Plant-Denoising-Net (PDN): A plant point cloud denoising network based on density gradient field learning

Plant phenotyping involves the measurements of plant traits to gain more insight into the interaction between the genotype (G), environment (E) and crop management strategies (M). To improve plant phenotyping, accurate measurements are crucial. Manual measurements are biased, time-intensive, and therefore limited to only a few plants. Especially measurements of 3D phenotypic traits, such as plant architecture, internode length, and leaf area are difficult to extract manually. To enhance the speed and accuracy of phenotyping, there is a need for automatic digital plant phenotyping solutions. The presented dataset contains 3D point clouds of tomato plants, which will enable researchers to develop novel methods to extract 3D phenotypic traits. Converting 3D point clouds to plant traits is also known as 3D plant phenotyping. This process can be subdivided into three steps: point cloud segmentation, skeletonisation to extract plant architecture, and plant-traits extraction. Those three steps need to be analysed properly to indicate bottlenecks and improve 3D phenotyping algorithms. Currently, the development of 3D phenotyping algorithms is inhibited by the availability of comprehensive datasets and algorithms to analyse all steps. To our best knowledge only five annotated datasets exist for testing and validating 3D phenotyping algorithms. However, these datasets mainly focus on the segmentation step. Skeletonisation and manual measured plant traits are frequently not included. To improve 3D plant phenotyping, a novel dataset, TomatoWUR, is presented. This comprehensive dataset consists of 44 point clouds of single tomato plants imaged by fifteen cameras to create a point cloud using the shape-from-silhouette methodology. The dataset includes annotated point clouds, skeletons, and manual reference measurements. In addition, the dataset includes software for comprehensive evaluation and comparison of phenotyping methods, which is expected to benefit the development of 3D phenotyping algorithms. The related software can be found our GIT: https://github.com/WUR-ABE/TomatoWUR.

The acquisition of point clouds is usually accompanied by noise due to imperfect laser scanning or image-based reconstruction techniques. Deep learning-based methods have achieved impressive performance in point cloud denoising. However, the features captured by a denoising network from noisy point clouds are usually contaminated by noise during training. The feature noise will lead to the oscillation of back-propagated gradients, which interferes with parameter optimization and reduces the denoising performance. In this paper, we propose to explicitly clean up feature noise for point cloud denoising from two aspects: feature noise cleaning and network training. From the first aspect, we propose the feature clean network (FCNet for short) to explicitly clean up the feature noise. From the second aspect, we train FCNet by a teacher-student learning model to learn the noise-free features under the guidance of feature domain losses. Specifically, FCNet is designed with emphasis on two modules: non-local self-similarity (NSS) and weighted average pooling (WAP). NSS module smooths features through a non-local filter based on the inherent non-local self-similarity of point clouds. WAP module applies original weights calculated by the statistical outlier removal algorithm to suppress the feature noise induced by outliers. In the teacher-student learning model, we introduce a clean input using the noisy point and its clean neighbors. The teacher network accepts the clean input to capture noise-free features. The student network is trained to imitate the teacher network to learn noise-free features by minimizing the feature loss. The experiments on synthetic and real scanned point clouds show that FCNet outperforms state-of-the-art point cloud denoising methods.

