Using k-NN to analyse images of diverse germination phenotypes and detect single seed germination in Miscanthus sinensis

Deep-learning-based automatic evaluation of rice seed germination rate

BACKGROUND

Rice is an important food crop plant in the world and is also a model plant for genetics and breeding research. The germination rate is an important indicator that measures the performance of rice seeds. Currently, solutions involving image processing techniques have substantial challenges in the identification of seed germination. The detection of rice seed germination without human intervention involves challenges because the rice seeds are small and densely distributed.

RESULTS

In this article, we develop a convolutional neural network (YOLO-r) that can detect the germination status of rice seeds and automatically evaluate the total number of germinations. Image partition, the Transformer encoder, a small target detection layer, and CDIoU loss are exploited in YOLO-r to improve the detection accuracy. A total of 21 429 seeds were collected, which have different phenotypic characteristics in length, shape, and color. The results show that the mean average precision of YOLO-r is 0.9539, which is higher than the compared models. Moreover, the average detection time per image of YOLO-r was 0.011 s, which meets the real-time requirements. The experimental results demonstrate that YOLO-r is robust to complex situations such as water stains, impurities, awns, adhesion, and so on. The results also show that the mean absolute error of the predicted germination rate mainly exists within 0.1.

CONCLUSIONS

Numerous experimental studies have demonstrated that YOLO-r can predict rice germination rate in a fast, easy, and accurate manner. © 2022 Society of Chemical Industry.

WSVAS: A YOLOv4 -based phenotyping platform for automatically detecting the salt tolerance of wheat based on seed germination vigour

Salt stress is one of the major environmental stress factors that affect and limit wheat production worldwide. Therefore, properly evaluating wheat genotypes during the germination stage could be one of the effective ways to improve yield. Currently, phenotypic identification platforms are widely used in the seed breeding process, which can improve the speed of detection compared with traditional methods. We developed the Wheat Seed Vigour Assessment System (WSVAS), which enables rapid and accurate detection of wheat seed germination using the lightweight convolutional neural network YOLOv4. The WSVAS system can automatically acquire, process and analyse image data of wheat varieties to evaluate the response of wheat seeds to salt stress under controlled environments. The WSVAS image acquisition system was set up to continuously acquire images of seeds of four wheat varieties under three types of salt stress. In this paper, we verified the accuracy of WSVAS by comparing manual scoring. The cumulative germination curves of wheat seeds of four genotypes under three salt stresses were also investigated. In this study, we compared three models, VGG16 + Faster R-CNN, ResNet50 + Faster R-CNN and YOLOv4. We found that YOLOv4 was the best model for wheat seed germination target detection, and the results showed that the model achieved an average detection accuracy (mAP) of 97.59%, a recall rate (Recall) of 97.35% and the detection speed was up to 6.82 FPS. This proved that the model could effectively detect the number of germinating seeds in wheat. In addition, the germination rate and germination index of the two indicators were highly correlated with germination vigour, indicating significant differences in salt tolerance amongst wheat varieties. WSVAS can quantify plant stress caused by salt stress and provides a powerful tool for salt-tolerant wheat breeding.

Genome-Wide Identification and Expression Profile of the HD-Zip Transcription Factor Family Associated with Seed Germination and Abiotic Stress Response in Miscanthus sinensis

