China
Africa
Explore chapters and articles related to this topic
Potential of Thermal Imaging to Detect Complications in Diabetes
Published in U. Snekhalatha, K. Palani Thanaraj, Kurt Ammer, Artificial Intelligence-Based Infrared Thermal Image Processing and Its Applications, 2023
U. Snekhalatha, K. Palani Thanaraj, Kurt Ammer
Machine learning approaches rely on feature extraction processes for object detection, recognition, and classification. For computer vision problems and more specifically for image classification problems such as diabetes foot screening, there are a plethora of spatial and image features that are reported in literature. Some of the widely used image features are based on computing gray level co-occurrence matrix (GLCM) and extracting textural features such as contrast, correlation, moments, variance, entropy, homogeneity, and dissimilarity (Ashwin Kumaar and Thanaraj, 2015). Following that, a few researches also concentrated on extracting feature points or interest points for common computer vision applications such as image stitching, image matching, object detection, and recognition. One of the widely used and validated feature point extractor methods is the SIFT (Lowe, 2004b; Clemons, n.d.) and SURF (H. Bay et al., 2008). Though these methods provided excellent feature extraction performance, they are computation-intensive algorithms. A viable and alternate open-source feature extractor named ORB was proposed by Rublee et al. (2011), which provided good feature extraction performance with fewer computation subroutines. In this chapter, we review the application of ORB feature extractors for diabetes classification from anatomical thermograms.
A novel system for the monitoring of fatigue cracks in orthotropic steel-box girder
Published in Airong Chen, Xin Ruan, Dan M. Frangopol, Life-Cycle Civil Engineering: Innovation, Theory and Practice, 2021
Y.Q. Dong, Y. Pan, T. Ma, R.J. Ma, D.L.Wang
A novel machine vision-based system for monitoring of fatigue cracks in U-rib-to-deck weld seams is proposed in this study. The IoT based image acquisition device is firstly developed to observe the fatigue cracks, but only part-view crack images obtained. In order to acquire a panoramic crack image, we also design a specific coded calibration board. The panoramic crack image can be obtained after the developed image stitching algorithm. Afterwards, a crack recognition method is developed, containing crack semantic segmentation and the skeleton extraction algorithm. In the end, the area, length, width and direction of cracks are measured from the segmented masks and skeleton masks. By applying the whole system to monitor 14 fatigue cracks in a real OSG bridge, its feasibility is validated with satisfied results.
Crack identification and measurement of bridges by using CNN models
Published in Hiroshi Yokota, Dan M. Frangopol, Bridge Maintenance, Safety, Management, Life-Cycle Sustainability and Innovations, 2021
In practical applications, the images were obtained from an independently designed bridge inspection system, and the CNN model is used to identify the area where the crack exists. Due to the limitation of shooting distance and image definition, only the images in size of up to 15×10 cm can be acquired in a single time. When the image is captured, the distance between the camera and the surface to be detected remains fixed, which indicates the size of the acquired single image area is constant. However, in order to ensure that the detection information is not missed, there will be overlapped at the edges of adjacent images. As shown in Figure 6, the overlapping portion is clipped based on the position information of the image, and then the image is stitched . In the image mosaic process, the SIFT algorithm is also used, which not only has the invariance of scale, rotation, affine, angle of view and illumination, but also maintains a good matching degree for the effects of target motion, occlusion, noise, etc. The image stitching method of location information provides comparative verification.
Research and analysis on precise matching method for multi-feature of fuzzy digital image
Published in International Journal of Computers and Applications, 2020
In recent years, with the development of computer technology, fuzzy digital image multi-feature matching technology has been widely used and developed from the military field to people’s daily life and industrial production. Image feature matching method lays a good foundation for image stitching, scene recognition, image retrieval, and other technologies. However, image multi-feature matching has been the focus and difficulty in the fuzzy digital image processing research area. And the current method, during the matching process of fuzzy digital image multi-feature, cannot guarantee the image feature matching efficiency and accuracy. In this case, how to ensure the efficiency of image feature matching, while ensuring the accuracy of fuzzy digital image multi-feature matching, has become the focus of the current study [1]. The method uses the angle matching algorithm to match the multi-feature points of the fuzzy digital image, and performs the basic processing on the image before the feature matching, uses the mean filtering method to denoise to smooth the image, and retains the edge information of the fuzzy digital image [2]. Absolute gradient method and relative gradient method are used to strengthen the image. Afterwards, the fuzzy digital image gradient field function is utilized to calculate and amend the fuzzy digital image gradient field [1], and the K-means clustering algorithm is used to segment the fuzzy digital image to facilitate the extraction of fuzzy digital image feature, and then the pyramid principle is employed to extract the feature points of fuzzy digital image. Finally, the method and sparse matching method are adopted to match the multi-feature points of fuzzy digital images [3,4]. Thus, the fuzzy digital image multi-feature matching method based on ACDSEE is an effective method to solve the above problems, and it has great significance and value for research, and some research results have been obtained [5,6].
Intelligent driving system at opencast mines during foggy weather
Published in International Journal of Mining, Reclamation and Environment, 2022
Sushma Kumari, Monika Choudhary, Khushboo Kumari, Virendra Kumar, Abhishek Chowdhury, Swades Kumar Chaulya, Girendra Mohan Prasad, Sujit Kumar Mandal
The algorithm for real-time image stitching is depicted in Figure 6. The significant steps involved for image stitching are colour transfer, feature points detection, sorting of good feature points, feature point matching, warp perspective transforms for stitching, etc.