Background Illumination Research Articles

Real-Time Rain Detection and Wiper Control Employing Embedded Deep Learning

A state-of-the-art real-time rain detection and wiper control method is proposed in this article. Currently, commercial models adopt electronic sensors that can only sample the humidity of a small region of the windshield. The existing computer vision methods primarily focus on the detection and counting of raindrops and provide a recall rate of less than 70%. Here we adopted a holistic-view deep learning approach to build a visual classifier that is robust to large varieties of background scenes, illumination, and water forms. Specifically, Deep Residual Network (ResNet) was adopted as the visual classifier that distinguishes between rainy and fair street scenes and controls the wipers accordingly. To verify the practicality of the proposed deep learning framework, we tested the network on various embedded computing systems, including an embedded computing cluster. The results show that the deep learning rain detector outperforms previous state-of-the-art methods with higher rain recall and precision. It was also found that with the help of some graphic computation-enhancing components, commonly available embedded systems in the market can provide comparable performance to personal computers. While using the enhanced embedded systems to build a cluster, a performance superior to PC was witnessed. As the embedded system is cost-effective, small, and lighter, normalized performances for various aspects clearly show the competitive edge of the embedded systems and confirm the practicality of the proposed system. We also release the dataset of 160 k images used for training the visual rain detector.

Comparison of Absorbed and Intercepted Fractions of PAR for Individual Trees Based on Radiative Transfer Model Simulations

The fraction of absorbed photosynthetically active radiation (fAPAR) is a key parameter for estimating the gross primary production (GPP) of trees. For continuous, dense forest canopies, fAPAR, is often equated with the intercepted fraction, fIPAR. This assumption is not valid for individual trees in urban environments or parkland settings where the canopy is sparse and there are well-defined tree crown boundaries. Here, the distinction between fAPAR and fIPAR can be strongly influenced by the background and large illumination variations due to multi-scattering and shadows of buildings. This study investigates the radiative budget of PAR bands using a coupled leaf-canopy radiative transfer model (PROSPECT-DART), considering a suite of tropical tree species over a wide range of assumed leaf chlorophyll contents. The analyses simulate hyperspectral images (5 nm bandwidth) of individual tree crowns for the selected background (concrete vs. grass) and illumination conditions. We then use an artificial neural network-based method to partition sunlit vs. shaded leaves within each crown, as the latter have lower fAPAR and fIPAR values. Our results show fAPAR of sunlit leaves decreases with the ratio of diffuse to direct scene irradiance (SKYL), while SKYL has minimal influence for shaded leaves. Both fAPAR and fIPAR decrease at more oblique solar zenith angles (SZA). Higher values of fAPAR and fIPAR occur with concrete backgrounds and the influence of the background is larger at higher diffuse ratio and solar zenith angles. The results show that fIPAR is typically 6–9% higher than fAPAR, and up to 14% higher for sunlit leaves with a concrete background at SKYL = 0. The differences between the fIPAR and fAPAR also depend on the health condition of the leaves, such as chlorophyll content. This study can improve the understanding of urban individual trees fAPAR/fIPAR and facilitate the development of protocols for fAPAR field measurements.

The Search of Diffusive Properties in Ambient Seismic Noise

ABSTRACTAmbient seismic noise (ASN) is becoming of interest for geophysical exploration and engineering seismology, because it is possible to exploit its potential for imaging. Theory asserts that the Green’s function can be retrieved from correlations within a diffuse field. Surface waves are the most conspicuous part of Green’s function in layered media. Thus, the velocities of surface waves can be obtained from ASN if the wavefield is diffuse. There is widespread interest in the conditions of emergence and properties of diffuse fields. In the applications, useful approximations of the Green’s function can be obtained from cross correlations of recorded motions of ASN. An elastic field is diffuse if the background illumination is azimuthally uniform and equipartitioned. It happens with the coda waves in earthquakes and has been verified in carefully planned experiments. For one of these data sets, the 1999 Chilpancingo (Mexico) experiment, there are some records of earthquake pre-events that undoubtedly are composed of ASN, so that the processing for coda can be tested on them. We decompose the ASN energies and study their equilibration. The scheme is inspired by the original experiment and uses the ASN recorded in an L-shaped array that allows the computation of spatial derivatives. It requires care in establishing the appropriate ranges for measuring parameters. In this search for robust indicators of diffusivity, we are led to establish that under certain circumstances, the S and P energy equilibration is a process that anticipates the diffusion regime (not necessarily isotropy), which justifies the use of horizontal-to-vertical spectral ratio in the context of diffuse-field theory.

