WO2017193701A1 - 共享单车的倒地检测方法 - Google Patents

共享单车的倒地检测方法 Download PDF

Info

Publication number
WO2017193701A1
WO2017193701A1 PCT/CN2017/077195 CN2017077195W WO2017193701A1 WO 2017193701 A1 WO2017193701 A1 WO 2017193701A1 CN 2017077195 W CN2017077195 W CN 2017077195W WO 2017193701 A1 WO2017193701 A1 WO 2017193701A1
Authority
WO
WIPO (PCT)
Prior art keywords
bicycle
image
identification code
sub
distance
Prior art date
Application number
PCT/CN2017/077195
Other languages
English (en)
French (fr)
Inventor
黄安武
Original Assignee
黄安武
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 黄安武 filed Critical 黄安武
Priority to CN201780001007.7A priority Critical patent/CN107851325A/zh
Publication of WO2017193701A1 publication Critical patent/WO2017193701A1/zh

Links

Images

Classifications

    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F01MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
    • F01NGAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR MACHINES OR ENGINES IN GENERAL; GAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR INTERNAL COMBUSTION ENGINES
    • F01N9/00Electrical control of exhaust gas treating apparatus
    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F01MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
    • F01NGAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR MACHINES OR ENGINES IN GENERAL; GAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR INTERNAL COMBUSTION ENGINES
    • F01N3/00Exhaust or silencing apparatus having means for purifying, rendering innocuous, or otherwise treating exhaust
    • F01N3/08Exhaust or silencing apparatus having means for purifying, rendering innocuous, or otherwise treating exhaust for rendering innocuous
    • F01N3/0807Exhaust or silencing apparatus having means for purifying, rendering innocuous, or otherwise treating exhaust for rendering innocuous by using absorbents or adsorbents
    • F01N3/0828Exhaust or silencing apparatus having means for purifying, rendering innocuous, or otherwise treating exhaust for rendering innocuous by using absorbents or adsorbents characterised by the absorbed or adsorbed substances
    • F01N3/0857Carbon oxides
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/70Determining position or orientation of objects or cameras
    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F01MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
    • F01NGAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR MACHINES OR ENGINES IN GENERAL; GAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR INTERNAL COMBUSTION ENGINES
    • F01N2370/00Selection of materials for exhaust purification
    • F01N2370/22Selection of materials for exhaust purification used in non-catalytic purification apparatus
    • FMECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F01MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
    • F01NGAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR MACHINES OR ENGINES IN GENERAL; GAS-FLOW SILENCERS OR EXHAUST APPARATUS FOR INTERNAL COMBUSTION ENGINES
    • F01N2560/00Exhaust systems with means for detecting or measuring exhaust gas components or characteristics
    • F01N2560/02Exhaust systems with means for detecting or measuring exhaust gas components or characteristics the means being an exhaust gas sensor
    • F01N2560/022Exhaust systems with means for detecting or measuring exhaust gas components or characteristics the means being an exhaust gas sensor for measuring or detecting CO or CO2
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02TCLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00Road transport of goods or passengers
    • Y02T10/10Internal combustion engine [ICE] based vehicles
    • Y02T10/40Engine management systems

