Actinidia deliciosa is a globally important economic fruit crop, and its fruit quality and yield are profoundly influenced by light and environmental conditions. Sucrose phosphate synthase (SPS), a key rate-limiting enzyme in the sucrose biosynthesis pathway, plays a central role in regulating carbon metabolism and sucrose accumulation in plants. However, comprehensive studies of the SPS gene family in A. deliciosa are still lacking, particularly regarding its expression in response to different light qualities. In this study, genome-wide identification of the SPS gene family in A. deliciosa was conducted using bioinformatics approaches. A total of 31 SPS genes were identified and named AdSPS1 to AdSPS31 on the basis of their chromosomal positions. The encoded proteins were predicted to be acidic, hydrophilic, and primarily localized in the chloroplast. All the AdSPS proteins contained the conserved domains Sucrose_synth, Glyco_trans_1, and S6PP, indicating potential roles in sucrose metabolism. Phylogenetic analysis classified the 31 AdSPS members into three subfamilies, A, B, and C, comprising 20, 5, and 6 members, respectively. Collinearity analysis revealed extensive syntenic relationships among AdSPS genes across different chromosomes, suggesting that gene duplication events contributed to the expansion of this gene family. Promoter cis-acting element analysis revealed that light-responsive elements were the most abundant among all the detected elements in the upstream regions of the AdSPS genes, implying potential regulation by light signals. Different light qualities significantly affected the contents of sucrose, glucose, and fructose, as well as SPS activity in kiwifruit leaves, with the highest activity observed under the R3B1 (red–blue light 3:1) treatment. Spearman’s correlation analysis indicated that AdSPS3 was significantly negatively correlated with sucrose, fructose, glucose, and SPS activity, suggesting a potential role in negatively regulating sugar accumulation in kiwifruit leaves, whereas AdSPS12 showed positive correlations with these parameters, implying a role in promoting sucrose synthesis. To further explore the light response of the AdSPS genes, eight representative members were selected for qRT‒PCR analysis under red light, blue light, and combined red‒blue light treatments. These results demonstrated that light quality significantly influenced SPS gene expression. Specifically, AdSPS6 and AdSPS24 were highly responsive to R1B1 (1:1 red‒blue light), AdSPS9 was significantly upregulated under R6B1 (6:1 red‒blue light), AdSPS21 was strongly induced by blue light, and AdSPS12 expression was suppressed. This study systematically identified and analyzed the SPS gene family in A. deliciosa, revealing its structural characteristics and light-responsive expression patterns. These findings suggest that AdSPS genes may play important roles in light-regulated carbon metabolism. These results provide a theoretical foundation and valuable genetic resources for further elucidating the molecular mechanisms of sucrose metabolism and light signal transduction in kiwifruit.

To investigate the effects of different seedless treatments on grape coloring and fruit quality, Vitis vinifera × Vitis labrusca cv. ‘Jumeigui’ were treated with different concentrations of forchlorfenuron (CPPU) (0.5, 1 and 1.5 mg/L), thidiazuron (TDZ) (0.5, 1 and 1.5 mg/L), and 6-benzyladenine (6-BA) (10, 20 and 30 mg/L) in combination with 18 mg/L gibberellic acid (GA3) during the seedless-fruit-setting period. After the grapes ripened, multiple quality indicators were measured to analyze and evaluate the effects of different treatments on the fruit coloration and quality of ‘Jumeigui’ grapes. The results showed that increasing concentrations of CPPU and TDZ gradually reduced the comprehensive fruit quality of ‘Jumeigui’ grapes. The treatments with 18 mg/L GA3 + 0.5 mg/L CPPU/TDZ were relatively effective in improving the comprehensive quality of ‘Jumeigui’ grapes. With increasing concentrations of 6-BA, the comprehensive effect initially increased and then decreased. The treatment with 18 mg/L GA3 + 20 mg/L 6-BA resulted in a soluble solids content of 20.03% and a coloring index of 4.10, demonstrating the best overall improvement in the comprehensive quality of ‘Jumeigui’ grapes. Based on practical production considerations, it is recommended to apply 18 mg/L GA3 + 20 mg/L 6-BA during the seedless-fruit-setting period of ‘Jumeigui’ grapes to enhance coloring effects and improve fruit quality.

