An optical and non-contact continuous measurement method to detect human blood pressure through a high-speed camera is discussed in this paper. With stable ambient light, photoplethysmographic (PPG) signals of face and palm area are obtained simultaneously from the video captured by high-speed camera, whose frame rate should be higher than 100 frames per second. Pulse transit time (PTT) is measured from the R-wave distance between the two PPG signals. The Partial least squares regression( PLSR) model was established to train the samples, and the relationship between PTT and blood pressure, including intra-arterial systolic pressure (SBP) and diastolic pressure (DBP), was established to obtain blood pressure. Compared with the output of traditional sphygmomanometer, the blood pressure data collected from non-contact system has little error and meets the fitting conditions. We first proposed an accurate video-based method for non-contact blood pressure measurement using machine learning, and the average error of SBP is 0.148mmHg and of DBP is 0.359mmHg.
Haze is the result of the interaction between specific climate and human activities. When observing objects in hazy conditions, optical system will produce degradation problems such as color attenuation, image detail loss and contrast reduction. Image haze removal is a challenging and ill-conditioned problem because of the ambiguities of unknown radiance and medium transmission. In order to get clean images, traditional machine vision methods usually use various constraints/prior conditions to obtain a reasonable haze removal solutions, the key to achieve haze removal is to estimate the medium transmission of the input hazy image in earlier studies. In this paper, however, we concentrated on recovering a clear image from a hazy input directly by using Generative Adversarial Network (GAN) without estimating the transmission matrix and atmospheric scattering model parameters, we present an end-to-end model that consists of an encoder and a decoder, the encoder is extracting the features of the hazy images, and represents these features in high dimensional space, while the decoder is employed to recover the corresponding images from high-level coding features. And based perceptual losses optimization could get high quality of textural information of haze recovery and reproduce more natural haze-removal images. Experimental results on hazy image datasets input shows better subjective visual quality than traditional methods. Furthermore, we test the haze removal images on a specialized object detection network- YOLO, the detection result shows that our method can improve the object detection performance on haze removal images, indicated that we can get clean haze-free images from hazy input through our GAN model.
The blood oxygen saturation (SpO2) detection is of great importance in medical science due to its relationship with cardiopulmonary function. Determination of SpO2 by infrared spectroscopy is a commonly used method. In this paper, a novel non-invasive blood oxygen saturation detection method with video image is proposed. Firstly, the index fingertip is illuminated by LEDs with four wavelengths. Then, the reflected light is collected by an image sensor. Meanwhile, a finger clip pulse oximeter is adopted to monitor SpO2 changes. Finally, a color regression model is established with the obtained RGB values and the corresponding SpO2. Furthermore, the influence of quantization noise and other factors is analyzed. Experiments with multiple samples are carried out using the proposed method. The results show that the relative errors between the predicted and reference values are within 5%. Compared with traditional methods, the proposed method can effectively detect the SpO2 in a specific area instead of a single point. In addition, it provides an alternative approach that guides an SpO2 detection device for daily use.
Non-contact, imaging photoplethysmography (IPPG) uses video sequence to measure variations in light absorption, caused by blood volume pulsations, to extract cardiopulmonary parameters including heart rate (HR), pulse rate variability, and respiration rate. Previous researches most focused on extraction of these vital signs base on the focus video, which require a static and focusing environment. However, little has been reported about the influence of defocus blur on IPPG signal’s extraction. In this research, we established an IPPG optical model in defocusing motion conditions. It was found that the IPPG signal is not sensitive to defocus blur by analysis the light intensity distribution in the defocus images. In this paper, a real-time measurement of heart rate in defocus and motion conditions based on IPPG was proposed. Automatically select and track the region of interest (ROI) by constructing facial coordinates through facial key points detection, obtained the IPPG signal. The signal is de-noised to obtain the spectrum by the wavelet filtering, color-distortion filter (CDF) and fast Fourier transform (FFT). The peak of the spectrum is corresponded to heartbeats. Experimental results on a data set of 30 subjects show that the physiological parameters include heart rate and pulse wave, derived from the defocus images captured by the IPPG system, exhibit characteristics comparable to conventional the blood volume pulse (BVP) sensor. Contrast experiment show that the difference between the results measured by both methods is within 3 beat per minute (BPM). This technology has significant potential for advancing personal health care and telemedicine in motion situation.
