Yan-Fu Kuo

Quantifying floral shape variation in 3D using microcomputed tomography: a case study of a hybrid line between actinomorphic and zygomorphic flowers

The quantification of floral shape variations is difficult because flower structures are both diverse and complex. Traditionally, floral shape variations are quantified using the qualitative and linear measurements of two-dimensional (2D) images. The 2D images cannot adequately describe flower structures, and thus lead to unsatisfactory discrimination of the flower shape. This study aimed to acquire three-dimensional (3D) images by using microcomputed tomography (μCT) and to examine the floral shape variations by using geometric morphometrics (GM). To demonstrate the advantages of the 3D-μCT-GM approach, we applied the approach to a second-generation population of florists gloxinia (Sinningia speciosa) crossed from parents of zygomorphic and actinomorphic flowers. The flowers in the population considerably vary in size and shape, thereby served as good materials to test the applicability of the proposed phenotyping approach. Procedures were developed to acquire 3D volumetric flower images using a μCT scanner, to segment the flower regions from the background, and to select homologous characteristic points (i.e., landmarks) from the flower images for the subsequent GM analysis. The procedures identified 95 landmarks for each flower and thus improved the capability of describing and illustrating the flower shapes, compared with typically lower number of landmarks in 2D analyses. The GM analysis demonstrated that flower opening and dorsoventral symmetry were the principal shape variations of the flowers. The degrees of flower opening and corolla asymmetry were then subsequently quantified directly from the 3D flower images. The 3D-μCT-GM approach revealed shape variations that could not be identified using typical 2D approaches and accurately quantified the flower traits that presented a challenge in 2D images. The approach opens new avenues to investigate floral shape variations.

Strawberry foliar anthracnose assessment by hyperspectral imaging

Hyperspectral imaging has proven to be an effective non-destructive method for assessing strawberry foliar anthracnose.The incubation stage, in which symptoms are not yet visible, can be distinguished.Several hyperspectral imaging analysis methods were investigated using 3 duplicate sets of experiments.Significant wavelengths for strawberry foliar anthracnose are 551, 706, 750 and 914nm. Hyperspectral imaging provides comprehensive spectral and spatial information about observed objects. This technology has been applied to many fields, such as geology, mining, surveillance and agriculture. Strawberry qualities have been examined using hyperspectral imaging in several studies. However, none of the previous literature presented a non-destructive method for diagnosing the infection stages of anthracnose, a devastating disease for strawberries. This study examined strawberry foliar anthracnose using three different hyperspectral imaging analyzing methods: spectral angle mapper (SAM), stepwise discriminant analysis (SDA) and self-developed correlation measure (CM). Three different infection stages, including healthy, incubation and symptomatic stages, were investigated using these methods. The incubation stage is a stage at which the symptoms are still not yet visible. The three infection stage classification results were promising, with a classification accuracy of approximately 80%. For two infection stage classification (healthy and symptomatic stages), an average accuracy of high 80% was attained. In fact, an average accuracy of 93% was achieved by SDA for two-stage classification. This study not only proves the feasibility of hyperspectral imaging for diagnosing strawberry foliar anthracnose infection, but also identifies a smaller set of significant wavelengths at which similar classification performance was accomplished. For significant wavelength selection, partial least squares (PLS) regression is an standard wavelength selection method and it was applied to be compared with SDA and CM. Wavelengths of 551, 706, 750 and 914nm formed the multispectral imaging observing bands that showed an accuracy of 80% when classifying the three infection stages. Therefore, using either hyperspectral or multispectral imaging to detect anthracnose infected foliar areas is more practical and efficient than classic destructive methods. In particular, early detection (the incubation stage), something that cannot be seen via naked eyes, reaches 80% classification accuracy with both SDA and CM. Strawberry farmers could profit greatly from this technology.