Pinching is a summer pruning practice in kiwifruit cultivation that involves permanently removing the shoot's growing tip after the distal bud. Automating this practice requires accurate positioning of pinching points, which remains challenging due to dense canopies and variability in shoot type. This study proposed a novel automated method for positioning pinching points by combining YOLO11s-seg for segmenting branches and shoots with a morphology-guided kiwifruit pinching point positioning algorithm (KPPPA). KPPPA integrated two sequential sub-algorithms: the direction-aware mask mending (MaskMend) reconstructed the occlusion-induced fragmented mask of shoots by estimating the local search direction for candidate reconnection extreme endpoints; then, proportion-based point locating (PPL) utilized the reconstructed mask for computing shoot geometric features to distinguish non-terminating from terminating shoots and ultimately position optimal pinching points. Experimental results show that YOLO11s-seg model achieved a mean average precision of 91.7% for detection and 61.1% for mask segmentation, while KPPPA positioned pinching points obtaining success rate of 87.0% and Average Euclidean error of 5.8 pixels with Mean absolute errors of 3.6 and 2.7 pixels along x, and y-axis, respectively. The findings of the study show that the proposed method can effectively position pinching points for kiwifruit in summer at a minimal computation speed of 136.6 ms per frame, making it feasible for real-time robotic pinching operation in kiwifruit orchards.

Accurate detection and pose estimation of kiwifruit are essential for robotic harvesting in modern orchards, which remains challenging due to complex lighting conditions and dense canopy in natural environments. In this study, a dual-view camera-LiDAR fusion methodology based on Depth-Iterative Canopy Filtering (DICF) method was proposed for precise pose estimation of kiwifruits. The proposed methodology aimed to overcome limitations of existing camera-based fruit pose estimation research by enhancing robustness under variable lighting and complex canopy conditions. The methodology integrated a lightweight end-to-end stereo matching network with the DICF, leveraging binocular disparity maps to guide LiDAR point-cloud adhesion noise suppression. YOLO11m was employed for kiwifruit detection, dual-view fusion increased detection accuracy of occluded fruits to 92.8%, representing a 28.5% improvement over single-view detection. Experiments conducted in commercial orchards demonstrated that the methodology achieved a mean absolute localization error of 6.54 mm and an orientation angle error of 7.70 degrees. Moreover, filtering time was reduced to 1.12 s for point clouds containing 700,000 points, compared to 3.56 s required by the TSPFA, meeting the requirements of robotic harvesting. This study established a novel and efficient methodology for kiwifruit pose estimation and provided a foundation for end-to-end multimodal sensor fusion in future agricultural robotics.

Kiwifruit winter pruning is a laborious, seasonal, and irreversible operation. The key to achieving automatic winter pruning of kiwifruit lies in the recognition of pruning position. In this study, a method based on kiwifruit vine segmentation and dormant bud detection with their planar distribution was developed to recognize the pruning position of kiwifruit vine. Firstly, YOLO11x-seg and YOLO11x were applied to segment vines and detect dormant buds, respectively. Secondly, the DBSCAN algorithm was applied to remove wrongly segmented vine, and vine skeleton was then extracted. Thirdly, dormant buds were matched with their corresponding canes based on the overlap ratio, and then ordered from the cane base. Finally, the pruning position was recognized based on the planar distribution of vines and dormant buds. Results showed that YOLO11x-seg and YOLO11x achieved an outstanding performance with segmentation and detection mAP of 82.5% and 89.0%, respectively. The method proposed in this study attained pruning position recognition accuracies of 77.3%, 81.8%, and 88.9% under the three pruning strategies of cane short-truncation, lateral cane removal, and cane renewal, respectively. These results indicate that the proposed method is promising for robotic winter pruning of kiwifruit.