Atmospheric turbulence is an irregular form of motion in the atmosphere. Because of turbulence interference, when the optical system through the atmosphere of the target imaging, the observed image will appear point intensity diffusion, image blur, image drift and other turbulence effects. Digital recovery of the turbulence-degraded images technique is a classical ill-conditioned problem by removing the blurring effect and suppressing the noise. Traditional approaches relying on image heuristics suffer from high frequency noise amplification and processing artifacts. In this paper, the image degradation models of the turbulent flow are given, the point spread function of turbulence is simulated by the similar Gaussian function model, and investigated a general framework of neural networks for restoring turbulence-degraded images. The blur and additive noise are considered simultaneously. Two solutions respectively exploiting fully convolutional networks (FCN) and conditional Generative Adversarial Networks (CGAN) are presented. The FCN based on minimizing the mean squared reconstruction error (MSE) in pixel space gets high PSNR. On the other side, the CGAN based on perceptual loss optimization criterion retrieves more textures. We conduct comparison experiments to demonstrate the performance at different degree of turbulence intensity from the training configuration. The results indicate that the proposed networks outperform traditional approaches for restoring high frequency details and suppressing noise effectively.
We present a new method to reduce the image artifacts in defocused hybrid imaging systems with a cubic phase mask. Image artifacts are caused by a mismatch of phase during decoding. According to the corresponding relationship between the defocus amount and the object distance, we analyze the feasibility of identifying the defocus amount according to distance information to optimize the decoding process. Experimental results show that our method can improve image quality over a wide range of defocus.
Synthetic aperture radar (SAR) is a microwave imaging equipment based on the principle of synthetic aperture, which has all kinds of characteristics such as all-time, all-weather, high resolution and wide breadth. It also has high research value and applied foreground in the area of military and civilian. In particular, worldwide, a great deal of researches on SAR target classification and identification based Deep Learning are ongoing, and the obtained results are highly effective. However, it is well known that Deep Learning requires a large amount of data, and it is costly and inaccessible to acquire SAR samples through field experiment, so image simulation research for expanding SAR dataset is essential. In this paper, we concentrated on generating highly realistic SAR simulated images for several equipment models using Generative Adversarial Network (GAN) without construction of terrain scene model and RCS material mapping. Then we tested the SAR simulated images on a specialized SAR classification model pretrained on MSTAR dataset. The results showed that simulated targets could be identified and classified accurately, demonstrating the high similarity of SAR simulated images with real samples. Our work could provide a greater variety of available SAR images for target classification and identification study.
Recent advances in convolution neural networks have shown promising results for the challenging task of filling large missing regions in an image with semantically plausible and context aware details. These learning-based methods are significantly more effective in capturing high-level features than prior techniques, but often create distorted structures or blurry textures inconsistent with existing areas. This is mainly due to ineffectiveness of convolutional neural networks in explicitly borrowing or copying information from distant locations. Motivated by these observations, we use a convolution neural networks architecture with Atrous Spatial Pyramid Pooling module, which can obtain multi-scale objection information, to be our inpainting network. We also use global and local Wasserstein discriminators that are jointly trained to distinguish real images from completed ones. We evaluate our approach on four datasets including faces (CelebA) and natural images (Paris Streetview, COCO, ImageNet) and achieved state-of-the-art inpainting accuracy.
Convolution neural network (CNN) has made great success in image classification tasks. Even in the field of synthetic aperture radar automatic target recognition (SAR-ATR), state-of-art results has been obtained by learning deep representation of features on the MSTAR benchmark. However, the raw data of MSTAR have shortcomings in training a SAR-ATR model because of high similarity in background among the SAR images of each kind. This indicates that the CNN would learn the hierarchies of features of backgrounds as well as the targets. To validate the influence of the background, some other SAR images datasets have been made which contains the simulation SAR images of 10 manufactured targets such as tank and fighter aircraft, and the backgrounds of simulation SAR images are sampled from the whole original MSTAR data. The simulation datasets contain the dataset that the backgrounds of each kind images correspond to the one kind of backgrounds of MSTAR targets or clutters and the dataset that each image shares the random background of whole MSTAR targets or clutters. In addition, mixed datasets of MSTAR and simulation datasets had been made to use in the experiments. The CNN architecture proposed in this paper are trained on all datasets mentioned above. The experimental results shows that the architecture can get high performances on all datasets even the backgrounds of the images are miscellaneous, which indicates the architecture can learn a good representation of the targets even though the drastic changes on background.
