Box Recognition Research Articles

Computer vision based cargo boxes pose adjustment system for two-dimensional conveyor platform

For enhancing the accuracy of cargo box position and angle recognition on the conveyor platform, this paper proposes a cargo box attitude detection and adjustment method based on instance segmentation and image processing. This approach involves generating Mask data through target detection of the cargo box using Mask R-CNN. Using image processing algorithms to generate a minimum rectangle according to the Mask data, and the minimum rectangle data is aligned with the Bbox data of Maks R-CNN. The position and angle of the cargo box are detected based on the minimum rectangular data, and the conveyor platform is adjusted to control the cargo box attitude using the Bbox data. Nine attitude acquisition and comparison experiments were conducted on the cargo box using an angle sensor, and the deviation of the method was consistently <0.6∘\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$<0.6^{\\circ }$$\\end{document}, with a relative error of 1.27% for the nine total changes. Importantly, the image processing technique in this study avoids external image processing, reducing ambient light impact and enhancing cargo box recognition and attitude feedback functionality on the conveyor platform. Throughout the entire cargo box adjustment experiment, the cargo box’s can be stabilized to reach the set angle.

Extraction of building footprint using MASK-RCNN for high resolution aerial imagery

Extracting individual buildings from satellite images is crucial for various urban applications, including population estimation, urban planning, and other related fields. However, Extracting building footprints from remote sensing data is a challenging task because of scale differences, complex structures and different types of building. Addressing these issues, an approach that can efficiently detect buildings in images by generating a segmentation mask for each instance is proposed in this paper. This approach incorporates the Regional Convolutional Neural Network (MASK-RCNN), which combines Faster R-CNN for object mask prediction and boundary box recognition and was evaluated against other models like YOLOv5, YOLOv7 and YOLOv8 in a comparative study to assess its effectiveness. The findings of this study reveals that our proposed method achieved the highest accuracy in building extraction. Furthermore, we performed experiments on well-established datasets like WHU and INRIA, and our method consistently outperformed other existing methods, producing reliable results.

EGCY-Net: An ELAN and GhostConv-Based YOLO Network for Stacked Packages in Logistic Systems

Dispatching, receiving, and transporting goods involve a large amount of manual effort. Within a logistics supply chain, a wide variety of transported goods need to be handled, recognized, and checked at many different points. Effective planning of automated guided vehicle (AGV) transportation can reduce equipment energy consumption and shorten task completion time. As the need for efficient warehouse logistics has increased in manufacturing systems, the use of AGVs has also increased to reduce working time. These processes hold automation potential, which we can exploit by using computer vision techniques. We propose a method for the complete automation of box recognition, covering both the types and quantities of boxes. To do this, an ELAN and GhostConv-based YOLO network (EGCY-Net) is proposed with a Conv-GhostConv Stack (CGStack) module and an ELAN-GhostConv Network (EGCNet). To enhance inter-channel relationships, the CGStack module captures complex patterns and information in the image by using ghost convolution to increase the model inference speed while retaining the ability to capture spatial features. EGCNet is designed and constructed based on ELAN and the CGStack module to capture and utilize hierarchical features efficiently in layer aggregation. Additionally, the proposed methodology involves the creation of a dataset comprising images of boxes taken in warehouse settings. The proposed system is realized on the NVIDIA Jetson Nano platform, using an Arducam IMX477 camera. To evaluate the proposed model, we conducted experiments with our own dataset and compared the results with some state-of-the-art (SOTA) models. The proposed network achieved the highest detection accuracy with the fewest parameters compared to other SOTA models.

3D Human Pose Estimation from multi-view thermal vision sensors

Human Pose Estimation from images allows the recognition of key daily activity patterns in Smart Environments. Current State-of-the-art (SOTA) 3D pose estimators are built on visible spectrum images, which can lead to privacy concerns in Ambient Assisted Living solutions. Thermal Vision sensors are being deployed in these environments, as they preserve privacy and operate in low brightness conditions. Furthermore, multi-view setups provide the most accurate 3D pose estimation, as the occlusion problem is overcome by having images from different perspectives. Nevertheless, no solutions in the literature use thermal vision sensors following a multi-view scheme.In this work, a multi-view setup consisting of low-cost devices is deployed in the Smart Home of the University of Almería. Thermal and visible images are paired using homography, and SOTA solutions such as YOLOv3 and Blazepose are used to annotate the bounding box and 2D pose in the thermal images. ThermalYOLO is built by fine-tuning YOLOv3 and outperforms YOLOv3 by 5% in bounding box recognition and by 1% in IoU value. Furthermore, InceptionResNetV2 is found as the most appropriate architecture for 2D pose estimation.Finally, a 3D pose estimator was built comparing input approaches and convolutional architectures. Results show that the most appropriate architecture is having three single-channel thermal images processed by independent convolutional backbones (ResNet50 in this case). After these, the output is fused with the 2D poses. The resulting convolutional neural network shows excellent behaviour when having occlusions, outperforming single-view SOTA approaches in the visible spectrum.

