A novel splat feature classification method is presented with application to retinal hemorrhage detection in fundus images. Reliable of hemorrhages important the development automated screening systems which can be translated into practice. Under our supervised approach, color images are partitioned nonoverlapping segments covering entire image. Each segment, i.e., splat, contains pixels similar and spatial location. set features extracted from each describe its characteristics relative...

Global localization in 3D point clouds is a challenging problem of estimating the pose vehicles without any prior knowledge. In this paper, solution to presented by achieving place recognition and metric estimation global map. Specifically, we present semi-handcrafted representation learning method for LiDAR using siamese LocNets, which states similarity modeling problem. With final learned representations LocNet, framework with range-only observations proposed. To demonstrate performance...

Global localization in 3D point clouds is a challenging task for mobile vehicles outdoor scenarios, which requires the vehicle to localize itself correctly given map without prior knowledge of its pose. This critical component autonomous or robots on road handling failures. In this paper, based reduced dimension scan representations learned from neural networks, solution global proposed by achieving place recognition first and then metric pose estimation map. Specifically, we present...

In the process of fabric production, various types defects affect quality a fabric. However, due to wide variety defects, complexity textures, and concealment small target current defect detection algorithms suffer from issues such as having slow speed, low accuracy, recognition rate defects. Therefore, developing an efficient accurate system has become urgent problem that needs be addressed in textile industry. Addressing aforementioned issues, this paper proposes improved YOLOv8n-LAW...

The 3-D spectral-domain optical coherence tomography (SD-OCT) images of the retina often do not reflect true shape and are distorted differently along x y axes. In this paper, we propose a novel technique that uses thin-plate splines in two stages to estimate correct distinct axial artifacts SD-OCT images. method was quantitatively validated using nine pairs OCT scans obtained with orthogonal fast-scanning axes, where segmented surface compared after both datasets had been corrected. mean...

LiDAR-camera calibration is a precondition for many heterogeneous systems that fuse data from LiDAR and camera. However, the constraint common field of view requirement strict time synchronization make challenging problem. In this paper, we propose novel method aiming to eliminate these two constraints. Specifically, capture scan 3D when both environment sensors are stationary, then move camera reconstruct using sequentially obtained images. Finally, align visual points laser based on...

Long-term visual localization in outdoor environment is a challenging problem, especially faced with the cross-seasonal, bi-directional tasks and changing environment. In this paper we propose novel inertial framework that localizes against LiDAR-built map. Based on geometry information of laser map, hybrid bundle adjustment proposed, which estimates poses cameras respect to prior map as well optimizes state variables online odometry system simultaneously. For more accurate crossmodal data...

Large scale 3D maps constructed via LiDAR sensor are widely used on intelligent vehicles for localization in outdoor scenes. However, loading, communication and processing of the original dense time consuming onboard computing platform, which calls a more concise representation to reduce complexity but keep performance localization. In this paper, we propose teacher-student learning paradigm compress point cloud map. Specifically, first find subset points with high number observations...

Millimeter wave (MMW) imaging systems have been widely used for security screening in public places due to their advantages of being able detect a variety suspicious objects, non-contact operation, and harmlessness the human body. In this study, we propose an innovative, multi-dimensional information fusion YOLO network that can aggregate capture multimodal cope with challenges low resolution susceptibility noise MMW images. particular, data aggregation module is developed adaptively...

Dried tangerine peel ("Chenpi"), has numerous clinical and nutritional benefits, with its quality being significantly influenced by storage age, referred to as "Chen Jiu Zhe Liang" in Chinese. Concequently, the rapid accurate identification of Chenpi's age is important for consumers. In this study, Fourier transform infrared spectroscopy (FTIR) was employed capture spectral images Chenpi. These FTIR were then analyzed using a two-dimensional convolutional neural networks (2D-CNN) model,...

A robust multiscale stereo matching algorithm is proposed to find reliable correspondences between low contrast and weakly textured retinal image pairs with radiometric differences. Existing algorithms designed deal piecewise planar surfaces distinct features Lambertian reflectance do not apply in applications such as 3D reconstruction of medical images including images. In this paper, pixel feature vectors are formulated extract discriminative the presence noise scale space, through which...

Autonomous mobile vehicles are expected to perform persistent and accurate localization with low-cost equipment. To achieve this goal, we propose a stereo camera based visual method using modified laser map, which takes the advantage of both low cost camera, high geometric precision data long-term performance. Considering that LiDAR give measurements same environment in different modalities, cross-modal invariance is investigated modify map for localization. Specifically, learning algorithm...

When robots move autonomously for long-term, varied appearance such as the transition from day to night and seasonal variation brings challenges visual place recognition. Defining an condition (e.g. a season, kind of weather) domain, we consider that desired representation recognition (i) should be domain-unrelated so images different time can matched regardless appearance, (ii) learned in self-supervised manner without need massive manually labeled data, (iii) able train among multiple...

Radar and lidar, provided by two different range sensors, each has pros cons of various perception tasks on mobile robots or autonomous driving. In this paper, a Monte Carlo system is used to localize the robot with rotating radar sensor 2D lidar maps. We first train conditional generative adversarial network transfer raw data data, achieve reliable points from generator. Then an efficient odometry included in system. Combining initial guess odometry, measurement model proposed match prior...

LiDAR based place recognition is popular for loop closure detection and re-localization. In recent years, deep learning brings improvements to by learnable feature extraction. However, these methods degenerate when the robot re-visits previous places with a large perspective difference. To address challenge, we propose DeepRING learn roto-translation invariant representation from scan, so that visiting same different can have similar representations. There are two keys in DeepRING: extracted...

Traversable area segmentation is important for safe navigation of mobile robot in outdoor environment. To address this problem, we propose a unified framework to register data across sessions, on which an unsupervised method presented traversable intended unstructured environments. With collected vehicle equipped with camera and laser, the proposed can generate massive label images obstacle without any human intervention, are fed as training samples pixel-wise semantic neural network. In...

Loop closure detection in 3D LIDAR data is an essential but challenging problem SLAM system. It important to reduce global inconsistency or re-localize the robot that loses localization, while difficult for lack of prior information. We present a semi-handcrafted representation learning method point cloud using siamese convolution neural network, which states loop similarity modeling problem. With learned representation, between two scans transformed as Euclidean distance representations...

Transcriptional enhancers commonly work over long genomic distances to precisely regulate spatiotemporal gene expression patterns. Dissecting the promoters physically contacted by these distal regulatory elements is essential for understanding developmental processes as well role of disease-associated risk variants. Modern proximity-ligation assays, like HiChIP and ChIA-PET, facilitate accurate identification long-range contacts between promoters. However, assays are technically challenging,...

Autonomous navigation based on precise localization has been widely developed in both academic research and practical applications. The high demand for accuracy essential safe robot planing while it makes the current geometric solutions less robust to environmental changes. Recent end-to-end methods handle raw sensory data with forms of instructions directly output command control. However, lack intermediate semantics system more rigid unstable use. To explore these issues, this paper...