Developing an Accurate and Fast Non-Destructive Single Leaf Area Model for Loquat (Eriobotrya japonica Lindl) Cultivars

Scaling the leaf length-times-width equation to predict total leaf area of shoots

BACKGROUND AND AIMS

An individual plant consists of different-sized shoots, each of which consists of different-sized leaves. To predict plant-level physiological responses from the responses of individual leaves, modelling this within-shoot leaf size variation is necessary. Within-plant leaf trait variation has been well investigated in canopy photosynthesis models but less so in plant allometry. Therefore, integration of these two different approaches is needed.

METHODS

We focused on an established leaf-level relationship that the area of an individual leaf lamina is proportional to the product of its length and width. The geometric interpretation of this equation is that different-sized leaf laminas from a single species share the same basic form. Based on this shared basic form, we synthesized a new length-times-width equation predicting total shoot leaf area from the collective dimensions of leaves that comprise a shoot. Furthermore, we showed that several previously established empirical relationships, including the allometric relationships between total shoot leaf area, maximum individual leaf length within the shoot and total leaf number of the shoot, can be unified under the same geometric argument. We tested the model predictions using five species, all of which have simple leaves, selected from diverse taxa (Magnoliids, monocots and eudicots) and from different growth forms (trees, erect herbs and rosette herbs).

KEY RESULTS

For all five species, the length-times-width equation explained within-species variation of total leaf area of a shoot with high accuracy (R2 > 0.994). These strong relationships existed despite leaf dimensions scaling very differently between species. We also found good support for all derived predictions from the model (R2 > 0.85).

CONCLUSIONS

Our model can be incorporated to improve previous models of allometry that do not consider within-shoot size variation of individual leaves, providing a cross-scale linkage between individual leaf-size variation and shoot-size variation.

TMSCNet: A three-stage multi-branch self-correcting trait estimation network for RGB and depth images of lettuce

Growth traits, such as fresh weight, diameter, and leaf area, are pivotal indicators of growth status and the basis for the quality evaluation of lettuce. The time-consuming, laborious and inefficient method of manually measuring the traits of lettuce is still the mainstream. In this study, a three-stage multi-branch self-correcting trait estimation network (TMSCNet) for RGB and depth images of lettuce was proposed. The TMSCNet consisted of five models, of which two master models were used to preliminarily estimate the fresh weight (FW), dry weight (DW), height (H), diameter (D), and leaf area (LA) of lettuce, and three auxiliary models realized the automatic correction of the preliminary estimation results. To compare the performance, typical convolutional neural networks (CNNs) widely adopted in botany research were used. The results showed that the estimated values of the TMSCNet fitted the measurements well, with coefficient of determination (R ²) values of 0.9514, 0.9696, 0.9129, 0.8481, and 0.9495, normalized root mean square error (NRMSE) values of 15.63, 11.80, 11.40, 10.18, and 14.65% and normalized mean squared error (NMSE) value of 0.0826, which was superior to compared methods. Compared with previous studies on the estimation of lettuce traits, the performance of the TMSCNet was still better. The proposed method not only fully considered the correlation between different traits and designed a novel self-correcting structure based on this but also studied more lettuce traits than previous studies. The results indicated that the TMSCNet is an effective method to estimate the lettuce traits and will be extended to the high-throughput situation. Code is available at https://github.com/lxsfight/TMSCNet.git.

Application of image processing and soft computing strategies for non-destructive estimation of plum leaf area

Plant leaf area (LA) is a key metric in plant monitoring programs. Machine learning methods were used in this study to estimate the LA of four plum genotypes, including three greengage genotypes (Prunus domestica [subsp. italica var. claudiana.]) and a single myrobalan plum (prunus ceracifera), using leaf length (L) and width (W) values. To develop reliable models, 5548 leaves were subjected to experiments in two different years, 2019 and 2021. Image processing technique was used to extract dimensional leaf features, which were then fed into Linear Multivariate Regression (LMR), Support Vector Regression (SVR), Artificial Neural Networks (ANN), and the Adaptive Neuro-Fuzzy Inference System (ANFIS). Model evaluation on 2019 data revealed that the LMR structure LA = 0.007+0.687 L×W was the most accurate among the various LMR structures, with R² = 0.9955 and Root Mean Squared Error (RMSE) = 0.404. In this case, the linear kernel-based SVR yielded an R² of 0.9955 and an RMSE of 0.4871. The ANN (R² = 0.9969; RMSE = 0.3420) and ANFIS (R² = 0.9971; RMSE = 0.3240) models demonstrated greater accuracy than the LMR and SVR models. Evaluating the models mentioned above on data from various genotypes in 2021 proved their applicability for estimating LA with high accuracy in subsequent years. In another research segment, LA prediction models were developed using data from 2021, and evaluations demonstrated the superior performance of ANN and ANFIS compared to LMR and SVR models. ANFIS, ANN, LMR, and SVR exhibited R² values of 0.9971, 0.9969, 0.9950, and 0.9948, respectively. It was concluded that by combining image analysis and modeling through ANFIS, a highly accurate smart non-destructive LA measurement system could be developed.

