Object Size Research Articles

Deep Learning Enhanced Volumetric Photoacoustic Imaging of Vasculature in Human.

The development of high-performance imaging processing algorithms is a core area of photoacoustic tomography. While various deep learning based image processing techniques have been developed in the area, their applications in 3D imaging are still limited due to challenges in computational cost and memory allocation. To address those limitations, this work implements a 3D fully-dense (3DFD) U-net to linear array based photoacoustic tomography and utilizes volumetric simulation and mixed precision training to increase efficiency and training size. Through numerical simulation, phantom imaging, and in vivo experiments, this work demonstrates that the trained network restores the true object size, reduces the noise level and artifacts, improves the contrast at deep regions, and reveals vessels subject to limited view distortion. With these enhancements, 3DFD U-net successfully produces clear 3D vascular images of the palm, arms, breasts, and feet of human subjects. These enhanced vascular images offer improved capabilities for biometric identification, foot ulcer evaluation, and breast cancer imaging. These results indicate that the new algorithm will have a significant impact on preclinical and clinical photoacoustic tomography.

CAMRL: A Joint Method of Channel Attention and Multidimensional Regression Loss for 3D Object Detection in Automated Vehicles

Fully automated vehicles collect information about their road environments to adjust their driving actions, such as braking and slowing down. The development of artificial intelligence (AI) and the Internet of Things (IoT) has improved the cognitive abilities of vehicles, allowing them to detect traffic signs, pedestrians, and obstacles for increasing the intelligence of these transportation systems. Three-dimensional (3D) object detection in front-view images taken by vehicle cameras is important for both object detection and depth estimation. In this paper, a joint channel attention and multidimensional regression loss method for 3D object detection in automated vehicles (called CAMRL) is proposed to improve the average precision of 3D object detection by focusing on the model’s ability to infer the locations and sizes of objects. First, channel attention is introduced to effectively learn the yaw angles from the road images captured by vehicle cameras. Second, a multidimensional regression loss algorithm is designed to further optimize the size and position parameters during the training process. Third, the intrinsic parameters of the camera and the depth estimate of the model are combined to reduce the object depth computation error, allowing us to calculate the distance between an object and the camera after the object’s size is confirmed. As a result, objects are detected, and their depth estimations are validated. Then, the vehicle can determine when and how to stop if an object is nearby. Finally, experiments conducted on the KITTI dataset demonstrate that our method is effective and performs better than other baseline methods, especially in terms of 3D object detection and bird’s-eye view (BEV) evaluation.

Classification Size of Underground Object from Ground Penetrating Radar Image using Machine Learning Technique

Ground Penetrating Radar (GPR) is a useful tool in detecting subsurface object or hidden structure defects However, the time-consuming problems and high requirement of professional manpower is required to analyse the GPR data. Machine learning is a tool that endowed with the ability to learn, and it can reduce time taken for the GPR data analysing. To simplify the identification process, a framework is proposed to classify the size of underground metallic pipe by using Histogram of Oriented Gradient (HOG) as a feature extraction algorithm. Two machine learning algorithms namely Support Vector Machines (SVM) and Backpropagation Neural Network were proposed to classify the size of the underground metallic pipe. As a result, the accuracy from the identification is more than 98% for both classifier algorithm.

The size of a novel object is learned rapidly, and unlearned slowly, for purposes of computing apparent distance.

The size of a novel object is learned rapidly, and unlearned slowly, for purposes of computing apparent distance.

Cortical areas involved in imagery and haptic exploration of object size

Cortical areas involved in imagery and haptic exploration of object size

MSSPA-GC: Multi-Scale Shape Prior Adaptation with 3D Graph Convolutions for Category-Level Object Pose Estimation

Category-level object pose estimation aims to predict the 6D object pose and size of arbitrary objects from known categories. It remains a challenge due to the large intra-class shape variation. Recently, the introduction of the shape prior adaptation mechanism into the normalized canonical coordinates (i.e., NOCS) reconstruction process has been shown to be effective in mitigating the intra-class shape variation. However, existing shape prior adaptation methods simply map the observed point cloud to the normalized object space, and the extracted object descriptors are not sufficient for the perception of the object pose. As a result, they fail to predict the pose of objects with complex geometric structures (e.g., cameras). To this end, this paper proposes a novel shape prior adaption method named MSSPA-GC for category-level object pose estimation. Specifically, our main network takes the observed instance point cloud converted from the RGB-D image and the prior shape point cloud pre-trained on the object CAD models as inputs. Then, a novel 3D graph convolution network and a PointNet-like MLP network are designed to extract pose-aware object features and shape-aware object features from these two inputs, respectively. After that, the two-stream object features are aggregated through a multi-scale feature propagation mechanism to generate comprehensive 3D object descriptors that maintain both pose-sensitive geometric stability and intra-class shape consistency. Finally, by leveraging object descriptors aware of both object pose and shape when reconstructing the NOCS coordinates, our approach elegantly achieves state-of-the-art performance on the widely used REAL275 and CAMERA25 datasets using only 25% of the parameters compared with existing shape prior adaptation models. Moreover, our method also exhibits decent generalization ability on the unconstrained REDWOOD75 dataset.

