SPY I SMART PREDICTIVE POLICING

Preview: In previous years, our country has seen a fair share of crimes in public as well as in private sectors. From the year 2013, there has been a reasonable yet an insufficient decline in these criminal activities, so a monitoring system which can detect any sus

Project Title

SPY I SMART PREDICTIVE POLICING

Project Area of Specialization

Artificial Intelligence

Project Summary

Preview:

In previous years, our country has seen a fair share of crimes in public as well as in private sectors. From the year 2013, there has been a reasonable yet an insufficient decline in these criminal activities, so a monitoring system which can detect any suspicious individual or group of individuals during busiest hours of our daily routine may be the answer to the question that our authorities have been seeking for a quite a long time.

The Idea:

In order to prevent any potential future criminal activity, we have pitched the idea of a surveillance system that does not require any man power but merely a small fitted hardware namely, “SPY I: Smart Predictive Policing”. Given how criminal activities in larger crowds have exponentially increased in our country, such a system that can potentially detect any future happenings can come in handy under numerous situations. Another effective application of this project is to differentiate a dubious individual or individuals among a jam-packed crowd.

Previously, Crowd Monitoring systems have been able to detect an individual’s suspiciousness using algorithms of Machine Learning, but they were unable to do so among a large group of people. Our project is a prime candidate for all authorities, corporations/organizations and federations.

Existing Crowd Monitoring techniques heavily focus on analyzing the crowd as a single entity usually in terms of their movement patterns. These shortcomings are to be compensated in the making of our project. In this work of ours we have proposed a novel Crowd Monitoring system that uses Facial Expression Recognition (FER). By isolating different emotions in a crowd, we aim to predict the overall emotion of a crowd and the anomalies which lie within it.

Additional Features:

In addition to violence and suspicious activity detection, some additional features of the project include:

Crowd Density Estimation:

Crowd Density Estimation could help build an environment more conducive to pedestrian safety.

Race Detection:

The detection of people of different races such as Asian, African, American and British.

Overhead Crowd Estimation:

An approximate count of the crowd using a spider cam like view to generate an estime of how many people are present in a crowd.

Crowd Flow and Behavioural Prediction:

Imitation of the decision making skills of a crowd and violent flow prediction.

Project Objectives

Objectives:

We shall go through the objectives in accordance with each od the applications of the project.

1) To have a potential decrease in crime activities:

.The detction of violent and suspicious individuals in a crowd is the first step towards the prevention of a potential street crime, theft or even a terrorist activity.

2) To build a condusive pedestrian safety environment:

Crowd Density Estimation could help build an environment more conducive to pedestrian safety. Crowd data such as density is an important factor in the planning, design and management of public facilities subject to dense pedestrian traffic.

3) To give behavioral prediction of a suspicious individual:

There can be many applications where the imitation of the decision making skills of a person can come in handy such as psychological tests, mental and street awareness and possible research works on how a human mind makes decisions particulrly when some suspicious activity is in their mind.

4) To accurately predict an individul's race/nationality:

Crimes by short-term migrants, such as tourists, exchange students and transient workers, are counted as crimes by immigrants or foreigners, and gives the impression that a higher share of the migrant population commits crimes. There has been a fair share of these crimes lately and for this purpose race detection can be a pivotal tool.

5) To give an approximate count of a large crowd:

An approximate count of the number of individuals in a large crowd is an important aspect of crowd analysis and has been a challenging problem for automatic visual surveillance over many years.

6) To provide a smart security system to our country:

Our beloved country in recent years has faced many terrorist and criminal activities and has had a fare share of dealing with these things. The most important aim of this project is to build a smart security system which could serve the nation in terms of defence, better life style and share a positive and secyure image of the country to the outside world.

Project Implementation Method

Methodology:

The basic methodology behid the project is as follows:

Project Design:

It cosist of two main parts:

• Software Design
• Hardware Design

Software Implementation:

The software implementation consists of the following steps:

Procedure and Analysis

• Real time capturing images from our designed hardware.
• Pass them through openCV algorithms
• Detect the faces in the crowd using Viola and Jones Algorithm
• Extraction of features using Gradient Local Ternary Pattern (GLTP) and Deep Belief Networks (DBN)
• Estimating overall emotion of a crowd using minimum spanning tree and chains of emotions.
• Isolation of a suspicious Individual.
• Using face textures to detect individuals of a particular race.
• Deep Learning Models for overhead crowd counting.

Data Sets

 We shall analyze the following data sets in our research:

• Japanese female facial expressions (JAFFE)
• Extended Cohn Kanade (CK+)

Viola and Jones algorithm will differentiate objects and faces from a specific image.

Gradient Local Ternary Pattern will encode the local texture of an image by computing the gradient magnitudes of local neighborhoods in the image and quantizing the values into three discrimination levels.

Minimum Spanning Tree will represent each face’s closest neighbor and predict overall emotion of the crowd

Support Vector machines will take the variables from our final image as an input and encode output according to that specific emotion.