Currently, the methods for identifying agro-products origin usually focus on a single algorithm and overemphasize predictive performance at the expense of interpretability, which greatly hampers the development of accurate and effective traceability methods as well as insights into the intrinsic mechanisms of the models. Therefore, using kiwifruit from six regions in China, this study combines hyperspectral imaging (HSI) technology with ensemble learning methods to construct a kiwifruit origin traceability model and utilizes explainable artificial intelligence (AI) to hierarchically and holistically interpret the optimal stacking ensemble model, which aims to improve the traceability accuracy while providing an in-depth understanding of the logic behind origin prediction using spectral features. The results revealed the stacking ensemble model, with Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Random Forest (RF) as base-learners, and Logistic Regression (LR) as the meta-learner, demonstrated the best discrimination performance with 95.24% accuracy, 95.79% precision, 95.24% recall, and 95.19% F1-score, the classification accuracy of whose was improved by 2.38% than that of the best single model. Furthermore, SHapley Additive exPlanations (SHAP) analysis elucidated the decision mechanism of the optimal traceability model by quantifying the contribution of spectral features, with wavelengths of 730 nm (SHAP 0.59) and 893 nm (SHAP 0.56) being dominant for origin prediction under the kiwifruit sample conditions of this experiment. These findings serve as a significant scientific groundwork for further optimizing the application of spectral techniques in kiwifruit traceability systems, and provide a perspective for applying this technology to distinguish the geographical origin of other agro-products.

Grid-distributed wires in T-trellis kiwifruit orchards provide structural support for dense canopy growth and enable mechanized operations such as robotic harvesting. However, these wires also introduce potential risks of collision and damage to the robot end-effector, particularly when target fruits are partially occluded by thin and visually inconspicuous wire structures. Therefore, accurate identification of wire segments and reconstruction of wire layouts are essential for safe and efficient robotic picking strategies. In this study, an identification and reconstruction method for grid-distributed wires in dense kiwifruit canopies was presented based on Geometric Feature Grouping-connection (GFG). First, fragmented wire segments were extracted from canopy images using a YOLO11x-seg model combined with an Image Overlap-partitioning and Stitching (IOS) strategy. Then, the proposed GFG approach grouped and connected discrete wire segments through directional classification, spatial grouping, and boundary extension, enabling complete reconstruction of grid-distributed wire structures. Experimental results show that the model trained on the overlap-partitioned dataset achieved an average precision of 28.50% at an Intersection over Union (IoU) threshold of 0.5, representing a 13.50% improvement over the model trained on the original dataset. With the integration of IOS strategy, the model achieved an IoU of 21.19% and a Pixel Accuracy (PA) of 21.00%, corresponding to improvements of 9.54% and 10.00%, respectively, compared with the baseline without IOS. In addition, the proposed GFG approach achieved a reconstruction accuracy of 78.28% for grid-distributed wire structures. Overall, these findings indicate that the presented method enables the reliable identification of fragmented wire segments and reconstruction of continuous grid-distributed wire structures in dense canopy environments. This study highlights the importance of reconstructing non-target support structures for improving orchard scene understanding and providing useful geometric information for future obstacle-aware robotic harvesting.

