An embedded smart camera is a stand-alone unit that not only captures images, but also includes a processor, memory and communication interface. Battery-powered, embedded smart cameras introduce many additional challenges since they have very limited resources, such as energy, processing power and memory. When camera sensors are added to an embedded system, the problem of limited resources becomes even more pronounced. Hence, computer vision algorithms running on these camera boards should be light-weight and efficient. This thesis is about designing and developing computer vision algorithms, which are aware and successfully overcome the limitations of embedded platforms (in terms of power consumption and memory usage). Particularly, we are interested in object detection and tracking methodologies and the impact of them on the performance and battery life of the CITRIC camera (embedded smart camera employed in this research). This thesis aims to prolong the life time of the Embedded Smart platform, without affecting the reliability of the system during surveillance tasks. Therefore, the reader is walked through the whole designing process, from the development and simulation, followed by the implementation and optimization, to the testing and performance analysis. The work presented in this thesis carries out not only software optimization, but also hardware-level operations during the stages of object detection and tracking. The performance of the algorithms introduced in this thesis are comparable to state-of-the-art object detection and tracking methods, such as Mixture of Gaussians, Eigen segmentation, color and coordinate tracking. Unlike the traditional methods, the newly-designed algorithms present notable reduction of the memory requirements, as well as the reduction of memory accesses per pixel. To accomplish the proposed goals, this work attempts to interconnect different levels of the embedded system architecture to make the platform more efficient in terms of energy and resource savings. Thus, the algorithms proposed are optimized at the API, middleware, and hardware levels to access the pixel information of the CMOS sensor directly. Only the required pixels are acquired in order to reduce the unnecessary communications overhead. Experimental results show that when exploiting the architecture capabilities of an embedded platform, 41.24% decrease in energy consumption, and 107.2% increase in battery-life can be accomplished. Compared to traditional object detection and tracking methods, the proposed work provides an additional 8 hours of continuous processing on 4 AA batteries, increasing the lifetime of the camera to 15.5 hours.
","abstract_html":"<p>An embedded smart camera is a stand-alone unit that not only captures images, but also includes a processor, memory and communication interface. Battery-powered, embedded smart cameras introduce many additional challenges since they have very limited resources, such as energy, processing power and memory. When camera sensors are added to an embedded system, the problem of limited resources becomes even more pronounced. Hence, computer vision algorithms running on these camera boards should be light-weight and efficient. This thesis is about designing and developing computer vision algorithms, which are aware and successfully overcome the limitations of embedded platforms (in terms of power consumption and memory usage). Particularly, we are interested in object detection and tracking methodologies and the impact of them on the performance and battery life of the CITRIC camera (embedded smart camera employed in this research). This thesis aims to prolong the life time of the Embedded Smart platform, without affecting the reliability of the system during surveillance tasks. Therefore, the reader is walked through the whole designing process, from the development and simulation, followed by the implementation and optimization, to the testing and performance analysis. The work presented in this thesis carries out not only software optimization, but also hardware-level operations during the stages of object detection and tracking. The performance of the algorithms introduced in this thesis are comparable to state-of-the-art object detection and tracking methods, such as Mixture of Gaussians, Eigen segmentation, color and coordinate tracking. Unlike the traditional methods, the newly-designed algorithms present notable reduction of the memory requirements, as well as the reduction of memory accesses per pixel. To accomplish the proposed goals, this work attempts to interconnect different levels of the embedded system architecture to make the platform more efficient in terms of energy and resource savings. Thus, the algorithms proposed are optimized at the API, middleware, and hardware levels to access the pixel information of the CMOS sensor directly. Only the required pixels are acquired in order to reduce the unnecessary communications overhead. Experimental results show that when exploiting the architecture capabilities of an embedded platform, 41.24% decrease in energy consumption, and 107.2% increase in battery-life can be accomplished. Compared to traditional object detection and tracking methods, the proposed work provides an additional 8 hours of continuous processing on 4 AA batteries, increasing the lifetime of the camera to 15.5 hours.</p>","abstract_has_math":false,"creators":["Casares, Mauricio"],"institution":null,"degree_name":"Doctor of Philosophy (PhD)","degree_level":"Dissertation","degree_discipline":"Electrical Engineering and Computer Science","degree_department":null,"school":null,"contributors":["Senem Velipasalar","Cliff Davidson"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2014,"date_issued":"2014-05-30T07:00:00Z","date_published":"2014-05-30T07:00:00Z","updated_at":"2026-07-24T04:54:51Z","subjects":["Adaptive tracking","Background subtraction","Device drivers","Feedback tracking","Hardware-level operations","Middleware","Engineering"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://surface.syr.edu/etd/97","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Senem Velipasalar","Cliff Davidson"]},{"key":"dc:creator","label":"Author","values":["Casares, Mauricio"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical Engineering and Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Adaptive