Differential involvement of the anterior and posterior hippocampus, parahippocampus, and retrosplenial cortex in making precise judgments of spatial distance and object size for remotely acquired memories of environments and objects.

The hippocampus is known to support processing of precise spatial information in recently learned environments. It is less clear, but crucial for theories of systems consolidation, to know whether it also supports processing of precise spatial information in familiar environments learned long ago and whether such precision extends to objects and numbers. In this fMRI study, we asked participants to make progressively more refined spatial distance judgments among well-known Toronto landmarks (whether landmark A is closer to landmark B or C) to examine hippocampal involvement. We also tested whether the hippocampus was similarly engaged in estimating magnitude regarding sizes of familiar animals and numbers. We found that the hippocampus was only engaged in spatial judgment. Activation was greater and lasted longer in the posterior than anterior hippocampus, which instead showed greater modulation as discrimination between spatial distances became more fine grained. These findings suggest that the anterior and posterior hippocampus have different functions which are influenced differently by estimation of differential distance. Similarly, parahippocampal-place-area and retrosplenial cortex were involved only in the spatial condition. By contrast, activation of the intraparietal sulcus was modulated by precision in all conditions. Therefore, our study supports the idea that the hippocampus and related structures are implicated in retrieving and operating even on remote spatial memories whenever precision is required, as posted by some theories of systems consolidation.

Microfluidic biochip for target tumor cell and cell-cluster sorting

This study designed an integrated microfluidic biochip to separate target tumor cells and their clusters from whole blood. The first-stage sorting mechanism in the top channels was based on sorted object size by taking advantage of the in-series Deterministic Lateral Displacement (DLD). In the downstream bottom channel, the target tumor cells/cell clusters were trapped and sorted out by the antibody-coating lateral filter at the second stage. In the top-channel DLD cluster-sorting session, U-shaped pillars and two-section continuous sorting areas with different critical diameters could disperse the displacement of target tumor clusters and effectively rotate the non-symmetry tumor clusters to increase their effective size. In the downstream bottom channel, the selective coating on the lateral filter posts was used to coat antibodies exclusively on the side surface of micro-pillars to prevent clogging. The suspension of solid tumors from the 4T1-bearing BALB/c mouse was spiked in the blood from the same mouse to serve as the experimental blood sample. The DLD cluster-sorting session recovered 88.58% of tumor cell clusters with a purity of 92.20%. The lateral filter in the bottom chip captured 89.54% of tumor cells with a purity of 89.44%. This two-stage microfluidic chip developed in this research aims to study CTC clusters in cancer research and clinical applications.

Accuracy of Structure-from-Motion/Multiview Stereo Terrain Models: A Practical Assessment for Applications in Field Geology

We assess the accuracy of Structure-from-Motion/Multiview stereo (SM) terrain models acquired ad hoc or without high-resolution ground control to analyze their usage as a base for inexpensive 3D bedrock geologic mapping. Our focus is on techniques that can be utilized in field projects without the use of heavy and/or expensive equipment or the placement of ground control in logistically challenging sites (e.g., steep cliff faces or remote settings). We use a Terrestrial Light Detection and Ranging (LiDAR) survey as a basis for the comparison of two types of SM models: (1) models developed from images acquired in a chartered airplane flight with ground control referenced by natural objects located on Google Earth scenes; and (2) drone flights with a georeference established solely from camera positions located by conventional, differentially corrected Global Navigation Satellite systems (GNSS). We find that all our SM models are indistinguishable in scale from the LiDAR reference model. The SM models do, however, show rigid body translations and rotations, with translations generally within the 1–5 m size of the natural objects used for ground control, the resolution of the GNSS receivers, or both. The rigid body rotations can be attributed to a poor imaging plan, which can be avoided with survey planning. Analyses of point densities in various models show a limitation of Terrestrial LiDAR point clouds as a mapping base due to the rapid falloff of resolution with distance. In contrast, SM models are characterized by relatively uniform point densities controlled by camera optics, the numbers of images, and the distance from the target. This uniform density is the product of the Multiview stereo step in SM processing that fills areas between key points and is important for bedrock geologic mapping because it affords direct interpretation on a point cloud at a relatively uniform scale throughout a model. Our results indicate that these simple methods allow SM model construction to be accurate to the range of conventional GNSS with resolutions to the submeter, even cm, scale depending on data acquisition parameters. Thus, SM models can, and should, serve as a base for high-resolution geologic mapping, particularly in a steep terrain where conventional techniques fail. Our SM models appear to provide accurate visualizations of geologic features over km scales that allow detailed geologic mapping in 3D with a relative accuracy to the decimeter or centimeter level and absolute positioning in the 2–5 m precision of GNSS; a geometric precision that will allow unprecedented new studies of any geologic system where geometry is the fundamental data.

