Srijith Rajeev
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Srijith Rajeev has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Sos Agaian
College of Staten Island, USA
5 shared publications - Karen Panetta
Tufts University
4 shared publications - Shreyas Kamath K. M.
Department of Electrical and Computer Engineering, Tufts University, Medford, MA, USA
4 shared publications Karen Panetta
Department of Electrical and Computer Engineering, Tufts University, Medford, MA, USA
3 shared publications- Arash Samani
MITRE Corporation,Bedford,MA,USA
2 shared publications Sos S. Agaian
Department of Computer Science, The City University of New York, New York, NY, USA
2 shared publications
Latest Publications
- 15 June 2023
- Published by SPIE-Intl Soc Optical Eng
- Vol. 12526, 23-138
Instant-level vehicle speed and traffic density estimation using deep neural network
The growing network of highway video surveillance cameras generates an immense amount of data that proves tedious for manual analysis. Automated real-time analysis of such data may provide many solutions, including traffic monitoring, traffic incident detection, and smart-city planning. More specifically, assessing traffic speed and density is critical in determining dynamic traffic conditions and detecting slowdowns, traffic incidents, and traffic alerts. However, despite several advancements, there are numerous challenges in estimating vehicle speed and traffic density, which are integral parts of ITS. Some of these challenges include variations in road networks, illumination constraints, weather, structure occlusion, and vehicle user-driving behavior. To address these issues, this paper proposes a novel deep learning-based framework for instant-level vehicle speed and traffic flow density estimation to effectively harness the potential of existing large-scale highway surveillance cameras to assist in real-time traffic analysis. This is achieved using the state-of-the-art region-based Siamese MOT network, SiamMOT which detects and associates object instances for multi-object tracking (MOT), to accurately estimate instant level vehicle speed in live video feeds. The UA-DETRAC dataset is used to train the speed estimation model. Computer simulations show that the proposed framework a) allows the classifying of traffic density into light, medium, or heavy traffic flows, b) is robust to different types of road networks and illuminations without prior road information, and c) shows good performance when compared to current state-of-the-art methods using adequate performance metrics.
- 15 June 2023
- Published by SPIE-Intl Soc Optical Eng
- Vol. 12526, 23-138
Instant-level vehicle speed and traffic density estimation using deep neural network
The growing network of highway video surveillance cameras generates an immense amount of data that proves tedious for manual analysis. Automated real-time analysis of such data may provide many solutions, including traffic monitoring, traffic incident detection, and smart-city planning. More specifically, assessing traffic speed and density is critical in determining dynamic traffic conditions and detecting slowdowns, traffic incidents, and traffic alerts. However, despite several advancements, there are numerous challenges in estimating vehicle speed and traffic density, which are integral parts of ITS. Some of these challenges include variations in road networks, illumination constraints, weather, structure occlusion, and vehicle user-driving behavior. To address these issues, this paper proposes a novel deep learning-based framework for instant-level vehicle speed and traffic flow density estimation to effectively harness the potential of existing large-scale highway surveillance cameras to assist in real-time traffic analysis. This is achieved using the state-of-the-art region-based Siamese MOT network, SiamMOT which detects and associates object instances for multi-object tracking (MOT), to accurately estimate instant level vehicle speed in live video feeds. The UA-DETRAC dataset is used to train the speed estimation model. Computer simulations show that the proposed framework a) allows the classifying of traffic density into light, medium, or heavy traffic flows, b) is robust to different types of road networks and illuminations without prior road information, and c) shows good performance when compared to current state-of-the-art methods using adequate performance metrics.
