Data Visualization And Neural Network In Defense Industry – Engineering Assignment Help

Assignment Task

Task 

Abstract
In modern defence and security operations, analysts are faced with significant information overload problems and prevent developing good situation awareness. So, Data Visualization and Visual Analytics (VA) is an efficient way of handling massive data sets by using interactive visualization technologies. Developments in artificial intelligence (AI) have resulted in a breakthrough for many AIapplications, such as computer vision, natural language processing. Therefore, there are manyefforts to use these developments for military applications.

Introduction
In the era of the information age, first responders in Defence and Security are faced with increasing amounts of dynamic information originating from different source and in a various format, which needed to be analysed in order to understand a situation and react promptly. Artificial intelligence (AI), specifically the subfields machine learning (ML) and deep learning (DL), has moved from research institutes and universities to industry and real-world application. In a military context, AI is present in all domains (land, sea, air, space and information) and all levels of warfare (i.e. political, strategic, operational and tactical). Artificial Intelligence can improve autonomous control in unmanned systems so that operators can operate unmanned systems more efficiently to, ultimately, increase battlefield impact.

Data Visualization
Data visualization is the graphical representation of information and data. Visualization is not merely used to display information using pictures. The visual representation should be in a meaningful way in order to provide information to the user. The choices made by the operators highly dependent on the data involved and the task to be performed.

Visual Analytics
“Visual analytics is the science of analytical reasoning facilitated by interactive visual interfaces”[1]. This human-information discourse is between the data and his analyst. It supports three goals: assessment, forecasting and planning. Analysts use reasoning techniques to achieve these goals.

Figure 1: Visual Analytics Interaction
The goal of vision of visual analytics is to turn the information overload into an opportunity, to make our way of processing data and information transparent for an analytic discourse. The visualization of these processes will provide the means of communicating about them, instead of being left with the results. Visual Analytics did the constructive evaluation, correction and improvement of our processes and models. It helps in the improvement of our knowledge and our decisions. Visual analytics is a technology that combines the strengths of human and electronic data processing. It becomes a semi-automated analytical process, where humans and machines cooperate using their respective distinct capabilities for the most effective results.

Deep learning and Neural NetworksDeep Learning refer to machine learning models consisting of multiple of layers processing units. These models are represented by artificial neural networks. So, a neuron refers to a single computation unit. DNNs refer to systems with a large number of serially connected layers of parallel-connected neurons. DNN is a shallow neural network that has only one layer of parallel-connected neurons.
The parameters of the layer- by-layer-trained deep networks were finally fine-tuned using stochastic gradient methods to maximize the classification accuracy[2]. High performance of DNNs is mostly depends on representation learning. During the training of a DNN discriminating features are automatically added.

Objective of the Project:
1. Creating Training Simulations.
2. Maritime Awareness.
3. Surveillance.
4. Saving Human Life

Creating Training Simulations
Forces operate in complex, dynamic and stressful environments where wrong decisions have fatal consequences. Live, Virtual and Constructive training environments all provide elements of best practices for this type of training. Virtual training systems offer a safe, effective and efficient training that has practical and economic advantages over more traditional training approaches.

Maritime Awareness
Maritime applications have a strong geospatial component and visualization of this aspect is critical to Maritime Domain Awareness (MDA). “Maritime Domain Awareness is the effective understanding of everything on, under, related to,adjacent to, or bordering a sea,ocean or other navigable waterway, including all maritime-related activities, infrastructure,people, cargo, vessels, or other conveyances”[1].Operators/analystsof maritime domain have a mandate to be aware of all that is happening in maritime approaches. This mandate is based on the need to protect fromattack, defend sovereignty, detect illegal activities, and support search and rescue activities.

Surveillance
Surveillance is performed using fixed radar stations, patrol aircrafts, ships, and electronic tracking for vessels using the automatic identification system (AIS). These information sources provide information about vessel or person movement that may reveal illegal, unsafe, threatening, and anomalous behavior. ML-approaches are used to generate normality models from the gathered information [4].

Saving Human 
LifeCreating automated weapons used for both offensive and defensive purposes. Which could be used as frontline weaponry in case of any threat or can be used in search and rescue operations saving countless human life. These weapons based on Neural networks and using many machine learning algorithms can assess the threat percentage and act accordingly. These can be used in extreme weather condition.

Working Principle
Using a task-driven design based on a formative user study with various experts, an interactive prototype can be designed and implemented.

Figure -2: framework of task driven design
Exploratory Analysis: In this stage, the primary task is to integrate multiple data sets in a common reference frame while providing filtering and clutter reduction capabilities. It enables the expert to understand the strengths and limitations of different models, and help to choose most informative one to include in a “reference model”.
Tracking: Output from the previous stage is used in the tracking subsystem to act as a prediction for tracking. So, the positions of ground truth measurements can be tracked at any desired time instant. This effectively fills the gaps in the data and provides a more complete understanding of how the detected points must have evolved over time [5].
Visual Feedback: Results obtained from tracking are fed to the rendering subsystem to provide visual feedback. Tracking calculations and visual mappings are performed in GPU shaders to maintain interactive rates

Figure -3: neural network working
Many algorithms are used for interpreting DNNs. These algorithms provide a better option to act with high accuracy, reliability and success percentage. To get better results, we transform the DNN processing into the original input space in order to visualize discriminating features.
Typically, two general approaches are used for feature visualization, activation maximation and DNN explanation[3]. Activation maximation computes which inputs features that will maximally activate possible system recommendations. DNN explanation explains system recommendations by highlighting discriminating input features. AI-planning is based on models of domain dynamics.