ABOships—An Inshore and Offshore Maritime Vessel Detection Dataset with Precise Annotations

Availability of domain-specific datasets is an essential problem in object detection. Datasets of inshore and offshore maritime vessels are no exception, with a limited number of studies addressing maritime vessel detection on such datasets. For that reason, we collected a dataset consisting of images of maritime vessels taking into account different factors: background variation, atmospheric conditions, illumination, visible proportion, occlusion and scale variation. Vessel instances (including nine types of vessels), seamarks and miscellaneous floaters were precisely annotated: we employed a first round of labelling and we subsequently used the CSRT tracker to trace inconsistencies and relabel inadequate label instances. Moreover, we evaluated the out-of-the-box performance of four prevalent object detection algorithms (Faster R-CNN, R-FCN, SSD and EfficientDet). The algorithms were previously trained on the Microsoft COCO dataset. We compared their accuracy based on feature extractor and object size. Our experiments showed that Faster R-CNN with Inception-Resnet v2 outperforms the other algorithms, except in the large object category where EfficientDet surpasses the latter.

Evaluation of accurate iris center and eye corner localization method in a facial image for gaze estimation

Accurate estimation of eye-related information is important for many applications such as gaze estimation, face alignment, driver drowsiness detection, etc. Earlier works fail to estimate eye information in low-resolution images captured by a regular camera or webcam. This paper is aimed at developing an Iris Center (IC) and Eye Corner (EC) localization method in low-resolution facial images with an application of gaze estimation. A three-stage method is proposed for IC and EC localization. In the first stage, a circular gradient-intensity-based operator is proposed for rough ICs estimation and a CNN model is used in the second stage to find true ICs. In the third stage, Explicit Shape Regression (ESR) method is used for EC localization where initialization is done taking the ICs as a reference point to the mean eye contour shape model. The proposed IC localization method is evaluated on BioID and Gi4E database and it shows better accuracy compare to some of the state-of-the-art methods. This method further evaluated for gaze estimation based on IC and EC which does not require any prior calibrations unlike earlier infrared illumination-based gaze trackers. Here, the experiment for gaze estimation is performed in our proposed NITSGoP database that prepared under indoor conditions with complex background and uneven illuminations. The experimental results suggest that the proposed method can be used for gaze estimation with better accuracy both in still images and videos.

Development of smartphone-based imaging techniques for the estimation of chlorophyll content in lettuce leaves

Leaf color is a good indicator of plant’s health status. In this study, a new image acquisition technique was developed to estimate chlorophyll content of lettuce leaves. The images of lettuce leaves grown under artificial light were acquired using a smartphone. Leaves images was captured by directly attached the leaves to the camera lens with the aid of background illumination from SMD LED. Red, green, blue (RGB) color indices were extracted from leaves color images and some vegetation indices were also calculated. Then, the correlation between these indices and chlorophyll content obtained from SPAD502 chlorophyll meter were evaluated. Significant correlation was found between all the image indices and chlorophyll content with the R2 ranging from 0.63 to 0.85 except for G and B indices from RGB component. Highly significant correlation was found between vegetation indices (VI) and chlorophyll content (R2 = 0.85) with the lowest root mean square error (RMSE) of 8.07 g of chlorophyll/100 g fresh tissue. This demonstrated that the chlorophyll content of lettuce leaves can be successfully estimated using regular smartphone with added background light illumination from SMD LED.