KEYWORDS: Target detection, Digital signal processing, Detection and tracking algorithms, Image filtering, Image processing, Field programmable gate arrays, Signal processing, Video, Image compression, Target recognition
In order to solve the target fast tracking problem in embedded system, a moving target detecting and tracking algorithm based on a combination of three-frame difference and template matching is proposed. The system utilizes DSP to design a set of image processing equipment and DSP uses TI company’s DM6437.Three-frame difference can detect a initial position of the target, then Mean Normalized Product Correlation(NNPROD) template matching algorithm was utilized in a partial area to achieve a precise position and reduce the amount of calculation. The algorithm utilized four templates and image compression to fit pose and scale changes when moving. To meet the real-time requirement, an improved algorithm of NNPROD was proposed under certain lighting conditions, what ’ s more the C language code was optimized and TI company’s highly optimized VLIB vision library was reasonably utilized. After several tests, the results showed that NNPROD can fit the changing of environmental light well, but more time was needed. The improved method can still work well with the changes of pose and scale when the light changes less intensely , and the processing speed of the improved method increased from the previous 11F / s to 23F / s.
Many convolution neural networks(CNN) architectures have been proposed to strengthen the performance on synthetic aperture radar automatic target recognition (SAR-ATR) and obtained state-of-art results on targets classification on MSTAR database, but few methods concern about the estimation of depression angle and azimuth angle of targets. To get better effect on learning representation of hierarchies of features on both 10-class target classification task and target posture estimation tasks, we propose a new CNN architecture with spatial pyramid pooling(SPP) which can build high hierarchy of features map by dividing the convolved feature maps from finer to coarser levels to aggregate local features of SAR images. Experimental results on MSTAR database show that the proposed architecture can get high recognition accuracy as 99.57% on 10-class target classification task as the most current state-of-art methods, and also get excellent performance on target posture estimation tasks which pays attention to depression angle variety and azimuth angle variety. What’s more, the results inspire us the application of deep learning on SAR target posture description.
A method of keyhole imaging based on camera array is realized to obtain the video image behind a keyhole in shielded space at a relatively long distance. We get the multi-angle video images by using a 2×2 CCD camera array to take the images behind the keyhole in four directions. The multi-angle video images are saved in the form of frame sequences. This paper presents a method of video frame alignment. In order to remove the non-target area outside the aperture, we use the canny operator and morphological method to realize the edge detection of images and fill the images. The image stitching of four images is accomplished on the basis of the image stitching algorithm of two images. In the image stitching algorithm of two images, the SIFT method is adopted to accomplish the initial matching of images, and then the RANSAC algorithm is applied to eliminate the wrong matching points and to obtain a homography matrix. A method of optimizing transformation matrix is proposed in this paper. Finally, the video image with larger field of view behind the keyhole can be synthesized with image frame sequence in which every single frame is stitched. The results show that the screen of the video is clear and natural, the brightness transition is smooth. There is no obvious artificial stitching marks in the video, and it can be applied in different engineering environment .
Imaging Photoplethysmography (IPPG) is an emerging technique for the extraction of vital signs of human being using video recordings. IPPG technology with its advantages like non-contact measurement, low cost and easy operation has become one research hot spot in the field of biomedicine. However, the noise disturbance caused by non-microarterial area cannot be removed because of the uneven distribution of micro-arterial, different signal strength of each region, which results in a low signal noise ratio of IPPG signals and low accuracy of heart rate.
In this paper, we propose a method of improving the signal noise ratio of camera-based IPPG signals of each sub-region of the face using a weighted average. Firstly, we obtain the region of interest (ROI) of a subject’s face based camera. Secondly, each region of interest is tracked and feature-based matched in each frame of the video. Each tracked region of face is divided into 60x60 pixel block. Thirdly, the weights of PPG signal of each sub-region are calculated, based on the signal-to-noise ratio of each sub-region. Finally, we combine the IPPG signal from all the tracked ROI using weighted average. Compared with the existing approaches, the result shows that the proposed method takes modest but significant effects on improvement of signal noise ratio of camera-based PPG estimated and accuracy of heart rate measurement.