Improved Lightweight YOLOv4 Foreign Object Detection Method for Conveyor Belts Combined with CBAM

During the operation of the belt conveyor, foreign objects such as large gangue and anchor rods may be mixed into the conveyor belt, resulting in tears and fractures, which affect transportation efficiency and production safety. In this paper, we propose a lightweight target detection algorithm, GhostNet-CBAM-YOLOv4, to resolve the problem of the difficulty of detecting foreign objects at high-speed movement in an underground conveyor belt. The Kmeans++ clustering method was used to preprocess the data set to obtain the anchor box suitable for the foreign object size. The GhostNet lightweight module replaced the backbone network, reducing the model’s parameters. The CBAM attention module was introduced to enhance the ability of feature extraction facing the complex environment under the mine. The depth separable convolution was used to simplify the model structure and reduce the number of parameters and calculations. The detection accuracy of the improved method on the foreign body data set reached 99.32%, and the detection rate reached 54.7 FPS, which was 6.83% and 42.1% higher than the original YOLOv4 model, respectively. The improved method performed better than the original model on the other two datasets and could effectively avoid misdetection and omission detection. In comparison experiments with similar methods, our proposed method also demonstrated good performance, verifying its effectiveness.

A systematic task-based image quality assessment of photon-counting and energy integrating CT as a function of reconstruction kernel and phantom size.

Image quality of photon-counting and energy integrating CT scanners changes with object size, dose to the object, and kernel selection. To comprehensively compare task-generic image quality of photon-counting CT (PCCT) and energy integrating CT (EICT) systems as a function of phantom size, dose, and reconstruction kernel. A size-variant phantom (Mercury Phantom 3.0) was used to characterize the image quality of PCCT and EICT systems as a function of object size. The phantom contained five cylinders attached by slanted tapered sections. Each cylinder contained two sections: one uniform for noise, and the other with inserts for spatial resolution and contrast measurements. The phantom was scanned on Siemens' SOMATOM Force and NAEOTOM Alpha at 1.18 and 7.51mGy without tube current modulation. CTDIvol was matched across two systems by setting the required tube currents, else, all other acquisition settings were fixed. CT sinograms were reconstructed using FBP and iterative (ADMIRE2 - Force; QIR2 - Alpha) algorithms with Body regular (Br) kernels. Noise Power Spectrum (NPS), Task Transfer Function (TTF), contrast-to-noise ratio (CNR), and detectability index (d') for a task of identifying 2-mm disk were evaluated based on AAPM TG-233 metrology using imQuest, an open-source software package. Averaged noise frequency (fav ) and 50% cut-off frequency for TTF (f50 ) were used as scalar metrics to quantify noise texture and spatial resolution, respectively. The difference between image quality metrics' measurements was calculated as IQPCCT - IQEICT . From Br40 (soft) to Br64 (sharp), f50 for air insert increased from 0.35mm-1 ±0.04 (standard deviation) to 0.76mm-1 ±0.17, 0.34mm-1 ±0.04 to 0.77mm-1 ±0.17, 0.37mm-1 ±0.02 to 0.95mm-1 ±0.17 for PCCT-T3D-QIR2, PCCT-70keV-QIR2, and EICT-ADMIRE2, respectively, when averaged over all sizes and dose levels. Similarly, from Br40 to Br64, noise magnitude increased from 10.86 HU±7.12 to 38.61 HU±18.84, 10.94 HU±7.08 to 38.82 HU±18.70, 13.74 HU±11.02 to 52.11 HU±26.22 for PCCT-T3D-QIR2, PCCT-70keV-QIR2, and EICT-ADMIRE2, respectively. The difference in fav was consistent across all sizes and dose levels. PCCT-70keV-VMI showed better consistency in contrast measurements for iodine and bone inserts than PCCT-T3D and EICT; however, PCCT-T3D had higher contrast for both inserts. From Br40 to Br64, considering all sizes and dose levels, CNR for iodine insert decreased from 52.30±46.44 to 12.18±10.07, 40.42±33.42 to 9.48±7.16, 39.94±37.60 to 7.84±6.67 for PCCT-T3D-QIR2, PCCT-70keV-QIR2, and EICT-ADMIRE2, respectively. Both PCCT image types, T3D and 70-keV-VMI exhibited similar or better noise, contrast, CNR than EICT when comparing kernels with similar names. For 512×512 matrix, PCCT's sharp kernels had lower resolution than EICT's sharp kernels. For all image quality metrics, except extreme low, every dose condition had a similar set of IQ-matching kernels. It suggests that considering patient size and dose level to determine IQ-matching kernel pairs across PCCT and EICT systems may not be imperative while translating protocols, except when the signal to the detector is extremely low.