Kiwifruit (Actinidia Lindl.) acclaimed as the“king of fruits”due to its rich nutritional composition, is an important horticultural crop with high economical and nutritional value. It is rich in sugar, vitamin C, dietary fiber, polyphenols, and various bioactive compounds, which contributes to its strong antioxidant capacity and numerous health benefits. In response to the growing market demand for high-quality, safe, and convenient fruit products, the modernization of kiwifruit harvesting and postharvest handling has become essential for industrial advancement and sustainable development. This review summarizes recent progress in harvest maturity evaluation, storage and preservation techniques, ripening control, and quality regulation of kiwifruit. Studies have indicated that determining harvest maturity based on soluble solids content (SSC), fruit firmness, and dry matter content is critical for maintaining postharvest quality and ensuring optimal flavor potential. SSC is closely related to sugar accumulation and sweetness, firmness determines texture and damage resistance during transport, and dry matter serves as an effective predictor of flavor development. Scientific evaluation of these parameters provides a foundation for defining the optimal harvest period and achieving a balance between storability and eating quality. Postharvest handling has attracted considerable research attention in recent years.Low-temperature storage remains the most fundamental and effective method for slowing fruit respiration, suppressing ethylene biosynthesis, and extending shelf life. Variable-temperature storage simulates natural temperature fluctuations to reduce chilling injury, maintain cell membrane integrity, and preserve flavor compounds. Controlled-atmosphere storage further enhances postharvest stability by adjusting gas composition to delay senescence. The application of 1-methylcyclopropene (1-MCP), an ethylene receptor inhibitor, has also been shown to effectively delay softening and maintain firmness, and nutritional integrity during long-term storage and transport. Ripening regulation technologies are equally important for achieving the desired eating quality of kiwifruit. Ethylene treatment promotes uniform softening and aroma development, while temperature-controlled ripening enables coordinated regulation of ripening and flavor formation. The integration of ethylene management with temperature control allows a more precise and controllable ripening process, thereby improving consumer satisfaction and commercial value. In addition, emerging quality regulation techniques, such as antioxidant application, calcium-based firmness regulation, and natural or edible coating treatments, have shown promising results in delaying senescence and maintaining the visual and textural quality of fruit during storage. The integration of diverse preservation and ripening technologies establishes a comprehensive postharvest handling system that supports large-scale, standardized, and high-value production of kiwifruit. Furthermore, the development of digital traceability, intelligent monitoring, and predictive modeling technologies offers new opportunities to improve supply chain transparency and optimize storage and distribution management. By combining traditional postharvest physiology with modern intelligent management tools, the kiwifruit industry can enhance product quality while promoting sustainability. In summary, this review emphasizes the current progress and future perspectives of kiwifruit harvesting standard and postharvest handling technologies. It emphasizes the importance of integrating scientific harvest standards with advanced storage, preservation, and ripening strategies to maintain fruit quality and economic value. Continued research on the molecular mechanisms of ripening, digital management systems, and environmentally friendly preservation technologies will further promote the sustainable and high-quality development of the kiwifruit industry.

The vigorous growth of new shoots can significantly reduce grape yield and compromise fruit quality. In order to explore the effects of prohexadione calcium (Pro-Ca) and mepiquat chloride (MC) on the control effect of new shoot growth and fruit quality of grape, ‘Shine Muscat’ grapevine (Vitis labruscana × V. vinifera) was used as the test material, and different concentrations of Pro-Ca and a combination of Pro-Ca and MC were sprayed four times before flowering of ‘Shine Muscat’ grapevines, and the effects of the different treatments on the new shoot growth and fruit quality of ‘Shine Muscat’ grape were analyzed and evaluated. The results demonstrated that low concentrations of Pro-Ca had limited efficacy in controlling shoot growth. However, the combined treatment of Pro-Ca 300 mg/L + MC 300 mg/L not only effectively inhibited shoot elongation but also significantly enhanced the chlorophyll content of the leaves opposite to the clusters and increased branch density. Additionally, this treatment improved berry size (single berry weight, vertical and horizontal diameter) and elevated the soluble solids content (SSC). These findings suggest that the combined application of Pro-Ca (300 mg/L) and MC (300 mg/L) is the most effective strategy for balancing vegetative growth and enhancing fruit quality in ‘Shine Muscat’ grapevines.

Biostimulants have been increasingly investigated as eco-friendly alternatives to synthetic fruit enlargement agents in horticulture. In this study, two commercially cultivated kiwifruit cultivars, Zhongmi No. 2 and Jintao, were used as experimental materials. Three biostimulant products with distinct functional compositions were investigated: Benefit PZ (BPZ), which is rich in potassium humate and organic nitrogen; Shengcai A (SCA), containing amino acids and trace elements; and Puluosaiting (PLST), a natural seaweed extract-based formulation rich in bioactive compounds. Their effects on fruit development and internal quality attributes were compared with those of two fruit enlargement agents, 6-benzylaminopurine (6-BA) and forchlorfenuron (CPPU). Field experiments were conducted in two orchards located in Dancheng and Xixia, Henan Province, China, and treatments were applied during early fruit development. Growth traits (longitudinal and transverse diameters, single-fruit weight and firmness) and quality indicators (Soluble Solids Content, Titratable Acidity and Dry Matter Content) were measured at commercial maturity. CPPU and 6-BA substantially increased fruit size and weight compared with the control, whereas biostimulants produced moderate improvements without excessive enlargement. Notably, biostimulant treatments consistently enhanced internal quality attributes, indicating their potential to improve fruit quality without the drawbacks of excessive enlargement. Environmental and management differences between sites may also have contributed to treatment variability. These results suggest that biostimulants can improve internal quality traits while avoiding excessive fruit enlargement, representing a promising option for sustainable kiwifruit production.

