LED Image Research Articles

Mobile recognition and positioning for multiple visible light communication cells using a convolutional neural network

The industrial Internet of Things (IIoT) environment involves multiple production items, such as robots and automated guided vehicles (AGVs), among others. The practical industrial scenario requires communication of production items while also considering mobile recognition and positioning. Hence the perception approach requires not only combining communications but also realizing the recognition and positioning of multiple communication cells. This Letter proposes a multi-optical cell recognition and positioning framework based on LED image features. The LED images are obtained by a CMOS image sensor. This framework utilizes convolutional neural networks (CNN) to train LED images for recognition between multiple optical cells and locates precise positions through region recognition within the optical cells. The experimental results show that the mean accuracy of the CNN model for two LED cells is above 99%, and the mean accuracy of region recognition within the optical cell is as high as 100%, which is significantly better than other traditional recognition algorithms. Therefore, the proposed framework can provide location-aware services for visible light communication and has a wide application prospect in IIoT.

Method for measuring bubble size under low-light conditions for mass transfer enhancement in industrial-scale systems

Shadowgraphy-based image processing studied in the literature on multiphase processes has led to meaningful advances in mass transfer enhancement via interfacial area enlargement. However, the industrial applications of shadowgraphy have been limited due to the requirement of an additional light source at specific locations. To overcome this limitation, in this study, a new bubble size measurement technique in low-light conditions is proposed. The technique uses reflected LED image on the bubble surface to estimate the bubble size in low-light conditions and includes a newly derived measurement correlation model, which was validated with lab-scale experimental data. Furthermore, the proposed model was applied to industrial-scale bubble systems for hydrogen fluoride (HF) removal. Using the bubble properties identified through analysis, the overall mass transfer coefficient (OMTC) was determined as an indicator of HF mass transfer enhancement. The optimal conditions for HF mass transfer were determined by identifying the system with the highest OMTC. By manipulating the pressure difference and flow rate, OMTC was increased by ∼78% of the base case.

Deep Learning-Based Robust Visible Light Positioning for High-Speed Vehicles

Robustness is a key factor for real-time positioning and navigation, especially for high-speed vehicles. While visible light positioning (VLP) based on LED illumination and image sensors is widely studied, most of the VLP systems still suffer from the high positioning latency and the image blurs caused by high-speed movements. In this paper, a robust VLP system for high-speed vehicles is proposed based on a deep learning and data-driven approach. The proposed system can significantly increase the success rate of decoding VLP-LED user identifications (UID) from blurred images and reduce the computational latency for detecting and extracting VLP-LED stripe image regions from captured images. Experimental results show that the success rate of UID decoding using the proposed BN-CNN model could be higher than 98% when that of the traditional Zbar-based decoder falls to 0, while the computational time for positioning is decreased to 9.19 ms and the supported moving speed of our scheme can achieve 38.5 km/h. Therefore, the proposed VLP system can enhance the robustness against high-speed movement and guarantee the real-time response for positioning and navigation for high-speed vehicles.

Investigation of photobiological safety of blue light of LED image projector of household use

Abstract
 Light emitted in the wavelength range of 400–500 nm causes a very significant photochemical effect on the retina. At certain doses of blue light energy can cause damage to the retina, so the study of the safety of blue light projectors used in domestic conditions is an urgent task. The aim of this work is to study the photobiological safety of light image projectors with light emitting diodes. The paper analyzes the requirements for photobiological safety of tube systems and methods for assessing the danger of blue light of projectors that use LED light sources. The information is provided on the equipment for measurement of spectroradiometric and photometric parameters of beams of projectors, as well as features of radiance and irradiance measurement of lamp systems in which lens optics is used.
 The results of research of photobiological danger levels of blue light of image projectors intended for use in household conditions are presented. It is shown that at a distance of 1 m from the projector (the distance for which the level of photobiological safety of blue light is normalized), the investigated type of projector belongs to the risk group RG1 (insignificant risk). When the distance is reduced to 0.5 m and 0.2 m, the risk of blue light increases to the RG2 group (medium risk).
 The features of the requirements for photobiological safety of blue light of image projectors used in everyday life, educational institutions and in other conditions are analyzed. In particular, the risk group at a distance of 1 m from the collimator lens should not exceed RG1, and it is recommended to put an additional label on the projector indicating the risk group RG1 and mark this information in the user manual.
 Conclusions are made on the compliance of the studied low-power image projectors with the safety conditions of blue light, established by the international standard IEC 62471-5 (Photobiological Safety of Lamps and Lamp Systems – Part 5 Image Projectors).