Accurate and high-throughput determination of plant morphological traits is essential for phenotyping studies. Nowadays, there are many approaches to acquire high-quality three-dimensional (3D) point clouds of plants. However, it is difficult to estimate phenotyping parameters accurately of the whole growth stages of maize plants using these 3D point clouds. In this paper, an accurate skeleton extraction approach was proposed to bridge the gap between 3D point cloud and phenotyping traits estimation of maize plants. The algorithm first uses point cloud clustering and color difference denoising to reduce the noise of the input point clouds. Next, the Laplacian contraction algorithm is applied to shrink the points. Then the key points representing the skeleton of the plant are selected through adaptive sampling, and neighboring points are connected to form a plant skeleton composed of semantic organs. Finally, deviation skeleton points to the input point cloud are calibrated by building a step forward local coordinate along the tangent direction of the original points. The proposed approach successfully generates accurately extracted skeleton from 3D point cloud and helps to estimate phenotyping parameters with high precision of maize plants. Experimental verification of the skeleton extraction process, tested using three cultivars and different growth stages maize, demonstrates that the extracted matches the input point cloud well. Compared with 3D digitizing data-derived morphological parameters, the NRMSE of leaf length, leaf inclination angle, leaf top length, leaf azimuthal angle, leaf growth height, and plant height, estimated using the extracted plant skeleton, are 5.27, 8.37, 5.12, 4.42, 1.53, and 0.83%, respectively, which could meet the needs of phenotyping analysis. The time required to process a single maize plant is below 100 s. The proposed approach may play an important role in further maize research and applications, such as genotype-to-phenotype study, geometric reconstruction, functional structural maize modeling, and dynamic growth animation.

Plant phenotyping analysis serves as a cornerstone of agricultural research. 3D point clouds greatly improve the problem of overlapping and occlusion of leaves in two-dimensional images and have become a popular field of plant phenotyping research. The realization of faster and more effective plant point cloud segmentation is the basis and key to the subsequent analysis of plant phenotypic parameters. To balance lightweight design and segmentation precision, we propose a Graph Convolutional Attention Synergistic Segmentation Network (GCASSN) specifically for plant point cloud data. The framework mainly comprises (1) Trans-net, which normalizes input point clouds into canonical poses; (2) Graph Convolutional Attention Synergistic Module (GCASM), which integrates graph convolutional networks (GCNs) for local feature extraction and self-attention mechanisms to capture global contextual dependencies. Complementary advantages are realized. On plant 3D point cloud segmentation via the Plant3D and Phone4D datasets, the model achieves state-of-the-art performance with 95.46% mean accuracy and 90.41% mean intersection-over-union (mIoU), surpassing mainstream methods (PointNet, PointNet++, DGCNN, PCT, and Point Transformer). The computational efficiency is competitive, with the inference time and parameter quantity slightly exceeding that of the DGCNN. Without parameter tuning, it attains 85.47% mIoU and 82.9% mean class IoU on ShapeNet, demonstrating strong generalizability. The method proposed in this article can fully extract the local detail features and overall global features of plants, and efficiently and robustly complete the segmentation task of plant point clouds, laying a solid foundation for plant phenotype analysis. The code of the GCASSN can be found in https://github.com/fallovo/GCASSN.git.

Ground 3D laser point cloud denoising processing is the key to realize terrain 3D modeling of debris flow ditch. Most point cloud denoising methods have poor effect on the complex terrain, do not ideally preserve the terrain features either. This study put forward a point cloud denoising processing method based on terrain feature refinement. Firstly, based on the relief amplitude and slope information reflected by the original point cloud, the terrain niche index model was established, and the natural discontinuity method was used to block the point cloud in the experimental area. Secondly, a multi-scale virtual grid was constructed based on the terrain and vegetation growth of each block to extract more accurate initial ground points, and the initial terrain surface was constructed based on these points. Finally, a set of ground point judgment parameters is determined, and the ground point is extracted in combination with the initial terrain surface, to realize the denoising processing of the debris flow valley point cloud. The results quality is assessed with the traditional triangulated irregular network denoising processing method. The analysis results show that the denoising method based on terrain feature refinement not only improve the quality of point cloud denoising and effectively remove the vegetation with different height characteristics under different slope environments in the debris flow valley, but also keep the terrain structure features very well.