TranSDet: Toward Effective Transfer Learning for Small-Object Detection

Small-object detection is a challenging task in computer vision due to the limited training samples and low-quality images. Transfer learning, which transfers the knowledge learned from a large dataset to a small dataset, is a popular method for improving performance on limited data. However, we empirically find that due to the dataset discrepancy, directly transferring the model trained on a general object dataset to small-object datasets obtains inferior performance. In this paper, we propose TranSDet, a novel approach for effective transfer learning for small-object detection. Our method adapts a model trained on a general dataset to a small-object-friendly model by augmenting the training images with diverse smaller resolutions. A dynamic resolution adaptation scheme is employed to ensure consistent performance on various sizes of objects using meta-learning. Additionally, the proposed method introduces two network components, an FPN with shifted feature aggregation and an anchor relation module, which are compatible with transfer learning and effectively improve small-object detection performance. Extensive experiments on the TT100K, BUUISE-MO-Lite, and COCO datasets demonstrate that TranSDet achieves significant improvements compared to existing methods. For example, on the TT100K dataset, TranSDet outperforms the state-of-the-art method by 8.0% in terms of the mean average precision (mAP) for small-object detection. On the BUUISE-MO-Lite dataset, TranSDet improves the detection accuracy of RetinaNet and YOLOv3 by 32.2% and 12.8%, respectively.

Vitrimeric shape memory polymer-based fingertips for adaptive grasping.

The variability in the shapes and sizes of objects presents a significant challenge for two-finger robotic grippers when it comes to manipulating them. Based on the chemistry of vitrimers (a new class of polymer materials that have dynamic covalent bonds, which allow them to reversibly change their mechanical properties under specific conditions), we present two designs as 3D-printed shape memory polymer-based shape-adaptive fingertips (SMP-SAF). The fingertips have two main properties needed for an effective grasping. First, the ability to adapt their shape to different objects. Second, exhibiting variable rigidity, to lock and retain this new shape without the need for any continuous external triggering system. Our two design strategies are: 1) A curved part, which is suitable for grasping delicate and fragile objects. In this mode and prior to gripping, the SMP-SAFs are straightened by the force of the parallel gripper and are adapted to the object by shape memory activation. 2) A straight part that takes on the form of the objects by contact force with them. This mode is better suited for gripping hard bodies and provides a more straightforward shape programming process. The SMP-SAFs can be programmed by heating them up above glass transition temperature (54°C) via Joule-effect of the integrated electrically conductive wire or by using a heat gun, followed by reshaping by the external forces (without human intervention), and subsequently fixing the new shape upon cooling. As the shape programming process is time-consuming, this technique suits adaptive sorting lines where the variety of objects is not changed from grasp to grasp, but from batch to batch.