Definitions

  • the invention relates to the field of bicycle field and computer vision processing, and in particular to a method for detecting a ground fault of a shared bicycle.
  • the invention provides a method for detecting the falling of a shared bicycle, which can improve the efficiency of detecting whether the bicycle is thrown onto the ground.
  • the method for detecting the falling of a shared bicycle of the present invention comprises the following steps:
  • the bicycle identification code image is an image pre-applied to a specific part of the bicycle, and the image includes several independent and Sub-identification code images with different attribute values;
  • the output is displayed to inform the corresponding staff that there is a need to deal with the bicycle falling to the ground.
  • the distance between the center point of all other sub-identification image images and the center point of the reference image is calculated using the following formula: In the formula, Representing the distance, a vector representing the reference image, A vector representing other sub-ID images, and m represents the sum of the numbers of all sub-ID images.
  • the standard distance coefficient is calculated using the following formula: Where ⁇ denotes the sub-identification image mean, X denotes a standard distance coefficient, and N denotes the total amount of attribute values.
  • the process of calculating the deflection angle of the bicycle according to the offset angle and the offset direction comprises: calculating the rotation coefficient according to the offset angle ⁇ and the offset direction (f x , f y , f z ) H, and calculating a deflection angle ⁇ of the bicycle corresponding to the reference angle according to the rotation coefficient H; using the following formula:
  • a sub-identification code image is re-selected as the updated reference image, and the calculation returns to calculate all other sub-identification code image centers. The step of the distance between the point and the center point of the updated reference image.
  • the bicycle identification code further includes direction information.
  • the method for detecting the falling of the shared bicycle of the present invention has a bicycle identification code attached to the bicycle in advance, and then obtains the bicycle identification code image through the surveillance cameras everywhere, and calculates each child.
  • the distance between the identification code images and the standard distance coefficient are selected, and the sub-identification code image corresponding to the distance of the standard distance coefficient is less than the set threshold value is selected to compare with the pre-stored original identification code image, and the sub-identification code image is calculated.
  • the offset angle and the offset direction are used to calculate the deflection angle of the entire bicycle, thereby identifying whether the bicycle is down to the ground, and if so, automatically outputting an alarm prompt message to the corresponding staff for reminding.
  • the solution of the present invention is completely implemented by a software algorithm, and does not need to add additional hardware sensors, avoids the blind search of manual detection, and can find and remind the first time where there is a bicycle falling to the ground and realize the cost.
  • Low, and high accuracy greatly improve the efficiency of bicycle down detection, and is conducive to the sharing and use of bicycles in the community, to facilitate the public's green travel.
  • FIG. 1 is a schematic flow chart of a method for detecting a falling of a shared bicycle provided by the present invention.
  • a schematic flowchart of a preferred embodiment of a method for detecting a fall of a shared bicycle includes the following steps:
  • Step S1 acquiring a video image captured by the surveillance camera in real time, and extracting a bicycle identification code image on the bicycle from the video image, and then proceeding to step S2;
  • the bicycle identification code image is an image pre-applied to a specific part of the bicycle, and the image is
  • the image includes a plurality of sub-identification image images that are independent of each other and have different attribute values.
  • a bicycle identification code is attached to the bicycle in advance (for example, it can be attached to the handlebar and the seatbar, etc.), and the bicycle identification code is specially made, similar to the form of the two-dimensional code, and each manufacturer can According to the specific needs of the bicycle identification code, the most basic requirement is to add direction information to the bicycle identification code, so that the direction of the original identification code can be obtained, so that the direction of the original identification code can be called in subsequent comparison. In addition to this, it is also possible to add information unique to each manufacturer in the bicycle identification code to distinguish it. At present, there are some similar identification codes in the field of computer vision computing, which are not described in the present invention.
  • Step S2 performing image segmentation processing on the bicycle identification code image, obtaining each sub-identification code image, and arbitrarily selecting one sub-identification code image as a reference image, and calculating all other sub-identification code image center points and the reference image center point. The distance between them then proceeds to step S3.
  • the distance between all other sub-identification image image center points and the reference image center point may be calculated using the following formula: In the formula, Representing the distance, a vector representing the reference image, A vector representing other sub-ID images, and m represents the sum of the numbers of all sub-ID images.
  • step S3 the total amount of attribute values of each sub-ID image is calculated, and the standard distance coefficient is calculated according to the total amount of the attribute values, and then proceeds to step S4.
  • the standard distance coefficient can be calculated using the following formula:
  • denotes the sub-identification image mean
  • X denotes the standard distance coefficient
  • N denotes the total amount of the attribute value
  • the attribute refers to the dimension information such as the direction value and the feature value
  • each sub-ID image is given a different from the other sub
  • the attribute value of the identification code image is added, and the attribute values of all the sub-ID images are added to obtain the total value of the attribute values.
  • Step S4 determining whether there is a distance from the standard distance coefficient that is less than the set threshold, and if so, comparing the sub-identification code image corresponding to the distance with the original identification code image stored in the database in advance, and determining by comparison The offset angle and the offset direction of the sub-ID image with respect to the original ID image, and then proceeds to step S5.
  • is a constant coefficient, which may be 2 in the embodiment of the present invention
  • is a constant coefficient, which may be 2 in the embodiment of the present invention
  • step S5 the deflection angle of the bicycle is calculated according to the offset angle and the offset direction, and then proceeds to step S6.
  • the process of calculating the deflection angle of the bicycle according to the offset angle and the offset direction in this step may specifically include: according to the offset angle ⁇ and the offset direction (f x , f y , f z ) calculating the rotation coefficient H, and calculating the deflection angle ⁇ of the bicycle corresponding to the reference angle according to the rotation coefficient H; the following formula can be adopted:
  • v f (0, 0, 1) is the reference vector.
  • step S6 it is determined whether the deflection angle of the bicycle exceeds a set angle threshold. If yes, it is determined that the bicycle has fallen to the ground.
  • the local GPS positioning information of the surveillance camera can be automatically invoked, and the monitoring range of the surveillance camera can be A surveillance camera captures that the grounded bicycle is in a certain area near the surveillance camera, so the local GPS positioning information of the surveillance camera is defaulted to an inverted single.
  • the approximate location of the vehicle is edited and the local GPS positioning information is edited into the first alarm prompt information for output display to notify the corresponding staff that the bicycle has to be disposed of.
  • the efficiency is higher, the downhill bicycle can be found and notified at the first time; in addition, compared with the scheme of adding a hardware sensor on the bicycle to detect whether the bicycle is fell to the ground, On the one hand, it does not increase the hardware cost of the bicycle, and the maintenance cost is lower than that of the hardware sensor.
  • the scheme passes the verification of the sub-identification image. In order to improve the detection accuracy, it can be measured to achieve a detection accuracy of more than 99.9%, which is more advantageous than the hardware sensor solution.
  • a sub-identification code image may be re-selected as the updated reference image, and the process returns to the calculation in step S2. The step of the distance between the center point of the other all sub-identification code images and the updated reference image center point until the condition is satisfied.
  • the method includes the following steps: if the ratio of the distance to the standard distance coefficient is less than one than the set threshold, the distances corresponding to the sub-identification code images satisfying the condition are sorted in ascending order And selecting the first sub-identification code image to perform the step of comparing with the original identification code image.
  • the bicycle identification code may further include direction information.
  • the following steps may be further included:
  • the structure and shape of the bicycle are represented by using the geometric model or structure of the bicycle in advance. Shape, and by extracting the characteristics of the object of the bicycle, establish a correspondence between the model and the image, and then through the geometric method for bicycle recognition, the recognition efficiency is higher and more accurate.
  • the distance between the first key feature point and the second key feature point calculated in the above implementation does not satisfy the pre-preparation.
  • the distance threshold is set, it indicates that there is a fault in the bicycle identification code, such as loss or damage.
  • an alarm prompt is issued to notify the corresponding staff to replace the bicycle identification code.
  • the method for detecting the falling of the shared bicycle of the present invention has a bicycle identification code attached to the bicycle in advance, and then obtains the bicycle identification code image through the surveillance cameras located everywhere, and calculates each child.
  • the distance between the identification code images and the standard distance coefficient are selected, and the sub-identification code image corresponding to the distance of the standard distance coefficient is less than the set threshold value is selected to compare with the pre-stored original identification code image, and the sub-identification code image is calculated.
  • the offset angle and the offset direction are used to calculate the deflection angle of the entire bicycle, thereby identifying whether the bicycle is down to the ground, and if so, automatically outputting an alarm prompt message to the corresponding staff for reminding.
  • the solution of the present invention is completely implemented by a software algorithm, and does not need to add additional hardware sensors, avoids the blind search of manual detection, and can find and remind the first time where there is a bicycle falling to the ground and realize the cost.
  • Low, and high accuracy greatly improve the efficiency of bicycle down detection, and is conducive to the sharing and use of bicycles in the community, to facilitate the public's green travel.