When an aircraft flies at a hypersonic speed within the atmosphere, the temperature of the infrared window (IRW) on the aircraft will rise rapidly due to the high-speed incoming flow will produce a severe aerodynamic heating to its optical detection window. The infrared (IR) radiation of the high-temperature gas and optical window will generate severe pneumatic thermal radiation effect upon the detection system, with the performance of the IR detector possibly being reduced or even destroyed.
To evaluate the influence on the target imaging made by the IRW radiation, the experiment on the basis of building a simulating model is conducted by the means of ray tracing so that the accurate transmittance of the IRW can be observed under the different temperature. And then the radiation distribution of the thermal radiation on the detector generated by the IRW radiation noise and target signal can finally be obtained.
This paper also records the different parameters in the detection system being set in the experiment, and analyzes the different influences brought by various factors to the Signal to Noise Ratio (SNR). It is also expected that it will provide a data reference to the following research of radiation noise suppression and design of IR detection system.
Wave-front coding has a great prospect in extending the depth of the optical imaging system and reducing optical aberrations, but the image quality and noise performance are inevitably reduced. According to the theoretical analysis of the wave-front coding system and the phase function expression of the cubic phase plate, this paper analyzed and utilized the feature that the phase function expression would be invariant in the new coordinate system when the phase plate rotates at different angles around the z-axis, and we proposed a method based on the rotation of the phase plate and image fusion. First, let the phase plate rotated at a certain angle around the z-axis, the shape and distribution of the PSF obtained on the image surface remain unchanged, the rotation angle and direction are consistent with the rotation angle of the phase plate. Then, the middle blurred image is filtered by the point spread function of the rotation adjustment. Finally, the reconstruction images were fused by the method of the Laplacian pyramid image fusion and the Fourier transform spectrum fusion method, and the results were evaluated subjectively and objectively. In this paper, we used Matlab to simulate the images. By using the Laplacian pyramid image fusion method, the signal-to-noise ratio of the image is increased by 19%~27%, the clarity is increased by 11%~15% , and the average gradient is increased by 4%~9% . By using the Fourier transform spectrum fusion method, the signal-to-noise ratio of the image is increased by 14%~23%, the clarity is increased by 6%~11% , and the average gradient is improved by 2%~6%. The experimental results show that the image processing by the above method can improve the quality of the restored image, improving the image clarity, and can effectively preserve the image information.
Wavefront coding technique as a means of athermalization applied to infrared imaging system, the design of phase plate is the key to system performance. This paper apply the externally compiled programs of ZEMAX to the optimization of phase mask in the normal optical design process, namely defining the evaluation function of wavefront coding system based on the consistency of modulation transfer function (MTF) and improving the speed of optimization by means of the introduction of the mathematical software. User write an external program which computes the evaluation function on account of the powerful computing feature of the mathematical software in order to find the optimal parameters of phase mask, and accelerate convergence through generic algorithm (GA), then use dynamic data exchange (DDE) interface between ZEMAX and mathematical software to realize high-speed data exchanging. The optimization of the rotational symmetric phase mask and the cubic phase mask have been completed by this method, the depth of focus increases nearly 3 times by inserting the rotational symmetric phase mask, while the other system with cubic phase mask can be increased to 10 times, the consistency of MTF decrease obviously, the maximum operating temperature of optimized system range between -40℃-60℃. Results show that this optimization method can be more convenient to define some unconventional optimization goals and fleetly to optimize optical system with special properties due to its externally compiled function and DDE, there will be greater significance for the optimization of unconventional optical system.
This paper proposes a new salient region detection algorithm to detect and recognize ship on the sea in a shaky field of
view. Based on this situation, the approach this paper adopts to solve the problem is detecting the salient region of each
frame separately instead of using tracking algorithm and the salient region detection is based on local and global
contrast. The result shows that the interference can be restrained and the shape of target can be detected correctly. It
proves that the algorithm in this paper is a highly efficient target detection algorithm for ship detection.