Refined feature fusion for in-field high-density and multi-scale rice panicle counting in UAV images

The yield of rice crops is strongly correlated to the number of panicles per unit area. Computer vision techniques have been utilized in previous studies to count the number of panicles automatically. However, when using images captured by Unmanned Aerial Vehicles (UAVs), the height and coverage area of the UAVs can cause shrinkage of the rice panicle objects, leading to errors in feature learning and reduced counting efficacy. In addition, there is a considerable relative size difference in rice ears, which can further affect counting accuracy. To address these issues, this paper proposes an algorithm named Refined Feature Fusion for Panicle Counting (RFF-PC). This algorithm utilizes a more refined scale division by extracting and fusing optimal features according to the object size distribution. Firstly, the number of rice ears with different sizes is quantified, and a fine division of the scale is made to calculate the receptive field size of different features that need to be fused. Secondly, multi-scale convolution generates ]features at each layer and feature pyramid fusion fuses more appropriate features at different layers to improve the ability to represent objects in complex multi-scale. Redundant information is removed through channel attention. Additionally, a refined Gaussian is used to generate ground truth close to the shape of real rice panicles. Over the UFPC2019 dataset captured at the height of 5 meters, the average counting accuracy of RFF-PC is 89.80%, and the counting accuracy only decreases from 94.33% to 90.58% over the images with an increased number of rice ears from 180 to 260, which outperforms several state-of-the-art algorithms, including MCNN, CSRNet, TasselNetV2+, DSNet, and DMCount.

Fractal nature of human gastrointestinal system: Exploring a new era.

The morphological complexity of cells and tissues, whether normal or pathological, is characterized by two primary attributes: Irregularity and self-similarity across different scales. When an object exhibits self-similarity, its shape remains unchanged as the scales of measurement vary because any part of it resembles the whole. On the other hand, the size and geometric characteristics of an irregular object vary as the resolution increases, revealing more intricate details. Despite numerous attempts, a reliable and accurate method for quantifying the morphological features of gastrointestinal organs, tissues, cells, their dynamic changes, and pathological disorders has not yet been established. However, fractal geometry, which studies shapes and patterns that exhibit self-similarity, holds promise in providing a quantitative measure of the irregularly shaped morphologies and their underlying self-similar temporal behaviors. In this context, we explore the fractal nature of the gastrointestinal system and the potential of fractal geometry as a robust descriptor of its complex forms and functions. Additionally, we examine the practical applications of fractal geometry in clinical gastroenterology and hepatology practice.

Multi-Branch Deep Learning Framework for Land Scene Classification in Satellite Imagery

Land scene classification in satellite imagery has a wide range of applications in remote surveillance, environment monitoring, remote scene analysis, Earth observations and urban planning. Due to immense advantages of the land scene classification task, several methods have been proposed during recent years to automatically classify land scenes in remote sensing images. Most of the work focuses on designing and developing deep networks to identify land scenes from high-resolution satellite images. However, these methods face challenges in identifying different land scenes. Complex texture, cluttered background, extremely small size of objects and large variations in object scale are the common challenges that restrict the models to achieve high performance. To tackle these challenges, we propose a multi-branch deep learning framework that efficiently combines global contextual features with multi-scale features to identify complex land scenes. Generally, the framework consists of two branches. The first branch extracts global contextual information from different regions of the input image, and the second branch exploits a fully convolutional network (FCN) to extract multi-scale local features. The performance of the proposed framework is evaluated on three benchmark datasets, UC-Merced, SIRI-WHU, and EuroSAT. From the experiments, we demonstrate that the framework achieves superior performance compared to other similar models.

Empirical optimization of energy bin weights for compressing measurements with realistic photon counting x-ray detectors.

Photon counting detectors (PCDs) provide higher spatial resolution, improved contrast-to-noise ratio (CNR), and energy discriminating capabilities. However, the greatly increased amount of projection data in photon counting computed tomography (PCCT) systems becomes challenging to transmit through the slip ring, process, and store. This study proposes and evaluates an empirical optimization algorithm to obtain optimal energy weights for energy bin data compression. This algorithm is universally applicable to spectral imaging tasks including 2 and 3 material decomposition (MD) tasks and virtual monoenergetic images (VMIs). This method is simple to implement while preserving spectral information for the full range of object thicknesses and is applicable to different PCDs, for example, silicon detectors and CdTe detectors. We used realistic detector energy response models to simulate the spectral response of different PCDs and an empirical calibration method to fit a semi-empirical forward model for each PCD. We numerically optimized the optimal energy weights by minimizing the average relative Cramér-Rao lower bound (CRLB) due to the energy-weighted bin compression, for MD and VMI tasks over a range of material area density (0-40g/cm2 water, 0-2.16g/cm2 calcium). We used Monte Carlo simulation of a step wedge phantom and an anthropomorphic head phantom to evaluate the performance of this energy bin compression method in the projection domain and image domain, respectively. The results show that for 2 MD, the energy bin compression method can reduce PCCT data size by 75% and 60%, with an average variance penalty of less than 17% and 3% for silicon and CdTe detectors, respectively. For 3 MD tasks with a K-edge material (iodine), this method can reduce the data size by 62.5% and 40% with an average variance penalty of less than 12% and 13% for silicon and CdTe detectors, respectively. We proposed an energy bin compression method that is broadly applicable to different PCCT systems and object sizes, with high data compression ratio and little loss of spectral information.