UAV ranging method based on monocular vision

Aiming at the problem that the unmanned aerial vehicle (UAV) cannot complete the positioning and ranging in the complex background and the signal is interfered, we proposed a monocular measurement scheme of assigning 4 LED fiducials (also called landmarks) with 3 colors in total on the target UAV. Firstly, feature areas are screened out by using high saturation, high brightness and specific hue range of LED in HSV (Hue, Saturation, Value) color space. After morphological processing, the background that does not meet the features is removed. Then, the mapping relationship between target LED fiducials and image fiducials is established by using the average hue of connected domain and special geometric layout. The improved Lambda-Twist is used to find the distance from the target UAV to the center of the camera, and the performance of the UAV recognition and ranging from 90m to 245m is tested. The experimental results show that the method can quickly and accurately measure the distance to the target, and can be used for UAV auxiliary measurement.

Ambient light noise filtering technique for multimedia high speed transmission system in MIMO-VLC

In this paper, we analyze the degradation of system performance due to the variation of ambient illumination in MIMO-VLC environment for high-speed multimedia transmission and propose an adaptive filter technique to solve this problem. Conventional PD-based visible light communication systems are not suitable for multimedia high-speed transmission. It is vulnerable to optical interference in the presence of a large number of LEDs, and it is difficult to implement MIMO. For large-capacity high-speed transmission, it is necessary to increase the transmission capacity. MIMO-VLC using m * m LED matrix and image sensor enables high-speed transmission through data parallel transmission. However, the received power is greatly influenced by the surrounding light source environment and the receiving distance. Illumination changes caused by non-scattered lights in the indoor environment, which deteriorates the performance of the system. In order to solve this problem, adaptive filter coefficients according to the ambient illuminance are applied to improve on / off recognition rate of LED matrix. By using this technique to adjust the filter values for image processing according to the step-by-step illumination, the multi-LED pixel can be accurately recognized and the MISO-VLC QoS can be maintained.

Improved Smartphone-Based Indoor Pedestrian Dead Reckoning Assisted by Visible Light Positioning

Aiming at the problem of heavy preparation work and high cost of indoor positioning technology, this paper proposes a visible light positioning (VLP)-assisted smartphone-based pedestrian dead reckoning (PDR) indoor positioning technology. This technology combines the advantages of both PDR and VLP algorithms for indoor positioning. During the pedestrian position calculation, the method of using the averaging of mode to process the heading angle data is creatively used, which improves the positioning accuracy and increases reliable operation time. At the same time, the two-level verification method consisting of cyclic redundancy check and range constraint is implemented for VLP, which uses a LED carrying a modulation code as a beacon for absolute position and the camera of smartphone to capture/decode the LED image to obtain the absolute position and periodically correct the position of the PDR positioning, thereby avoiding the accumulation of PDR positioning technology errors. The scope and reliability of PDR are demonstrated. The experiments show that this method is a kind of low cost, high robust, and high precision indoor positioning technology.

