Recognition of Plain Objects Using Local Region Matching

Mansur, Al; Sakata, Katsutoshi; Das, Dipankar; Kuno, Yoshinori

doi:10.1093/ietisy/e91-d.7.1906

Cited by 2 publications

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, performance of SIFT greatly depends on the object type. The approach used for SIFT is not robust enough to extract stable key points from a plain object Al Mansur [9].…”

Section: B Digital Methodsmentioning

confidence: 99%

Area measurement of seed from distorted images for quality seed selection

Chaugule¹,

Mali²

2013

2013 Nirma University International Conference on Engineering (NUiCONE)

View full text Add to dashboard Cite

Measurement of seed area is an important trait in studies of developmental physiology. However, direct measurement of this feature is difficult, not very precise and most of the time is destructive. The objective of the study presented in this paper is to develop a computer vision technique to determine the projected area of a seed for quality seed selection, irrespective of how the image is distorted. When a seed area is measured it is difficult to keep camera's optical axis vertical with seed plane. So the plane of camera is not superposed with image plane, and the seed image is distorted because of some geometric distortions. This may affect the precision of seed area measurement. Therefore the geometric distortion must be corrected before calculating the area. For improving the inconsistent feature problem caused by the distorted images, feature extraction strategies are proposed. The seed area is one of the features used for discriminant analysis. It is determined by using image processing and analysis technique using a conversion factor, for improving the inconsistent feature problem caused by different distortions. The imaging system developed, acquires and stores images of paddy seeds. Image digitization methodology was adapted and compared with the traditional manual method. For a digital imaging system to be able to predict the quality of the image, from large varieties of the image quality metrics available, six metrics are used. Correlation coefficient and Root Mean Square Error were used to compare the two methods. The method based on image digitization was found to be faster than the manual method.

show abstract

“…Moreover, performance of SIFT greatly depends on the object type. The approach used for SIFT is not robust enough to extract stable key points from a plain object Al Mansur [9].…”

Section: B Digital Methodsmentioning

confidence: 99%