Landscapes

  • Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Combustion & Propulsion (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Treating Waste Gases (AREA)
  • Exhaust Gas After Treatment (AREA)
  • Image Analysis (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)
  • Exhaust Gas Treatment By Means Of Catalyst (AREA)

Abstract

一种共享单车的倒地检测方法,包括:获取监控摄像头实时拍摄到的视频图像,并从视频图像中提取单车上的单车识别码图像(S1);对单车识别码图像进行图像分割处理,获得各子识别码图像,并任意挑选一个子识别码图像为基准图像,计算其他所有子识别码图像中心点与该基准图像中心点之间的距离(S2);计算各子识别码图像的属性值总量,并根据属性值总量计算标准距离系数(S3);判断是否存在与标准距离系数的比值小于设定阈值的距离,若是,则确定子识别码图像相对于原始识别码图像的偏移角度和偏移方向(S4);根据偏移角度和偏移方向计算单车的偏转角度(S5);判断单车的偏转角度是否超过设定角度阈值,若是则判定为单车已经倒地(S6)。该共享单车的倒地检测方法能提高单车倒地检测的效率。

Description

共享单车的倒地检测方法 技术领域
本发明涉及自行车领域和计算机视觉处理领域,尤其涉及一种共享单车的倒地检测方法。
背景技术
近来,大城市中“随停随走”的共享单车(或称“公共自行车”)非常受欢迎。随之而来的共享单车乱停乱放的问题不断出现,甚至有些共享单车被人使用完之后随意丢弃在地面上,倒在地面上的共享单车很容易弄脏,这样后面的人即使看到这辆弄脏的共享单车也不会有人愿意使用,从而极大影响了市容和行人的出行。
为了解决这个问题,有些共享单车运营机构不得不雇佣专门的工作人员来进行维护和整治,通过工作人员分工负责一片区域并采用定期巡逻的方式来寻找和发现这些乱丢并倒在地面上的共享单车。但是采用这种人工检测的方法,其效率非常低,往往不能及时发现哪里有被乱丢倒在地面上的共享单车。
发明内容
本发明提出一种共享单车的倒地检测方法,能够提高检测单车是否被乱丢倒在地面上的效率。
本发明的一种共享单车的倒地检测方法,包括如下步骤:
获取监控摄像头实时拍摄到的视频图像,并从视频图像中提取单车上的单车识别码图像;所述单车识别码图像为预先贴在单车特定部位的图像,且该图像中包括若干个彼此独立且属性值各不相同的子识别码图像;
对所述单车识别码图像进行图像分割处理,获得各子识别码图像,并任意挑选一个子识别码图像为基准图像,计算其他所有子识别码图像中心点与该基 准图像中心点之间的距离;
计算各子识别码图像的属性值总量,并根据所述属性值总量计算标准距离系数;
判断是否存在与标准距离系数的比值小于设定阈值的距离,若是,则将该距离所对应的子识别码图像与预先存储的原始识别码图像进行对比,确定所述子识别码图像相对于所述原始识别码图像的偏移角度和偏移方向;
根据所述偏移角度和偏移方向计算单车的偏转角度;
判断所述单车的偏转角度是否超过设定角度阈值,若是,则判定为单车已经倒地,此时自动获取本地GPS定位信息,并将所述本地GPS定位信息编辑到第一告警提示信息中进行输出显示,以通知相应的工作人员有单车倒地需要处理。
在其中一个实施例中,采用如下公式来计算其他所有子识别码图像中心点与所述基准图像中心点之间的距离:
Figure PCTCN2017077195-appb-000001
式中,
Figure PCTCN2017077195-appb-000002
表示所述距离,
Figure PCTCN2017077195-appb-000003
表示基准图像的向量,
Figure PCTCN2017077195-appb-000004
表示其他子识别码图像的向量,m表示所有子识别码图像的数量总和。
在其中一个实施例中,采用如下公式来计算所述标准距离系数:
Figure PCTCN2017077195-appb-000005
式中,α表示子识别码图像均值,X表示标准距离系数,N表示属性值总量。
在其中一个实施例中,根据所述偏移角度和偏移方向计算单车的偏转角度的过程包括:根据所述偏移角度α和偏移方向(fx,fy,fz)计算旋转系数H,并根据所述旋转系数H计算单车相当于基准角度的偏转角度β;采用如下公式:
Figure PCTCN2017077195-appb-000006
式中,I是3x3的单位矩阵,
Figure PCTCN2017077195-appb-000007
vf=(0,0,1)。
在其中一个实施例中,若不存在与标准距离系数的比值小于设定阈值的距离,则重新挑选一个子识别码图像作为更新后的基准图像,并返回计算所述其他所有子识别码图像中心点与该更新后的基准图像中心点之间的距离的步骤。
