Computer Vision Approaches for Plant Phenotypic Parameter Determination
KRISHI: Publication and Data Inventory Repository
View Archive InfoField | Value | |
Title |
Computer Vision Approaches for Plant Phenotypic Parameter Determination
Not Available |
|
Creator |
Alka Arora
Tanuj Misra Mohit Kumar Sudeep Marwaha Sudhir Kumar Viswanathan Chinnusamy |
|
Subject |
Computer vision
Deep learning Image analysis Machine learning m—Senescencica Plant phenomics |
|
Description |
Not Available
Climate change and the growing population are major challenges in the global agriculture scenario. High-quality crop genotypes are essential to counter the challenges. In plant breeding, phenotypic trait measurement is necessary to develop improved crop varieties. Plant phenotyping refers to studying the plant's morphological and physiological characteristics. Plant phenotypic traits like the number of spikes/panicle in cereal crops and senescence quantification play an important role in assessing functional plant biology, growth analysis, and net primary production. However, conventional plant phenotyping is time-consuming, labor-intensive, and error-prone. Computer vision-based techniques have emerged as an efficient method for non-invasive and non-destructive plant phenotyping over the last two decades. Therefore to measure these traits in high-throughput and non-destructive way, computer vision-based methodologies are proposed. For recognition and counting of number of spikes from visual images of wheat plant, a deep learning-based encoder-decoder network is developed. The precision, accuracy, and robustness (F1-score) of the approach for spike recognition are found as 98.97%, 98.07%, and 98.97%, respectively. For spike counting, the average precision, accuracy, and robustness are 98%, 93%, and 97%, respectively. The performance of the approach demonstrates that the encoder-decoder network-based approach is effective and robust for spike detection and counting. For senescence quantification, machine learning-based approach has been proposed which segments the wheat plant into different senescence and greenness classes. Six machine learning-based classifiers: decision tree, random forest, KNN, gradient boosting, naïve Bayes, and artificial neural network (ANN) are trained to segment the senescence portion from wheat plants. All the classifiers performed well, but ANN outperformed with 97.28% accuracy. After senescence segmentation, percentage of senescence area is also calculated. A GUI-based desktop application, m—Senescencica has been developed, which processes the input images and generates output for senescence percentage, plant height, and plant area. Not Available |
|
Date |
2023-06-06T13:15:47Z
2023-06-06T13:15:47Z 2023-05-20 |
|
Type |
Book chapter
|
|
Identifier |
Arora, A., Misra, T., Kumar, M., Marwaha, S., Kumar, S., Chinnusamy, V. (2023). Computer Vision Approaches for Plant Phenotypic Parameter Determination. In: Chaudhary, S., Biradar, C.M., Divakaran, S., Raval, M.S. (eds) Digital Ecosystem for Innovation in Agriculture. Studies in Big Data, vol 121. Springer, Singapore. https://doi.org/10.1007/978-981-99-0577-5_13
Not Available http://krishi.icar.gov.in/jspui/handle/123456789/78054 |
|
Language |
English
|
|
Relation |
Not Available;
|
|
Publisher |
Springer, Singapore
|
|