Bernd Stock

ORCID: 0000-0003-3642-7885
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Robotics and Sensor-Based Localization
  • Remote Sensing and LiDAR Applications
  • Advanced Vision and Imaging
  • 3D Surveying and Cultural Heritage
  • Medical Image Segmentation Techniques
  • Digital Imaging for Blood Diseases
  • Cell Image Analysis Techniques
  • Smart Agriculture and AI
  • 3D Shape Modeling and Analysis
  • AI in cancer detection
  • Advanced Electron Microscopy Techniques and Applications
  • Remote-Sensing Image Classification
  • Animal Behavior and Welfare Studies
  • Human-Animal Interaction Studies
  • Radiomics and Machine Learning in Medical Imaging
  • Advanced Image and Video Retrieval Techniques
  • Effects of Environmental Stressors on Livestock
  • Advanced Fluorescence Microscopy Techniques
  • Computer Graphics and Visualization Techniques

University of Applied Sciences and Arts Hildesheim/Holzminden/Göttingen
2021

HAW Hamburg
2014-2015

We present a general purpose, open-source software library for estimation of non-linear parameters by the Levenberg-Marquardt algorithm. The software, Gpufit, runs on Graphics Processing Unit (GPU) and executes computations in parallel, resulting significant gain performance. measured speed increase up to 42 times when comparing Gpufit with an identical CPU-based algorithm, no loss precision or accuracy. is designed such that it easily incorporated into existing applications adapted new...

10.1038/s41598-017-15313-9 article EN cc-by Scientific Reports 2017-11-10

The monitoring of farm animals and the automatic recognition deviant behavior have recently become increasingly important in animal science research practical agriculture. aim this study was to develop an approach automatically predict posture sows by using a 2D image-based deep neural network (DNN) for detection localization relevant sow pen features, followed hierarchical conditional statement based on human expert knowledge behavior/posture classification. body parts equipment trained...

10.3389/fanim.2021.758165 article EN cc-by Frontiers in Animal Science 2021-11-03

We present a novel method for photogrammetric wood pile surveying, which runs on mobile devices as well desktop computers. The demand measurement techniques piles has strongly increased in the last years. Unlike existing methods, our is not limited to single image and uses 3D reconstruction set of images taken with smartphone or digital camera. reconstructed model then used identify individual logs perform surveying entire pile. An extensive evaluation conducted 246 data sets (7655 images)...

10.1109/mva.2015.7153101 article EN 2015-05-01

A novel stitching method is presented that capable of automatically determining the panorama type, planar or rotational, form a set input images. The proposed distinguishes between two kinds panoramas, which we achieve through homography parametrization for estimation global camera motion. algorithm part complete image pipeline, where determined type utilized adaptive bundle adjustment with matching number degrees freedom (DOFs). For scenes homography's full eight DOFs are used, while three...

10.1145/2557642.2578222 article EN 2014-03-19

We present a general purpose, open-source software library for estimation of non-linear parameters by the Levenberg-Marquardt algorithm. The software, Gpufit, runs on Graphics Processing Unit (GPU) and executes computations in parallel, resulting significant gain performance. measured speed increase up to 42 times when comparing Gpufit with an identical CPU-based algorithm, no loss precision or accuracy. is designed such that it easily incorporated into existing applications adapted new...

10.1101/174110 preprint EN bioRxiv (Cold Spring Harbor Laboratory) 2017-08-10

We propose a novel method capable of detecting and segmenting quasi-planar surfaces based on homograph decomposition Semi-Global-Matching without the need for extrinsic calibration or stereo rectification. Existing methods require co planarity all points dominant plane are thus unsuited unconstrained surfaces. In contrast to state art we can account local depth variations. This is achieved by introducing planar inter-image rectification technique, which enables us perform block matching...

10.1109/crv.2015.19 article EN 2015-06-01
Coming Soon ...