在其中一个实施例中,在判断得出存在与标准距离系数的比值小于设定阈值的距离之后、将该距离所对应的子识别码图像与预先存储的原始识别码图像进行对比之前,还包括步骤:若所述与标准距离系数的比值小于设定阈值的距离的个数多于1个,则对满足此条件下的各子识别码图像对应的距离按从小到大的顺序进行排序,并选择排序第一位的子识别码图像来执行与原始识别码图像进行对比的步骤。
在其中一个实施例中,所述单车识别码中还包括方向信息。
在其中一个实施例中,在从视频图像中提取单车上的单车识别码图像之前,还包括如下步骤:
将所述视频图像输入到预先建立的单车模型中进行匹配,提取所述视频图像中的物体特征进行相似度计算;
当相似度计算的结果超过设定特征阈值时,判定为检测到视频图像中存在单车,并提取出与所述物体特征相近的纹理进行保存;否则判定为视频图像中不存在单车,并返回摄像头实时拍摄视频图像的步骤。
在其中一个实施例中,在提取出单车上的单车识别码图像之后,还包括如下步骤:
从所述单车的车把上贴着的单车识别码图像中提取出第一关键特征点;以及从所述单车的坐杆上贴着的单车识别码图像中提取出第二关键特征点;
计算所述第一关键特征点和第二关键特征点之间的距离;
判断所述距离是否在预设的距离阈值内,若否,则输出第二告警提示信息,以提示相应的工作人员所述单车识别码出现故障。
从以上方案可以看出,本发明的一种共享单车的倒地检测方法,预先在单车上贴有单车识别码,然后通过遍布各处的监控摄像头来获取到单车识别码图像,并计算各子识别码图像之间的距离以及标准距离系数,选取与标准距离系数的比值小于设定阈值的距离对应的子识别码图像来与预先存储的原始识别码图像进行对比,计算出子识别码图像的偏移角度和偏移方向,进而计算出整个单车的偏转角度,以此识别出单车是否倒地,如果是则自动输出告警提示信息到相应的工作人员进行提醒。相比于现有的技术,本发明的方案完全采用软件算法实现,不需要增加额外的硬件传感器,避免了人工检测的盲目性查找,能第一时间发现并提醒哪里有单车倒地,实现成本低,且准确度高,极大提高了单车倒地检测的效率,有利于共享单车在社会上的推广使用,方便市民的绿色出行。
附图说明
图1是本发明提供的一种共享单车的倒地检测方法流程示意图。
具体实施方式
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作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。
如图1所示,为本发明提供的一种共享单车的倒地检测方法一个优选的实施例的流程示意图,包括如下步骤:
步骤S1,获取监控摄像头实时拍摄到的视频图像,并从视频图像中提取单车上的单车识别码图像,然后进入步骤S2;所述单车识别码图像为预先贴在单车特定部位的图像,且该图像中包括若干个彼此独立且属性值各不相同的子识别码图像。
目前,监控摄像头遍布市区各处,尤其是大城市中监控摄像头的应用更加 广泛。但是传统的监控摄像头主要用来拍摄视频并进行存档,以便事后可进行追溯;由于缺少智能的算法,传统的监控摄像头对于突发状况的处理并不是很理想。采用本发明提供的算法,可以有效提升传统监控摄像头的智能程度,让传统的监控摄像头发挥出更大的作用。
本发明中,预先在单车上贴有单车识别码(例如可以贴在车把和坐杆上等),且该单车识别码是特殊制作而成的,类似于二维码的形态,各厂商可以根据具体需求制作属于自己的单车识别码,最基本的要求是需要在单车识别码中添加方向信息,这样就可以获取到原始识别码的方向,以便后续对比时可以调用该原始识别码的方向。除此之外还可以在单车识别码添加各厂商独有的信息等以便进行区分。目前在计算机视觉计算领域已经有一些类似的识别码的应用,本发明中不作过多描述。
由于单车识别码中的特征点较为明显,较易识别,因此对这些特征点进行识别的难度较低,相对于实物识别来讲,进行图像对比和计算的运算量成几何级数降低,可以有效提高运算效率,节省计算机的资源。
步骤S2,对所述单车识别码图像进行图像分割处理,获得各子识别码图像,并任意挑选一个子识别码图像为基准图像,计算其他所有子识别码图像中心点与该基准图像中心点之间的距离,然后进入步骤S3。
在其中一个实施例中,可以采用如下公式来计算其他所有子识别码图像中心点与所述基准图像中心点之间的距离:
Figure PCTCN2017077195-appb-000008
式中,
Figure PCTCN2017077195-appb-000009
表示所述距离,
Figure PCTCN2017077195-appb-000010
表示基准图像的向量,
Figure PCTCN2017077195-appb-000011
表示其他子识别码图像的向量,m表示所有子识别码图像的数量总和。
步骤S3,计算各子识别码图像的属性值总量,并根据所述属性值总量计算标准距离系数,然后进入步骤S4。
在其中一个实施例中,可以采用如下公式来计算所述标准距离系数:
Figure PCTCN2017077195-appb-000012
