设时间内沿斑马线方向横向运动时，对此时所检测到的车辆位置及时空行为进行判断，在当前车辆进入斑马线区域后，该车辆与移动行人的距离小于设置的阈值且车辆在前t秒时间内沿车道方向移动，判断该车辆未礼让行人；The behavior determination module performs motion detection on vehicles and pedestrians entering the detection area. When pedestrians move laterally in the direction of the zebra crossing within a preset time, the detected vehicle position and space-time behavior are determined at this time. When the current vehicle enters the zebra crossing After entering the area, the distance between the vehicle and the moving pedestrian is less than the set threshold and the vehicle moves in the direction of the lane within the first t seconds, and it is judged that the vehicle has not yielded to the pedestrian;
当行人穿过斑马线移动时，此时在斑马线区域内检测到车辆，且该车辆在前t秒内沿车道方向以超过阈值速度行驶，基于车辆无法在1秒内停车的先验条件，判断该车辆未礼让行人。When a pedestrian moves across the zebra crossing, a vehicle is detected in the zebra crossing area, and the vehicle travels along the lane at a speed exceeding the threshold within the first t seconds. Based on the prior condition that the vehicle cannot stop within 1 second, determine the Vehicles fail to yield to pedestrians.
如图2所示，为本发明实施例斑马线礼让行人视觉检测系统的工作流程图：As shown in FIG. 2 , it is a working flowchart of the visual detection system for pedestrians at zebra crossings according to an embodiment of the present invention:
1)、AI边缘计算平台通过视频拍摄设备实时获取道路交通场景的视频流。1) The AI edge computing platform obtains the video stream of the road traffic scene in real time through the video shooting device.
2)、对获取的视频流进行抽帧，将抽取的图像转换为RGB格式。2), extracting frames from the acquired video stream, and converting the extracted images into RGB format.
3)、检测当前视频帧中的人、车实例，包括车辆及行人的数量，对车辆及行人进行追踪。3) Detect the instances of people and vehicles in the current video frame, including the number of vehicles and pedestrians, and track the vehicles and pedestrians.
4)、根据车辆未礼让行人的判定条件判定车辆是否礼让行人，将车辆未礼让行人的图像进行细粒度识别。4), according to the judgment condition that the vehicle fails to yield to the pedestrian, determine whether the vehicle yields to the pedestrian, and performs fine-grained identification on the image of the vehicle without yielding to the pedestrian.
5)、对于属于待检类别的车辆，识别车辆的车牌并上传至服务器。5) For vehicles belonging to the category to be inspected, identify the license plate of the vehicle and upload it to the server.
本发明提供的基于深度学习的斑马线礼让行人视觉检测方法，该方法通过深度学习技术，对在复杂交通场景下的行人、机动车等进行实时的目标识别和追踪，对于机动车不礼让行人、机动车超速行驶等违反道路交通规范章程的行为进行实时的捕捉和判别，并且能够根据实际需要对相关车辆的车牌进行准确识别，协助交通监管者进行高效、便捷、智能的交通监管，提高了识别的效率和准确率，减少了漏检、误检的发生。The present invention provides a visual detection method for pedestrians and pedestrians at zebra crossing based on deep learning. The method uses deep learning technology to perform real-time target recognition and tracking for pedestrians, motor vehicles, etc. in complex traffic scenes. Real-time capture and identification of behaviors that violate road traffic regulations, such as motor vehicles speeding, can accurately identify the license plates of relevant vehicles according to actual needs, assist traffic supervisors in efficient, convenient and intelligent traffic supervision, and improve the recognition efficiency. Efficiency and accuracy reduce the occurrence of missed detection and false detection.