A Tilt Receiver Correction Method for Visible Light Positioning Using Machine Learning Method

In recent years, a visible light positioning (VLP) technology based on complementary metal-oxide-semiconductor sensors has been widely studied due to its high precision and high robustness. However, the existing VLP algorithm based on image sensors often fails to achieve a good positioning effect when the camera is tilted. In order to solve this problem, we propose a neural network algorithm to correct the error caused by the tilt angle of the camera. Because when the tilt angle is different, the LED image captured by the camera will be different and produce different characteristics. By extracting these features and using neural networks to establish the relationship between the characteristics of the LED image and the distance between the receiving and sending terminal, we finally achieve the positioning of the camera by the triangulation algorithm. Experiments demonstrate that our positioning algorithm can achieve high-precision positioning and can be applied to most indoor positioning systems.

The LED-ID Detection and Recognition Method Based on Visible Light Positioning Using Proximity Method

Complementary metal-oxide-semiconductor (CMOS) sensor based visible light positioning (VLP) has been widely studied in recent years due to its high robustness and high precision. In most researches about CMOS sensor based VLP, researchers always focus on the high-precision positioning algorithm but ignore that the accuracy of LED-ID detection and recognition plays a more important role in a VLP system. Without the correct recognition of LED-ID, the positioning algorithm would be meaningless no matter how effective it is. In addition, high-precision positioning is not required in most applications since it is enough for people to know just the approximate location. To solve these problems, in this paper, an LED-ID detection and recognition method based on visible light positioning using proximity method is propose. Different from the traditional LED-ID coding and decoding method, we create different features for different LED-ID, and use a machine learning method to identify the LED-ID once the feature extraction and selection of the LED image is achieved with an image processing method. It is the first time the machine learning method is used for LED-ID recognition in VLP. Moreover, we use a proximity-based positioning method to get the approximate location since it is easy to obtain once the LED-ID is recognized. The studies we have demonstrated shows that the proposed method can achieve high LED-ID recognition rate, and provide enough unique LED-ID for variable large-scale indoor VLP system. Furthermore, with the development of camera technology, the number of the unique LED-ID and the maximum recognizable distance would increase. Therefore, this scheme may be considered as one of the useful LED-ID detection and recognition method for visible light positioning in the future.

A Single LED Positioning System Based on Circle Projection

A simple and efficient visible light positioning (VLP) system based on a single LED lamp is proposed and demonstrated. In our system, a common circular LED lamp with a point marker is used as a transmitter and a smartphone camera serves as a receiver. The LED image is no longer treated as a point as in the existing works, but as an image whose geometric features are exploited to determine the receiver's orientation and location relative to the reference LED lamp. The expressions for determining the receiver's orientation and location are derived in terms of the geometric parameters. Our algorithm enhances the system robustness by overcoming the possible interruptions in conventional VLP systems, which require multiple LEDs. Meanwhile, it avoids the usage of angular sensors, which are unable to provide reliable and accurate measurements on receiver's orientation in mobile indoor environment. We experimentally evaluate and analyze the performance of the proposed VLP system. A centimeter-level positioning accuracy is achieved in an area of 3 m × 3 m.

Data rate enhancement of optical camera communications by compensating inter-frame gaps

Optical camera communications (OCC) is a convenient way of transmitting data between LED lamps and image sensors that are included in most smart devices. Although many schemes have been suggested to increase the data rate of the OCC system, it is still much lower than that of the photodiode-based LiFi system. One major reason of this low data rate is attributed to the inter-frame gap (IFG) of image sensor system, that is, the time gap between consecutive image frames. In this paper, we propose a way to compensate for this IFG efficiently by an interleaved Hamming coding scheme. The proposed scheme is implemented and the performance is measured.

Demonstration of 136 dB dynamic range capability for a simultaneous dual optical band CAOS camera.

