Saumil Patel Email & Phone Number

Sunnyvale, California, United States

Munich, Bavaria, Germany

Munich Area, Germany

3D Point Cloud Processing

Munich Area, Germany

Object Recognition in 3D Point Cloud Data of Indoor Environments1. The main goal of the master thesis was to retrieve similarly shaped objects from the point cloud of an indoor environment (e.g. a building) based on the user selected query object.2. In this master thesis, I evaluated the performance of state-of-the-art point cloud segmentation techniques: Euclidean clustering, Region-growing segmentation, Graph-cut segmentation and Conditional euclidean clustering on TUM-building point cloud.3. Then I performed object matching of the segmented clusters using local shape descriptors: PFH, FPFH, SHOT and global shape descriptors: VFH, CVFH, Shape functions and ESF. The addition of object size (obtained using Principal Component Analysis) to ESF provided the best recognition results.4. I was successfully able to perform shape retrieval of typical objects found in human environment such as chair, monitor, dustbin, door, etc. 5. With a little modification in the framework, it is also possible to provide semantic labeling to the point cloud if a reference database is available.

Unterschleißheim

Munich Area, Germany

Dense Map Reconstruction using Dense Tracking and Mapping (DTAM) AlgorithmPart A: Dense depth map using stereo camera1. Perform camera calibration to determine the camera intrinsic and extrinsic parameters.2. Generate left and right side images using a stereo camera.3. Removal of lens distortion from the captured images4. Generate dense depth map using DTAM algorithmPart B: Dense depth map using multiple images1. Perform camera calibration to determine the camera intrinsic and extrinsic parameters.2. Capture multiple images using a monocular camera.3. The camera position and orientation information is available along with each captured image4. Removal of lens distortion from the captured images5. Generate dense depth map using DTAM algorithm

Ahmedabad Area, India

Temperature and Motion Control for Optical Parametric Oscillators1. A laser beam with frequency f incident on a PPLN (Periodically Poled Lithium Niobate) crystal gets split into two beams with frequencies f1 and f2 (f=f1+f2). The output frequencies f1 and f2 depends on the temperature and position of PPLN w.r.t. laser beam.2. The goal of the project was to design a GUI on a display based Rabbit controller(OP 7200) and program the temperature and motion controllers.3. Based on this GUI, the user can simultaneously control the temperature controller (Oven Industries 5r7-001) and the motion controller (PI C-863 and PI M-122)

Master'S Degree, Communications Engineering

B.Tech, Electronics And Communications

Education record