Razin Bin Issa

Mr. Razin Bin Issa

BRAC University, Bangladesh
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
20 February 2021 in Sensors
Verified
Journal Article
Open access
PDF file
Razin Bin IssaRazin Bin Issa
Razin Bin Issa
Modhumonty DasModhumonty Das
Modhumonty Das
Md. Saferi RahmanMd. Saferi Rahman
Md. Saferi Rahman
Monika BaruaMonika Barua
Monika Barua
MD KHALILUR RHAMANMD KHALILUR RHAMAN
MD KHALILUR RHAMAN
Md. Golam Rabiul AlamMd. Golam Rabiul Alam
Md. Golam Rabiul Alam
Autonomous vehicle navigation in an unknown dynamic environment is crucial for both supervised- and Reinforcement Learning-based autonomous maneuvering. The cooperative fusion of these two learning approaches has the potential to be an effective mechanism to tackle indefinite environmental dynamics. Most of the state-of-the-art autonomous vehicle navigation systems are trained on a specific mapped model with familiar environmental dynamics. However, this research focuses on the cooperative fusion of supervised and Reinforcement Learning technologies for autonomous navigation of land vehicles in a dynamic and unknown environment. The Faster R-CNN, a supervised learning approach, identifies the ambient environmental obstacles for untroubled maneuver of the autonomous vehicle. Whereas, the training policies of Double Deep Q-Learning, a Reinforcement Learning approach, enable the autonomous agent to learn effective navigation decisions form the dynamic environment. The proposed model is primarily tested in a gaming environment similar to the real-world. It exhibits the overall efficiency and effectiveness in the maneuver of autonomous land vehicles.

1 January 2020
Verified
Conference Paper
Closed access
Razin Bin IssaRazin Bin Issa
Razin Bin Issa
Md. Saferi RahmanMd. Saferi Rahman
Md. Saferi Rahman
Modhumonty DasModhumonty Das
Modhumonty Das
MBMonika Barua
Monika Barua
Md. Golam Rabiul AlamMd. Golam Rabiul Alam
Md. Golam Rabiul Alam
This research focuses on autonomous traversal of land vehicles through exploring undiscovered tracks and overcoming environmental barriers. Most of the existing systems can only operate and traverse in a distinctive mapped model especially in a known area. However, the proposed system which is trained by Deep Reinforcement learning can learn by itself to operate autonomously in extreme conditions. The dynamic double deep Q-learning (DDQN) model enables the proposed system not be confined to only in known environments. The ambient environmental obstacles are identified through Faster R-CNN for smooth movement of the autonomous vehicle. The exploration and exploitation strategies of DDQN enables the autonomous agent to learn proper decisions for various dynamic environments and tracks. The proposed model is tested in a gaming environment. It shows the overall effectiveness in traversing of autonomous land vehicles. The goal is to integrate Deep Reinforcement learning and Faster R-CNN to make the system effective to traverse through undiscovered paths by detecting obstacles.

5 October 2020
Verified
Conference Paper
Closed access
SASayantan Roy Arko
Sayantan Roy Arko
Razin Bin IssaRazin Bin Issa
Razin Bin Issa
Modhumonty DasModhumonty Das
Modhumonty Das
MRMohammed Saferi Rahman
Mohammed Saferi Rahman
In this project, we have tried to develop a system to monitor water quality data of some water bodies in Dhaka city using a small autonomous hovering boat. It is a matter of concern that the water we have access to in Bangladesh is polluted and lacks a proper monitoring system. To manage and protect water, reliable and dynamic water monitoring is a must. Water monitoring systems that exist today require human intervention and expensive methodologies. Moreover, it is difficult to gather, analyze and index water quality data using these conventional systems. This problem can be solved by automating the process and reducing the dependency on existing flawed systems. The proposed project uses a miniature boat shaped automated hovering platform for mobility in water which is guided by a GPS and compass. The device consists of several water quality sensors for on-board data collection. At the same time, a small pump is used to collect water samples for laboratory tests and further analysis. Further, the system is integrated in a communication system to autonomously store data on real time servers through internet connectivity. Being lightweight and rechargeable the portable device can be taken to remote places conveniently. Therefore, this autonomous GPS guided water monitoring system can have a positive impact on water quality monitoring as well as assessment of water quality parameters.