Hardware Implementation:

The hardware design would require certain specific smart devices/Microcontrollers for their proper functioning.

The details of these are as follows:

NVIDIA Jetson Nano developer kit:

VIDIA Jetson Nano developer kit is a low-cost AI computer. It delivers the compute performance to run modern AI workloads at unprecedented size. It is incredibly power-efficient, consuming as little as 5 watts. It is supported by NVIDIA Jetpack, used across the entire NVIDIA Jetson family of products, reducing complexity and overall effort for developers, learners, and makers. NVIDIA Jetson Nano Developer Kit delivers the compute performance to run modern AI workloads at unprecedented size, power, and cost. The developer kit can be powered by micro-USB and comes with extensive I/Os, ranging from GPIO to CSI. This makes it simple for developers to connect a diverse set of new sensors to enable a variety of AI applications.This proven software stack reduces complexity and overall effort for developers.

GPU:

MSI Geforce GTX 1660 Armor 6G OC Video Graphics Card 6GB GDDR5 for fast processing of the video.

Final Product:

A security surveillance camera for smart predictive policing.

Benefits of the Project

A Smart Security System:

In computer vision crowd surveillance in an emerging interest largely bond from the desire to monitor the nature of a group of individuals in packed areas where conventional methods of image processing would not suffice. In this project we proposed a real time crowd monitoring system that can be deployed in many areas including sport stadiums, airport and public transport terminals. Due to the difficulty associated with extracting individuals from a crowd, these monitoring systems rely on many different methods including Holistic and Object Level based method of Crowd Monitoring but a more intriguing way of achieving precise results is by the method of Emotion Estimation of that group of Individuals.

Crowd Monitoring through AI:

People Monitoring is an active area of research and has been developing since the emergence of deep learning, as you can imagine limited surveillance technologies can not devise a simpler way of detecting possible threats and is prone to errors. Thankfully a wide range of technological solutions for crowd monitoring is now available utilizing a range of cameras and modern algorithms. We will have to rely on installing our software into existing CCTV systems because current CCTV systems suffer from reduce accuracy due to a lack of depth perception and interference from outside effects. Installations will be typically aimed at security concerns and will be positioned in the optimal way for detecting a doubtable individual and differentiating the rest of the crowd from him/her.

Summary of Benefits:

Using our product, we can achieve an efficient solution to the problem that crowd monitoring typically suffers from. The benefits we shall gain from it will be

• A decline in the number of terrorist activities.
• Contribution to the national safety.
• Contribution to the social and economical betterment of the country.

Crowd Detection uses Facial Expression Recognition (FER) to estimate emotions of a crowd as a single entity as well as on individual basis under both panic and non-panic conditions.

Isolation of a Dubious Individual among a crowd can be achieved by comparing the deviation of that person’s emotions from the rest of the crowd.

Race Detection to solve many problems that tourism industry faces in case of an unwanted or unwilling happening.

Technical Details of Final Deliverable

Technical Details:

Software Details:

The technical details involving the algorithms and the technology required for each of the steps and softwares is as follows:

1. Violence/Non Violence Detection:

Prediction of violent/non-violent activities using statistics of how flow-vector magnitudes change over time. These statistics, collected for short frame sequences, are represented using the Violent Flows (ViF) descriptor. ViF descriptors are then classified as either violent or non-violent using linear SVM.

2. Crowd Density Estimation

Crowd Density Estimation could help build an environment more conducive to pedestrian safety. To estimate the crowd density, crowd features need to be designed first and then a classifier needs to be trained to discriminate the crowd density.

3. Behavioral Prediction:

Imitation of the decision making skills of a person using deep convolutional networks to figure out human intention apart from human-human interaction can be done. The behavioral prediction will be done by anticipating the future path of people.

4. Race Detection:

Race Recognition Framework (RRF) is proposed that includes information collector (IC), face detection and preprocessing (FD&P), and RR modules. For the RR module, this study proposes two independent models. The first model is RR using a deep convolutional neural network (CNN) (the RR-CNN model). The second model (the RR-VGG model) is a fine-tuning model for RR based on VGG, the famous trained model for object recognition.

5. Crowd Counting:

By using a combination of deep and shallow, fully convolutional neural networks. This feature helps in capturing both the low-level and high-level features. The dataset is augmented to learn scale-invariant representations. The deep network is similar to the well-known VGG-16 network. It captures the high-level semantics needed for crowd counting and returns the density maps.

Hardware Details:

A SPY Cam that uses interfacing via NVIDIA Jetson Nano Development kit and differentiates suspicious activities and individuals in a crowd and performes all the software activities that have been described above in the software section. A Graphics Processing Unit for the purpose of increasing the real time video capturing at a higher resolution and a more clear imaging to perform these desired tasks will also be used. 

As a whole, this project involves a very high understanding of Deep Convolutional Networks, Microprocessor Systems, Image Processing and Embedded Systems.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Security

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Industry, Innovation and Infrastructure, Sustainable Cities and Communities

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com