本文中应用了具体个例对本发明的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本发明的方法及其核心思想；同时，对于本领域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围上均会有改变之处。综上所述，本说明书内容不应理解为对本发明的限制。In this paper, specific examples are used to illustrate the principles and implementations of the present invention. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the present invention; meanwhile, for those skilled in the art, according to the present invention There will be changes in the specific implementation and application scope. In conclusion, the contents of this specification should not be construed as limiting the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210027599.2ACN114387549A (en) | 2022-01-11 | 2022-01-11 | A visual detection system and method for pedestrians at zebra crossing based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210027599.2ACN114387549A (en) | 2022-01-11 | 2022-01-11 | A visual detection system and method for pedestrians at zebra crossing based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114387549Atrue CN114387549A (en) | 2022-04-22 |
Family
ID=81201003
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210027599.2APendingCN114387549A (en) | 2022-01-11 | 2022-01-11 | A visual detection system and method for pedestrians at zebra crossing based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114387549A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114708643A (en)* | 2022-06-02 | 2022-07-05 | 杭州智诺科技股份有限公司 | Computing power improving method for edge video analysis device and edge video analysis device |
| CN115861907A (en)* | 2023-03-02 | 2023-03-28 | 山东华夏高科信息股份有限公司 | Helmet detection method and system |
| WO2025107599A1 (en)* | 2023-11-21 | 2025-05-30 | 北京百度网讯科技有限公司 | Method and apparatus for detecting failure to yield to pedestrians of vehicle |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104680133A (en)* | 2015-02-04 | 2015-06-03 | 上海依图网络科技有限公司 | Real-time detecting method for pedestrian avoidance behavior of illegal vehicle |
| CN107730906A (en)* | 2017-07-11 | 2018-02-23 | 银江股份有限公司 | Zebra stripes vehicle does not give precedence to the vision detection system of pedestrian behavior |
| CN109712406A (en)* | 2019-02-12 | 2019-05-03 | 合肥极光科技股份有限公司 | A kind of pedestrian running red light and motor vehicle do not give precedence to pedestrian and monitor capturing system |
| CN109741608A (en)* | 2019-01-29 | 2019-05-10 | 浙江浩腾电子科技股份有限公司 | Motor vehicle based on deep learning, which is turned right, gives precedence to pedestrian's analysis capturing system and method |
| KR102131716B1 (en)* | 2020-03-27 | 2020-07-08 | 정병석 | Method for image monotoring employing car number recognition technology |
- 2022
- 2022-01-11CNCN202210027599.2Apatent/CN114387549A/enactivePending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104680133A (en)* | 2015-02-04 | 2015-06-03 | 上海依图网络科技有限公司 | Real-time detecting method for pedestrian avoidance behavior of illegal vehicle |
| CN107730906A (en)* | 2017-07-11 | 2018-02-23 | 银江股份有限公司 | Zebra stripes vehicle does not give precedence to the vision detection system of pedestrian behavior |
| CN109741608A (en)* | 2019-01-29 | 2019-05-10 | 浙江浩腾电子科技股份有限公司 | Motor vehicle based on deep learning, which is turned right, gives precedence to pedestrian's analysis capturing system and method |
| CN109712406A (en)* | 2019-02-12 | 2019-05-03 | 合肥极光科技股份有限公司 | A kind of pedestrian running red light and motor vehicle do not give precedence to pedestrian and monitor capturing system |
| KR102131716B1 (en)* | 2020-03-27 | 2020-07-08 | 정병석 | Method for image monotoring employing car number recognition technology |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114708643A (en)* | 2022-06-02 | 2022-07-05 | 杭州智诺科技股份有限公司 | Computing power improving method for edge video analysis device and edge video analysis device |
| CN115861907A (en)* | 2023-03-02 | 2023-03-28 | 山东华夏高科信息股份有限公司 | Helmet detection method and system |
| WO2025107599A1 (en)* | 2023-11-21 | 2025-05-30 | 北京百度网讯科技有限公司 | Method and apparatus for detecting failure to yield to pedestrians of vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114387549A (en) | A visual detection system and method for pedestrians at zebra crossing based on deep learning | |
| CN105744232B (en) | A kind of method of the transmission line of electricity video external force damage prevention of Behavior-based control analytical technology | |
| CN107705552B (en) | Emergency lane occupation behavior detection method, device and system | |
| CN103235938B (en) | The method and system of car plate detection and indentification | |
| KR101671428B1 (en) | Intelligent Monitoring System For Violation Vehicles in crossroads | |
| CN100557657C (en) | A Vehicle Detection Method Based on Video Image Features | |
| CN103021186B (en) | Vehicle monitoring method and vehicle monitoring system | |
| CN103366569B (en) | The method and system of real-time grasp shoot traffic violation vehicle | |
| CN107301777A (en) | Vehicle peccancy lane change detection method based on video detection technology | |
| CN105741559B (en) | A kind of illegal occupancy Emergency Vehicle Lane detection method based on track line model | |
| CN111126171A (en) | Vehicle reverse running detection method and system | |
| CN110580808B (en) | Information processing method and device, electronic equipment and intelligent traffic system | |
| CN101964145A (en) | Automatic license plate recognition method and system | |
| CN108269407B (en) | Security robot capable of automatically managing people stream and logistics | |
| CN110738857A (en) | vehicle violation evidence obtaining method, device and equipment | |
| KR101780320B1 (en) | Intelligent type of system for monitoring with a plural functions | |
| CN112381014A (en) | Illegal parking vehicle detection and management method and system based on urban road | |
| CN102142193A (en) | Automatic detection system and method for traffic violation | |
| CN106384513A (en) | Fake-licensed car capturing system and method based on intelligent traffic | |
| CN113112813B (en) | Illegal parking detection method and device | |
| CN205281786U (en) | alarm system | |
| CN109176513A (en) | A kind of method for inspecting and cruising inspection system of intelligent inspection robot | |
| CN116311166A (en) | Traffic obstacle recognition method and device and electronic equipment | |
| KR100948382B1 (en) | Security service method and system | |
| CN113903008A (en) | Ramp exit vehicle violation identification method based on deep learning and trajectory tracking |
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 |