Identifying rice grains using image analysis and sparse-representation-based classification

A microscope system was developed for acquiring high resolution grain images of 30 rice varieties.The morphological, textural, and color traits of the rice grains were quantified using image processing.Trait discrepancies among varieties were observed and explained.Sparse-representation-based classifier was developed to identify the varieties of the grains.The classification achieved an accuracy of 89.1% and a standard deviation of 7.0%. Rice (Oryza sativa L.) is a major staple food worldwide, and is traded extensively. The objective of this study is to distinguish the rice grains of 30 varieties nondestructively using image processing and sparse-representation-based classification (SRC). SRC uses over-complete bases to capture the representative traits of rice grains. In the experiments, rice grain images were acquired by microscopy. The morphological, color, and textural traits of the grain body, sterile lemmas, and brush were quantified. An SRC classifier was subsequently developed to identify the varieties of the grains using the traits as the inputs. The proposed approach could discriminate rice grain varieties with an accuracy of 89.1%.

Computer-aided assessment of tumor grade for breast cancer in ultrasound images.

This study involved developing a computer-aided diagnosis (CAD) system for discriminating the grades of breast cancer tumors in ultrasound (US) images. Histological tumor grades of breast cancer lesions are standard prognostic indicators. Tumor grade information enables physicians to determine appropriate treatments for their patients. US imaging is a noninvasive approach to breast cancer examination. In this study, 148 3-dimensional US images of malignant breast tumors were obtained. Textural, morphological, ellipsoid fitting, and posterior acoustic features were quantified to characterize the tumor masses. A support vector machine was developed to classify breast tumor grades as either low or high. The proposed CAD system achieved an accuracy of 85.14% (126/148), a sensitivity of 79.31% (23/29), a specificity of 86.55% (103/119), and an A Z of 0.7940.

Association between Petal Form Variation and CYC2-like Genotype in a Hybrid Line of Sinningia speciosa

This study used three-dimensional (3D) micro-computed tomography (μCT) imaging to examine petal form variation in a hybrid cross of Sinningia speciosa between a cultivar with actinomorphic flowers and a variety with zygomorphic flowers. The major objectives were to determine the genotype–phenotype associations between the petal form variation and CYCLOIDEA2-like alleles in S. speciosa (SsCYC) and to morphologically investigate the differences in petal types between actinomorphic and zygomorphic flowers. In this study, μCT was used to accurately acquire 3D floral images. Landmark-based geometric morphometrics (GM) was applied to evaluate the major form variations of the petals. Nine morphological traits of the petals were defined according to the form variations quantified through the GM analysis. The results indicated that the outward curvature of dorsal petals, the midrib asymmetry of lateral petals, and the dilation of ventral region of the tube were closely associated with the SsCYC genotype. Multiple analyses of form similarity between the petals suggested that the dorsal and ventral petals of actinomorphic plants resembled the ventral petals of zygomorphic plants. This observation indicated that the transition from zygomorphic to actinomorphic flowers in S. speciosa might be caused by the ventralization of the dorsal petals. We demonstrated that the 3D-GM approach can be used to determine genotype–phenotype associations and to provide morphological evidence for the transition of petal types between actinomorphic and zygomorphic flowers in S. speciosa.

Observation and Measurement of Residual Bran on Milled Rice Using Hyperspectral Imaging

ABSTRACT Residual bran on milled rice is directly related to its quality. This study proposes a method to measure the residual bran patterns on a single rice grain by using hyperspectral imaging (HSI). HSI is a sensing technique that combines both spatial and spectral information and may be used for chemical compound identification and quantification. In this study, HSI was applied to assess rice bran residue nondestructively. In the experiment, rice samples were milled and scanned with an HSI system. Afterward, the rice samples were dyed to enable the residual bran to be identified with optical microscopy and image processing algorithms. Classifiers were then developed to predict the rice bran residue by using the HSI measurements as inputs. The predicted images were compared with the micrograph images for classifier performance evaluation. The proposed approach can estimate the residual bran distribution on milled rice surface with an accuracy of 93.5%.