Actinidia arguta is a newly emerged, commercially cultivated Actinidia species. A. arguta has a beautiful appearance and is rich in anthocyanin, and is thus highly welcomed by consumers. However, the mechanism of anthocyanin regulation in A. arguta remains unclear. In this study, we assembled the nearly complete genome of the first red A. arguta cultivar, ‘Tianyuanhong’, with an N50 of 21 Mb. Comparative genome analysis revealed a role of the expansion/contraction of gene families in the species-specific trait formation of A. arguta. Through verification of transient overexpression and stable transformation, RNA-seq analysis revealed a key bHLH transcription factor, AaBEE1, which negatively regulates anthocyanin biosynthesis. DAP-seq analysis combined with Y1H, EMSA, Chip-qPCR and LUC suggested that AaBEE1 binds to the G-box of the AaLDOX promoter and suppresses its expression. Overall, we assembled the genome of A. arguta and clarified its AaBEE1-AaLDOX module-mediated molecular mechanism of anthocyanin regulation.

Digital Twins (DTs) represent a new tool for enabling data-driven orchard management, with the goal of optimizing resource allocation and decision-making. However, unlike general agriculture or annual horticultural systems, fruit tree production is constrained by fixed seasonal cycles, making it difficult to conduct experiments or interventions outside these periods. In addition, the perennial nature of fruit trees, along with complex canopy structures and long production cycles, further increases the complexity and variability of orchard management. An orchard DT consists of a virtual counterpart of the physical orchard entity, with real-time integration of sensor data and a model allowing predictions of system behavior. DTs have the potential to cover all stages of tree-fruit production, from cultivation to post-harvest. This review article of the literature up to 2025 systematically summarizes and indicates that the application of DTs in orchard management remains in an exploratory stage, reflecting the state of development and adoption of enabling technologies, such as the Internet of Things, artificial intelligence, cloud computing, edge computing, extended reality, communications, and blockchain. DTs have been developed for orchard establishment, operations, harvest forecast and optimization, robotic harvesting, natural disaster response, and orchard inventory, with the dominant focus being on harvest operations. Additionally, DTs have been applied to optimize specific orchard processes, such as modelling spray droplet movement within canopies in support of the design of spraying equipment. Very few applications have involved a control system. Commonalities observed in the development of existing orchard DT models suggest the potential for a standardized or universal DT model to support expanded automation operations. Beyond routine orchard management, the potential application of DTs in areas such as natural disaster response is also highlighted, offering opportunities for cost sharing and broader cross-sector benefits.

Actinidia arguta has become popular with consumers recently because of its edible and colorful fruit skin. The 3D spatial organization of its genome plays a key role in the formation of various biological traits. However, the function of 3D genome reorganization during fruit skin color formation is poorly understood in A. arguta. In this study we constructed the 3D genome of the red-skinned A. arguta cultivar 'Zhonghongbei' (ZHB) and the green-skinned cultivar 'Zhonglvbei' (ZLB), and performed chromatin structure comparisons between them at compartment, topologically associating domain (TAD), and loop levels. Global compartment comparisons at whole 3D genome level between red-skinned and green-skinned A. arguta showed that A-B compartment transition specifically occurred in chromosome 7 and chromosome 16, based on which all genes within 3 Mb upstream and downstream of A-B compartment transition were retrieved to construct a four-way Venn diagram, which showed that AaCBP60B-like, encoding calmodulin-binding protein 60 B-like, is the key candidate gene negatively correlating with fruit color. Exogenous calcium chloride treatments enhancing AaCBP60B-like expression to repress anthocyanin biosynthesis proved a negative role of AaCBP60B-like in anthocyanin biosynthesis. Overexpression and virus-induced gene silencing assays of AaCBP60B-like revealed the inhibition of anthocyanin biosynthesis derived from differential expression of AaCBP60B-like resulting from a 346-bp InDel variation located at the AaCBP60B-like promoter resulting in activity differences in red- and green-skinned A. arguta. ATAC-seq results proved that the 346-bp InDel variation affects 3D genome organization. Our study provides the first 3D chromosome organization in red- and green-skinned A. arguta, based on which a candidate gene, AaCBP60B-like, involved in anthocyanin regulation is identified.