Implementation and Evaluation of Novel Architecture Using Optical Wireless for WLAN Control Plane

We propose a novel wireless control system architecture that divides the wireless service area into smaller optical cells and centrally controls the user equipment (UE) connections under each optical cell. The proposal transmits via IoT smart lighting an optical identifier (ID) specifying connection information to the UE with illuminance sensor. The received optical ID indicates the optimum connection destination. Two solutions are proposed to overcome the hardware restrictions faced by optical ID transmission/reception. Oversampled edge-excluded receiving scheme (OE) reduces the error rate; the signal is oversampled and each window subjected to majority decision. Optical ID reception with lower error rate than conventional approach is realized with a minimum received illuminance of just 8.5 lx (background illuminance 226.67 lx) and a modulation factor of 7 %. Fast optical ID authentication (FA) reduces the authentication cycle preceding optical ID reception by performing correlation calculation between the cyclic matrix and an optical ID list. FA shortens the authentication cycle by 62.5 % (1200 ms to 450 ms). Furthermore, an optical cell control algorithm (ScanLine-based/MinDist-based) is proposed to offset the deterioration in network quality created by non-uniform UE distribution. ScanLine-based algorithm controls optical cells in scanning line manner, while MinDist-based favors optical cells with the minimum distance from the access point (AP). It is confirmed that with an optical cell radius of 5 m or less, the capacity difference per user is reduced compared to the existing RSSI-based alternative while reducing the deterioration of total capacity in extremely biased user distribution scenarios.

Non-Contact Heart Rate Estimation via Adaptive RGB/NIR Signal Fusion.

We propose a non-contact heart rate (HR) estimation method that is robust to various situations, such as bright, low-light, and varying illumination scenes. We utilize a camera that records red, green, and blue (RGB) and near-infrared (NIR) information to capture the subtle skin color changes induced by the cardiac pulse of a person. The key novelty of our method is the adaptive fusion of RGB and NIR signals for HR estimation based on the analysis of background illumination variations. RGB signals are suitable indicators for HR estimation in bright scenes. Conversely, NIR signals are more reliable than RGB signals in scenes with more complex illumination, as they can be captured independently of the changes in background illumination. By measuring the correlations between the lights reflected from the background and facial regions, we adaptively utilize RGB and NIR observations for HR estimation. The experiments demonstrate the effectiveness of the proposed method.

Models of Direct Time-of-Flight Sensor Precision That Enable Optimal Design and Dynamic Configuration

Direct time-of-flight (dToF) sensors that measure depth by pulsing a laser and timing the photon return are used in many applications, including consumer electronics for proximity sensing and depth map generation. A histogram of photon return times is measured and then processed to estimate object depth. By collecting many photons that span multiple bins of the histogram the final depth estimate interpolates between time-to-digital converter (TDC) bins to produce a result that is more precise than the converter resolution. The precision of this interpolation depends on the temporal spread of the measurement, the resolution of the TDC, and the number of signal and background photons measured. There is a need for dToF depth precision models to guide design and predict and tune performance during use. In this article, we present models that estimate sensor depth precision versus dToF design parameters and photons measured. We use Monte Carlo simulations and experimental measurements to prove the accuracy of the models. With proven models in hand, we investigate a dToF sensor design by first presenting the dependence of precision upon the TDC resolution and the signal-to-noise ratio. Second, we experimentally measure the depth precision versus the intensity of background illumination. The models closely match the measurements of background susceptibility and locate a transition point of background intensity below which precision is constant and above which the precision continuously degrades. Finally, experimental measurements demonstrate how the modeling enables dynamic tuning: from a single histogram we estimate precision, thus enabling sensor exposure time tuning for a target precision or prediction of the precision given a change in object distance or background illumination. This work presents straightforward models verified by simulation and measurement. These models guide dToF design and enable dynamic adjustments that benefit power-constrained usage scenarios.