Miscanthus sinensis is an ornamental grass, non-food bioenergy crop, and forage with high feeding value. It can adapt to many kinds of soil conditions due to its high level of resistance to various abiotic stresses. However, a low level of seed germination restricts the utilization and application of M. sinensis. It is reported that the Homeodomain-leucine zipper (HD-Zip) gene family participates in plant growth and development and ability to cope with outside environment stresses, which may potentially regulate seed germination and stress resistance in M. sinensis. In this study, a complete overview of M. sinensis HD-Zip genes was conducted, including gene structure, conserved motifs, chromosomal distribution, and gene duplication patterns. A total of 169 members were identified, and the HD-Zip proteins were divided into four subgroups. Inter-chromosomal evolutionary analysis revealed that four pairs of tandem duplicate genes and 72 segmental duplications were detected, suggesting the possible role of gene replication events in the amplification of the M. sinensis HD-Zip gene family. There was an uneven distribution of HD-Zip genes on 19 chromosomes of M. sinensis. Also, evolutionary analysis showed that M. sinensis HD-Zip gene family members had more collinearity with monocotyledons and less with dicotyledons. The gene structure analysis showed that there were 93.5% of proteins with motif 1 and motif 4, while motif 10 was only found in group IV. Based on the cis-elements analysis, it appeared that most of the genes were related to plant growth and development, various hormones, and abiotic stress. Furthermore, qRT-PCR analysis showed that Misin06G303300.1 was significantly expressed in seed germination and Misin05G030000.1 and Misin06G303300.1 were highly expressed under chromium, salt, and drought stress. Results in this study will provide a basis for further exploring the potential role of HD-Zip genes in stress responses and genetic improvement of M. sinensis seed germination.

Automatic Monitoring System for Seed Germination Test Based on Deep Learning

Germination test is an irreplaceable step in seed selection and breeding. The current traditional germination test method must rely on experienced professional technicians to repeatedly classify and count the germination status of seeds and count the germination rate at different moments during the whole test process (usually takes 2 to 10 days). Currently, only the German seed germination detection system (Germination Scanalyzer) can solve this problem, but it is so expensive that it has not been practically promoted. In order to improve breeding efficiency, an automatic monitoring system for seed germination tests based on deep learning was designed. It includes a modified germination thermostat, connected with a three-dimensional movable camera bin with built-in camera; a multifunctional software system capable of online, offline, and sentinel mode monitoring; a dense distributed small target detection algorithm (DDST-CenterNet) for seed germination monitoring systems. The system test results show that the seed germination test automatic monitoring system is low cost, does not depend on the seed background, light, camera model, and other usage environments, and has high scalability. The DDST-CenterNet algorithm proposed in this paper can still maintain high accuracy and good stability in the process of seed target detection and classification as the number and density of seeds increase, which is suitable for a special application scenario of seed germination test. In addition, the algorithm has high computational efficiency and can give detection results at a frame rate of not less than 10fps, which can be used in practical applications.

Accurate machine learning-based germination detection, prediction and quality assessment of three grain crops

BACKGROUND

Assessment of seed germination is an essential task for seed researchers to measure the quality and performance of seeds. Usually, seed assessments are done manually, which is a cumbersome, time consuming and error-prone process. Classical image analyses methods are not well suited for large-scale germination experiments, because they often rely on manual adjustments of color-based thresholds. We here propose a machine learning approach using modern artificial neural networks with region proposals for accurate seed germination detection and high-throughput seed germination experiments.

RESULTS

We generated labeled imaging data of the germination process of more than 2400 seeds for three different crops, Zea mays (maize), Secale cereale (rye) and Pennisetum glaucum (pearl millet), with a total of more than 23,000 images. Different state-of-the-art convolutional neural network (CNN) architectures with region proposals have been trained using transfer learning to automatically identify seeds within petri dishes and to predict whether the seeds germinated or not. Our proposed models achieved a high mean average precision (mAP) on a hold-out test data set of approximately 97.9%, 94.2% and 94.3% for Zea mays, Secale cereale and Pennisetum glaucum respectively. Further, various single-value germination indices, such as Mean Germination Time and Germination Uncertainty, can be computed more accurately with the predictions of our proposed model compared to manual countings.

CONCLUSION

Our proposed machine learning-based method can help to speed up the assessment of seed germination experiments for different seed cultivars. It has lower error rates and a higher performance compared to conventional and manual methods, leading to more accurate germination indices and quality assessments of seeds.