Logistics box recognition in robotic industrial de-palletising procedure with systematic RGB-D image processing supported by multiple deep learning methods

In an automated box de-palletisation system that utilises robots, vision-based box recognition on the pallet plays the main role in providing picking guidelines. The complexity of the working condition and the target object, particularly the cluttered arrangement and various outer surfaces of the boxes, significantly affect the quality of the outcome. Typically, a large-scale vision dataset is required to train a deep learning object-detection model. However, considerable effort and time is required to achieve this. Therefore, this study proposes a Mask R-CNN-based detection approach for box objects, which is supported by a cycle generative adversarial network (Cycle GAN). The purpose of the Cycle-GAN is to optimise the outer surfaces of boxes by automatically erasing tags, stickers, labels, and symbols that exist on the boxes before loading them to the Mask R-CNN for detection. Subsequently, the obtained result was combined with the output from the developed boundary-enhancing technique that was applied to a depth map. Consequently, the box detection performance was significantly improved, and it was confirmed through experiments with a practical robot system in picking tasks. In the experiments, the success rate of the proposed method was validated using 200 cases of orderly and disorderly arrangements of boxes, respectively. Furthermore, the metric of the mean absolute error between the predicted picking point and the ground truth values for the test cases in the implementation process for the robot operation was also researched.

Pill Box Text Identification Using DBNet-CRNN.

The recognition process of natural scenes is complicated at present, and images themselves may be complex owing to the special features of natural scenes. In this study, we use the detection and recognition of pill box text as an application scenario and design a deep-learning-based text detection algorithm for such natural scenes. We propose an end-to-end graphical text detection and recognition model and implement a detection system based on the B/S research application for pill box recognition, which uses DBNet as the text detection framework and a convolutional recurrent neural network (CRNN) as the text recognition framework. No prior image preprocessing is required in the detection and recognition processes. The recognition result from the back-end is returned to the front-end display. Compared with traditional methods, this recognition process reduces the complexity of preprocessing prior to image detection and improves the simplicity of the model application. Experiments on the detection and recognition of 100 pill boxes demonstrate that the proposed method achieves better accuracy in text localization and recognition results than the previous CTPN + CRNN method. The proposed method is significantly more accurate and easier to use than the traditional approach in terms of both training and recognition processes.

Automated Segmentation of Cervical Cell Images Using IMBMDCR-Net

Early screening of cervical lesions is of great significance in pathological diagnosis. Owing to the complexity of cell morphological changes and the limitations of medical images, accurate segmentation of cervical cells is still a challenging task. In this paper, an isomorphic multi-branch modulation deformable convolution residual model is proposed to extract features for enhancing the segmentation of small cells and overlapping cytoplasmic boundaries. Then the regional feature extraction, boundary box recognition, and adding a single pixel-level mask at the last level are integrated and optimized based on the cascade regional convolution neural network (Cascade R-CNN) to complete the segmentation of cervical cells for getting better accuracy. The proposed framework was evaluated on the ISBI2014 cervical cell segmentation competition public dataset. Experimental results show that the average accuracy of the network model in cervical cell segmentation is 81.1%, and the accuracy of small targets is 77%. To some extent, it can assist pathologists in screening cervical cancer in the early phase.