Plant Leaf Segmentation and Phenotypic Analysis Based on Fully Convolutional Neural Network

HighlightsModify the U-Net segmentation network to reduce the loss of segmentation accuracy.Reducing the number of layers U-net network, modifying the loss function, and the increase in the output layer dropout.It can be well extracted after splitting blade morphological model and color feature.Abstract. From the perspective of computer vision, the shortcut to extract phenotypic information from a single crop in the field is image segmentation. Plant segmentation is affected by the background environment and illumination. Using deep learning technology to combine depth maps with multi-view images can achieve high-throughput image processing. This article proposes an improved U-Net segmentation network, based on small sample data enhancement, and reconstructs the U-Net model by optimizing the model framework, activation function and loss function. It is used to realize automatic segmentation of plant leaf images and extract relevant feature parameters. Experimental results show that the improved model can provide reliable segmentation results under different leaf sizes, different lighting conditions, different backgrounds, and different plant leaves. The pixel-by-pixel segmentation accuracy reaches 0.94. Compared with traditional methods, this network achieves robust and high-throughput image segmentation. This method is expected to provide key technical support and practical tools for top-view image processing, Unmanned Aerial Vehicle phenotype extraction, and phenotype field platforms. Keywords: Deep learning, Full convolution neural network, Image segmentation, Phenotype analysis, U-Net.

Evolutionary Rear-Lamp Tracking at Nighttime

Rear-lamp tracking at nighttime plays a momentous role in the advanced driver assistance system (ADAS), involving collision mitigation, automatic cruise control, automatic headlamp dimming, etc. Most of the existing tracking methods based on monocular camera leverage on color features. However, such tracking methods can be easily influenced by background clutter, illumination change, distance variation, and occlusion. In this paper, we propose an evolutionary adaptive rear-lamp tracking method at nighttime, in which a novel genetic algorithm powered by the probabilistic bitwise operation (PBO) is utilized. Also, to improve the robustness against various environments, a balanced fitness function is proposed by taking color information, symmetry, spatial relationship, and rigidity into account. Especially, a series of adaptive thresholds based on rear data in HSV color space is proposed to exploit color information reasonably with respect to our task. A strategy to deal with occlusion is also proposed, which relies on color information and rigidity. Moreover, to our knowledge, there is no publicly available dataset for rear-lamp tracking at nighttime. To fill the gap between the real-world application and the theoretical research, we create a novel dataset, which contains diverse traffic conditions at nighttime. The experimental results indicate that our method outperforms comparative online tracking methods in terms of success rate and center location error.

Noncontact Heart Rate Measurement Using a Webcam, Based on Joint Blind Source Separation and a Skin Reflection Model: For a Wide Range of Imaging Conditions

Remote photoplethysmography (rPPG) can be used for noncontact and continuous measurement of the heart rate (HR). Currently, the main factors affecting the accuracy and robustness of rPPG‐based HR measurement methods are the subject’s skin tone, body movement, exercise recovery, and variable or inadequate illumination. In response to these challenges, this study is aimed at investigating a rPPG‐based HR measurement method that is effective under a wide range of conditions by only using a webcam. We propose a new approach, which combines joint blind source separation (JBSS) and a projection process based on a skin reflection model, so as to eliminate the interference of background illumination and enhance the extraction of pulse rate information. Three datasets derived from subjects with different skin tones considering six environmental scenarios are used to validate the proposed method against three other state‐of‐the‐art methods. The results show that the proposed method can provide more accurate and robust HR measurement for all three datasets and is therefore more applicable to a wide range of scenarios.

Two-stage classifier for face recognition using HOG features

Facerecognition ismainlyused in biometric applications, surveillance systems and computer vision.Themajorproblem while recognizingfacesarisesdue topose variations, background illumination invariants and facial expressions.Theproposed two-stage classifier aims to recognize given inputfaceimages usingk-NNand SVM. The use of two separate classifiers, one after the other, improves the system recognition accuracy.The pre processed face images are then utilized toextractprominent features by Histogram of Oriented Gradients (HOG).In the initial stage, k-NN classifier was used and afterwards, unrecognized face images were tested with SVM classifier.It is found that recognition accuracy increased by cascading k-NN followed bySVM.The two-stage classifierachieved a recognition accuracy of 95.2%.

Robust 3D Reconstruction of Dynamic Scenes From Single-Photon Lidar Using Beta-Divergences.