Instruments of Devotion in Close Bodily Contact – Jewellery, Costume Details and Other Personal Belongings in Relation to Late Medieval Piety

Summary Late medieval personal belongings such as jewellery, costume details, weapons and tools are sometimes provided with direct and indirect prayers for help and protection, images of Christ and saints relating to motifs in the church interior and in prayer books, and image composition giving associations to altar-screens and shrines. Taken altogether these objects explicitly have a devotional character, which has been further strengthened through analysis of written sources such as exempla and preserved more extensive prayers. This point towards the relevance of talking about these objects as instruments of small-scale and everyday devotion. People were in constant need of help and protection from everyday threats. By relating to preserved prayers and exempla from late medieval Sweden, the article shows that these objects with prayers and holy images were regarded as transmitting divine help, protection, consolation and blessing. Due to the small size and portability of the personal objects carried on the body, divine intervention could take place irrespective of time and space. Lived religion and everyday devotion could therefore be performed anywhere, including on the move.

A fine-tuned YOLOv5 deep learning approach for real-time house number detection.

Detection of small objects in natural scene images is a complicated problem due to the blur and depth found in the images. Detecting house numbers from the natural scene images in real-time is a computer vision problem. On the other hand, convolutional neural network (CNN) based deep learning methods have been widely used in object detection in recent years. In this study, firstly, a classical CNN-based approach is used to detect house numbers with locations from natural images in real-time. Faster R-CNN, MobileNet, YOLOv4, YOLOv5 and YOLOv7, among the commonly used CNN models, models were applied. However, satisfactory results could not be obtained due to the small size and variable depth of the door plate objects. A new approach using the fine-tuning technique is proposed to improve the performance of CNN-based deep learning models. Experimental evaluations were made on real data from Kayseri province. Classic Faster R-CNN, MobileNet, YOLOv4, YOLOv5 and YOLOv7 methods yield f1 scores of 0.763, 0.677, 0.880, 0.943 and 0.842, respectively. The proposed fine-tuned Faster R-CNN, MobileNet, YOLOv4, YOLOv5, and YOLOv7 approaches achieved f1 scores of 0.845, 0.775, 0.932, 0.972 and 0.889, respectively. Thanks to the proposed fine-tuned approach, the f1 score of all models has increased. Regarding the run time of the methods, classic Faster R-CNN detects 0.603 seconds, while fine-tuned Faster R-CNN detects 0.633 seconds. Classic MobileNet detects 0.046 seconds, while fine-tuned MobileNet detects 0.048 seconds. Classic YOLOv4 and fine-tuned YOLOv4 detect 0.235 and 0.240 seconds, respectively. Classic YOLOv5 and fine-tuned YOLOv5 detect 0.015 seconds, and classic YOLOv7 and fine-tuned YOLOv7 detect objects in 0.009 seconds. While the YOLOv7 model was the fastest running model with an average running time of 0.009 seconds, the proposed fine-tuned YOLOv5 approach achieved the highest performance with an f1 score of 0.972.

Design and fabrication of large area vat photopolymerization 3D printing system using a 32-inch quasi-collimated visible backlight module with local dimming control

In this paper, we accomplished the first 32-inch liquid crystal display (LCD) type vat photopolymerization 3D printing system based on the quasi-collimated visible backlight module with local dimming control. The large building area (40x70cm) of the 3D printing system can greatly increase the size of printing objects and the production rate. We focused on the design of quasi-collimated backlights and analyzed their optical properties and the appearance of 3D printing objects fabricated by them. The optical performances of the quasi-collimated backlight using stacks of optical films feature no large angle leakage light, 86% intensity uniformity, and less than ±15° angular distribution of emitting light. After the optimization process by combining a variety of optical films, the uniformity, and collimation of the backlight module are greatly improved. Moreover, combining a local dimming algorithm with 3D printing technology can completely suppress unexpected photopolymerization residues, increase resin utilization rate, save energy, and greatly improve the probability of success in the 3D printing process, especially in large printing objects.