Person Search by a Bi-Directional Task-Consistent Learning Model

Two-stage person search methods achieve the state-of-the-art performance by separate detection and re-ID stages, but neglect the consistency needs between these two stages. The re-ID stage needs more accurate query bounding boxes and fewer boxes of distractors; The detection stage needs the re-ID stage to have robustness against unavailable detection errors. In this paper, we introduce a novel Bi-directional Task-Consistent Learning (BTCL) person search framework, including a Target-Specific Detector (TSD) and a re-ID model with Dynamic Adaptive Learning Structure (DALS). For the former consistency need, we add a verification head for predicting the similarity scores between query and proposals in parallel with the existing heads for bounding box recognition. Thus, TSD generates accurate boxes for the query-like pedestrians, which are suitable for the re-ID stage. For the re-ID robustness need, DALS dynamically generates a large number of possible detection results in line with the real distribution. By training the re-ID model on data with different types of detection errors, DLAS improves the model robustness to detection inputs. Experimental results show our framework achieves state-of-the-art performance on two widely-used person search datasets.

Parcel Classification and Positioning of Intelligent Parcel Storage System Based on YOLOv5

Parcel storage provides last-mile delivery services as part of the logistics process. In order to build an intelligent system for parcel storage, we conducted a study on parcel box recognition using AI’s deep learning technology. Box detection and location estimation studies were conducted using the YOLOv5 model for parcel recognition, and this model is presently being applied to many studies because it has excellent object recognition and is faster than previous models. The YOLOv5 model is classified into small, medium, large, and xlarge according to the size and performance of the model. In this study, these four models were compared and analyzed to perform an experiment showing the optimal parcel box recognition performance. As a result of the experiment, it was determined that the precision, recall, and F1 of the YOLOv5large model were 0.966, 0.899, and 0.932, respectively, showing a stronger performance than the other models. Additionally, the size of the YOLOv5large is half that of YOLOv5xlarge, and the YOLOv5large showed the optimal performance in object recognition of the parcel box. Therefore, it seems that the basis for building an intelligent parcel storage system, which shows optimal efficiency in real time using the YOLOv5large model, can be laid through the parcel object recognition experiment conducted in this study.

Object Detection for Cargo Unloading System Based on Fuzzy C Means

With the recent increase in the utilization of logistics and courier services, it is time for research on logistics systems fused with the fourth industry sector. Algorithm studies related to object recognition have been actively conducted in convergence with the emerging artificial intelligence field, but so far, algorithms suitable for automatic unloading devices that need to identify a number of unstructured cargoes require further development. In this study, the object recognition algorithm of the automatic loading device for cargo was selected as the subject of the study, and a cargo object recognition algorithm applicable to the automatic loading device is proposed to improve the amorphous cargo identification performance. The fuzzy convergence algorithm is an algorithm that applies Fuzzy C Means to existing algorithm forms that fuse YOLO(You Only Look Once) and Mask R-CNN(Regions with Convolutional Neuron Networks). Experiments conducted using the fuzzy convergence algorithm showed an average of 33 FPS(Frames Per Second) and a recognition rate of 95%. In addition, there were significant improvements in the range of actual box recognition. The results of this study can contribute to improving the performance of identifying amorphous cargoes in automatic loading devices.

Assessing the Preference of Rabbit Does to Social Contact or Seclusion: Results of Different Investigations.

Simple SummaryMany studies were developed to improve welfare of domestic rabbits and on the possibility to increase contacts between co-specifics using animal group housing. However, a lot of behavior disorder in rabbit does was reported. This study presents two experiments carried out, respectively, in Italy (experiment 1) and in Germany (experiment 2), to evaluate the does’ motivation towards social contact. The Italian trial was based on a choice of test cages to investigate the preferences of nulliparous rabbit does for contacts with co-specifics (in a group of four animals) or seclusion. The German trial tested a different group-housing system constituted of four does and their kits to evaluate the same behavioral aspect. Experiment 1 showed that the time spent in seclusion or in group was almost the same (49.61% vs. 50.39%, respectively) whereas, experiment 2 showed that female does with kits preferred to stay alone (71.90%) rather than in a group, probably due to a hierarchical response stimulation. Besides, it showed a different individual preference; not all does like to share space with the others. This study suggests the need to find a cage adapted to the different physiological phase of does, for example, with the possibility to modify it during the breed cycle.The aim of this study was to verify the motivation of rabbit does to social contact or seclusion. The results of two different research activities assessed in Italy (experiment 1) and Germany (experiment 2) through the use of motivational cages are reported. In experiment 1, only the average time of occupation of the group or seclusion zone was recorded of four nulliparous does, while, in experiment 2, the group-housing system provided space for does with kits and consisted of four single areas (nest boxes with individual electronic nest box recognition systems). Experiment 1 showed that does spent a similar amount of time in seclusion or in group (49.61% vs 50.39%, respectively). On the contrary, in experiment 2, does with kits appeared to prefer spending time alone (71.90%) rather than in groups. The presence of kits probably stimulates a hierarchical and aggressive response of the dominant does, with the low-ranking does staying secluded to avoid violent interactions. In fact, in each reproductive cycle, one doe did stay in the group area whereas the other three does used this area in different percentages of time. Further researches are needed to find a good combination of the cage with the does’ physiological phases.