Study on Image De-Noising and Technique

Зроблено акцент на аерокосмічних зображеннях, методах їх отримання та застосування. Якість отриманих первинних зображень часто не відповідає потребам кінцевих користувачів. Якщо спотворювальним чинником для зображення є шум, то використовуються передусім фільтри, що з різним успіхом усувають різні типи шумів. Виділено групу сучасних систем та засобів програмного забезпечення, що використовують для отримання та оброблення аерокосмічних зображень. Охарактеризовано особливості таких зображень у різних спектральних діапазонах. Подано особливості типових спотворень для такого типу зображень. Окрему увагу при цьому приділено шумам різних типів. Оцінювати рівень шуму можна на базі одного та багатьох зображень. Для цього існують різні методи. Проаналізовано наявні методи оцінювання шуму для графічних зображень, зокрема такі типи методів: визначення функції рівня шуму з єдиного зображення, що використовує залежність дисперсії шуму від інтенсивності зображення і передбачає визначення ділянок однорідності (гомогенності); визначення типу та рівня шуму з гістограми яскравості зображення; оцінка на базі окремого зображення з використанням кусково-гладкої функції попередньої моделі зображень та функції відгуку камер із зарядовим зв'язком; оцінка на базі математичних співвідношень щодо залежності автокореляційної функції зображень від дисперсії адитивного шуму та інші. Експериментальним чином застосовано кілька методів для оцінки шуму для зображення з відкритого датасету DOTA.

Plant phenotyping is a central task in crop science and plant breeding. It involves measuring plant traits to describe the anatomy and physiology of plants and is used for deriving traits and evaluating plant performance. Traditional methods for phenotyping are often time-consuming operations involving substantial manual labor. The availability of 3D sensor data of plants obtained from laser scanners or modern depth cameras offers the potential to automate several of these phenotyping tasks. This automation can scale up the phenotyping measurements and evaluations that have to be performed to a larger number of plant samples and at a finer spatial and temporal resolution. In this paper, we investigate the problem of registering 3D point clouds of the plants over time and space. This means that we determine correspondences between point clouds of plants taken at different points in time and register them using a new, non-rigid registration approach. This approach has the potential to form the backbone for phenotyping applications aimed at tracking the traits of plants over time. The registration task involves finding data associations between measurements taken at different times while the plants grow and change their appearance, allowing 3D models taken at different points in time to be compared with each other. Registering plants over time is challenging due to its anisotropic growth, changing topology, and non-rigid motion in between the time of the measurements. Thus, we propose a novel approach that first extracts a compact representation of the plant in the form of a skeleton that encodes both topology and semantic information, and then use this skeletal structure to determine correspondences over time and drive the registration process. Through this approach, we can tackle the data association problem for the time-series point cloud data of plants effectively. We tested our approach on different datasets acquired over time and successfully registered the 3D plant point clouds recorded with a laser scanner. We demonstrate that our method allows for developing systems for automated temporal plant-trait analysis by tracking plant traits at an organ level.

PST: Plant segmentation transformer for 3D point clouds of rapeseed plants at the podding stage

Random screening-based feature aggregation for point cloud denoising

The development of the distributed acoustic sensing (DAS) technique enables us to record seismic data at a significantly improved spatial sampling rate at meter scales, which offers new opportunities for high-resolution subsurface imaging. However, DAS recordings are often characterized by a low signal-to-noise ratio (S/N) due to the presence of data noise, significantly degrading the reliability of imaging and interpretation. Current DAS data noise reduction methods remain insufficient in simultaneously preserving weak signals and eliminating various types of noise. Particularly when dealing with DAS data that are contaminated by four types of noise (i.e., high-frequency noise, high-amplitude erratic noise, horizontal noise, and random background noise), it becomes challenging to attenuate the strong noise while maintaining fine-scale features. To address these issues, we develop an integrated local orthogonalization (LO) method that can remove a mixture of different types of noise while protecting the useful signal. Our LO method effectively eliminates the aforementioned noise by concatenating multiple denoising operators including a band-pass filter, a structure-oriented, spatially varying median filter, a dip filter in the frequency-wavenumber domain, and a curvelet filter. Next, the local orthogonalization weighting operator is applied to extract signal energy from the removed noise section. We demonstrate the robustness of our LO method on various challenging DAS data sets from the Frontier Observatory for Research in Geothermal Energy geothermal field. The denoising results demonstrate that our LO method can successfully minimize the levels of different types of noise while preserving the energy of weak signals.