式中,α表示子识别码图像均值,X表示标准距离系数,N表示属性值总量,属性是指方向值、特征值等维度信息,每个子识别码图像均事先赋予了一个不同于其他子识别码图像的属性值,将所有子识别码图像的属性值相加,即可得到属性值总量。
步骤S4,判断是否存在与标准距离系数的比值小于设定阈值的距离,若是,则将该距离所对应的子识别码图像与预先存储在数据库中的原始识别码图像进行对比,通过对比可确定所述子识别码图像相对于所述原始识别码图像的偏移角度和偏移方向,然后进入步骤S5。
例如,若存在
Figure PCTCN2017077195-appb-000013
(ε为一常数系数,本发明实施例中可取值为2),则认为存在在与标准距离系数的比值小于设定阈值的距离。
步骤S5,根据所述偏移角度和偏移方向计算单车的偏转角度,然后进入步骤S6。
在其中一个实施例中,本步骤中根据所述偏移角度和偏移方向计算单车的偏转角度的过程具体可以包括如下:根据所述偏移角度α和偏移方向(fx,fy,fz)计算旋转系数H,并根据所述旋转系数H计算单车相当于基准角度的偏转角度β;可以采用如下公式:
Figure PCTCN2017077195-appb-000014
式中,I是3x3的单位矩阵,
Figure PCTCN2017077195-appb-000015
vf=(0,0,1)为参考向量。
步骤S6,判断所述单车的偏转角度是否超过设定角度阈值,若是,则判定为单车已经倒地,此时可以自动调用监控摄像头的本地GPS定位信息,由于监控摄像头的监控范围有限,能被某一监控摄像头拍摄到说明倒地单车就在该监控摄像头附近一定区域,因此把监控摄像头的本地GPS定位信息默认为倒地单 车的大致位置,并将所述本地GPS定位信息编辑到第一告警提示信息中进行输出显示,以通知相应的工作人员有单车倒地需要处理。
采用本发明的方案,首先相比于人工检测的效率更高,能第一时间发现倒地单车并进行通知;此外,相比于在单车上增加硬件传感器以检测单车是否倒地的方案来说,一方面不会增加单车的硬件成本,而且后期维护成本相比于硬件传感器的方案来说也更低;另一方面,从检测的准确性来讲,本方案通过子识别码图像的校验来提升检测准确率,经测算能达到99.9%以上的检测准确率,相比硬件传感器的方案更具有优势。
在其中一个实施例中,若步骤S4中判断出不存在与标准距离系数的比值小于设定阈值的距离,则可以重新挑选一个子识别码图像作为更新后的基准图像,并返回步骤S2中计算所述其他所有子识别码图像中心点与该更新后的基准图像中心点之间的距离的步骤,直到满足条件为止。
在其中一个实施例中,在判断得出存在与标准距离系数的比值小于设定阈值的距离之后、将该距离所对应的子识别码图像与预先存储的原始识别码图像进行对比之前,还可以包括如下步骤:若所述与标准距离系数的比值小于设定阈值的距离的个数多于1个,则对满足此条件下的各子识别码图像对应的距离按从小到大的顺序进行排序,并选择排序第一位的子识别码图像来执行与原始识别码图像进行对比的步骤。
在其中一个实施例中,所述单车识别码中还可以包括方向信息。
在其中一个实施例中,在从视频图像中提取单车上的单车识别码图像之前,还可以包括如下步骤:
将所述视频图像输入到预先建立的单车模型中进行匹配,提取所述视频图像中的物体特征进行相似度计算;
当相似度计算的结果超过设定特征阈值时,判定为检测到视频图像中存在单车,并提取出与所述物体特征相近的纹理进行保存;否则判定为视频图像中不存在单车,并返回摄像头实时拍摄视频图像的步骤。
本发明实施例中,事先利用单车的几何模型或结构来表示单车的结构和形 状,并通过提取单车的物体特征,在模型和图像之间建立起对应关系,然后通过几何方法来进行单车识别,识别效率更高、更准确。
在其中一个实施例中,在提取出单车上的单车识别码图像之后,还可以包括如下步骤:
从所述单车的车把上贴着的单车识别码图像中提取出第一关键特征点;以及从所述单车的坐杆上贴着的单车识别码图像中提取出第二关键特征点;
计算所述第一关键特征点和第二关键特征点之间的距离;
判断所述距离是否在预设的距离阈值内,若否,则输出第二告警提示信息,以提示相应的工作人员所述单车识别码出现故障。
因为考虑到单车识别码可能会被人为损坏的情况出现,为了确保本发明方案计算的准确性,当上述实施中计算出来的第一关键特征点和第二关键特征点之间的距离不满足预设的距离阈值时,说明有可能单车识别码出现了故障,例如丢失或损坏了等,此时进行告警提示,通知相应的工作人员更换单车识别码。
通过以上方案可以看出,本发明的一种共享单车的倒地检测方法,预先在单车上贴有单车识别码,然后通过遍布各处的监控摄像头来获取到单车识别码图像,并计算各子识别码图像之间的距离以及标准距离系数,选取与标准距离系数的比值小于设定阈值的距离对应的子识别码图像来与预先存储的原始识别码图像进行对比,计算出子识别码图像的偏移角度和偏移方向,进而计算出整个单车的偏转角度,以此识别出单车是否倒地,如果是则自动输出告警提示信息到相应的工作人员进行提醒。相比于现有的技术,本发明的方案完全采用软件算法实现,不需要增加额外的硬件传感器,避免了人工检测的盲目性查找,能第一时间发现并提醒哪里有单车倒地,实现成本低,且准确度高,极大提高了单车倒地检测的效率,有利于共享单车在社会上的推广使用,方便市民的绿色出行。