Yield prediction by machine learning from UAS-based mulit-sensor data fusion in soybean

BACKGROUND

Nowadays, automated phenotyping of plants is essential for precise and cost-effective improvement in the efficiency of crop genetics. In recent years, machine learning (ML) techniques have shown great success in the classification and modelling of crop parameters. In this research, we consider the capability of ML to perform grain yield prediction in soybeans by combining data from different optical sensors via RF (Random Forest) and XGBoost (eXtreme Gradient Boosting). During the 2018 growing season, a panel of 382 soybean recombinant inbred lines were evaluated in a yield trial at the Agronomy Center for Research and Education (ACRE) in West Lafayette (Indiana, USA). Images were acquired by the Parrot Sequoia Multispectral Sensor and the S.O.D.A. compact digital camera on board a senseFly eBee UAS (Unnamed Aircraft System) solution at R4 and early R5 growth stages. Next, a standard photogrammetric pipeline was carried out by SfM (Structure from Motion). Multispectral imagery serves to analyse the spectral response of the soybean end-member in 2D. In addition, RGB images were used to reconstruct the study area in 3D, evaluating the physiological growth dynamics per plot via height variations and crop volume estimations. As ground truth, destructive grain yield measurements were taken at the end of the growing season.

RESULTS

Algorithms and feature extraction techniques were combined to develop a regression model to predict final yield from imagery, achieving an accuracy of over 90.72% by RF and 91.36% by XGBoost.

CONCLUSIONS

Results provide practical information for the selection of phenotypes for breeding coming from UAS data as a decision support tool, affording constant operational improvement and proactive management for high spatial precision.

Utilization of computer vision and multispectral imaging techniques for classification of cowpea (Vigna unguiculata) seeds

BACKGROUND

The traditional methods for evaluating seeds are usually performed through destructive sampling followed by physical, physiological, biochemical and molecular determinations. Whilst proven to be effective, these approaches can be criticized as being destructive, time consuming, labor intensive and requiring experienced seed analysts. Thus, the objective of this study was to investigate the potential of computer vision and multispectral imaging systems supported with multivariate analysis for high-throughput classification of cowpea (Vigna unguiculata) seeds. An automated computer-vision germination system was utilized for uninterrupted monitoring of seeds during imbibition and germination to identify different categories of all individual seeds. By using spectral signatures of single cowpea seeds extracted from multispectral images, different multivariate analysis models based on linear discriminant analysis (LDA) were developed for classifying the seeds into different categories according to ageing, viability, seedling condition and speed of germination.

RESULTS

The results revealed that the LDA models had good accuracy in distinguishing 'Aged' and 'Non-aged' seeds with an overall correct classification (OCC) of 97.51, 96.76 and 97%, 'Germinated' and 'Non-germinated' seeds with OCC of 81.80, 79.05 and 81.0%, 'Early germinated', 'Medium germinated' and 'Dead' seeds with OCC of 77.21, 74.93 and 68.00% and among seeds that give 'Normal' and 'Abnormal' seedlings with OCC of 68.08, 64.34 and 62.00% in training, cross-validation and independent validation data sets, respectively. Image processing routines were also developed to exploit the full power of the multispectral imaging system in visualizing the difference among seed categories by applying the discriminant model in a pixel-wise manner.

CONCLUSION

The results demonstrated the capability of the multispectral imaging system in the ultraviolet, visible and shortwave near infrared range to provide the required information necessary for the discrimination of individual cowpea seeds to different classes. Considering the short time of image acquisition and limited sample preparation, this stat-of-the art multispectral imaging method and chemometric analysis in classifying seeds could be a valuable tool for on-line classification protocols in cost-effective real-time sorting and grading processes as it provides not only morphological and physical features but also chemical information for the seeds being examined. Implementing image processing algorithms specific for seed quality assessment along with the declining cost and increasing power of computer hardware is very efficient to make the development of such computer-integrated systems more attractive in automatic inspection of seed quality.