- Open Access
- 26 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 145212-145227
FTNet: Feature Transverse Network for Thermal Image Semantic Segmentation
Thermal imaging is a process of using infrared radiation and thermal energy to collect information about objects. It is superior to visible imaging for its ability to operate in darkness and tolerate illumination variations. In addition, it has potential to penetrate smoke, aerosol, dust, and mist, which are critical inhibitors for visible imaging applications, including semantic segmentation. Unfortunately, current state-of-the-art image semantic segmentation methods (i) mainly concentrate on visible spectrum images, which do not adequately capture the context of corresponding pixels, particularly edge details in thermal images, and (ii) accept a trade-off between higher accuracy and lower speed, or vice-versa. Here, a novel end-to-end trainable convolutional neural network architecture, feature transverse network (FTNet), has been proposed to solve the aforementioned problems. FTNet captures and optimizes feature representation at the multi-scale resolution, thereby improving the capability to process high-resolution images and producing quality output with a lower computational cost. Extensive computer experimentations were conducted on publicly available benchmarking thermal datasets, including SODA, MFNet, and SCUT-Seg, to demonstrate the effectiveness of the proposed FTNet compared to state-of-the-art methods. This comparison includes multiple aspects, including the quantitative accuracy and speed of the various approaches. The source code is available at https://github.com/shreyaskamathkm/FTNet.
- 26 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 9, 145212-145227
Open AccessFTNet: Feature Transverse Network for Thermal Image Semantic Segmentation
Thermal imaging is a process of using infrared radiation and thermal energy to collect information about objects. It is superior to visible imaging for its ability to operate in darkness and tolerate illumination variations. In addition, it has potential to penetrate smoke, aerosol, dust, and mist, which are critical inhibitors for visible imaging applications, including semantic segmentation. Unfortunately, current state-of-the-art image semantic segmentation methods (i) mainly concentrate on visible spectrum images, which do not adequately capture the context of corresponding pixels, particularly edge details in thermal images, and (ii) accept a trade-off between higher accuracy and lower speed, or vice-versa. Here, a novel end-to-end trainable convolutional neural network architecture, feature transverse network (FTNet), has been proposed to solve the aforementioned problems. FTNet captures and optimizes feature representation at the multi-scale resolution, thereby improving the capability to process high-resolution images and producing quality output with a lower computational cost. Extensive computer experimentations were conducted on publicly available benchmarking thermal datasets, including SODA, MFNet, and SCUT-Seg, to demonstrate the effectiveness of the proposed FTNet compared to state-of-the-art methods. This comparison includes multiple aspects, including the quantitative accuracy and speed of the various approaches. The source code is available at https://github.com/shreyaskamathkm/FTNet.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
3D Navigational Insight using AR Technology
Navigation, the ability to relocate from one place to another, is a critical skill for any individual or group. Navigating safely across unknown environments is a critical factor in determining the success of a mission. While there is an existing body of applications in the field of land navigation, they primarily rely on GPS-enabled technologies. Moreover, there is limited research on Augmented Reality (AR) as a tool for navigation in unknown environments. This research proposes to develop an AR system to provide 3-dimensional (3D) navigational insights in unfamiliar environments. This can be accomplished by generating 3D terrestrial maps leveraging Synthetic Aperture Radar (SAR) data, Google earth imagery and sparse knowledge of GPS coordinates of the region. Furthermore, the 3D terrestrial images are converted to navigational meshes to make it feasible for path-finding algorithms to work. The proposed method can be used to create an iteratively refined 3D landscape knowledge-database that can assist personnel in navigating novel environments or assist in mission planning for other operations. It can also be used to help plan/access the best strategic vantage points in the landscape.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
3D Navigational Insight using AR Technology
Navigation, the ability to relocate from one place to another, is a critical skill for any individual or group. Navigating safely across unknown environments is a critical factor in determining the success of a mission. While there is an existing body of applications in the field of land navigation, they primarily rely on GPS-enabled technologies. Moreover, there is limited research on Augmented Reality (AR) as a tool for navigation in unknown environments. This research proposes to develop an AR system to provide 3-dimensional (3D) navigational insights in unfamiliar environments. This can be accomplished by generating 3D terrestrial maps leveraging Synthetic Aperture Radar (SAR) data, Google earth imagery and sparse knowledge of GPS coordinates of the region. Furthermore, the 3D terrestrial images are converted to navigational meshes to make it feasible for path-finding algorithms to work. The proposed method can be used to create an iteratively refined 3D landscape knowledge-database that can assist personnel in navigating novel environments or assist in mission planning for other operations. It can also be used to help plan/access the best strategic vantage points in the landscape.