Growth monitoring of greenhouse lettuce based on a convolutional neural network

Growth-related traits, such as aboveground biomass and leaf area, are critical indicators to characterize the growth of greenhouse lettuce. Currently, nondestructive methods for estimating growth-related traits are subject to limitations in that the methods are susceptible to noise and heavily rely on manually designed features. In this study, a method for monitoring the growth of greenhouse lettuce was proposed by using digital images and a convolutional neural network (CNN). Taking lettuce images as the input, a CNN model was trained to learn the relationship between images and the corresponding growth-related traits, i.e., leaf fresh weight (LFW), leaf dry weight (LDW), and leaf area (LA). To compare the results of the CNN model, widely adopted methods were also used. The results showed that the values estimated by CNN had good agreement with the actual measurements, with R2 values of 0.8938, 0.8910, and 0.9156 and normalized root mean square error (NRMSE) values of 26.00, 22.07, and 19.94%, outperforming the compared methods for all three growth-related traits. The obtained results showed that the CNN demonstrated superior estimation performance for the flat-type cultivars of Flandria and Tiberius compared with the curled-type cultivar of Locarno. Generalization tests were conducted by using images of Tiberius from another growing season. The results showed that the CNN was still capable of achieving accurate estimation of the growth-related traits, with R2 values of 0.9277, 0.9126, and 0.9251 and NRMSE values of 22.96, 37.29, and 27.60%. The results indicated that a CNN with digital images is a robust tool for the monitoring of the growth of greenhouse lettuce.

Analysis of Cultivar-Specific Variability in Size-Related Leaf Traits and Modeling of Single Leaf Area in Three Medicinal and Aromatic Plants: Ocimum basilicum L., Mentha Spp., and Salvia Spp

In this study, five allometric models were used to estimate the single leaf area of three well-known medicinal and aromatic plants (MAPs) species, namely basil (Ocimum basilicum L.), mint (Mentha spp.), and sage (Salvia spp.). MAPs world production is expected to rise up to 5 trillion US$ by 2050 and, therefore, there is a high interest in developing research related to this horticultural sector. Calibration of the models was obtained separately for three selected species by analyzing (a) the cultivar variability-i.e., 5 cultivars of basil (1094 leaves), 4 of mint (901 leaves), and 5 of sage (1103 leaves)-in the main two traits related to leaf size (leaf length, L, and leaf width, W) and (b) the relationship between these traits and single leaf area (LA). Validation of the chosen models was obtained for each species using an independent dataset, i.e., 487, 441, and 418 leaves, respectively, for basil (cv. 'Lettuce Leaf'), mint (cv. 'Comune'), and sage (cv. 'Comune'). Model calibration based on fast-track methodologies, such as those using one measured parameter (one-regressor models: L, W, L2, and W2) or on more accurate two-regressors models (L × W), allowed to achieve different levels of accuracy. This approach highlighted the importance of considering intra-specific variability before applying any models to a certain cultivar to predict single LA. Eventually, during the validation phase, although modeling of single LA based on W2 showed a good fitting (R2basil = 0.948; R2mint = 0.963; R2sage = 0.925), the distribution of the residuals was always unsatisfactory. On the other hand, two-regressor models (based on the product L × W) provided the best fitting and accuracy for basil (R2 = 0.992; RMSE = 0.327 cm2), mint (R2 = 0.998; RMSE = 0.222 cm2), and sage (R2 = 0.998; RMSE = 0.426 cm2).