
Lightweight deep learning architectures are attractive for diagnosing plant diseases on low-cost field devices, but recent designs have rarely been evaluated on crop-specific tasks. We compared five recent lightweight architectures (MambaOut-Femto, FastViT-S12, MobileOne-S0, MobileNetV4-Conv-Small and EfficientViT-M0) with two conventional baselines (EfficientNet-B0 and ResNet-50) in two independent avocado (Persea americana) case studies. In Scenario 1, 3983 laboratory fruit images (healthy, scab and anthracnose), corresponding to 569 original photographs with seven flipped or rotated versions each, were grouped by photograph and evaluated with photograph-level stratified 5-fold cross-validation. In Scenario 2, 1897 field images of avocado leaves affected by three pests were evaluated with the same protocol. Differences in the Matthews correlation coefficient (MCC) were assessed with frequentist and Bayesian tests. On fruit images, MambaOut-Femto obtained the best performance (sensitivity 85.39%, specificity 92.78%, precision 86.00%, F1-score 85.44%, MCC 78.71% ± 2.05%, AUC 96.17%), followed by MobileOne-S0 (MCC 76.22%) and ResNet-50 (MCC 72.64%), with no significant differences among them, whereas the other four architectures obtained MCC values of 64.35–67.65%. On leaf images, all architectures exceeded an MCC of 93%; FastViT-S12 (sensitivity 97.70%, specificity 98.84%, precision 97.76%, MCC 96.55%) and MambaOut-Femto (97.60%, 98.78%, 97.69% and 96.42%, respectively) ranked highest, and the rankings of the two case studies were only partially consistent (Spearman’s ρ=0.64). MambaOut-Femto was the only architecture among the two best in both case studies, and MobileOne-S0, with 1.06 M parameters and 0.54 GFLOPs after reparameterization, is a lower-cost alternative for the fruit task.
The flow demand of multi-branch pipeline systems in air-suction seeders directly affects airflow stability at seed metering devices and fan capacity matching. To clarify the quantitative relationship between pipeline structural parameters and main pipe flow demand, and to avoid fan overcapacity and increased energy consumption caused by empirical estimation or single-branch extrapolation, the pneumatic system of an air-suction precision hill-drop rice seeder was investigated using mechanistic analysis, numerical simulation, and experimental validation. The effects of main pipe inner diameter and length, header inner diameter, branch pipe inner diameter and length, branch pipe spacing, and inlet flow rate of branch pipe 1 were examined. Results showed that these factors mainly altered branch pipe flow distribution by changing header pressure. Flow turning, jet mixing, and local resistance losses induced by successive confluence at tee junctions, together with frictional losses, governed the pressure-gradient variation along the header. A main pipe flow prediction model was established using dimensional analysis and interpolation. Predicted and experimental values agreed well (R2 > 0.99), and prediction errors were within ±10% over the investigated parameter range. The model effectively characterized main pipe flow demand under different pipeline configurations and operating conditions, providing technical support for low-energy pneumatic-system design and fan capacity matching in air-suction seeders.
Delayed photosynthetic induction can restrict carbon assimilation during the morning transition from darkness to light in greenhouse strawberry. The potential of light pretreatment to accelerate this process and improve fruit performance requires evaluation. We examined the effects of temperature, light quality, and light intensity on photosynthetic induction in Benihoppe strawberry (Fragaria × ananassa) and used the screening results to evaluate morning red-light supplementation during fruit development. The greenhouse treatment supplied red light at 40 μmol m−2 s−1 for 1 h before blanket opening over approximately 35 days. In the leaf-chamber experiments, monochromatic red light and a red–blue photon ratio of 6:1 produced significantly higher steady-state stomatal conductance than blue light. A 35 min red-light pretreatment at 40 μmol m−2 s−1, the lowest effective intensity tested, shortened subsequent photosynthetic induction by approximately 30 min compared with pretreatment at 20 μmol m−2 s−1. At the first harvest, total plant dry mass, average individual fruit weight, and soluble solid content increased significantly by 10.73%, 11.54%, and 12.75%, respectively, compared with the control without supplementary lighting. These findings support a morning red-light pretreatment strategy for accelerating photosynthetic induction and improving first-harvest fruit weight and quality under the tested greenhouse conditions.
Soybean is a major global source of vegetable oil and plant protein, making efficient harvesting an important component of crop production. Header losses remain a major limitation in mechanical soybean harvesting, particularly where low pods and mature, brittle plants increase the risk of pod shattering. This study designed, fabricated, and field-evaluated a 4.2 m soybean header that integrates a mechanically driven tine reel with distributed pneumatic assistance. The key design feature is a stationary central reel shaft that also functions as an air manifold, allowing 15 fixed nozzles to direct airflow into the reel–crop interaction zone while the outer six-tube reel rotates mechanically. A forward-curved centrifugal blower supplied air at approximately 8.5 m s−1 at the fan outlet and 8.0 m s−1 at the nozzles. The header was mounted on a John Deere 1055 combine and evaluated in four commercial fields in Ardabil Province, Iran, following national combine-test standards. Mean grain yield was 1346 kg ha−1. Mean header, Rear-Discharge Grain, and total losses were 2.42% (SD 0.24), 2.03% (SD 0.21), and 4.45% (SD 0.28), respectively. Effective field capacity was 1.065 ha h−1, and mean field efficiency was 80.25% (range 63–89%), indicating that actual field capacity averaged 80.25% of theoretical field capacity. Ambient relative humidity was not associated with header loss (Spearman rs = −0.26, p = 0.742), although inference was limited by the small sample and narrow humidity range. The prototype operated for approximately 20 h without observed structural failure, confirming its short-term structural integrity and retrofit compatibility. These results establish the field operability of the integrated architecture. However, the experiment did not include a fan-off treatment or a matched conventional-header control; therefore, the measured losses characterize the complete prototype, and the isolated pneumatic contribution to loss reduction could not be quantified. Controlled, replicated comparisons are required before any causal reduction in loss can be attributed to the pneumatic assistance. This study provides the necessary baseline for such future optimization and validation.
Rotor downwash influences crop motion during UAV spraying. Binocular high-speed photography was used to characterize the mean kinematic response of three 51-day-old cotton seedling plants at three positions beneath a quadrotor UAV: directly below a single rotor (P1), the midpoint between adjacent rotors (P2), and the fuselage center (P3). Orientation angles, accelerations, and trajectories were analyzed across the upper, middle, and lower canopy layers. Main-stem vibration amplitude decreased monotonically from apex to base. Combined peak-to-trough displacements at B/C/D were 25%/17%/8% of the apex value at P1 and 38%/16%/5% at P2; at P3, B/C retained 79%/69%, whereas D decreased to approximately 24%. Petiole trajectories were elongated ellipses with position- and layer-dependent axis ratios. Leaf motion was strongest at P2: maximum upper-canopy travel distance per cycle reached 1709.97 mm, compared with 304.38 mm at P1 and 205.35 mm at P3. Maximum middle-canopy accelerations were 12.59 m/s2 at P1 and 26.16 m/s2 at P2, approximately 1.7 times the upper- and lower-canopy values at P1 and 2.3 times the upper-canopy value at P2. At P3, maximum acceleration occurred in the upper canopy (16.35 m/s2). Overall, responses depended on rotor-relative position and canopy layer, with pronounced disturbance at P2 and in the middle canopy. The results suggest following the P2–P3 line and, where possible, the crop-row direction during spraying, potentially enhancing canopy motion and spray access to the middle and lower canopy and both leaf surfaces.