Real-time recursive filter to identify weather events using traffic CCTV video
专利汇可以提供Real-time recursive filter to identify weather events using traffic CCTV video专利检索，专利查询，专利分析的服务。并且A program is implemented on a processor for executing the following steps: receiving images of a scene comprising successive frames from a camera, each frame including an array of pixels; filtering intensities of the array of pixels in a respective frame by using a Bessel Filter to compute a constant value for the respective frame; and summing a predetermined number of respective frames from the camera to determine rapid motion in the scene. The filter includes a Bessel filter, which is an n-pole recursive digital filter, where n is a positive integer.，下面是Real-time recursive filter to identify weather events using traffic CCTV video专利的具体信息内容。
What is claimed:
The present invention relates, in general, to monitoring and tracking extreme weather events. More specifically, the present invention relates to using a real-time recursive filter to identify extreme weather events in video footage obtained from closed circuit television (CCTV).
Forecasters use a combination of science and technology to predict the state of the atmosphere for a given location. A mix of radar, satellite, forecast models and observational data are all used to assist forecasters. Historically, weather observation networks focus on atmospheric conditions and ignore the area below the horizon where people live and drive. Thus, most weather data is primarily tailored to aviation, and not ground traffic. An example of this is forecasting a tornado for a specific area. Forecasting that atmospheric conditions are favorable for this event is possible; however, confirming this event is not possible from current weather forecasting technology. Forecasters can predict specific atmospheric conditions which are consistent with a specific weather event, but unfortunately cannot verify that the event is actually taking place, or will take place. Confirmation is left to local storm-chasers, or amateur footage to confirm the event by visual means.
Current predictive techniques for weather patterns are generally performed on a national, or multi-state regional scale. This large scale weather prediction is largely insufficient for predicting the location of specific weather conditions with enough precision to ultimately assist Emergency Response Management personnel. Prior art weather monitoring stations are generally spaced 100-200 miles apart due to expense and complexity. The wide spacing and skyward focus of these monitoring stations largely prevents them from monitoring ground conditions, and provides weather data accurate enough only to predict general weather patterns.
In order to mitigate the inability to verify local weather events, the industry has begun developing observation systems focusing on surface/ground observation. Some of these systems are able to view weather conditions using a CCTV camera system. The National Mesonet Program Alliance is one program developed specifically for ground-based observations. Another example of emerging ground-based detection systems is lightning detection, which enables real-time verification of lightning strikes.
State and local governments operate ground-based observation systems which include CCTV capabilities using weather information for winter road maintenance and traffic control. Local transportation departments usually collect road and weather data from two or more sources, such as the National Weather Service (NWS) and Road Weather Information Systems (RWIS), and fuse these data to generate information for winter road maintenance. Traffic CCTV cameras are used primarily for traffic control, although some are used to assess weather conditions.
Mesonet systems offer a significant advance in small-scale weather data. These systems, however, do not provide a coordinated visual observation capability for tracking weather events. Much of the weather data focuses on air temperature, dew point, wind speed, and data which may be augmented to atmospheric data.
Very few of any of the above mentioned systems have the ability to archive visual observations which could be used for forensics and research forecasting efforts.
Finally, the above approaches fail to exploit an inherent benefit of visually being able to observe an event in real-time, which is a critical aspect of providing validation to atmospheric models and forecasts.
To meet this and other needs, and in view of its purposes, the present invention provides a program implemented on a processor for executing the following steps: (a) receiving images of a scene comprising successive frames from a camera, each frame including an array of pixels; (b) filtering intensities of the array of pixels in a respective frame by using a Bessel Filter to compute a constant value for the respective frame, and (c) summing a predetermined number of respective frames from the camera to determine rapid motion in the scene.
The filtering step includes using a Bessel filter, which is an n-pole recursive digital filter, where n is a positive integer. The Bessel filter is an 8-pole recursive digital filter.
Each frame is defined by a 2-dimensional tensor of Ii,j(t), where i and j are a row and column, respectively, in the array of pixels, and I are the intensities in the array of pixels at time t.
The scene is captured by a closed circuit television (CCTV) camera, and the rapid motion in the scene is a weather-related occurrence. The weather-related occurrence includes motion of a rain fall, a hail storm, a snow storm, or a tornado.
The Bessel filter is defined by a ratio between (a) an (n−1) order polynomial of a complex number and (b) an n order polynomial of a complex number, where n is a positive number.