Excessive shoot vigor in grapevines negatively impacts plant growth and fruit quality, necessitating the use of plant growth regulators (PGRs) for canopy management. This study investigated the effects of mepiquat chloride (MC) and chlormequat chloride (CCC) on shoot growth (including new shoot length, relative chlorophyll content, leaf area, etc.) and fruit quality in Vitis vinifera cv. ‘Shine Muscat’. Different concentrations of MC (100, 300, 500, 700 mg/L) and CCC (100, 300, 500, 700 mg/L) were applied via foliar spraying at multiple stages before flowering. The results demonstrated that both PGRs effectively suppressed shoot elongation, with CCC exhibiting superior inhibitory efficacy compared to MC. However, high concentration of either compound also restricted leaf and cluster development. Optimal treatments MC (500 mg/L) and CCC (100 mg/L) significantly enhanced berry size, soluble solids content (SSC), and solid–acid ratio while maintaining effective shoot control. For practical application, we recommend spraying MC (500 mg/L) or CCC (100 mg/L) during the new shoot growth, flower-cluster separation, and flowering stages of ‘Shine Muscat’ grapevines to improve the new shoot control effect and fruit quality.

ABSTRACTManual pollination of kiwifruit flowers is a labor‐intensive work that is highly desired to be replaced by robotic operations. In this research, a pollination robot was developed to achieve precision pollination of clustered kiwifruit flowers in the orchard. The pollination robot consists of five systems, including a multinozzle end‐effector, a mechanical arm, a vision system, a crawler‐type chassis, and a control system. The robot can select preferential flowers and then target their pistil to achieve precision pollination. First, statistical analysis of the dimensions of flower clusters and individual flowers was conducted to fit normal distribution curves, which guided the design of the spray coverage and combination intervals for the multinozzle end‐effector. Second, optimal spray parameters were determined based on a three‐factor, five‐level quadratic orthogonal experiment, that is, air pressure of 70.4 kPa, rate of flow of 86.0 mL/min, and spray distance of 27.8 cm. A targeted pollination strategy was developed based on the preferential flower selection strategy and structure of the multinozzle end‐effector. Field experiments were conducted in a commercial kiwifruit orchard to evaluate its feasibility and performance, and an average success targeting rate of 93.4% at an average speed of 1.0 s per flower was achieved. Furthermore, compared with artificial assisted pollination methods, it can improve the utilization rate of kiwifruit pollen with an average consumption of 0.20 g in every 60 flowers with an average fruit set rate of 88.9%. The validations demonstrated that the pollination robot can efficiently pollinate kiwifruit flowers and save pollen.

