Recently, disease prevention in jute plants has become an urgent topic as a result of the growing demand for finer quality fiber. This research presents a deep learning network called YOLO-JD for detecting jute diseases from images. In the main architecture of YOLO-JD, we integrated three new modules such as Sand Clock Feature Extraction Module (SCFEM), Deep Sand Clock Feature Extraction Module (DSCFEM), and Spatial Pyramid Pooling Module (SPPM) to extract image features effectively. We also built a new large-scale image dataset for jute diseases and pests with ten classes. Compared with other state-of-the-art experiments, YOLO-JD has achieved the best detection accuracy, with an average mAP of 96.63%.

Phenotyping of plant growth improves the understanding of complex genetic traits and eventually expedites the development of modern breeding and intelligent agriculture. In phenotyping, segmentation of 3D point clouds of plant organs such as leaves and stems contributes to automatic growth monitoring and reflects the extent of stress received by the plant. In this work, we first proposed the Voxelized Farthest Point Sampling (VFPS), a novel point cloud downsampling strategy, to prepare our plant dataset for training of deep neural networks. Then, a deep learning networkPSegNet, was specially designed for segmenting point clouds of several species of plants. The effectiveness of PSegNet originates from three new modules including the Double-Neighborhood Feature Extraction Block (DNFEB), the Double-Granularity Feature Fusion Module (DGFFM), and the Attention Module (AM). After training on the plant dataset prepared with VFPS, the network can simultaneously realize the semantic segmentation and the leaf instance segmentation for three plant species. Comparing to several mainstream networks such as PointNet++, ASIS, SGPN, and PlantNet, the PSegNet obtained the best segmentation results quantitatively and qualitatively. In semantic segmentation, PSegNet achieved 95.23, 93.85, 94.52, and 89.90 for the mean Prec, Rec, F1, and IoU, respectively. In instance segmentation, PSegNet achieved 88.13, 79.28, 83.35, and 89.54 for the mPrec, mRec, mCov, and mWCov, respectively.

Learning and extracting high-level features from point cloud is the key to improving the segmentation performances on point clouds for many networks. At present, many networks present very deep structures to extract high-level features for 3D perception. However, we argue that even better results can be achieved by (i) building feature vectors that integrates multi-scale geometric features, and (ii) exerting discriminative constraints on the learning of mid-levels features. In this paper, we propose a Multi-scale Neighborhood Feature Extraction and Aggregation Model (MNFEAM) to enhance feature extraction for point cloud learning. We try to first extract multi-scale neighborhood information for each input point and then aggregate local information of a mid-level locality feature space, and finally integrate the aggregated local and the global feature vectors. A new discriminative loss function is designed to strengthen the coarse semantics on mid-levels features so that the semantic abstraction process can be improved and accelerated. We improve the performances of three popular networks for point cloud segmentation using the proposed MNFEAM on standard 3D datasets.

Nowadays, 3D imaging of plants not only contributes to monitoring and managing plant growth, but is also becoming an essential part of high-throughput plant phenotyping. In this paper, an inexpensive (less than 70 USD) and portable platform with binocular stereo vision is established, which can be controlled by a laptop. In the stereo matching step, an efficient cost calculating measure—AD-Census—is integrated with the adaptive support-weight (ASW) approach to improve the ASW’s performance on real plant images. In the quantitative assessment, our stereo algorithm reaches an average error rate of 6.63% on the Middlebury datasets, which is lower than the error rates of the original ASW approach and several other popular algorithms. The imaging experiments using the proposed stereo system are carried out in three different environments including an indoor lab, an open field with grass, and a multi-span glass greenhouse. Six types of greenhouse plants are used in experiments; half of them are ornamentals and the others are greenhouse crops. The imaging accuracy of the proposed method at different baseline settings is investigated, and the results show that the optimal length of the baseline (distance between the two cameras of the stereo system) is around 80 mm for reaching a good trade-off between the depth accuracy and the mismatch rate for a plant that is placed within 1 m of the cameras. Error analysis from both theoretical and experimental sides show that for an object that is approximately 800 mm away from the stereo platform, the measured depth error of a single point is no higher than 5 mm, which is tolerable considering the dimensions of greenhouse plants. By applying disparity refinement, the proposed methodology generates dense and accurate point clouds of crops in different environments including an indoor lab, an outdoor field, and a greenhouse. Our approach also shows invariance against changing illumination in a real greenhouse, as well as the capability of recovering 3D surfaces of highlighted leaf regions. The method not only works on a binocular stereo system, but is also potentially applicable to a SFM-MVS (structure-from-motion and multiple-view stereo) system or any multi-view imaging system that uses stereo matching.

Leaves account for the largest proportion of all organ areas for most kinds of plants, and are comprise the main part of the photosynthetically active material in a plant. Observation of individual leaves can help to recognize their growth status and measure complex phenotypic traits. Current image-based leaf segmentation methods have problems due to highly restricted species and vulnerability toward canopy occlusion. In this work, we propose an individual leaf segmentation approach for dense plant point clouds using facet over-segmentation and facet region growing. The approach can be divided into three steps: (1) point cloud pre-processing, (2) facet over-segmentation, and (3) facet region growing for individual leaf segmentation. The experimental results show that the proposed method is effective and efficient in segmenting individual leaves from 3D point clouds of greenhouse ornamentals such as Epipremnum aureum, Monstera deliciosa, and Calathea makoyana, and the average precision and recall are both above 90%. The results also reveal the wide applicability of the proposed methodology for point clouds scanned from different kinds of 3D imaging systems, such as stereo vision and Kinect v2. Moreover, our method is potentially applicable in a broad range of applications that aim at segmenting regular surfaces and objects from a point cloud.