Opinion dynamics are affected by cognitive biases and noise. While mathematical models have focused extensively on biases, we still know surprisingly little about how noise shapes opinion patterns. Here, we use an agent-based opinion dynamics model to investigate the interplay between confirmation bias-represented as bounded confidence-and different types of noise. After analysing where noise can enter social interaction, we propose a type of noise that has not been discussed so far, ambiguity noise. While previously considered types of noise acted on agents either before, after or independent of social interaction, ambiguity noise acts on communicated messages, assuming that socially transmitted opinions are inherently noisy. We find that noise can induce agreement when confirmation bias is moderate, but different types of noise require quite different conditions for this effect to occur. An application of our model to the climate change debate shows that at just the right mix of confirmation bias and ambiguity noise, opinions tend to converge to high levels of climate change concern. This result is not observed with the other types. Our findings highlight the importance of considering and distinguishing between the various types of noise and the unique role of ambiguity in opinion formation.

In industrial plants, piping and equipment are intricately interconnected, and there are many components with a variety of shapes. To use point clouds of industrial plants for simulation of maintenance work, it is necessary to extract the components. Deep learning is effective for recognizing components in point clouds of industrial plants. However, training classifiers is challenging due to the difficulty in acquiring diverse point cloud datasets and the labor-intensive process of annotating large-scale point clouds. A promising approach to address these issues is to train the classifier on virtual point clouds generated from CAD models. However, classifiers trained on these virtual point clouds often fail to achieve sufficient segmentation accuracy due to discrepancies between virtual point clouds and actual point clouds captured by terrestrial laser scanners. This paper proposes methods to improve segmentation accuracy by reducing these discrepancies. First, we introduce a method to incorporate features such as missing points, noise, and outliers observed in actual point clouds. Furthermore, we propose a data augmentation approach that applies up-sampling using a deep learning model trained on paired virtual and real point clouds to reduce the discrepancy between them. Our evaluation demonstrates that the proposed methods effectively improve the segmentation accuracy of point clouds of industrial plants.

Population systems are often subject to various different types of environmental noises. This paper considers a class of Kolmogorov-type systems perturbed by three different types of noise including Brownian motions, Markovian switching processes, and Poisson jumps, which is described by a regime-switching jump diffusion process. This paper examines these three different types of noises and determines their effects on the properties of the systems. The properties to be studied include existence and uniqueness of global positive solutions, boundedness of this positive solution, and asymptotic growth property, and extinction in the senses of the almost sure and the $p$th moment. Finally, this paper also considers a stochastic Lotka-Volterra system with regime-switching jump diffusion processes as a special case.

The microseismic signals released by rock mass fracture can be captured via microseismic monitoring to evaluate the development of geological disasters. This is crucial for underground engineering construction, underground mining, and earthquake and geological disaster evaluation. However, extracting information effectively is difficult due to the low signal-to-noise ratio of microseismic signals caused by complex environmental factors. Therefore, denoising and detection (onset time picking) are essential to processing microseismic signals and extracting source information. To improve the efficiency and accuracy of microseismic signal processing, we propose a parallel dual-tasking network, which is an advanced deep learning model that can simultaneously perform microseismic denoising and detection tasks. The network, comprising one encoder and two parallel decoders, is customised to extract input data features, and two outputs can be simultaneously generated to denoise and detect microseismic signals. The model exhibits excellent denoising and detection performance for microseismic signals containing various types of noise. Compared with traditional methods, the signal-to-noise ratio of the denoised signal is greatly improved, and the waveform distortion of the denoised signal is small. Even when the signal-to-noise ratio is low, the proposed model can maintain good onset time pickup performance. This method obviates the need for different denoising methods for different types of noise and precludes setting thresholds artificially to improve the denoising effect and detection accuracy. Moreover, the dual processing characteristics of the model facilitate simultaneous denoising and detection, which improves the processing efficiency of microseismic data and meets the demand for automatically processing massive microseismic data. Therefore, this method has excellent data processing potential in exploration seismology, and earthquake and disaster assessment.