The filtering is implemented using the following recursive relationship:
- where:
- Ii,j,k are pixel intensities of a kth frame captured by the camera,
- I′i,j,k are filtered pixel intensities of the kth frame,
- the kth frame is defined as a present frame,
- (k-1) to (k-8) are successive previous frames, and
- Γ0-Γ15 are constants.
- where:
The program may include the following step: (d) masking a first portion of the scene; and the step of filtering includes filtering intensities of the array of pixels in a second portion of the scene. The masking of the first portion of the scene is effective in removing motion of a non-weather-related occurrence in the scene.
Summing the predetermined number of respective frames includes summing eight successive frames from the camera.
Another embodiment of the present invention is a system for extracting weather-related events from a camera. The system includes a closed circuit television (CCTV) camera for monitoring a scene, and a filter for low pass filtering successive frames of the scene captured by the CCTV camera. The filter includes a Bessel function for computing constant values for the successive frames of the scene, and assigning a computed constant value to each of the successive frames, wherein a respective computed constant value multiplied by the intensities of a respective frame is defined as a filtered frame of the scene. A summer is included for summing a predetermined number of filtered frames of the scene, wherein the summer is effective in determining rapid motion in the scene. The filter includes a Bessel filter, which is an n-pole recursive digital filter, where n is a positive integer. The Bessel filter is defined by a ratio between (a) an (n−1) order polynomial of a complex number and (b) an n order polynomial of a complex number, where n is a positive number.
Yet another embodiment of the present invention is a method of extracting weather-related events from a camera. The method includes:
receiving successive frames captured by a closed circuit television (CCTV) camera of a scene,
filtering the successive frames of the scene captured by the CCTV camera by using a Bessel function for computing constant values for the successive frames of the scene,
assigning a computed constant value to each of the successive frames,
multiplying a respective computed constant value by intensities of the respective frame, wherein a result of the multiplying is defined as a filtered frame,
summing a predetermined number of filtered frames of the scene, and
determining rapid motion from the filtered frames of the scene.
The rapid motion in the scene is a weather-related event, including motion of a rain fall, a hail storm, a snow storm, or a tornado.
It is understood that the foregoing general description and the following detailed description are exemplary, but are not restrictive, of the invention.
The invention may be understood from the following detailed description when read in connection with the accompanying figures:
As will be explained, the present invention receives images from multiple cameras, such as closed circuit television (CCTV) cameras, and uses a recursive filter with an infinite frequency response to run in real-time on each CCTV image pixel in a frame. The recursive filter is executed in the time domain, rather than in the spatial domain. The recursive filter is effectively a low pass filter for identifying, in real-time, rapid changes in a scene being viewed by a CCTV camera.
The present invention examines scenes obtained from the cameras where typically no rapid changes are expected to occur. For example, the scenes may be that of a portion of the sky, a road intersection, or an underpass below a bridge. Since the filter operates very fast, the filter may raise a flag or an alert when a change occurs in an unexpected area. Once the flag or the alert is raised, a human observer, or a more complex computer analysis may be used to diagnose the type of event that is occurring, such as rain fall, hail storm, snow storm, tornado, or even forest fire.
Referring now to
It will be appreciated that although multiple CCTV cameras are shown in
Moreover, although one processor including a single filter 13 are shown in
In another embodiment of the present invention,
While satellite/aerial imaging depicts weather events from an atmospheric perspective, system 20 takes advantage of ground-level observations using existing terrestrial camera networks. Severe weather warning reports and other related weather event data (radar) may be used to geographically track local weather events and point camera network assets on a weather event from multiple vantage points for observation.
As shown, system 20 receives data from multiple sources, including third party weather data 21, ground-based environmental data 22, weather radar feeds 23, National Weather Service storm warning feeds 24, and CCTV feeds 25 from known locations. All inputted data are ingested and analyzed by module 26, with the aid of camera data and geospatial archive 27. A mapping and display interface 28 is available to visualize the weather events and enable a user or multiple users to control and enter data, by way of a mouse or keyboard (for example), generally shown as user control module 29.
The recursive digital filter 13, shown in
When designing filters, it is convenient to transfer Ii,j(t) to the Laplace domain to become Ii,j(s), where s is a complex frequency variable. A Bessel filter F(s) may be designed within the Laplace domain to low pass filter a signal Ii,j(s) using only multiplication as in Equation 1. The filter itself may be found from the reciprocal of a Bessel polynomial of order ‘n’. An 8th order filter is shown in Equations 2 and 3.
