The on non-triangulated localization methodologies. Environmentally, we expect to

The process of localizing objects has been on interest within the realm of computer
vision for the past decade. The idea of determining the spatial position of an object in time can
give rise to many applications in terms of surveillance, automated positioning systems, virtual
reality, and a digital storage medium. In the recent years, developments in object localization
have not been as user-friendly, flexible, or cost effective as most methodologies relied solely on
magnetic fields, radio waves, acoustic signals, and various other sensory acquisition devices
that also succumb to large quantization errors in localizing objects due to noise and varying
stochastic discrepancies . Under the algorithm proposed, accurate localization can be realized
for any user that has access to a camera(s) and can develop their own monitoring/guidance
In an ever growing society with concerns to safety as well as advances in automation
and virtual reality, accurate localization algorithms especially localizing with weak supervision
can fill vast needs in the marketplace. The general area of the market that we do see this
targeting is the security/surveillance sector, which could revolutionize the way modern day
surveillance footage is captured and analyzed. Market places that require automated robotic
systems could benefit greatly as the necessity for localization on board robots would be
completely eliminated and could mitigate quantization errors further seen by systems that
solely rely on non-triangulated localization methodologies. Environmentally, we expect to see
no major effect as all this system requires is a live video feed from the source of choice that has
cameras/webcams situated in a triangulated form. Firstly, we’ll analyze the main areas that this
algorithm will immediately affect.
Impact/Use? ?On? ?Indoor? ?Positioning? ?Systems? ?(IPS):
One of the main areas we see this impact immediately is in Indoor Positioning
Systems. Most modern day IPS systems are used for tracking objects and people that may
traverse in a room or space of a building. Such systems are subject to the discrepancies
mentioned previously where they solely rely on on magnetic fields, radio waves, acoustic
signals, and various other sensory acquisition devices that also succumb to large quantization
errors in localizing objects due to noise and varying stochastic discrepancies . Most of these
errors especially in indoor applications rise from signal attenuation and noise caused by the
materials used in constructing the building. Where GPS is effective within 10 feet of an
individual outdoors, this significantly reduced once moved indoors due to signal reflection
along surfaces causing multi-path propagation errors. Wi-Fi based positioning (WPS) systems
have been instilled in order to provide a better solution and utilize existing wireless architecture
however, the method solely relies on measuring the intensity of the signal and geolocating a
wireless access point as well as the SSID and MAC address of said point 3. The issue arises
when multiple positions are added into a database where a slight fluctuation in the signal can
provide an inaccurate path of the object/user that is to be localized. Utilizing triangulation with
webcams can eliminate this entirely as the localization of the user/object is solely reliant on
whether or not the user is passing between the cameras. The triangulation can be modified to
suit one’s needs such that cameras that already exist within a building can be used according to
the user’s needs. This entails segmenting an object desired in the event of analyzing footage
stored in a database, providing spatial coordinates for where the localized object/user is in the
respective room or space in the building, and a map of the object/user’s movement over time.
Impact/Use? on? ?Surveillance? ?Systems? ?(CCTV/IP):
As mass surveillance systems become more prevalent, so does the need in
updating current network and video infrastructure. Current day surveillance systems consist
majorly of CCTV/IP based cameras, but provide little to no analysis other than playback. Even
the majority of advanced surveillance systems provide at best data logging based on motion
detection, but yet fail to provide object segmentation and movement mapping. Implementing
the proposed system across surveillance systems that are currently instilled could link deep
convolutional neural networks to footage recorded. In the event of tracking the movement of a
specific object/user, color-based or cascade based tracking can provide object segmentation
that can exactly map the movement of a person/object within the set triangulated space over
time. In addition, the video data logged can be referenced in a database where the movement
of the said localized object/user can be monitored upon searching the interval between two
In larger areas where a tracking an object in a cluttered premise such as metropolitan
areas, airports, train stations, and entertainment establishments, object segmentation can be
used to filter and localize an object/individual with its/their initial reference position as a start
for localizing either static or non static behavior. By implementing such an algorithm that also
supports weak supervision in terms of learning objects from a sample of images, modern day
surveillance systems can become key in tracking things such as packages, criminal suspects,
workers, factory lines, and various other applications etc. Most of the CCTV/IP cameras can be
linked together using various LAN protocols in implementing the required triangulation
Impact/Use? on? ?Virtual? ?Reality:
Virtual Reality in its current form superimposes a stereoscopic image into a
headset that provides the user with an illusion that they are interacting in a 3D space. Most
objects that exist the the virtual reality space are generally computer generated which are then
superimposed into the environment that the user currently is immersed it. As with applications
in augmented reality where objects and graphics are overlaid and provide a more harmonious
relationship between both virtual and reality worlds, virtual reality has yet to see this. Methods
in imposed real objects into virtual reality have mostly involved photogrammetry and
converting real objects into CGI based 3D models, but haven’t provided direct interaction with
these objects in relation to real spatial surroundings.
In implementing the proposed form of localizing an object, a relationship between real
objects in a virtual space can be realized. Localizing and superimposing the object into a virtual
space and provide an even more immersive experience as the requirement of utilizing CGI
based models can be eliminated. By utilizing the method of triangulation, localization, and
filtering an object to be imposed in a virtual space, behaviors of said object can be accurately
portrayed by providing real data for trajectory and positioning. By implementing this algorithm,
we could see an increase of use in VR specifically in industries that require a lot of training. For
example, areas that require high level of training may involve the medical, construction,
transportation, and the military where trainees could expect to use actual equipment in various
scenarios. In superimposing equipments required in training into a virtual world, vast scenarios
can be introduced while the maintaining the same physicality in using such equipment.