MedGlasses: A Wearable Smart-Glasses-Based Drug Pill Recognition System Using Deep Learning for Visually Impaired Chronic Patients

Today, with the arrival of an aging society, the average age of the population is rising. It is known that the physiology of a person degrades with age. There are approximately 285 million visually impaired people in the world, of whom 140 million are elderly people over the age of 50, and 110 million of these visually impaired elderly people suffer from multiple chronic diseases. In the case of multiple medication usage, these 110 million vulnerable people will be more likely to take the wrong medicines or forget to take their medication. To solve this problem, this paper proposes a wearable smart-glasses-based drug pill recognition system using deep learning, named MedGlasses, for visually impaired people to improve their medication-use safety. The proposed MedGlasses system consists of a pair of wearable smart glasses, an artificial intelligence (AI)-based intelligent drug pill recognition box, a mobile device app, and a cloud-based information management platform. Experimental results show that a recognition accuracy of up to 95.1% can be achieved. Therefore, the proposed MedGlasses system can effectively mitigate the problem of drug interactions caused by taking incorrect drugs, thereby reducing the cost of medical treatment and providing visually impaired chronic patients with a safe medication environment.

A detection method for low-pixel ratio object

Low-pixel object detection is a kind of difficult program. Existing object detection benchmarks and methods mainly focus on standard detection task. However, these way cannot get good performance on low-pixel ratio object detection, which has a few pixel in high resolution images. In order to solve it, we propose a new deep learning framework. This framework improves Faster R-CNN by combining multiple level feature map and optimizing anchor size for bounding box recognition. In order to validate our approach, we collect and annotate a dataset for road garbage detection, which contains 801 images and 966 bounding boxes. Experiments demonstrate that our framework outperforms other state-of-the-art detection methods. What’s more, our method can apply on road garbage target.

Mask R-CNN.

We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without bells and whistles, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code has been made available at: https://github.com/facebookresearch/Detectron.

Compound Heterozygosity for Y Box Proteins Causes Sterility Due to Loss of Translational Repression.

The Y-box proteins YBX2 and YBX3 bind RNA and DNA and are required for metazoan development and fertility. However, possible functional redundancy between YBX2 and YBX3 has prevented elucidation of their molecular function as RNA masking proteins and identification of their target RNAs. To investigate possible functional redundancy between YBX2 and YBX3, we attempted to construct Ybx2 -/- ;Ybx3 -/- double mutants using a previously reported Ybx2 -/- model and a newly generated global Ybx3 -/- model. Loss of YBX3 resulted in reduced male fertility and defects in spermatid differentiation. However, homozygous double mutants could not be generated as haploinsufficiency of both Ybx2 and Ybx3 caused sterility characterized by extensive defects in spermatid differentiation. RNA sequence analysis of mRNP and polysome occupancy in single and compound Ybx2/3 heterozygotes revealed loss of translational repression almost exclusively in the compound Ybx2/3 heterozygotes. RNAseq analysis also demonstrated that Y-box protein dose-dependent loss of translational regulation was inversely correlated with the presence of a Y box recognition target sequence, suggesting that Y box proteins bind RNA hierarchically to modulate translation in a range of targets.

Composite aromatic boxes for enzymatic transformations of quaternary ammonium substrates.

Cation-π interactions to cognate ligands in enzymes have key roles in ligand binding and enzymatic catalysis. We have deciphered the key functional role of both charged and aromatic residues within the choline binding subsite of CTP:phosphocholine cytidylyltransferase and choline kinase from Plasmodium falciparum. Comparison of quaternary ammonium binding site structures revealed a general composite aromatic box pattern of enzyme recognition sites, well distinguished from the aromatic box recognition site of receptors.

