Underwater Image Enhancement and Mosaicking System Based on A-KAZE Feature Matching

Abstract

Feature extraction and matching is a key component in image stitching and a critical step in advancing image reconstructions, machine vision and robotic perception algorithms. This paper presents a fast and robust underwater image mosaicking system based on (2D)2PCA and A-KAZE key-points extraction and optimal seam-line methods. The system utilizes image enhancement as a preprocessing step to improve quality and allow for greater keyframe extraction and matching performance, leading to better quality mosaicking. The application focus of this paper is underwater imaging and it demonstrates the suitability of the developed system in advanced underwater reconstructions. The results show that the proposed method can address the problems of noise, mismatching and quality issues which are typically found in underwater image datasets. The results demonstrate the proposed method as scale-invariant and show improvements in terms of processing speed and system robustness over other methods found in the literature.

Research on the method of quickly identifying and reading pointer meter

“…Because most of the collected instrument images have certain angle problems, in order to make the instrument detection more accurate, this paper adopts perspective transformation to calibrate the instrument. Because perspective transformation needs to find four corresponding transformation points, this paper uses A-Kaze algorithm [21] to match features.…”

Wafer Defect Detection Technology Based on CTM-IYOLOv5 Network

“…Data Augmentation is an effective way to expand data size and improve detection performance. Multiple data enhancement strategies are chosen, which can increase the number of training datasets and enrich the diversity of training datasets while improving the robustness and generalization of detection models [27]. In this paper, some effective data processing methods including image scrambling, brightness, contrast, saturation, hue, added noise, random scaling, random cropping, flipping, etc.…”

“…In this paper, some common metrics to evaluate deep learning network models, including Precision, Mean average precision(mAP), Detection time, and model Weights are chosen. In addition, another excellent one-stage target detectors, SSD [27], were trained using the same dataset, which will be used to compare die detection effects under different algorithms. The comparison results are shown in Table 3.…”

Image Filtering to Increase the Efficiency in the Construction of Underwater Photomosaics

“…Some authors advanced in the image matching and stitching algorithms, a very important part to estimate the camera motion reliably, which in turn has a direct influence in the correct location of each image in the mosaic frame. For instance, 1) Elnashef et al Elnashef and Filin (2021) improved the image aligning minimizing also the local distortion, with testing datasets recorded from an AUV in rectilinear transects, 2) Abaspur,et al Kazerouni et al (2020) applied (2D) 2 PCA Zhang and Zhou (2005) and A-KAZE Fernández Alcantarilla ( 2013) key-points extraction to match images of underwater pipes enhanced with a noise removal procedure based on the Fast Fourier Transform, and destined to the construction of photo-mosaics, and 3) Garcia-Fidalgo et al Garcia-Fidalgo et al (2016) reduced the image matching time characterizing them with bags of binary words; however, the process required a complete cross-wise comparison of each image with all the rest.…”