For the first time, proposed and demonstrated is a simultaneous dual optical band coded access optical sensor (CAOS) camera design suited for extreme contrast multispectral bright target scenarios. Deploying a digital micromirror devices (DMDs)-based time-frequency agile pixels CAOS-mode within a two point detector spatially and spectrally isolating framework, this imager simultaneously and independently detects pixel selective image information for two different broad spectral bands that further undergo independent spectral image data extraction via finer-tuned wavelength filtering using all-optical or CAOS-mode electronic filters. A proof-of-concept visible-near infrared band CAOS imager is successfully demonstrated using a target scene containing LEDs and engaging narrowband optical filters. In addition, using the CAOS-mode, demonstrated is the RF domain simultaneous color content monitoring of a white light LED image pixel. Also proposed is the use of a higher bit count analog-to-digital converter (ADC) with both range and sampling duration parameter control along with a larger data set electronic DSP to extract higher DSP gain and realize additional noise suppression. Using a 16-bit ADC and 2,097,152 point fast Fourier transform (FFT) digital signal processing (DSP) for a 633 nm laser engaged test target scene that is subject to nearly 7 decades (107) of gradual optical attenuation, the experimental camera demonstrates an agile pixel extreme dynamic range of 136 dB, which is a 56 dB improvement over the previous CAOS-imaging demonstrations.

Close-up images of the final Philae landing site on comet 67P/Churyumov-Gerasimenko acquired by the ROLIS camera

After coming to rest on the night side of comet 67P/Churyumov-Gerasimenko, the ROLIS camera on-board Rosetta’s Philae lander acquired five images of the surface below the lander, four of which were with the aid of LED illumination of different colors. The images confirm that Philae was perched on a sloped surface. A local horizon is visible in one corner of the image, beyond which we can see the coma. Having spent a full day on the surface Philae was commanded to lift and rotate, after which a final, sixth, LED image was acquired. The change in perspective allowed us to construct a shape model of the surface. The distance to the foreground was about 80 cm, much larger than the nominal 30 cm. This caused stray light, rather than directly reflected LED light, to dominate the image signal, complicating the analysis. The images show a lumpy surface with a roughness of apparently fractal nature. Its appearance is completely different from that of the first landing site, which was characterized by centimeter to meter-sized debris (Mottola et al., 2015). We recognize neither particles nor pores at the image resolution of 0.8 mm per pixel and large color variations are absent. The surface has a bi-modal brightness distribution that can be interpreted in terms of the degree of consolidation, a hypothesis that we support with experimental evidence. We propose the surface below the lander to consist of smooth, cracked plates with unconsolidated edges, similar to terrain seen in CIVA images.

Ice Fog and Light Snow Measurements Using a High-Resolution Camera System

Ice fog, diamond dust, and light snow usually form over extremely cold weather conditions, and they affect both visibility and Earth’s radiative energy budget. Prediction of these hydrometeors using models is difficult because of limited knowledge of the microphysical properties at the small size ranges due to measurement issues. These phenomena need to be better represented in forecast and climate models; therefore, in addition to remote sensing accurate measurements using ground-based instrumentation are required. An imaging instrument, aimed at measuring ice fog and light snow particles, has been built and is presented here. The ice crystal imaging (ICI) probe samples ice particles into a vertical, tapered inlet with an inlet flow rate of 11 L min−1. A laser beam across the vertical air flow containing the ice crystals allows for their detection by a photodetector collecting the scattered light. Detected particles are then imaged with high optical resolution. An illuminating LED flash and image capturing are triggered by the photodetector. In this work, ICI measurements collected during the fog remote sensing and modeling (FRAM) project, which took place during Winter of 2010–2011 in Yellowknife, NWT, Canada, are summarized and challenges related to measuring small ice particles are described. The majority of ice particles during the 2-month-long campaign had sizes between 300 and 800 μm. During ice fog events the size distribution measured had a lower mode diameter of 300 μm compared to the overall campaign average with mode at 500 μm.

Different aerial image formation into two directions by crossed-mirror array

Different aerial images in two directions can be formed by using a crossed-mirror array (CMA). The array is composed of reflective imaging elements based on dihedral corner mirrors and is used to form an aerial LED image. However, as CMA is formed from one-sided mirrors for keeping down the cost, only two of the four aperture walls of CMA are mirrors and the other two are back sides of the mirrors. As a dihedral corner mirror needs double reflections, aerial image quality varies with the incidence direction of light on the CMA. In this paper, we propose a new CMA configuration for forming aerial images in two directions more clearly. Clear different aerial images in two directions can be successfully formed by using the new CMA configuration.

