The paper proposes a new approach for obstacle detection, which is a crucial ingredient for autonomous operation of unmanned surface vehicles. The new method applies a novel semantic statistical model based on a Markov random field with weak semantic priors. We propose a new optimization for minimizing the field energy that affords extremely efficient optimization in real time. The proposed approach was evaluated on a real life system and surpasses all related approaches in accuracy as well as speed.
COBISS.SI-ID: 1536310979
Multi-image photogrammetry can in favorable conditions even under water generate large clouds of 3D points which can be used for visualization of sunken heritage. For analysis of under-water archeological sites and comparison of artifacts, more compact shape models must be reconstructed from 3D points, where each object or a part of it is modeled individually. Volumetric models and superquadric models in particular are good candidates for such modeling since automated methods for their reconstruction and segmentation from 3D points exist. For the study case we use an underwater wreck site of a Roman ship from 2nd/3rd century AD located near Sutivan on island Brac in Croatia. We demonstrate how super- quadric models of sarcophagi and other stone blocks can be reconstructed from an unsegmented cloud of 3D points obtained by multi-image photogrammetry. We compare the dimensions of stone objects measured directly on the corresponding 3D point cloud with dimensions of the reconstructed superquadric models and discuss other advantages of these volumetric models. The average difference between point-to-point measurements of stone blocks and the dimensions of the corresponding superquadric model is on the order of few centimeters.
COBISS.SI-ID: 1536404675
Cell counting in microscopic images is one of the fundamental analysis tools in life sciences, but is usually tedious, time consuming and prone to human error. Several programs for automatic cell counting have been developed so far, but most of them demand additional training or data input from the user. Most of them do not allow the users to online monitor the counting results, either. Therefore, we designed two straightforward, simple-to-use cell-counting programs that also allow users to correct the detection results. In this paper, we present the Cellcounter and Learn123 programs for automatic and semiautomatic counting of objects in fluorescent microscopic images (cells or cell nuclei) with a user-friendly interface. Although Cellcounter is based on predefined and fine-tuned set of filters optimized on sets of chosen experiments, Learn123 uses an evolutionary algorithm to determine the adapt filter parameters based on a learning set of images. Cellcounter also includes an extension for analysis of overlaying images. The efficiency of both programs was assessed on images of cells stained with different fluorescent dyes by comparing automatically obtained results with results that were manually annotated by an expert. With both programs, the correlation between automatic and manual counting was very high (R^2 ( 0.9), although Cellcounter had some difficulties processing images with no cells or weakly stained cells, where sometimes the background noise was recognized as an object of interest. Nevertheless, the differences between manual and automatic counting were small compared to variations between experimental repeats. Both programs significantly reduced the time required to process the acquired images from hours to minutes. The programs enable consistent, robust, fast and accurate detection of fluorescent objects and can therefore be applied to a range of different applications in different fields of life sciences where fluorescent labelling is used for quantification of various phenomena. Moreover, Cellcounter overlay extension also enables fast analysis of related images that would otherwise require image merging for accurate analysis, whereas Learn123's evolutionary algorithm can adapt counting parameters to specific sets of images of different experimental settings.
COBISS.SI-ID: 11059028
In this paper we identify problems with excessive feature sharing and the lack of discriminative learning in hierarchical compositional models. We propose to overcome those issues by fully utilizing a discriminative features already present in the generative models of hierarchical compositions. A discriminative descriptor is formed from existing library of parts and added at the top of the compositional model for discriminative verification of generatively-generated compositional proposals of object regions. Our approach results in state-of-the-art performance in partial occlusions compared to the convolutional neural networks.
COBISS.SI-ID: 1536363971
An algorithm for automatic digitization of pluviograph strip charts is presented. The rainfall signal is incrementally extracted from the scanned image of a strip chart by combining the moving average method and the curve edge following method. The mechanical properties of float-based rain gauge were used as constraints in the algorithm design. The algorithm was tested on 58 strip chart images. The comparison between the data derived from the algorithm and the data from the Slovenian Environment Agency shows that the algorithm produces an accurate rainfall time series except for the charts that contain ink smudges. Thus, the algorithm is well suited as a main component of an interactive system that would enable visual inspection of the detected rainfall curve and its possible adjustment.
COBISS.SI-ID: 1536679363