Lines Of Interest Research Articles

Edge-Computing-Based People-Counting System for Elevators Using MobileNet–Single-Stage Object Detection

Existing elevator systems lack the ability to display the number of people waiting on each floor and inside the elevator. This causes an inconvenience as users cannot tell if they should wait or seek alternatives, leading to unnecessary time wastage. In this work, we adopted edge computing by running the MobileNet–Single-Stage Object Detection (SSD) algorithm on edge devices to recognize the number of people inside an elevator and waiting on each floor. To ensure the accuracy of people counting, we fine-tuned the SSD parameters, such as the recognition frequency and confidence thresholds, and utilized the line of interest (LOI) counting strategy for people counting. In our experiment, we deployed four NVIDIA Jetson Nano boards in a four-floor building as edge devices to count people when they entered specific areas. The counting results, such as the number of people waiting on each floor and inside the elevator, were provided to users through a web app. Our experimental results demonstrate that the proposed method achieved an average accuracy of 85% for people counting. Furthermore, when comparing it to sending all images back to a remote server for people counting, the execution time required for edge computing was shorter, without compromising the accuracy significantly.

Mapping artifacts generated in a tooth adjacent to titanium and zirconia implants located in the endomass and exomass in cone beam computed tomography: an ex vivo study

We evaluated artifact expression in a root adjacent to a dental implant located in the endomass or exomass in cone beam computed tomography (CBCT). We placed a single titanium or zirconia implant adjacent to a premolar root in a dried human mandible and acquired CBCT scans using an OP300 Maxio and a Picasso Trio 3D unit exposing a 5x5cm field of viewwith the implant in the endomass or exomass (implant groups) or without an implant (control group). We measured the mean gray values (MGVs) for 8 lines of interest (LOIs) around the root canal of the premolar and compared the MGVs by analysis of variance, with significance established at P < 0.05. For the OP300 Maxio scans, the MGVs of both implant groups in the endomass were lower than the MGVs of the control group. In the exomass, the titanium group had MGVs similar to the control group, but the zirconia group produced significantly lower MGVs than control and titanium groups (P <.0001), indicating the presence of hypodense artifacts. For the Picasso Trio 3D scans, the MGVs of both implant groups in the endomass were similar to the MGVs of the control group. In the exomass, the zirconia group generated lower MGVs than control and titanium groups (P <.0001), indicating hypodense artifacts. When performing CBCT examination, titanium implants produce less artifact expression in the exomass, and zirconia implants have less artifact expression in the endomass.

Classification of brain lesions using a machine learning approach with cross-sectional ADC value dynamics

Diffusion-weighted imaging (DWI) and its numerical expression via apparent diffusion coefficient (ADC) values are commonly utilized in non-invasive assessment of various brain pathologies. Although numerous studies have confirmed that ADC values could be pathognomic for various ring-enhancing lesions (RELs), their true potential is yet to be exploited in full. The article was designed to introduce an image analysis method allowing REL recognition independently of either absolute ADC values or specifically defined regions of interest within the evaluated image. For this purpose, the line of interest (LOI) was marked on each ADC map to cross all of the RELs’ compartments. Using a machine learning approach, we analyzed the LOI between two representatives of the RELs, namely, brain abscess and glioblastoma (GBM). The diagnostic ability of the selected parameters as predictors for the machine learning algorithms was assessed using two models, the k-NN model and the SVM model with a Gaussian kernel. With the k-NN machine learning method, 80% of the abscesses and 100% of the GBM were classified correctly at high accuracy. Similar results were obtained via the SVM method. The proposed assessment of the LOI offers a new approach for evaluating ADC maps obtained from different RELs and contributing to the standardization of the ADC map assessment.

The metal post material influences the performance of artefact reduction algorithms in CBCT images.