引入DDS算法的旋转LED空间稳定成像

引入DDS算法的旋转LED空间稳定成像

A simple and low-cost biofilm quantification method using LED and CMOS image sensor

A novel biofilm detection platform, which consists of a cost-effective red, green, and blue light-emitting diode (RGB LED) as a light source and a lens-free CMOS image sensor as a detector, is designed. This system can measure the diffraction patterns of cells from their shadow images, and gather light absorbance information according to the concentration of biofilms through a simple image processing procedure. Compared to a bulky and expensive commercial spectrophotometer, this platform can provide accurate and reproducible biofilm concentration detection and is simple, compact, and inexpensive. Biofilms originating from various bacterial strains, including Pseudomonas aeruginosa (P. aeruginosa), were tested to demonstrate the efficacy of this new biofilm detection approach. The results were compared with the results obtained from a commercial spectrophotometer. To utilize a cost-effective light source (i.e., an LED) for biofilm detection, the illumination conditions were optimized. For accurate and reproducible biofilm detection, a simple, custom-coded image processing algorithm was developed and applied to a five-megapixel CMOS image sensor, which is a cost-effective detector. The concentration of biofilms formed by P. aeruginosa was detected and quantified by varying the indole concentration, and the results were compared with the results obtained from a commercial spectrophotometer. The correlation value of the results from those two systems was 0.981 (N=9, P <0.01) and the coefficients of variation (CVs) were approximately threefold lower at the CMOS image-sensor platform.

A Single-Input Multiple-Output Optical System for Mobile Communication: Modeling and Validation

A single-input multiple-output optical communication system based on an LED transmitter and a receiving pixel grid of a camera is considered. The LED image received at a reference distance is modeled using a Gaussian mixture and the pixel values at any other desired distance are estimated using the reference image. Both optimal and suboptimal pixel combining techniques are analyzed. The experimental validation includes placement of the camera on a motorized translation stage to analyze the effects of random locations of the LED image on the pixel grid. Results show excellent agreement between proposed model and the experiment. When random image locations over the pixel grid due to mobility are considered, a receiver that uses only the best pixel is found to lose more than 50% of the received power over a maximal ratio combiner at a distance of 2 m.

Indoor Positioning by LED Visible Light Communication and Image Sensors

High power white LEDs are expected to replace the existing lighting technologies in near future which are also suggested for visible light communication (VLC). We proposed an algorithm for high precision indoor positioning using lighting LEDs, VLC and image sensors. In the proposed algorithm, four LEDs transmitted their three-dimensional coordinate information which were received and demodulated by two image sensors near the unknown position. The unknown position was then calculated from the geometrical relations of the LED images created on the image sensors. We described the algorithm in details. Simulation of the proposed algorithm was done and presented in this paper. This technique did not require any angular measurement which was needed in contemporary positioning algorithms using LED and image sensor. Simulation results showed that the proposed system could estimate the unknown position within the accuracy of few centimeters. Positioning accuracy could be increased by using high resolution image sensors or by increasing the separation between the image sensors. DOI: http://dx.doi.org/10.11591/ijece.v1i2.165

LED and CMOS image sensor based hemoglobin concentration measurement technique

This paper presents a technique that can provide accurate hemoglobin concentration measurement within a simple, cost-effective and compact platform. It is based on a LED and CMOS image sensor without using an environmentally harmful reagent and centrifugal process. The acquired standard curve and adequate image processing steps such as image calibration and noise suppression allow us to accurately estimate the hemoglobin concentration at an error rate of less than 2% for blind tests conducted with the 33 blood specimens of outpatients. The hemoglobin concentration measurement, one of the crucial part of the complete blood count (CBC) as well as the most commonly ordered test in clinical laboratory, within such a small, cheap and fast platform would be useful for wide applications in doctor's office and for applications in resource limited settings as a point-of-care testing tool.