- Open Access
- 30 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 104567-104576
LQM: Localized Quality Measure for Fingerprint Image Enhancement
Fingerprint is one of the most widely used biometric in law enforcement. However, low-quality fingerprint images can drastically degrade the performance of automated fingerprint identification systems (AFIS). AFIS can be substantially advanced by: 1) establishing a metric to evaluate the image quality accurately and 2) utilizing this metric to enable an automated enhancement process. This paper offers a novel localized quality measure (LQM) to evaluate the quality of fingerprint images, and a genetic localized quality measure enhancement (LQME) algorithm, which is tailored to iteratively enhance poor-quality fingerprint images. In addition, a method is introduced to automatically choose the enhancement algorithm's parameters based on the proposed measure such that it yields the best enhancement result. The presented LQM measure uses fingerprint image characteristics, which include sharpness, contrast, orientation certainty level, symmetry features, and imprints of friction ridge structure (minutiae) information. The FVC2004 Set B database containing fingerprint images from four different sensors and a total of 240 images (80 from each sensor) is used to evaluate the performance of the presented algorithms and methods. The computer simulations demonstrate that the LQM measure is useful in predicting the quality of the fingerprint images captured from various devices. Furthermore, the experiments show that LQME can recover retrievable-corrupt fingerprint regions.
- 30 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 104567-104576
Open AccessLQM: Localized Quality Measure for Fingerprint Image Enhancement
Fingerprint is one of the most widely used biometric in law enforcement. However, low-quality fingerprint images can drastically degrade the performance of automated fingerprint identification systems (AFIS). AFIS can be substantially advanced by: 1) establishing a metric to evaluate the image quality accurately and 2) utilizing this metric to enable an automated enhancement process. This paper offers a novel localized quality measure (LQM) to evaluate the quality of fingerprint images, and a genetic localized quality measure enhancement (LQME) algorithm, which is tailored to iteratively enhance poor-quality fingerprint images. In addition, a method is introduced to automatically choose the enhancement algorithm's parameters based on the proposed measure such that it yields the best enhancement result. The presented LQM measure uses fingerprint image characteristics, which include sharpness, contrast, orientation certainty level, symmetry features, and imprints of friction ridge structure (minutiae) information. The FVC2004 Set B database containing fingerprint images from four different sensors and a total of 240 images (80 from each sensor) is used to evaluate the performance of the presented algorithms and methods. The computer simulations demonstrate that the LQM measure is useful in predicting the quality of the fingerprint images captured from various devices. Furthermore, the experiments show that LQME can recover retrievable-corrupt fingerprint regions.
- Open Access
- 27 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 88174-88185
Unrolling Post-Mortem 3D Fingerprints Using Mosaicking Pressure Simulation Technique
Post-mortem fingerprints are a valuable biometric used to aid in the identification of a deceased individual. However, fingerprints from the deceased undergo decomposition leading to indefinite structure when compared to ante-mortem fingerprints. Moreover, the performance of the existing two-dimensional (2D) fingerprint recognition systems is still below the expected potential. These problems arise because fingerprints are generally captured by manipulating a finger against a plane. In post-mortem fingerprint recovery, the decedent’s finger must go through several reconditioning processes to prevent the rapid onslaught of decomposition. To address these deficiencies associated with the 2D systems, three-dimensional (3D) scanning systems have been employed to capture fingerprints. 3D technology is still in its transient phase, and is limited primarily by (i) the lack of existing 3D databases, (ii) the deficiency of 3D to 2D fingerprint image mapping algorithms, (iii) the incapacity to model and recreate the 2D fingerprint capturing procedure to improve 3D-2D fingerprint verification, and (iv) the inability to apply traditional fingerprint unrolling techniques on post-mortem 3D fingerprints. This paper presents a novel method to perform postmortem 3D fingerprint unrolling and pressure simulation to produce fingerprint images that are compatible with 2D fingerprint recognition systems. The thrust of this work strives to (i) develop a correspondence between 3D touchless and contact-based 2D fingerprint images, (ii) model fingerprints with deformities to provide a viable fingerprint image for matching, and (iii) develop a Mosaic Pressure Simulation (MPS) algorithm to recreate the effects of 2D fingerprint capturing procedure.
