sdPoster
1
Motivation Actual CSU Email The following is a message from Cleveland State University on Feb 24 th “A robbery was reported today to the Cleveland Police Department at approximately 7:20 pm on East 24th Street between Euclid and Prospect Avenue. “ “A female CSU student had her cell phone taken from her while she was talking on it.” “She was pushed to the ground while the suspects attempted to unsuccessfully take her purse. “ “One suspect was wearing an orange hoodie. The second suspect was wearing a white or gray hoodie or jacket. “ “Both suspects ran east on Euclid Avenue. No further information is available at this time.“ Technology and Innovation Front-end Application Real-time Feedback Using the mobile application on police tablets (developed by our team), police can add suspects to a chase, initiate a chase, view camera feeds, and perform many other tasks which give them eyes in nearly every corner of the city. System Architecture – Software Solution Cerebro HelpKey - Hardware Solution The goal of Cerebro is to utilize what we have learned as STEM students to create a scalable web based security system that will first and foremost secure campuses, cities , and eventually the world. Cerebro Real-time Security Built on a Distributed Cloud Architecture, Powered by Computer Vision Machine Learning Algorithms Nick White Mark Heller Andrew Fisher Brahm Powell Titus Lungu Robert Marshall Solutions Cerebro Solutions Hardware HelpKey Software & Data CV Algorithm Applications User Holds Button API Request Made Cerebro Initiates GPS 1 2 3 Cerebro block diagram and overview of major components What is Cerebro? Overview Cerebro uses a innovative human detection and recognition algorithm, based on a Computer Vision Machine Learning algorithm, to detect humans in the video streams of hundreds or thousands of cameras within a specified area. Using this algorithm, crimes can be detected in real time. As a crime is committed, the suspect is tagged in the system, and they are then tracked from camera to camera as they attempt to flee. Their location is reported to police in real-time, and is displayed on tablets in police cruisers as both video and as a pin on a map interface. Using this, police can track the suspect, update information about the chase, and apprehend the suspect both quicker and more efficiently than ever. Acknowledgements Organizations Parker Hannifin Corporation (Dr. Joseph Kovach) Cleveland State University EECS/ME Departments Individuals Dr. Sunnie Sun Chung (CSU EECS Department) Dr. Pong P. Chu (CSU EECS Department) Dr. Majid Rashidi (CSU ME Department) Computer Vision Algorithm Our proprietary algorithm makes detecting and tracking unique humans both fast, efficient, and accurate: 1. Individuals are detected using a Histogram of Oriented Gradients and a pre-trained Support Vector Machine which identifies humans in video frames. 2. Certain key identifying features such as location, RGB pattern, and unique feature vectors are extracted for every detected person and stored in our cloud database. 3. In successive frames, the detected people are compared to the previously detected ones in order to find matches and track them over multiple cameras. 4. Using this data, individuals cannot escape from local police, as they are tracked from camera to camera – having their location reported in real-time. We plan to improve our algorithm by continuing to make use of state-of-the-art artificial intelligence techniques and by re- training the Support Vector Machine on a recently acquired dataset of over 3 million unique pictures of people.
-
Upload
titus-lungu -
Category
Documents
-
view
13 -
download
1
Transcript of sdPoster
MotivationActual CSU Email
The following is a message from ClevelandState University on Feb 24th
“A robbery was reported today to theCleveland Police Department atapproximately 7:20 pm on East 24th Streetbetween Euclid and Prospect Avenue. ““A female CSU student had her cell phonetaken from her while she was talking on it.”“She was pushed to the ground while thesuspects attempted to unsuccessfully takeher purse. ““One suspect was wearing an orangehoodie. The second suspect was wearing awhite or gray hoodie or jacket. ““Both suspects ran east on Euclid Avenue.No further information is available at thistime.“
Technology and Innovation
Front-end Application
Real-time FeedbackUsing the mobile application on police tablets (developed byour team), police can add suspects to a chase, initiate a chase,view camera feeds, and perform many other tasks which givethem eyes in nearly every corner of the city.
System Architecture – Software Solution
Cerebro HelpKey - Hardware Solution
The goal of Cerebro is to utilize what we have learned as STEM students to create a scalable web based security system that will first and foremost secure campuses, cities , and eventually the world.
Cerebro Real-time SecurityBuilt on a Distributed Cloud Architecture, Powered by Computer Vision Machine Learning Algorithms
Nick White Mark Heller
Andrew Fisher
Brahm Powell
Titus Lungu
Robert Marshall
Solutions
Cerebro Solutions
Hardware
HelpKey
Software & DataCV Algorithm Applications
User Holds Button
API Request Made
Cerebro Initiates GPS
1
2
3
Cerebro block diagram and overview of major components
What is Cerebro?Overview
Cerebro uses a innovative human detection andrecognition algorithm, based on a Computer VisionMachine Learning algorithm, to detect humans in thevideo streams of hundreds or thousands of cameraswithin a specified area. Using this algorithm, crimes canbe detected in real time. As a crime is committed, thesuspect is tagged in the system, and they are thentracked from camera to camera as they attempt to flee.Their location is reported to police in real-time, and isdisplayed on tablets in police cruisers as both video andas a pin on a map interface. Using this, police can trackthe suspect, update information about the chase, andapprehend the suspect both quicker and moreefficiently than ever.
AcknowledgementsOrganizations Parker Hannifin Corporation (Dr. Joseph Kovach)
Cleveland State University EECS/ME Departments
Individuals Dr. Sunnie Sun Chung (CSU EECS Department)
Dr. Pong P. Chu (CSU EECS Department)
Dr. Majid Rashidi (CSU ME Department)
Computer Vision Algorithm
Our proprietary algorithm makes detecting and tracking uniquehumans both fast, efficient, and accurate:
1. Individuals are detected using a Histogram of OrientedGradients and a pre-trained Support Vector Machine whichidentifies humans in video frames.
2. Certain key identifying features such as location, RGBpattern, and unique feature vectors are extracted for everydetected person and stored in our cloud database.
3. In successive frames, the detected people are compared tothe previously detected ones in order to find matches andtrack them over multiple cameras.
4. Using this data, individuals cannot escape from local police,as they are tracked from camera to camera – having theirlocation reported in real-time.
We plan to improve our algorithm bycontinuing to make use of state-of-the-artartificial intelligence techniques and by re-training the Support Vector Machine on arecently acquired dataset of over 3 millionunique pictures of people.