Composite Aromatic Boxes for Enzymatic Transformations of Quaternary Ammonium Substrates

AbstractCation–π interactions to cognate ligands in enzymes have key roles in ligand binding and enzymatic catalysis. We have deciphered the key functional role of both charged and aromatic residues within the choline binding subsite of CTP:phosphocholine cytidylyltransferase and choline kinase from Plasmodium falciparum. Comparison of quaternary ammonium binding site structures revealed a general composite aromatic box pattern of enzyme recognition sites, well distinguished from the aromatic box recognition site of receptors.

Flexibility of PCNA-protein interface accommodates differential binding partners.

The expanding roles of PCNA in functional assembly of DNA replication and repair complexes motivated investigation of the structural and dynamic properties guiding specificity of PCNA-protein interactions. A series of biochemical and computational analyses were combined to evaluate the PIP Box recognition features impacting complex formation. The results indicate subtle differences in topological and molecular descriptors distinguishing both affinity and stoichiometry of binding among PCNA-peptide complexes through cooperative effects. These features were validated using peptide mimics of p85α and Akt, two previously unreported PCNA binding partners. This study characterizes for the first time a reverse PIP Box interaction with PCNA. Small molecule ligand binding at the PIP Box interaction site confirmed the adaptive nature of the protein in dictating overall shape and implicates allosterism in transmitting biological effects.

DNA replication origins in archaea.

DNA replication initiation, which starts at specific chromosomal site (known as replication origins), is the key regulatory stage of chromosome replication. Archaea, the third domain of life, use a single or multiple origin(s) to initiate replication of their circular chromosomes. The basic structure of replication origins is conserved among archaea, typically including an AT-rich unwinding region flanked by several conserved repeats (origin recognition box, ORB) that are located adjacent to a replication initiator gene. Both the ORB sequence and the adjacent initiator gene are considerably diverse among different replication origins, while in silico and genetic analyses have indicated the specificity between the initiator genes and their cognate origins. These replicator–initiator pairings are reminiscent of the oriC-dnaA system in bacteria, and a model for the negative regulation of origin activity by a downstream cluster of ORB elements has been recently proposed in haloarchaea. Moreover, comparative genomic analyses have revealed that the mosaics of replicator-initiator pairings in archaeal chromosomes originated from the integration of extrachromosomal elements. This review summarizes the research progress in understanding of archaeal replication origins with particular focus on the utilization, control and evolution of multiple replication origins in haloarchaea.

Characterization of the Replication Initiator Orc1/Cdc6 from the Archaeon Picrophilus torridus

Eukaryotic DNA replication is preceded by the assembly of prereplication complexes (pre-RCs) at or very near origins in G1 phase, which licenses origin firing in S phase. The archaeal DNA replication machinery broadly resembles the eukaryal apparatus, though simpler in form. The eukaryotic replication initiator origin recognition complex (ORC), which serially recruits Cdc6 and other pre-RC proteins, comprises six components, Orc1-6. In archaea, a single gene encodes a protein similar to both the eukaryotic Cdc6 and the Orc1 subunit of the eukaryotic ORC, with most archaea possessing one to three Orc1/Cdc6 orthologs. Genome sequence analysis of the extreme acidophile Picrophilus torridus revealed a single Orc1/Cdc6 (PtOrc1/Cdc6). Biochemical analyses show MBP-tagged PtOrc1/Cdc6 to preferentially bind ORB (origin recognition box) sequences. The protein hydrolyzes ATP in a DNA-independent manner, though DNA inhibits MBP-PtOrc1/Cdc6-mediated ATP hydrolysis. PtOrc1/Cdc6 exists in stable complex with PCNA in Picrophilus extracts, and MBP-PtOrc1/Cdc6 interacts directly with PCNA through a PIP box near its C terminus. Furthermore, PCNA stimulates MBP-PtOrc1/Cdc6-mediated ATP hydrolysis in a DNA-dependent manner. This is the first study reporting a direct interaction between Orc1/Cdc6 and PCNA in archaea. The bacterial initiator DnaA is converted from an active to an inactive form by ATP hydrolysis, a process greatly facilitated by the bacterial ortholog of PCNA, the β subunit of Pol III. The stimulation of PtOrc1/Cdc6-mediated ATP hydrolysis by PCNA and the conservation of PCNA-interacting protein motifs in several archaeal PCNAs suggest the possibility of a similar mechanism of regulation existing in archaea. This mechanism may involve other yet to be identified archaeal proteins.