In this article, we present a new algorithm for fast, online 3D reconstruction of dynamic scenes using times of arrival of photons recorded by single-photon detector arrays. One of the main challenges in 3D imaging using single-photon lidar in practical applications is the presence of strong ambient illumination which corrupts the data and can jeopardize the detection of peaks/surface in the signals. This background noise not only complicates the observation model classically used for 3D reconstruction but also the estimation procedure which requires iterative methods. In this work, we consider a new similarity measure for robust depth estimation, which allows us to use a simple observation model and a non-iterative estimation procedure while being robust to mis-specification of the background illumination model. This choice leads to a computationally attractive depth estimation procedure without significant degradation of the reconstruction performance. This new depth estimation procedure is coupled with a spatio-temporal model to capture the natural correlation between neighboring pixels and successive frames for dynamic scene analysis. The resulting online inference process is scalable and well suited for parallel implementation. The benefits of the proposed method are demonstrated through a series of experiments conducted with simulated and real single-photon lidar videos, allowing the analysis of dynamic scenes at 325 m observed under extreme ambient illumination conditions.

Reducing Photic Phenomena and Retinal Background Illumination by Using an Intraocular Lens

Background. After implantation of monofocal intraocular lenses (IOLs), the risk of developing light phenomena is 9%, and after implantation of multifocal IOLs the one is 41%. These effects not only cause discomfort, but also poor image quality. The existing IOLs have a number of deficiencies that cause different types of photic phenomena.Objective. The aim of the work is optical calculation and simulation of the parameters that an IOL should have in order to reduce photic phenomena and retinal background illumination while increasing the transmission contrast. We also are aimed to design a new IOL on the basis of the results obtained.Methods. To calculate the reflected rays, we used the Snell–Descartes law, according to which we obtained the value of the critical angle for a hydrophobic acrylic IOL with the refractive index of 1.55 at the wavelength of λ = 0.55 μm. This is consistent with aqueous humor, the refractive index of which is 1.33. The calculation of the transmitted light loss was determined by Fresnel reflection coefficient. We handled with optical system in which the aperture diaphragm fulfilled the role of a pupil. Using the Zemax 13 software environment, we simulated ray path for a human eye, the sagittal axis of which is 28 mm at different height of IOL placement. We applied the results obtained to design a new IOL in the SolidWorks environment.Results. The calculations made it possible to identify the shortcomings of modern IOLs and methods for their elimination. It was found that in order to reduce the risk of photic phenomena and, as a result, of increasing luminous transmission, an IOL should be placed at a distance of at least 4 mm from the iris. It should contain two or more optical layers, the refractive index of which changes towards the center of the lens, and have the surface roughness of 35 nm. Based on the calculations, we carried out simulation in the Zemax 13 environment, which confirmed their veracity. When simulated with these parameters, the standard deviation of an image fell completely within the Airy disk, which has a size of 3.598 μm with an image size of 2.972 μm. Thus, the optical system is considered diffraction limited and no further optical optimizations are possible. Using the Solidworks software and the results obtained, we proposed the proprietary IOL model called "NVision OP". This IOL has an optical part thickness of 1 mm with a diameter of 5 mm. In general, the hollow, volume-replacing IOL with a coating has a diameter of 10 mm and a thickness of 5 mm, the thickness of the coating is 150 μm.Conclusions. The study revealed a number of factors that require improvement and elimination to prevent the occurrence of various types of photic effects. These include: lens surface roughness, IOL refractive power, shape, lens edge thickness, depth of IOL implantation into an eye, phacodonesis and lens displacement, aperture diaphragm diameter. After data optimization, according to the calculated results, we carried out the simulation in the Zemax 13 and Solidworks environments. On the basis of this simulation we proposed the model of an intraocular lens "NVision OP"; the photic effects namely arcs, flare, flashes, glare and halo are eliminated as much as possible. The hollow, volume-replacing IOL "NVision OP" has elements on its coat that allow to use the suture fixation, which prevents the dislocation of the IOL. For the implantation of the proposed IOL "NVision OP", it is recommended to use a viscoelastic and the Alcon injector with the cartridge B. Due to the fact that the shape of the IOL corresponds to the native human lens, the lens is located in the place of the phacoemulsified substance, and the implantation does not take much time.

