China
Africa
Explore chapters and articles related to this topic
Segmentation
Published in Ravishankar Chityala, Sridevi Pudipeddi, Image Processing and Acquisition using Python, 2020
Ravishankar Chityala, Sridevi Pudipeddi
In 1992, F. Meyer proposed an algorithm to segment color images, [Mey92], [Mey94]. Internally, cv2.watershed uses Meyer’s flooding algorithm to perform watershed segmentation. Meyer’s algorithm is outlined below: The original input image and the marker image are given as inputs.For each region in the marker image, its neighboring pixels are placed in a ranked list according to their gray levels.The pixel with the highest rank (highest gray level) is compared with the labeled region. If the pixels in the labeled region have the same gray level as the given pixel, then the pixel is included in the labeled region. Then a new ranked list with the neighbors is formed. This step contributes to the growth of the labeled region.The above step is repeated until there are no elements in the list.
WSNs Routing Protocols
Published in Amine Dahane, Nasr-Eddine Berrached, Mobile, Wireless and Sensor Networks, 2019
Amine Dahane, Nasr-Eddine Berrached
Flooding and gossiping are two classical mechanisms to relay data in sensor networks without the need for any routing algorithms and topology maintenance. In flooding, each sensor node will transfer those messages received to all theneighbor nodes, and this process will be repeated until the messages arrive at sink node or is overtime due Time-To-Live(TTL, usually defined as the largest hop in WSNs). Gossiping improves flooding algorithm in some ways, and each sensor node only transfer the messages to a random neighbor node. However, even though flooding and gossiping is very simple and suitable for any network structure, both algorithms are not practical in application specified network, and they can easily bring implosion and overlap problems (see Fig. 4.2 and Fig. 4.3).
Models of Artificial Systems, Networking, and Communications
Published in Gabriel A. Wainer, Discrete-Event Modeling and Simulation, 2017
We start by putting the value 100.1 in the output node; that is, we locate a search agent on the cell with the minimum distance to the destination. The flooding algorithm modifies each cell with the distance to the output node. The exploration rules check for an exploration agent in the neighborhood (0.1). They then check to ensure the cell is on exploration mode (i.e., it is occupied, and a distance is stored). In this case, the agent is moved to the cell, and one more is added to the distance covered. We then decide in which direction the nodes should continue the flooding. The backtracking rules check for nodes within the minimum distance path. The idea is to see if the neighbors belong to the shortest path and, in such a case, incorporate the cells to the minimum distance path (marking them with the value 5001). We can find a minimal path, as seen in Figure 13.24.
Determination of 3D pore network structure of freeze-dried maltodextrin
Published in Drying Technology, 2022
M. Thomik, S. Gruber, P. Foerst, E. Tsotsas, N. Vorhauer-Huget
To identify individual pore bodies within the pore space a marker-based watershed algorithm was applied, using the function “skimage.morphology.watershed” from the Python libraries. This algorithm can be described as follows. Starting from the binary image, an Euclidean distance transform was used to determine the distance from each foreground voxel to its closest background voxel usually referred to as a topographical relief. Local minima are the basins of this relief and were used as markers from which the watershed algorithm started to flood. To avoid oversegmentation the Euclidean distance map was filtered with Gaussian smoothing using σ = 0.4.[14,28] As explained in Gostick[14], this step removes falsely determined flooding points at saddles and plateaus and reduces oversegmentation. After this step, the remaining minima of the topographical relief were selected as markers. The flooding algorithm started at the proposed markers and proceeded until the whole image was flooded utilizing a priority queue. All voxels in individual basins or pores were assigned with specific labels. The resulting image was finally masked by the binary image to identify the segmented pore bodies.
Flotation Froth Phase Bubble Size Measurement
Published in Mineral Processing and Extractive Metallurgy Review, 2022
Segmentation is the process where an image is decomposed into meaningful and separate regions containing pixels with similar visual features (Kornilov and Safonov 2018; Preim and Botha 2014). Segmentation methods/algorithms are functions of two basic characteristics of luminance which are discontinuity and similarity (Gonzalez and Woods 2008). Thus, they can be subdivided into two approaches (Preim and Botha 2014) viz. (1) the edge-based approach where discontinuities are identified in the data from an image. The discontinuities are likely to belong to edges of an image structure (2) region-based methods in which the image structure is taken to be a homogeneous region. Watershed segmentation is a region-based approach where a grayscale image is taken to be a topological surface with ridges (high-amplitude pixels) and valleys (low-amplitude pixels) (Pratt 2001). In the flooding algorithm, single-pixel holes are pierced at local minima of a topography surface and water enters the hole filling the catchment basin. When the catchment is about to overflow a conceptual dam (watershed) is built. The procedure continues until every catchment basin is flooded and the whole image is covered by the watershed lines (Kornilov and Safonov 2018).
Performance and robustness of discrete and finite time average consensus algorithms
Published in International Journal of Systems Science, 2018
Luca Faramondi, Roberto Setola, Gabriele Oliva
The Flooding algorithm is, therefore, prone to the above attack strategy, which effectively steers the average to in a stealth manner (assuming the values sent are within the admissible range). It should be noted, moreover, that the overall number of nodes calculated by each node is also altered (i.e. each node believes there are or nodes overall), and that the completion time is increased with respect to the nominal FLD. If the nodes have a good estimate of , however, the above strategy is not effective, as the agents are all able to spot the attack by noting that the counted for the number of nodes is above n.