5 November 2025 in Environmental Modelling & Software
Verified
Journal Article
Open access
SNSajan Neupane
Sajan Neupane
JHJeffery Horsburgh
Jeffery Horsburgh
Razin Bin IssaRazin Bin Issa
Razin Bin Issa
SYSierra Young
Sierra Young

21 January 2026 in Journal of Hydroinformatics
Verified
Journal Article
Open access
Razin Bin IssaRazin Bin Issa
Razin Bin Issa
SNSajan Neupane
Sajan Neupane
SKSafran Khan
Safran Khan
JHJeffery Horsburgh
Jeffery Horsburgh
SYSierra Young
Sierra Young
Accurate and cost-effective monitoring of river stage and discharge is essential for effective water resource management, yet conventional gaging stations are costly to install and maintain, particularly in remote areas. This study presents an integrated hardware–software system for camera-based hydrologic monitoring that combines deep learning image segmentation, ensemble regression, and edge computing for near real-time observations. Two hydrologically distinct sites in northern Utah were monitored using fixed cameras to collect imagery for model development and validation. The workflow segments water surfaces, extracts stage-dependent features, and predicts water levels using a stacked ensemble of support vector regression, random forest, and XGBoost models. Stage estimates achieved R2 values of 0.93–0.98 (MAE = 1.03–2.33 cm), and discharge estimates obtained through site-specific rating curves reached R2 = 0.84–0.99 (MAE = 0.27–0.30 m3/s). The pipeline was deployed on four edge devices, Raspberry Pi 5, Jetson Nano, Jetson Orin Nano, and LattePanda Sigma, to assess operational feasibility. Results showed that lightweight models such as MobileSAM enable efficient performance on low-power devices, while high-end hardware supports near real-time inference. The proposed approach offers a scalable, low-maintenance alternative to traditional gaging, advancing camera-based hydrologic monitoring in both developed and data-sparse regions.

17 March 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
SKSafran Khan
Safran Khan
Razin Bin IssaRazin Bin Issa
Razin Bin Issa
SYSierra Young
Sierra Young
Accurate estimation of river discharge from non-contact observations remains challenging, particularly due to spatially incomplete surface velocity fields and unreliable measurements near channel boundaries. This study presented a camera-based framework for estimating river surface velocity and discharge from fixed ground imaging systems using deep optical flow and camera calibration, without reliance on empirical stage–discharge relationships. To address persistent limitations in image-based methods, a lateral velocity reconstruction approach was introduced that integrated power-law, entropy-based, and parabolic formulations to recover near-boundary velocities that were missing or unreliable. Sensitivity analysis was performed to determine robust site-specific parameters for each formulation. Surface velocity fields were derived from image sequences and converted to physical units via camera calibration. These were then converted to depth-averaged flow and combined with surveyed bathymetry and stage observations to estimate discharge. The framework was evaluated at three instrumented river sites in northern Utah under low, moderate, and high flow conditions through a systematic comparison of seven deep optical flow models. Results were benchmarked against large-scale particle image velocimetry (LSPIV) and validated using an independent surface velocity radar sensor. Additional experiments assessed the influence of video duration, training dataset, ground control configuration, and challenging environmental conditions, including glare, snow, low light, and nighttime scenes. FlowFormer++ and MaskFlowNet achieved the highest velocity accuracy, with mean errors of 11.77% and 13.09% while LiteFlowNet2 provided competitive performance at a lower computational cost. Among the reconstruction methods, the power-law approach consistently produced the most accurate discharge estimates. Corrected discharge errors were within 10% of reference measurements in most cases, demonstrating that fixed camera systems can provide a practical and automated alternative to conventional stream gauging, particularly at ungauged or hazardous sites.