Fiber faceplate modulation was applied to read out the precise actuation of silicon-based, surface micro-fabricated cantilever mirrors array in optical imaging system. The faceplate was made by ordered bundles consisting of as many as ten thousands fibers. The transmission loss of an individual fiber in the bundles was 0.35dB/cm and the cross talk between neighboring fibers in the faceplate was about 15%. Micro-cantilever mirrors array (Focal-Plane Array (FPA)) which composed of two-level bi-material pixels, absorb incident infrared flux and result in a temperature increase. The temperature distribution of incident flux transformed to the deformation distribution in FPA which has a very big difference in coefficients of thermal expansion. FPA plays the roles of target sensing and has the characteristics of high detection sensitivity. Instead of general filter such as knife edge or pinhole, fiber faceplate modulate the beam reflected by the units of FPA. An optical readout signal brings a visible spectrum into pattern recognition system, yielding a visible image on monitor. Thermal images at room temperature have been obtained. The proposed method permits optical axis compact and image noise suppression.
In the research of optical synthetic aperture imaging system, phase congruency is the main problem and it is necessary to detect sub-aperture phase. The edge of the sub-aperture system is more complex than that in the traditional optical imaging system. And with the existence of steep slope for large-aperture optical component, interference fringe may be quite dense when interference imaging. Deep phase gradient may cause a loss of phase information. Therefore, it’s urgent to search for an efficient edge detection method. Wavelet analysis as a powerful tool is widely used in the fields of image processing. Based on its properties of multi-scale transform, edge region is detected with high precision in small scale. Longing with the increase of scale, noise is reduced in contrary. So it has a certain suppression effect on noise. Otherwise, adaptive threshold method which sets different thresholds in various regions can detect edge points from noise. Firstly, fringe pattern is obtained and cubic b-spline wavelet is adopted as the smoothing function. After the multi-scale wavelet decomposition of the whole image, we figure out the local modulus maxima in gradient directions. However, it also contains noise, and thus adaptive threshold method is used to select the modulus maxima. The point which greater than threshold value is boundary point. Finally, we use corrosion and expansion deal with the resulting image to get the consecutive boundary of image.
Multiple synthetic aperture imaging can enlarge pupil diameter of optical systems, and increase system resolution. Multiple synthetic aperture imaging is a cutting-edge topic and research focus in recent years, which is prospectively widely applied in fields like astronomical observations and aerospace remote sensing. In order to achieve good imaging quality, synthetic aperture imaging system requires phase extraction of each sub-aperture and co-phasing of whole aperture. In the project, an in-depth study about basic principles and methods of segments phase extraction was done. The study includes: application of sinusoidal extreme strip light irradiation phase shift method to extract the central dividing line to get segment phase extraction information, and the use of interference measurement to get the aperture phase extraction calibration coefficients of spherical surface. Study about influence of sinusoidal extreme strip phase shift on phase extraction, and based on sinusoidal stripe phase shift from multiple linear light sources of the illumination reflected image, to carry out the phase shift error for inhibiting the effect in the phase extracted frame.
Endoscopic imaging quality affects industrial safety and medical security. Rigid endoscope distortion is of great signification as one of optical parameters to evaluate the imaging quality. This paper introduces a new method of rigid endoscope distortion measurement, which is different from the common methods with low accuracy and fussy operation. It contains a Liquid Crystal Display (LCD) to display the target, a CCD to obtain the images with distortion, and a computer to process the images. The LCD is employed instead of common white screen. The autonomous control system of LCD makes it showing the test target designed for distortion, and its parameter is known. LCD control system can change the test target to satisfy the different demand for accuracy, which avoids replacing target frequently. The test system also contains a CCD to acquire images in the exit pupil position of rigid endoscope. Rigid endoscope distortion is regarded as centrosymmetric, and the MATLAB software automatically measures it by processing the images from CCD. The MATLAB software compares target images with that without distortion on LCD and calculates the results. Relative distortion is obtained at different field of view (FOV) radius. The computer plots the curve of relative distortion, abscissa means radius of FOV, ordinate means relative distortion. The industry standard shows that, the distortion at 70% field of view is pointed on the curve, which can be taken as an evaluation standard. This new measuring method achieves advantages of high precision, high degree of intelligence, excellent repeatability and gets calculation results quickly.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.