This study aimed to assess the effect of the MAR tool on the expression of artefacts in different regions of a tooth restored with different types of metal posts. Alveolar sockets (anterior, and posterior region) of a mandible and an unirradicular tooth were used. Cone beam computed tomography scans of the tooth without a metal post, and with cobalt-chromium (Co-Cr), nickel-chromium (Ni-Cr), or silver-palladium (Ag-Pd) were individually obtained, with 2 MAR conditions: disabled, and enabled. In an axial reconstruction, lines of interest (LOIs) were set around the canal: 4 in oblique (mesiobuccal, distobuccal, mesiolingual, distolingual) directions, and 4 in orthogonal (mesial, distal, buccal, lingual) directions. Beam-hardening artefacts expression was determined by calculating the difference in the mean of gray values (DMGV) between the experimental and control groups for each LOI. There was no significant difference in the DMGV values between "without MAR" and "with MAR" for any LOI, in neither anterior nor posterior mandible (p>0.05), for the Ni-Cr and Co-Cr groups. For the Ag-Pd, significant differences in the DMGV values were observed between "without MAR" and "with MAR" for most LOIs (p<0.05), mainly in oblique directions in the anterior region, and mesio-distal direction in the posterior region. MAR acted mostly in hypodense artefacts (negative DMGV). The effectiveness of the MAR tool of the OP300 CBCT unit varied according to the post material tested. It was effective in reducing the expression of artefacts raised by the Ag-Pd post, mainly in the tooth regions affected by hypodense artefacts, regardless of the mandibular region.

Real-time bi-directional people counting using an RGB-D camera

Purpose The purpose of this paper is to propose a new framework and improve a bi-directional people counting technique using an RGB-D camera to obtain accurate results with fast computation time. Therefore, it can be used in real-time applications. Design/methodology/approach First, image calibration is proposed to obtain the ratio and shift values between the depth and the RGB image. In the depth image, a person is detected as foreground by removing the background. Then, the region of interest (ROI) of the detected people is registered based on their location and mapped to an RGB image. Registered people are tracked in RGB images based on the channel and spatial reliability. Finally, people were counted when they crossed the line of interest (LOI) and their displacement distance was more than 2 m. Findings It was found that the proposed people counting method achieves high accuracy with fast computation time to be used in PCs and embedded systems. The precision rate is 99% with a computation time of 35 frames per second (fps) using a PC and 18 fps using the NVIDIA Jetson TX2. Practical implications The precision rate is 99% with a computation time of 35 frames per second (fps) using a PC and 18 fps using the NVIDIA Jetson TX2. Originality/value The proposed method can count the number of people entering and exiting a room at the same time. If the previous systems were limited to only one to two people in a frame, this system can count many people in a frame. In addition, this system can handle some problems in people counting, such as people who are blocked by others, people moving in another direction suddenly, and people who are standing still.

Mapping the expression of beam hardening artefacts produced by metal posts positioned in different regions of the dental arch.

To objectively assess the expression and direction of artefacts in the vicinity of metal posts positioned in different mandibular regions. A human mandible had two sockets prepared-anterior and posterior regions-to accommodate a single-rooted tooth. Two CBCT units, Picasso Trio and OP300, were used. CBCT images of the tooth without metal post, and with silver-palladium, nickel-chromium, or cobalt-chromium posts (experimental groups) were individually acquired. Then, 8 lines of interest (LOIs) were determined around the root canal, in an axial reconstruction: 4 in orthogonal (buccal, lingual, mesial, and distal) directions, and 4 in oblique (mesiobuccal, distobuccal, mesiolingual, and distolingual) directions. The mean of gray values was measured for each LOI. For the OP300, in general, all experimental groups showed greater expression of hyperdense streaks in orthogonal LOIs and greater expression of hypodense streaks in oblique LOIs. For the Picasso Trio, for both mandibular regions, the buccal LOI showed greater expression of hypodense streaks; conversely, greater expression of hyperdense streaks was observed in the distal and distobuccal LOIs in the anterior region and in the mesiolingual LOI in the posterior region. The silver-palladium group showed significantly greater expression of beam hardening artefacts in the posterior region of the mandible, regardless of the CBCT unit (p < 0.05). The mandibular posterior region showed significantly lower gray values than the anterior region (p < 0.05). The expression and direction of artefacts produced in the vicinity of metal posts vary according to the mandibular region, composition of the post, and CBCT unit. The mapping of artefacts produced by metal posts reveals the root regions with greater expression of hypodense and hyperdense artefacts, which may contribute to recognizing regions more likely to mimic or hide root fracture lines.

