Browsing by Author "Khan, Nazmuzzaman"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Clustering Algorithm Based Straight and Curved Crop Row Detection Using Color Based Segmentation
    (ASME, 2021-02) Khan, Nazmuzzaman; Rajendran, Veera P.; Al Hasan, Mohammad; Anwar, Sohel; Mechanical and Energy Engineering, School of Engineering and Technology
    Autonomous navigation of agricultural robot is an essential task in precision agriculture, and success of this task critically depends on accurate detection of crop rows using computer vision methodologies. This is a challenging task due to substantial natural variations in crop row images due to various factors, including, missing crops in parts of a row, high and irregular weed growth between rows, different crop growth stages, different inter-crop spacing, variation in weather condition, and lighting. The processing time of the detection algorithm also needs to be small so that the desired number of image frames from continuous video can be processed in real-time. To cope with all the above mentioned requirements, we propose a crop row detection algorithm consisting of the following three linked stages: (1) color based segmentation for differentiating crop and weed from background, (2) differentiating crop and weed pixels using clustering algorithm and (3) robust line fitting to detect crop rows. We test the proposed algorithm over a wide variety of scenarios and compare its performance against four different types of existing strategies for crop row detection. Experimental results show that the proposed algorithm perform better than the competing algorithms with reasonable accuracy. We also perform additional experiment to test the robustness of the proposed algorithm over different values of the tuning parameters and over different clustering methods, such as, KMeans, MeanShift, Agglomerative, and HDBSCAN.
  • Thumbnail Image
    Item
    Experimental and Numerical investigation of hot-jet ignition with shock effects in a constant-volume combustor
    (Office of the Vice Chancellor for Research, 2015-04-17) Paik, Kyong-Yup; Khan, Nazmuzzaman; Tarraf Kojok, Ali; Nalim, M. Razi
    A wave rotor, an array of channels arranged around the axis of a cylindrical drum, can be used as a combustor in gas turbine engines in order to reduce the consumption of the fuel by increasing the fuel efficiency. Since the wave rotor combustor consumes fuel in constant volume channels, the engine system derives benefit from not only high temperature of the combusted gas, but also high pressure by containing the hot gas in the channels. Combustion of gas mixture in one of channels ignited by hot jet penetration under the necessity of rapid ignition accompanies complex non-steady phenomena, such as shock wave propagation, shock-flame interaction, and vortex generation in the channel. Especially, when a shock wave passes through the flame surface, the heat release rate and fuel consumption rate can be suddenly increased by a deformation of the flame surface, which are closely related with the combustion time of the fuel mixture. This research aims to investigate the ignition process, and the shock-flame interaction in a constant volume combustor experimentally and numerically to extract useful information for future wave rotor combustor design. Varıous mixtures of CH4 and H2 with equivalence ratio 1.0 were set as fuel for the main chamber, providing variation in chemical kinetic timescale. The hot gas jet consists of combusted gas mixture of a fuel composed of 50% CH4+ 50% H2 (by volume), burned in the pre-chamber with air at equivalence ratio 1.1. For experimental research, three dynamic pressure transducers were installed on the main chamber to measure the pressure changes caused by shock waves and flame propagation in the main chamber. Time-dependent flame and shock wave images up to 20,000 fps were obtained by a high speed camera, and a Z-type schlieren system. The schlieren technique, an optimum system to capture shock waves in the channel, utilizes light deviation due to flow density gradient, visualizing flows which are invisible to the human eye. In numerical research, adaptive mesh refinement for velocity and temperature, and multi-zone reaction modeling to speed up the kinetics were used to analyze turbulent combustion with minimum computational cost. Advanced post-processing techniques were used to calculate flame surface area, heat release rate, and vorticity deposited on flame surface to understand the flame wrinkling and surface increase. Finally, pressure data in main chamber, flame propagation speed, and the large scale of vortices under different initial conditions obtained from the experimental study were compared to the numerical results under the same conditions in order to suggest reference data for designing future wave rotors.
  • Thumbnail Image
    Item
    Object Detection from a Vehicle Using Deep Learning Network and Future Integration with Multi-Sensor Fusion Algorithm
    (SAE, 2017-03) Dheekonda, Raja Sekhar Rao; Panda, Sampad K.; Khan, Nazmuzzaman; Al-Hasan, Mohammad; Anwar, Sohel; Mechanical Engineering, School of Engineering and Technology
    Accuracy in detecting a moving object is critical to autonomous driving or advanced driver assistance systems (ADAS). By including the object classification from multiple sensor detections, the model of the object or environment can be identified more accurately. The critical parameters involved in improving the accuracy are the size and the speed of the moving object. All sensor data are to be used in defining a composite object representation so that it could be used for the class information in the core object’s description. This composite data can then be used by a deep learning network for complete perception fusion in order to solve the detection and tracking of moving objects problem. Camera image data from subsequent frames along the time axis in conjunction with the speed and size of the object will further contribute in developing better recognition algorithms. In this paper, we present preliminary results using only camera images for detecting various objects using deep learning network, as a first step toward multi-sensor fusion algorithm development. The simulation experiments based on camera images show encouraging results where the proposed deep learning network based detection algorithm was able to detect various objects with certain degree of confidence. A laboratory experimental setup is being commissioned where three different types of sensors, a digital camera with 8 megapixel resolution, a LIDAR with 40m range, and ultrasonic distance transducer sensors will be used for multi-sensor fusion to identify the object in real-time.