- 27 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 88174-88185
Open AccessUnrolling Post-Mortem 3D Fingerprints Using Mosaicking Pressure Simulation Technique
Post-mortem fingerprints are a valuable biometric used to aid in the identification of a deceased individual. However, fingerprints from the deceased undergo decomposition leading to indefinite structure when compared to ante-mortem fingerprints. Moreover, the performance of the existing two-dimensional (2D) fingerprint recognition systems is still below the expected potential. These problems arise because fingerprints are generally captured by manipulating a finger against a plane. In post-mortem fingerprint recovery, the decedent’s finger must go through several reconditioning processes to prevent the rapid onslaught of decomposition. To address these deficiencies associated with the 2D systems, three-dimensional (3D) scanning systems have been employed to capture fingerprints. 3D technology is still in its transient phase, and is limited primarily by (i) the lack of existing 3D databases, (ii) the deficiency of 3D to 2D fingerprint image mapping algorithms, (iii) the incapacity to model and recreate the 2D fingerprint capturing procedure to improve 3D-2D fingerprint verification, and (iv) the inability to apply traditional fingerprint unrolling techniques on post-mortem 3D fingerprints. This paper presents a novel method to perform postmortem 3D fingerprint unrolling and pressure simulation to produce fingerprint images that are compatible with 2D fingerprint recognition systems. The thrust of this work strives to (i) develop a correspondence between 3D touchless and contact-based 2D fingerprint images, (ii) model fingerprints with deformities to provide a viable fingerprint image for matching, and (iii) develop a Mosaic Pressure Simulation (MPS) algorithm to recreate the effects of 2D fingerprint capturing procedure.
- 30 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 42 (3) , 509-520
A Comprehensive Database for Benchmarking Imaging Systems
Cross-modality face recognition is an emerging topic due to the wide-spread usage of different sensors in day-to-day life applications. The development of face recognition systems relies greatly on existing databases for evaluation and obtaining training examples for data-hungry machine learning algorithms. However, currently, there is no publicly available face database that includes more than two modalities for the same subject. In this work, we introduce the Tufts Face Database that includes images acquired in various modalities: photograph images, thermal images, near infrared images, a recorded video, a computerized facial sketch, and 3D images of each volunteer's face. An Institutional Research Board protocol was obtained and images were collected from students, staff, faculty, and their family members at Tufts University. The database includes over 10,000 images from 113 individuals from more than 15 different countries, various gender identities, ages, and ethnic backgrounds. The contributions of this work are: 1) Detailed description of the content and acquisition procedure for images in the Tufts Face Database; 2) The Tufts Face Database is publicly available to researchers worldwide, which will allow assessment and creation of more robust, consistent, and adaptable recognition algorithms; 3) A comprehensive, up-to-date review on face recognition systems and face datasets.