The demodulating and encoding heritage (DEH) platform for mobile digital culture heritage (M-DCH)

This work designed and developed an innovative Mobile Digital Culture Heritage (M-DCH) platform called Demodulating and Encoding Heritage (DEH) to allow users to have the on-site exploring of the culture heritage using handheld devices and wireless mobile networks. Since most of the current M-DCH platforms (i) view a heritage object as a point and (ii) do not have the group-like authoring scenario, main innovations of the DEH platform are twofold. (1) DEH interprets M-DCH’s contents from the viewpoint of Point of Interest (POI), Line of Interest (LOI), Area of Interest (AOI), and Story/Site of Interest (SOI). The authors’ roles of POIs/LOIs/AOIs/SOIs can be classified as player (regular users), expert and narrator. The work defines the logic rules of composing some POIs (POIs, LOIs and/or AOIs) into a LOI/AOI (SOI) according to (i) the author’s role and (ii) the status, which can be (i) public or private and (ii) successfully verified, unsuccessfully verified or un-verified of the corresponding LOI/AOI (SOI). (2) DEH defines a new grouping function that allows the group creator to have the privilege of modifying contents created by group members. Details of the functional specification, system design and usage examples of DEH’s APPs and web are presented.

Monodimensional enlargement of resolved field of view in line scan cameras by a single doubly symmetric mirror

The geometry of the sensed and thus optically relevant area in the image plane of line scan cameras suggests that improvement of the overall system is possible, by joint design of optics and electronics. By a technique named Linear Fovea (LF) proposed in this paper, the optical contrast along the Line-of-Interest (LOI) in the image plane i.e. the location of the opto-electronic sensor, is optimized at the expense of contrast elsewhere. In the LF method, employing a single doubly symmetric surface, the resolved field of view (RFOV) as the merit function, is maximized, by finding appropriate Zernike coefficients describing that surface. The technique is applied on a proof-of-concept single mirror telescope and both MTF analysis and standard target imaging simulation are performed. Results show that by this technique, the RFOV can be doubled, or alternatively the illumination can be increased 2.5 times keeping RFOV unchanged. Furthermore, a rather uniform response is kept along the LOI.

Quantifying Patterns of Smooth Muscle Motility in the Gut and Other Organs With New Techniques of Video Spatiotemporal Mapping.

The uses and limitations of the various techniques of video spatiotemporal mapping based on change in diameter (D-type ST maps), change in longitudinal strain rate (L-type ST maps), change in area strain rate (A-type ST maps), and change in luminous intensity of reflected light (I-maps) are described, along with their use in quantifying motility of the wall of hollow structures of smooth muscle such as the gut. Hence ST-methods for determining the size, speed of propagation and frequency of contraction in the wall of gut compartments of differing geometric configurations are discussed. We also discuss the shortcomings and problems that are inherent in the various methods and the use of techniques to avoid or minimize them. This discussion includes, the inability of D-type ST maps to indicate the site of a contraction that does not reduce the diameter of a gut segment, the manipulation of axis [the line of interest (LOI)] of L-maps to determine the true axis of propagation of a contraction, problems with anterior curvature of gut segments and the use of adjunct image analysis techniques that enhance particular features of the maps.