tracking","Background subtraction","Device drivers","Feedback tracking","Hardware-level operations","Middleware","Engineering"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://surface.syr.edu/etd/97"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["An embedded smart camera is a stand-alone unit that not only captures images, but also includes a processor, memory and communication interface. Battery-powered, embedded smart cameras introduce many additional challenges since they have very limited resources, such as energy, processing power and memory. When camera sensors are added to an embedded system, the problem of limited resources becomes even more pronounced. Hence, computer vision algorithms running on these camera boards should be light-weight and efficient. This thesis is about designing and developing computer vision algorithms, which are aware and successfully overcome the limitations of embedded platforms (in terms of power consumption and memory usage). Particularly, we are interested in object detection and tracking methodologies and the impact of them on the performance and battery life of the CITRIC camera (embedded smart camera employed in this research). This thesis aims to prolong the life time of the Embedded Smart platform, without affecting the reliability of the system during surveillance tasks. Therefore, the reader is walked through the whole designing process, from the development and simulation, followed by the implementation and optimization, to the testing and performance analysis. The work presented in this thesis carries out not only software optimization, but also hardware-level operations during the stages of object detection and tracking. The performance of the algorithms introduced in this thesis are comparable to state-of-the-art object detection and tracking methods, such as Mixture of Gaussians, Eigen segmentation, color and coordinate tracking. Unlike the traditional methods, the newly-designed algorithms present notable reduction of the memory requirements, as well as the reduction of memory accesses per pixel. To accomplish the proposed goals, this work attempts to interconnect different levels of the embedded system architecture to make the platform more efficient in terms of energy and resource savings. Thus, the algorithms proposed are optimized at the API, middleware, and hardware levels to access the pixel information of the CMOS sensor directly. Only the required pixels are acquired in order to reduce the unnecessary communications overhead. Experimental results show that when exploiting the architecture capabilities of an embedded platform, 41.24% decrease in energy consumption, and 107.2% increase in battery-life can be accomplished. Compared to traditional object detection and tracking methods, the proposed work provides an additional 8 hours of continuous processing on 4 AA batteries, increasing the lifetime of the camera to 15.5 hours.
"]},{"key":"dc:title","label":"Title","values":["ENERGY-EFFICIENT LIGHTWEIGHT ALGORITHMS FOR EMBEDDED SMART CAMERAS: DESIGN, IMPLEMENTATION AND PERFORMANCE ANALYSIS"]}]}],"canonical_facts":{"dc:contributor":["Senem Velipasalar","Cliff Davidson"],"dc:creator":["Casares, Mauricio"],"dc:description.abstract":["An embedded smart camera is a stand-alone unit that not only captures images, but also includes a processor, memory and communication interface. Battery-powered, embedded smart cameras introduce many additional challenges since they have very limited resources, such as energy, processing power and memory. When camera sensors are added to an embedded system, the problem of limited resources becomes even more pronounced. Hence, computer vision algorithms running on these camera boards should be light-weight and efficient. This thesis is about designing and developing computer vision algorithms, which are aware and successfully overcome the limitations of embedded platforms (in terms of power consumption and memory usage). Particularly, we are interested in object detection and tracking methodologies and the impact of them on the performance and battery life of the CITRIC camera (embedded smart camera employed in this research). This thesis aims to prolong the life time of the Embedded Smart platform, without affecting the reliability of the system during surveillance tasks. Therefore, the reader is walked through the whole designing process, from the development and simulation, followed by the implementation and optimization, to the testing and performance analysis. The work presented in this thesis carries out not only software optimization, but also hardware-level operations during the stages of object detection and tracking. The performance of the algorithms introduced in this thesis are comparable to state-of-the-art object detection and tracking methods, such as Mixture of Gaussians, Eigen segmentation, color and coordinate tracking. Unlike the traditional methods, the newly-designed algorithms present notable reduction of the memory requirements, as well as the reduction of memory accesses per pixel. To accomplish the proposed goals, this work attempts to interconnect different levels of the embedded system architecture to make the platform more efficient in terms of energy and resource savings. Thus, the algorithms proposed are optimized at the API, middleware, and hardware levels to access the pixel information of the CMOS sensor directly. Only the required pixels are acquired in order to reduce the unnecessary communications overhead. Experimental results show that when exploiting the architecture capabilities of an embedded platform, 41.24% decrease in energy consumption, and 107.2% increase in battery-life can be accomplished. Compared to traditional object detection and tracking methods, the proposed work provides an additional 8 hours of continuous processing on 4 AA batteries, increasing the lifetime of the camera to 15.5 hours.
"],"dc:identifier":["https://surface.syr.edu/etd/97"],"dc:subject":["Adaptive tracking","Background subtraction","Device drivers","Feedback tracking","Hardware-level operations","Middleware","Engineering"],"dc:title":["ENERGY-EFFICIENT LIGHTWEIGHT ALGORITHMS FOR EMBEDDED SMART CAMERAS: DESIGN, IMPLEMENTATION AND PERFORMANCE ANALYSIS"],"thesis:degree_discipline":["Electrical Engineering and Computer Science"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-24T04:54:51Z"}