Ii,j′(s)=F(s)×Ii,j(s) (1)
Since the Bessel polynomial has 8 complex roots of ν1→ν8, the filter may be written as the product of eight fractions, as shown in Equation 4. These are then converted into the partial fractions shown in Equation 5, which has the form that may be inverse Laplace transformed directly back to the time domain, as shown in Equation 6 (where
F(t)=Σp=18peν
Equation 7 is a rewrite of Equation 6 in the digital time domain for video frame k (where typically Δt=⅕ seconds). To create the recursive form of a filter with this time response, it is necessary to take the Uni-lateral Z transform, as shown in Equation 8 (where z is a general complex number).
Fk=Σp=18peν
F(z)=Σk=0∞Fk×z−k (8)
Since the Z transform is simply 8 infinite geometric summations with common factors eν
Similar to the Laplace domain, the filtered result S(z) in the Z domain may be found from the product of the signal S(z) and the filter impulse response F(z). This is shown in Equation 10. Expansion of the representation of F(z) in Equation 9 shows it to be a ratio between a 7th and an 8th order polynomial of z (as shown in Equation 11).
Advantage is then taken of the property of the Z transform shown in Equation 12. Re-arranging Equation 11 before performing an inverse Z transform to return to the digital time domain, where t=tk=kΔt (i.e. for frame k), provides the recursive relationship of Equation 13.
F(z)×z−p=Σk=0∞Fk-p×z−k (12)
From Equation 13, it will now be apparent that the filtering process includes a summation of successive frames from the incoming original video from any CCTV camera. In addition, each new filtered frame includes a summation of eight successive original frames and eight successive filtered frames. The constants, Γ0-Γ15, are computed using recursive computations of the Bessel Filter, shown above. This iterative process is depicted in
An example of the effectiveness of this filter is shown in
Due to its speed and simplicity, this recursive filter may be applied in real-time to all CCTV camera videos available to a weather detection and identification system.
Suitable masks for each CCTV frame may be used so that traffic motion may be ignored, while motion in the sky, or under an overpass may alert a computer or observer to a potential extreme weather event. In this manner, more observational/computational resources may be applied to further diagnose the event. A masking module is shown included, as an option, in processor 12, which is depicted in
Two additional examples of the effectiveness of the filter are shown in
Although the invention is illustrated and described herein with reference to specific embodiments, the invention is not intended to be limited to the details shown. Rather, various modifications may be made in the details within the scope and range of equivalents of the claims and without departing from the invention.
| 标题 | 发布/更新时间 | 阅读量 |
|---|---|---|
| 基于深度摄像和身高阈值像素标定的自动身高检测方法 | 2020-05-08 | 640 |
| 一种圆柱玻璃表面质量视觉检测系统与方法 | 2020-05-11 | 717 |
| 一种能自动筛选不合格品的轴承端面检测装置 | 2020-05-11 | 879 |
| 一种电子产品检测回收方法 | 2020-05-08 | 439 |
| 一种苹果内外品质在线无损检测方法 | 2020-05-08 | 532 |
| 基于ROS的灾后救援通道检测机器人 | 2020-05-11 | 297 |
| 一种使用残差注意力机制网络的同步定位与建图方法 | 2020-05-08 | 928 |
| 一种室外无人机的自主定位方法 | 2020-05-08 | 52 |
| 一种可调拍摄角度的工业图像采集装置 | 2020-05-08 | 795 |
| 一种弓网燃弧定点检测用的高速相机安装基座 | 2020-05-08 | 833 |
高效检索全球专利专利汇是专利免费检索,专利查询,专利分析-国家发明专利查询检索分析平台，是提供专利分析，专利查询，专利检索等数据服务功能的知识产权数据服务商。
我们的产品包含105个国家的1.26亿组数据，免费查、免费专利分析。
专利汇分析报告产品可以对行业情报数据进行梳理分析，涉及维度包括行业专利基本状况分析、地域分析、技术分析、发明人分析、申请人分析、专利权人分析、失效分析、核心专利分析、法律分析、研发重点分析、企业专利处境分析、技术处境分析、专利寿命分析、企业定位分析、引证分析等超过60个分析角度，系统通过AI智能系统对图表进行解读，只需1分钟，一键生成行业专利分析报告。