Claims (6)

  1. 一种共享单车的倒地检测方法,其特征在于,包括如下步骤:
    获取监控摄像头实时拍摄到的视频图像,并从视频图像中提取单车上的单车识别码图像;所述单车识别码图像为预先贴在单车特定部位的图像,且该图像中包括若干个彼此独立且属性值各不相同的子识别码图像;
    对所述单车识别码图像进行图像分割处理,获得各子识别码图像,并任意挑选一个子识别码图像为基准图像,计算其他所有子识别码图像中心点与该基准图像中心点之间的距离,采用如下公式:
    Figure PCTCN2017077195-appb-100001
    式中,
    Figure PCTCN2017077195-appb-100002
    表示所述距离,
    Figure PCTCN2017077195-appb-100003
    表示基准图像的向量,
    Figure PCTCN2017077195-appb-100004
    表示其他子识别码图像的向量,m表示所有子识别码图像的数量总和;
    计算各子识别码图像的属性值总量,并根据所述属性值总量计算标准距离系数,采用如下公式:
    Figure PCTCN2017077195-appb-100005
    式中,α表示子识别码图像均值,X表示标准距离系数,N表示属性值总量;
    判断是否存在与标准距离系数的比值小于设定阈值的距离,若是,则将该距离所对应的子识别码图像与预先存储的原始识别码图像进行对比,确定所述子识别码图像相对于所述原始识别码图像的偏移角度和偏移方向;
    根据所述偏移角度和偏移方向计算单车的偏转角度,具体为:根据所述偏移角度α和偏移方向(fx,fy,fz)计算旋转系数H,并根据所述旋转系数H计算单车相当于基准角度的偏转角度β;采用如下公式:
    Figure PCTCN2017077195-appb-100006
    式中,I是3x3的单位矩阵,
    Figure PCTCN2017077195-appb-100007
    vf=(0,0,1);
    判断所述单车的偏转角度是否超过设定角度阈值,若是,则判定为单车已 经倒地,此时自动获取本地GPS定位信息,并将所述本地GPS定位信息编辑到第一告警提示信息中进行输出显示,以通知相应的工作人员有单车倒地需要处理。
  2. 根据权利要求1所述的共享单车的倒地检测方法,其特征在于,若不存在与标准距离系数的比值小于设定阈值的距离,则重新挑选一个子识别码图像作为更新后的基准图像,并返回计算所述其他所有子识别码图像中心点与该更新后的基准图像中心点之间的距离的步骤。
  3. 根据权利要求1所述的共享单车的倒地检测方法,其特征在于,在判断得出存在与标准距离系数的比值小于设定阈值的距离之后、将该距离所对应的子识别码图像与预先存储的原始识别码图像进行对比之前,还包括步骤:若所述与标准距离系数的比值小于设定阈值的距离的个数多于1个,则对满足此条件下的各子识别码图像对应的距离按从小到大的顺序进行排序,并选择排序第一位的子识别码图像来执行与原始识别码图像进行对比的步骤。
  4. 根据权利要求1所述的共享单车的倒地检测方法,其特征在于,所述单车识别码中还包括方向信息。
  5. 根据权利要求1-4任意一项所述的共享单车的倒地检测方法,其特征在于,在从视频图像中提取单车上的单车识别码图像之前,还包括如下步骤:
    将所述视频图像输入到预先建立的单车模型中进行匹配,提取所述视频图像中的物体特征进行相似度计算;
    当相似度计算的结果超过设定特征阈值时,判定为检测到视频图像中存在单车,并提取出与所述物体特征相近的纹理进行保存;否则判定为视频图像中不存在单车,并返回摄像头实时拍摄视频图像的步骤。
  6. 根据权利要求5所述的共享单车的倒地检测方法,其特征在于,在提取出单车上的单车识别码图像之后,还包括如下步骤:
    从所述单车的车把上贴着的单车识别码图像中提取出第一关键特征点;以及从所述单车的坐杆上贴着的单车识别码图像中提取出第二关键特征点;
    计算所述第一关键特征点和第二关键特征点之间的距离;
    判断所述距离是否在预设的距离阈值内,若否,则输出第二告警提示信息,以提示相应的工作人员所述单车识别码出现故障。
PCT/CN2017/077195 2016-05-09 2017-03-18 共享单车的倒地检测方法 WO2017193701A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201780001007.7A CN107851325A (zh) 2016-05-09 2017-03-18 共享单车的倒地检测方法