Counting People Crossing a Line Using Integer Programming and Local Features

We propose an integer programming method for estimating the instantaneous count of pedestrians crossing a line of interest (LOI) in a video sequence. Through a line sampling process, the video is first converted into a temporal slice image. Next, the number of people is estimated in a set of overlapping sliding windows on the temporal slice image, using a regression function that maps from local features to a count. Given that the count in a sliding window is the sum of the instantaneous counts in the corresponding time interval, an integer programming method is proposed to recover the number of pedestrians crossing the LOI in each frame. Integrating over a specific time interval yields the cumulative count of pedestrians crossing the line. Compared with current methods for line counting, our proposed approach achieves state-of-the-art performance on several challenging crowd video data sets.

SU-D-500-07: Evaluation of the Spatial Resolution of Proton/carbon Computed Tomography Using Modulation Transfer Function

Purpose: To characterize the modulation transfer function (MTF) of proton/carbon computed tomography using Geant4 Monte Carlo simulations. To assess the spatial resolution of proton/carbon computed tomography (pCT/cCT). Methods: A cylindrical water (1.0 g/cm3) phantom was specifically modeled with a length of 20 cm and a diameter of 20 cm. Its inserts were composed of bone and lung tissue each with two different densities: bone (1.58 g/cm3), bone (1.29 g/cm3), lung (0.6 g/cm3), and lung (0.3 g/cm3). These inserts were modeled as 10 cm × 16 mm × 8 mm boxes located in the center of the water phantom. The energy of the beams were tested at 230 MeV and 330 MeV for protons and at 430 MeV per nucleon for carbon. The contrast‐to‐noise (CNR) ratio was evaluated along a line‐of‐interest (LOI) across each water‐bone/water‐lung interface in order to ultimately characterize the spatial resolution by the 10% point of the modulation transfer function (MTF10% or MTF10) in units of line‐pairs per mm (lp/mm). Results: The MTF was generated for interfaces of water‐bone(1.58 g/cm3), water‐bone(1.29 g/cm3), water‐lung(0.6 g/cm3), and water‐lung(0.3 g/cm3). Preliminary results indicate that, at 330 MeV, pCT can distinguish the above interfaces at 2.2 lp/mm, 2.5 lp/mm, 2.3 lp/mm and 2.5 lp/mm. These values are consistent with a previous study done for proton/carbon radiography [1], in which the average resolution was 2.3 lp/mm for bone and 2.4 lp/mm for lung. Conclusion: The sub‐millimeter resolution of pCT offers clinicians a potential tool in specific tumor diagnostics (such as in lung cancer), patient‐setup for daily proton therapy, the reduction of absorbed dose relative to photon imaging, and the reduction of range uncertainties in proton/carbon therapy.

SU-E-I-97: Characterizing the Modulation Transfer Function (MTF) of Proton Radiography.

To characterize the modulation transfer function (MTF) of proton radiography using GEANT4 Monte Carlo simulations. A phantom was specifically modeled using five main materials: bone (1.92 g/cm3 ), muscle (1.2 g/cm3 ), water (1.0 g/cm3 ), adipose tissue (0.9 g/cm3 ), and lung (0.3 g/cm3 ). The basic geometry of the phantom consists of cube-shaped inserts of biological materials placed in water. The thickness of the water, the size of the cube, the depth of the cube in the water, and the proton beam energy have all been varied and studied. The contrast-to-noise ratio (CNR) between the two materials was evaluated at multiple points along a line-of-interest (LOI) in order to ultimately characterize the spatial resolution by the 10% point of the modulation-transfer-function (MTF10% or MTF 10). The MTF was generated for interfaces of water-lung, water-bone, water-muscle, water-adipose. This study indicates that proton radiography can distinguish one material from another with a resolution better than 1 mm for water-adipose and water-muscle or sub-millimeter in cases of water-bone and water-lung interfaces. The sub-millimeter resolution of proton radiography offers clinicians a potentially tool in specific tumor diagnostics (such as in lung cancer), patient-setup for daily proton therapy, and the reduction of absorbed dose delivered when compared to photon imaging.