- 30 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 42 (3) , 509-520
A Comprehensive Database for Benchmarking Imaging Systems
Cross-modality face recognition is an emerging topic due to the wide-spread usage of different sensors in day-to-day life applications. The development of face recognition systems relies greatly on existing databases for evaluation and obtaining training examples for data-hungry machine learning algorithms. However, currently, there is no publicly available face database that includes more than two modalities for the same subject. In this work, we introduce the Tufts Face Database that includes images acquired in various modalities: photograph images, thermal images, near infrared images, a recorded video, a computerized facial sketch, and 3D images of each volunteer's face. An Institutional Research Board protocol was obtained and images were collected from students, staff, faculty, and their family members at Tufts University. The database includes over 10,000 images from 113 individuals from more than 15 different countries, various gender identities, ages, and ethnic backgrounds. The contributions of this work are: 1) Detailed description of the content and acquisition procedure for images in the Tufts Face Database; 2) The Tufts Face Database is publicly available to researchers worldwide, which will allow assessment and creation of more robust, consistent, and adaptable recognition algorithms; 3) A comprehensive, up-to-date review on face recognition systems and face datasets.
- 1 April 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Fingerprint authentication using geometric features
Biometric based authentication, particularly for fingerprint authentication systems play a vital role in identifying an individual. The existing fingerprint authentication systems depend on specific points known as minutiae for recognizing an individual. Designing a reliable automatic fingerprint authentication system is still very challenging, since not all fingerprint information is available. Further, the information obtained is not always accurate due to cuts, scars, sweat, distortion and various skin conditions. Moreover, the existing fingerprint authentication systems do not utilize other significant minutiae information, which can improve the accuracy. Various local feature detectors such as Difference-of-Gaussian, Hessian, Hessian Laplace, Harris Laplace, Multiscale Harris, and Multiscale Hessian have been extensively used for feature detection. However, these detectors have not been employed for detecting fingerprint image features. In this article, a versatile local feature fingerprint matching scheme is proposed. The local features are obtained by exploiting these local geometric detectors and SIFT descriptor. This scheme considers local characteristic features of the fingerprint image, thus eliminating the issues caused in existing fingerprint feature based matching techniques. Computer simulations of the proposed algorithm on specific databases show significant improvements when compared to existing fingerprint matchers, such as minutiae matcher, hierarchical matcher and graph based matcher. Computer simulations conducted on the Neurotechnology database demonstrates a very low Equal Error Rate (EER) of 0.8%. The proposed system a) improves the accuracy of the fingerprint authentication system, b) works when the minutiae information is sparse, and c) produces satisfactory matching accuracy in the case when minutiae information is unavailable. The proposed system can also be employed for partial fingerprint authentication.
- 1 April 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-7
Fingerprint authentication using geometric features
Biometric based authentication, particularly for fingerprint authentication systems play a vital role in identifying an individual. The existing fingerprint authentication systems depend on specific points known as minutiae for recognizing an individual. Designing a reliable automatic fingerprint authentication system is still very challenging, since not all fingerprint information is available. Further, the information obtained is not always accurate due to cuts, scars, sweat, distortion and various skin conditions. Moreover, the existing fingerprint authentication systems do not utilize other significant minutiae information, which can improve the accuracy. Various local feature detectors such as Difference-of-Gaussian, Hessian, Hessian Laplace, Harris Laplace, Multiscale Harris, and Multiscale Hessian have been extensively used for feature detection. However, these detectors have not been employed for detecting fingerprint image features. In this article, a versatile local feature fingerprint matching scheme is proposed. The local features are obtained by exploiting these local geometric detectors and SIFT descriptor. This scheme considers local characteristic features of the fingerprint image, thus eliminating the issues caused in existing fingerprint feature based matching techniques. Computer simulations of the proposed algorithm on specific databases show significant improvements when compared to existing fingerprint matchers, such as minutiae matcher, hierarchical matcher and graph based matcher. Computer simulations conducted on the Neurotechnology database demonstrates a very low Equal Error Rate (EER) of 0.8%. The proposed system a) improves the accuracy of the fingerprint authentication system, b) works when the minutiae information is sparse, and c) produces satisfactory matching accuracy in the case when minutiae information is unavailable. The proposed system can also be employed for partial fingerprint authentication.