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610300168.3A CN105840284A (zh) 2016-05-09 2016-05-09 处理尾气中一氧化碳的方法及系统
CN201610300168.3 2016-05-09

Publications (1)

Publication Number Publication Date
WO2017193701A1 true WO2017193701A1 (zh) 2017-11-16

Family

ID=56591314

Family Applications (2)

Application Number Title Priority Date Filing Date
PCT/CN2017/074042 WO2017193665A1 (zh) 2016-05-09 2017-02-20 处理尾气中一氧化碳的方法
PCT/CN2017/077195 WO2017193701A1 (zh) 2016-05-09 2017-03-18 共享单车的倒地检测方法

Family Applications Before (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/074042 WO2017193665A1 (zh) 2016-05-09 2017-02-20 处理尾气中一氧化碳的方法

Country Status (2)

Country Link
CN (2) CN105840284A (zh)
WO (2) WO2017193665A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113823040A (zh) * 2020-06-19 2021-12-21 中国移动通信集团福建有限公司 共享单车的管理方法及系统

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105840284A (zh) * 2016-05-09 2016-08-10 黄安武 处理尾气中一氧化碳的方法及系统
CN106014569A (zh) * 2016-05-09 2016-10-12 饶川辉 汽车尾气中的一氧化碳处理方法及系统
CN106014570A (zh) * 2016-05-09 2016-10-12 黄安武 处理汽车尾气中一氧化碳的方法及系统
CN111162931B (zh) * 2019-12-12 2022-09-16 上海钧正网络科技有限公司 共享车辆管理方法、装置、计算机设备和可读存储介质
CN115223092B (zh) * 2022-07-15 2023-11-14 广东万龙科技有限公司 一种大数据场景下的视频监控系统及方法
CN115601271B (zh) * 2022-11-29 2023-03-24 上海仙工智能科技有限公司 一种视觉信息防抖方法、仓储库位状态管理方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007269223A (ja) * 2006-03-31 2007-10-18 Nissin Kogyo Co Ltd 自動2輪車用システム
CN201856757U (zh) * 2010-10-18 2011-06-08 长安大学 一种惯性轮式二轮机器人稳定装置
CN204895677U (zh) * 2015-07-14 2015-12-23 于洪涛 一种电动自行车的倾倒报警装置
US20170021827A1 (en) * 2015-07-23 2017-01-26 Ford Global Technologies, Llc Bicycle stability control methods and systems
CN106447733A (zh) * 2016-09-28 2017-02-22 北京理工大学 颈椎活动度及活动轴线位置的确定方法、系统及装置