Automated kiwifruit counting in orchards delivers accurate, timely, and cost-effective insights into yield estimation, which is crucial for decision-making in harvesting, storage, and marketing operations. Although several studies have proposed methods for kiwifruit counting in orchard, these methods have mostly focused on a small section within the orchard row, which may provide insufficient data for practical application due to the complex and uneven fruit-growing condition. This study introduces a novel automatic approach for row-based accurate kiwifruit counting on video sequences. The sequences are captured by smartphones mounted on a stabilizer and an extension pole, from an upward perspective that spans the full length of each kiwifruit row, addressing the limitations of previous methods. The pipeline comprises of three key components: kiwifruit and support-post detection, video-based kiwifruit counting, and detection region adaptation. First, the performance of You Only Look Once (YOLO) detection models was compared based on the constructed dataset, indicating that the medium-scale model achieved a balanced performance in terms of parameters, model size, inference time and average precision (AP). Second, a two-containers verification (TCV) method was proposed and applied following fruit tracking to reduce over-estimation in video-based counting. Finally, the kiwifruits in neighboring rows were eliminated by detection region adaptation, which estimated the row boundaries and dynamically adapting the masks based on support-posts in orchards. YOLOv5m and YOLOv8m were considered as the most competitive frameworks in kiwifruit detection, achieving AP0.5:0.95 scores of 0.864 and 0.878, respectively. The proposed TCV method suppressed false counts and counting accuracy improved by 59.06 % (from 40.12 % to 94.20 %) on ByteTrack and 54.08 % (from 37.59 % to 96.65 %) on DeepSORT. Moreover, the detection region adaptation guided by support-post has eliminated most fruit counts in neighboring rows. The R-squared (R2) of the row-based kiwifruit counting was 0.9791, indicating the proposed approach has the potential to achieve yield estimation for kiwifruit orchards.

Kiwifruit is a dioecious woody liana fruit tree, and the non-fruitfulness of male plants leads to a great deal of blindness in the selection of male plants in crossbreeding. In this study, we induced the development of male plant ovary by externally applying plant growth regulators (PGRs) and performed histological observation, phytohormone content determination and transcriptome analysis on the abortive ovary of the male kiwifruit (Con), the ovary of the female kiwifruit (Fem) and the PGR-induced developing ovary of the male kiwifruit (PT). Histological analysis showed that the Con ovary was devoid of ovules and the carpels were atrophied, the Fem ovary had ovules and the PT ovary was devoid of ovules, but the carpels developed normally and were not atrophied. Endogenous phytohormone content measurements displayed higher levels of trans-zeatin (tZT) in PT and Fem than Con, and lower levels of gibberellin (GA3) and abscisic acid (ABA) than Con. Transcriptome analysis revealed significant differences in many key genes in the cytokinin and auxin pathways, which were consistent with the results of phytohormone content measurements. Meanwhile, the genes related to carpel development, SPT (DTZ79_04g03580) and SK41 (DTZ79_19g04340), were highly expressed in PT, suggesting that they may play a key role in PGR-induced development of the ovary in male kiwifruit. These results provide information for elucidating the potential regulatory network of PGR-induced ovary development in male flowers and contribute to further identification of valuable target genes.

Bud thinning is a critical operation in the early stage of kiwifruit production, which is currently performed by skilled workers and has an urgent need to develop bud thinning robots. Accurate detection of kiwifruit buds is the first step, which focuses on distinguishing main bud and lateral bud. Kiwifruit buds are small, with similar shapes and colors in the main and lateral buds. Therefore, two kiwifruit bud detection methodologies were proposed to distinguish them. One is two-stage kiwifruit bud detection methodology (T-SKBDM) with an enhanced algorithm that leverage kiwifruit bud growth characteristics after network training for precise detection of main and lateral buds, and another is one-stage kiwifruit bud detection methodology (O-SKBDM) that classifies buds during the training. These methodologies adopted a two-classes annotation strategy (T-CAS) and a five-classes annotation strategy (F-CAS), respectively. In addition, both utilized an overlap-partitioning algorithm (OPA) that partitions large images into small images with overlapping areas. YOLOv8l model was trained on the dataset with different annotation strategies before and after using the OPA. Results showed that the T-CAS achieved a mean average precision (mAP) of 66.4 % before employing the OPA, which was 17.5 % higher than the F-CAS. With the OPA, mAPs of the T-CAS and F-CAS increased by 15.8 % and 17.7 %, respectively. Furthermore, T-SKBDM improved by 12.0 % and 14.3 % in distinguishing main and lateral buds, respectively, compared with the average precisions of 69.2 % and 66.2 % for O-SKBDM. These results indicate that the T-SKBDM assists in detecting kiwifruit buds and distinguishing the main and lateral buds, thus laying the foundation for robotic bud thinning.