QR code detection using convolutional neural networks

Barcodes have been long used for data storage. Detecting and locating barcodes in images of complex background is an essential yet challenging step in the process of automatic barcode reading. This work proposed an algorithm that localizes and segments two-dimensional quick response (QR) barcodes. The localization involved a convolutional neural network that could detect partial QR barcodes. Majority voting was then applied to determine barcode locations. Then image processing algorithms were implemented to segment barcodes from the background. Experimental results shows that the proposed approach was robust to detect QR barcodes with rotation and deformation.

A Comparison of Machine Learning Methods on Hyperspectral Plant Disease Assessments

Abstract As plant diseases could cause agricultural production and economic loses, there is a need of fast and effective plant disease detection and assessment methods. Non-destructive methods have gained popularity among these methods as they do not affect plant growth while examining plant health conditions. Not only plant diseases can be detected but also production can be improved with proper quality controls. Hyperspectral imaging is one of the non-destructive examination techniques which have been widely applied in agriculture. Hyperspectral image analysis has been applied to different problems including plant disease detection and assessments. It provides not only spatial image but also spectral information of the observed object. This research has aimed to compare two hyperspectral image analysis methods: stepwise discriminant analysis (SDA) and spectral angle mapper (SAM) and the proposed Simple Slope Measure (SSM) method in strawberry foliage Anthracnose assessment. Anthracnose is one of the most devastating diseases for strawberries. Anthracnose disease can affect the whole plant and may result in 100 percent fruit loss from crown and fruit rot. Hence, an early detection of the Anthracnose disease will be beneficial to ensure production and quality of strawberries. This research has shown that the three different Anthracnose infection status (healthy, incubation and symptomatic) could be separated by the methods examined. The performance of these disease assessment models were evaluated and compared. The examination outcomes prove the feasibility to assess strawberry foliage Anthracnose nondestructively and as early as the symptoms not visible to naked eyes. As soon as early detections of the Anthracnose disease are achievable in the strawberry field, the damage to strawberry production due to the spread of Anthracnose disease could be reduced.

Developing a system for three-dimensional quantification of root traits of rice seedlings

Abstract A plant’s root system architecture (RSA) is the spatial configuration of its roots; the RSA of Oryza sativa L. (rice) shows a high degree of diversity. The RSA of rice should be quantified with high accuracy to understand the relationship between the RSA and functionality of rice roots. This study developed an imaging system for three-dimensional (3D) quantification of the RSA of rice. In this study, rice seedlings of 20 varieties were cultivated in glass tubes filled with transparent gellan gum for 10 days after germination. A servomotor-controlled camera captured two-dimensional side-view images of the seedlings at predetermined angles. A convolutional neural network classifier was then developed to segment the roots from the background. Subsequently, 3D images of the rice roots were constructed, and the phenotypic traits were quantified from the 3D images. Results displayed a high degree of diversity in the traits of the 20 varieties. A ground truth with designed parameters was used to validate the accuracy of this system. Analysis results indicated that the developed system was 98.3%, 97.6%, and 93.3% accurate regarding the primary root length, total root length, and root volume, respectively.

High-resolution phenotyping for investigating the genotype-phenotype association in rice grains

Abstract. Rice (Oryza sativa L.) shows remarkable variation in grains. The phenotypic information of the rice grains needs to be quantified accurately as the first step to investigate the association between the phenotypes and genotypes. This study proposed a procedure to phenotype rice grains of various cultivars in high resolution. In the process, rice seeds of more than 200 cultivars were acquired from Genetic Stocks Oryza. The genotypic variations (i.e., single nucleotide polymorphisms) of these cultivars are publicly available. The images of the grains were acquired in high resolution by microscopy (approximately 95 pixels per millimeter). Phenotypic traits were next quantified from the grain images automatically using image processing. The traits included morphological and color traits for the grain body, sterile lemmas, and brush. These traits were subsequently used for investigating the association between the phenotypic and genotypic variations of the cultivars using generalized linear model and mixed linear model. The analysis identified several candidate loci affecting morphological and color characters of the rice grains.