Dr Danish Arif

Links:

https://pk.linkedin.com/in/danish-arif-113456117

https://pu.edu.pk/faculty/description/1399/Engr-Dr-Danish-Arif.html

PhD Thesis:

Person Of Interest Recognition and Optical-based Tracking (POIROT) Intelligence Assistance System

  thesis by: Danish Arif

  Supervised by: Prof. Simon Winberg

Abstract

Pakistan has faced a multitude of security challenges over the past years due to its proximity to conflict zones such as Afghanistan. Pakistan has become one of the world's largest refugee-hosting countries, which has created financial and security problems for the country. These challenges have also led to the emergence of new terror groups and increased street crime in cities. In response to these challenges, the government has launched a project called "Punjab Safe Cities" that aims to increase surveillance capabilities through the installation of cameras. However, the current system has limitations and requires manual monitoring, which is not an efficient use of resources.

To address these limitations, an automation solution that uses artificial intelligence to track high-value targets has been proposed. The system is designed to process visual data feeds obtained from cameras installed frequently traversed locations. The first tier of the system includes a camera connected to a low-powered embedded platform that supports adequate computational complexity, to generate alerts for potentially recognized face and forms of potential suspects, so as to assist human detection. The next stage picks out faces more accurately from images that are sent to a central system that stores the data in a cloud-based repository. The system uses a highly secure, dedicated optical fibre network to transmit data. The main computing and detection of high-value targets are performed at the data centre, or at satellite security hubs, using a high-performance computer cluster equipped with GPU(s) to support parallel processing.

The system uses deep learning techniques to classify individuals by gender, ethnicity, and style of facial hair. Pakistan is located at the crossroads of the Middle East, Central Asia, and South Asia, which has led to a diverse population with distinct facial features that can be classified by the system. The cascaded classification system reduces the size of the search pool by detecting gender, face-type, and then facial hair type. Images of detected persons are stored in relevant databases hourly. When searching for a target image, only relevant databases are searched, reducing seek time. The system uses AlexNet, GoogLeNet, ResNet-50, and VGG-Face models trained with prepared datasets.

The system offers additional features to assist users, including generating hyper-real colour images from sketch-image input, transforming blurred or low-resolution images into high-resolution images, and approximating a frontal face image from the side pose of the target. The system uses a range of tools, including MATLAB, OCTAVE, Anaconda, Jupiter, and IP Video System Design Tool, for development and integration.

Once installed in a developing country like Pakistan, the system is expected to help decrease the crime rate and make cities safer to live. Smart-safe cities can attract foreign investors, boost economic growth, and increase business activities. Effective surveillance systems can also create a safe and secure environment for visitors, which can increase tourism-related income and assist the expansion of linked sectors like hospitality and transportation.

In conclusion, this proposed automation solution that uses artificial intelligence to track high-value targets has the potential to improve the security situation in Pakistan and other countries facing similar challenges. The system uses deep learning techniques to classify individuals by gender, ethnicity, and facial hair style, reducing seek time by searching relevant databases. The system also offers additional features to assist users, such as generating hyper-real colour images from sketched images and transforming blurred or low-resolution images into high-resolution images. Once installed, the aim of the system is to help create a safe and secure environment for visitors, attract foreign investors, and increase business activities, thus boosting economic growth and quality of life in areas it is installed.