A Method for the Evaluation of Image Quality According to the Recognition Effectiveness of Objects in the Optical Remote Sensing Image Using Machine Learning Algorithm

Yuan, Tao; Zheng, Xinqi; Hu, Xuan; Zhou, Wei; Wang, Wei

doi:10.1371/journal.pone.0086528

Cited by 27 publications

(12 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The image set simply lacked the quality and clarity to reveal finer details of interest, and therefore, failed to yield species level accuracy, especially for species with near-identical elytral patterns. This is consistent with similar studies on species identification or pattern recognition and highlights the necessity of a high-quality image set for achieving good prediction accuracies 25 – 28 . This compelled us to develop a better imaging method which could yield elytral images of sufficient detail to enable us to distinguish between species from the same family or genus .…”

Section: Introductionsupporting

confidence: 90%

See 1 more Smart Citation

Optimized imaging methods for species-level identification of food-contaminating beetles

Bera

Ding³

et al. 2021

Sci Rep

View full text Add to dashboard Cite

Identifying the exact species of pantry beetle responsible for food contamination, is imperative in assessing the risks associated with contamination scenarios. Each beetle species is known to have unique patterns on their hardened forewings (known as elytra) through which they can be identified. Currently, this is done through manual microanalysis of the insect or their fragments in contaminated food samples. We envision that the use of automated pattern analysis would expedite and scale up the identification process. However, such automation would require images to be captured in a consistent manner, thereby enabling the creation of large repositories of high-quality images. Presently, there is no standard imaging technique for capturing images of beetle elytra, which consequently means, there is no standard method of beetle species identification through elytral pattern analysis. This deficiency inspired us to optimize and standardize imaging methods, especially for food-contaminating beetles. For this endeavor, we chose multiple species of beetles belonging to different families or genera that have near-identical elytral patterns, and thus are difficult to identify correctly at the species level. Our optimized imaging method provides enhanced images such that the elytral patterns between individual species could easily be distinguished from each other, through visual observation. We believe such standardization is critical in developing automated species identification of pantry beetles and/or other insects. This eventually may lead to improved taxonomical classification, allowing for better management of food contamination and ecological conservation.

show abstract

Section: Introductionsupporting

confidence: 90%

“…Such randomness in image illumination ( i.e. , brightness and contrast) makes this setting very inconsistent, ( e.g., highly dependent on random, external factors) and rendering it ineffective for automated pattern analysis, particularly in developing computer models for pattern recognition 27 , 28 .…”

Section: Resultsmentioning

confidence: 99%

Optimized imaging methods for species-level identification of food-contaminating beetles

Bera

Ding³

et al. 2021

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Like in [40], Yuan et al [46] used a Haar feature-based model for comparing image quality from different sensor designs but observed overhead imagery instead of face images. The methods used in that work to simulate imagery from different sensors are less robust than pyBSM, NV-IPM, or our method, involving simple addition of noise, blurring, and image contrast reduction.…”

Section: E Cv-based Modelingmentioning

confidence: 99%

Remote Sensor Design for Visual Recognition With Convolutional Neural Networks

Jaffe

Zelinski

Sakla

2019

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

While deep learning technologies for computer vision have developed rapidly since 2012, modeling of remote sensing systems has remained focused around human vision. In particular, remote sensing systems are usually constructed to optimize sensing cost-quality tradeoffs with respect to human image interpretability. While some recent studies have explored remote sensing system design as a function of simple computer vision algorithm performance, there has been little work relating this design to the state of the art in computer vision: deep learning with convolutional neural networks. We develop experimental systems to conduct this analysis, showing results with modern deep learning algorithms and recent overhead image data. Our results are compared to standard image quality measurements based on human visual perception, and we conclude not only that machine and human interpretability differ significantly but also that computer vision performance is largely self-consistent across a range of disparate conditions. This paper is presented as a cornerstone for a new generation of sensor design systems that focus on computer algorithm performance instead of human visual perception.

show abstract

“…Besides contrast and brightness, sharpness feature is another important image feature, which includes sharpness of image plane and sharpness of image edge. More attention has been paid to the edge of the image when it comes to sharpness feature (Tao et al, 2014;Sheng et al, 2015), which also makes up for the lack of contrast sensitivity in this aspect of contrast. Image edge is a set of pixels connected by the boundary between two regions of an image.…”

Section: Multi-feature Fusionmentioning

confidence: 99%

Subjective and Objective Quality Assessment of Swimming Pool Images

Lei

Xie

et al. 2022

Front. Neurosci.

View full text Add to dashboard Cite

As the research basis of image processing and computer vision research, image quality evaluation (IQA) has been widely used in different visual task fields. As far as we know, limited efforts have been made to date to gather swimming pool image databases and benchmark reliable objective quality models, so far. To filled this gap, in this paper we reported a new database of underwater swimming pool images for the first time, which is composed of 1500 images and associated subjective ratings recorded by 16 inexperienced observers. In addition, we proposed a main target area extraction and multi-feature fusion image quality assessment (MM-IQA) for a swimming pool environment, which performs pixel-level fusion for multiple features of the image on the premise of highlighting important detection objects. Meanwhile, a variety of well-established full-reference (FR) quality evaluation methods and partial no-reference (NR) quality evaluation algorithms are selected to verify the database we created. Extensive experimental results show that the proposed algorithm is superior to the most advanced image quality models in performance evaluation and the outcomes of subjective and objective quality assessment of most methods involved in the comparison have good correlation and consistency, which further indicating indicates that the establishment of a large-scale pool image quality assessment database is of wide applicability and importance.

show abstract

A Method for the Evaluation of Image Quality According to the Recognition Effectiveness of Objects in the Optical Remote Sensing Image Using Machine Learning Algorithm

Cited by 27 publications

References 11 publications

Optimized imaging methods for species-level identification of food-contaminating beetles

Optimized imaging methods for species-level identification of food-contaminating beetles

Remote Sensor Design for Visual Recognition With Convolutional Neural Networks

Subjective and Objective Quality Assessment of Swimming Pool Images

Contact Info

Product

Resources

About