Family Cites Families (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0792684A3 (en) * 1996-02-29 1997-11-12 Mitsubishi Gas Chemical Company, Inc. Novel adsorbent for carbon monoxide and method
JP2001248426A (ja) * 2000-03-03 2001-09-14 Hideaki Tanaka 自動車排気ガス浄化器と浄化法
US9311540B2 (en) * 2003-12-12 2016-04-12 Careview Communications, Inc. System and method for predicting patient falls
JP2006169997A (ja) * 2004-12-14 2006-06-29 Nissan Motor Co Ltd 触媒の劣化判定装置
CN101785957B (zh) * 2010-02-10 2012-08-08 毛恒松 二氧化碳的分离和储存方法
CN101832167B (zh) * 2010-05-07 2012-07-11 东风汽车有限公司 一种scr催化器中的氨气吸附控制方法
CN102358364B (zh) * 2011-09-02 2014-12-31 深圳市宇恒互动科技开发有限公司 自行车骑行辅助装置、用该辅助装置的自行车及辅助方法
EP2618305B1 (de) * 2012-01-20 2014-04-30 ESG Elektroniksystem- und Logistik-GmbH Verfahren und Vorrichtung zur Online-Kalibrierung von Fahrzeug-Kameras
CN203240599U (zh) * 2013-04-26 2013-10-16 德阳市科瑞仪器设备厂 一种尾气处理装置
CN103604750A (zh) * 2013-11-25 2014-02-26 无锡俊达测试技术服务有限公司 一种红外汽车尾气检测报警装置
CN104239851B (zh) * 2014-07-25 2018-05-01 重庆科技学院 基于行为分析的智能小区巡检系统及其控制方法
CN105572166B (zh) * 2014-10-09 2019-01-15 通用电气公司 采用了基于催化剂的一氧化碳传感器的系统以及防止该传感器被硫和积炭损害的方法
CN105185050A (zh) * 2015-06-29 2015-12-23 柳州蚊敌香业有限公司 一种车载废气浓度监测报警装置
CN105319178B (zh) * 2015-10-27 2018-05-04 中国科学院合肥物质科学研究院 机动车尾气co和co2浓度实时检测系统及其控制方法
CN105822389A (zh) * 2016-05-09 2016-08-03 黄安武 汽车尾气中的一氧化碳净化方法
CN106014570A (zh) * 2016-05-09 2016-10-12 黄安武 处理汽车尾气中一氧化碳的方法及系统
CN106014569A (zh) * 2016-05-09 2016-10-12 饶川辉 汽车尾气中的一氧化碳处理方法及系统
CN105840284A (zh) * 2016-05-09 2016-08-10 黄安武 处理尾气中一氧化碳的方法及系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007269223A (ja) * 2006-03-31 2007-10-18 Nissin Kogyo Co Ltd 自動2輪車用システム
CN201856757U (zh) * 2010-10-18 2011-06-08 长安大学 一种惯性轮式二轮机器人稳定装置
CN204895677U (zh) * 2015-07-14 2015-12-23 于洪涛 一种电动自行车的倾倒报警装置
US20170021827A1 (en) * 2015-07-23 2017-01-26 Ford Global Technologies, Llc Bicycle stability control methods and systems
CN106447733A (zh) * 2016-09-28 2017-02-22 北京理工大学 颈椎活动度及活动轴线位置的确定方法、系统及装置

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113823040A (zh) * 2020-06-19 2021-12-21 中国移动通信集团福建有限公司 共享单车的管理方法及系统

Also Published As

Publication number Publication date
CN105840284A (zh) 2016-08-10
WO2017193665A1 (zh) 2017-11-16
CN107851325A (zh) 2018-03-27

Similar Documents

Publication Publication Date Title
WO2017193701A1 (zh) 共享单车的倒地检测方法
CN110084095B (zh) 车道线检测方法、车道线检测装置和计算机存储介质
Akagic et al. Pavement crack detection using Otsu thresholding for image segmentation
CN110135269B (zh) 一种基于混合颜色模型与神经网络的火灾图像检测方法
JP6904346B2 (ja) 画像処理装置、画像処理システム、および画像処理方法、並びにプログラム
WO2017193679A1 (zh) 自动检测单车是否倒地的方法
US9754160B2 (en) Method and device for detecting gathering of objects based on stereo vision as well as non-transitory computer-readable medium
US7916944B2 (en) System and method for feature level foreground segmentation
CN104303193B (zh) 基于聚类的目标分类
CN109598794B (zh) 三维gis动态模型的构建方法
US8462992B2 (en) Method of change detection for building models
TWI425454B (zh) 行車路徑重建方法、系統及電腦程式產品
WO2017084094A1 (zh) 烟雾检测装置、方法以及图像处理设备
CN111179279A (zh) 一种基于紫外、双目视觉的综合火焰检测方法
CN111369495B (zh) 一种基于视频的全景图像的变化检测方法
CN105426863B (zh) 检测车道线的方法和装置
CN112270381B (zh) 基于深度学习的人流量检测方法
CN108960115A (zh) 基于角点的多方向文本检测方法
TW201140502A (en) System and method for monitoring objects
CN110490150A (zh) 一种基于车辆检索的违章图片自动审核系统及方法
TWI493510B (zh) 跌倒偵測方法
Tao et al. Smoky vehicle detection based on range filtering on three orthogonal planes and motion orientation histogram
JP2003141546A (ja) 画像処理方法
CN110992361A (zh) 基于代价平衡的发动机紧固件检测系统及检测方法
CN107832732B (zh) 基于三叉树遍历的车道线检测方法

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17795319

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 17795319

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 24.05.2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17795319

Country of ref document: EP

Kind code of ref document: A1