Comprehensive analysis of a diverse group of features and development of vision-based two-level hand detector under practical environment conditions

Developing a bare-hand detection system for practical environment conditions is a complex and challenging task. Factors such as change in appearance, uneven illumination, and complex background add up to the difficulty in detecting the target hand. Present study newly explored 13 colour-texture and integrates them with texture models to develop robust two-level hand detector under practical conditions mentioned above. Colour-texture and texture models are assessed using multiple classification tools and employed in two subsequent levels such that the second level only classifies the optimal sub-windows classified in the first level. The analysis showed that the proposed two-level detection system detects the hand with 53.4% higher accuracy than the baseline model which the integrated motion detection and skin filtering method, under the practical conditions. With five times lower time-complexity than the baseline model, the proposed system can be used to detect hand in both static as well as dynamic gesture systems.

Limited Spatial Spread Explains the Dependence of Visual Response Adaptation on Stimulus Size in Rat Superior Colliculus Neurons

Although adaptation to light occurs in the eye and mainly preserves the full dynamic range of neuronal responses during changing background illumination, it affects the entire visual system and helps to optimize visual information processing. We have shown recently that in rat superior colliculus (SC) neurons adaptation to light acts as a local low-pass filter because, in contrast to the primate SC, in rat collicular neurons adaptation to small stimuli is largely limited to the vicinity of the adaptor stimulus. However, it was unclear whether large visual stimuli would induce the same spatially limited adaptation. We addressed this question by evaluating the effects of 1.8°, 6.2° and 20.8° wide adaptor stimuli on test stimuli of variable size. Single unit recordings in the adult rat SC were employed to estimate the response amplitude. Small, 1.8° and 6.2° adaptors habituated visual responses only to stimuli smaller than the adaptive stimuli. However, the 20.8° adaptor dramatically reduced responses even to test stimuli >3 times wider than the adaptor (up to 70° wide). The latter result may be explained by a nearly complete occlusion by a large adaptor of the neuron’s receptive field (RF). All these results are consistent with the idea of a limited spatial spread of adaptation in rat SC neurons that is the consequence of high convergence of retinal inputs, in which small RFs limit the spatial spread of adaptation. It is concluded that, in this limited spatial spread of adaptation, rodent SC resembles higher visual system areas in primates and indicates potential differences in visual information processing between rodents and primates.

A high performance integrated readout circuit for wavefront sensors

This work presents an integrated pixel topology that promises to offer superior performance in a Hartmann–Shack Wavefront Sensor (WFS) with an orthogonal array of Quad Cells serving as Position-Sensitive Detectors. The readout integrated circuit for each photodiode is fully compliant to any standard CMOS microelectronics technology and is advantageously tolerant to high background illumination levels whereas maintaining both high linearity and high sensitivity. To assess the operation of this pixel on the focal-plane array of the WFS, we developed a computational platform encompassing a full detection chain comprising wavefront sampling, photodetection, electronic circuitry and wavefront reconstruction. It couples an algorithm written in C to SPICE (Simulated Program with Integrated Circuits Emphasis). The platform is technology agnostic and flexible, enabling easy modification to represent different detection or reconstruction methods. The results obtained with the proposed pixel have been compared to those obtained with a conventional pixel in CMOS image sensor.

Energy-Efficient Image Recognition System for Marine Life

This article focuses on designing an energy-efficient image recognition system for marine monitoring. One of the main challenges of an underwater imaging system is the strict power consumption constraints due to the limited on-site resources. Considering the need for continuous operation in different water turbidity levels and background illumination conditions, an energy-efficient approach is needed for the effective utilization of the resources. In this work, we propose a recognition framework that will adaptively adjust the system parameters, such as camera frame rate and LED illumination level, based on the environmental conditions to optimize the energy consumption while ensuring a high recognition accuracy. The first part of the proposed decision system contains the convolutional neural network (CNN)-based animal recognition block which is used for obtaining the confidence level for a single frame. The second part is the adaptive decision block that dynamically changes the system parameters and combines the results of the recognition block for multiple frames based on the environmental conditions. In our experiments, we have used nearly 8000 underwater images for training and testing the single frame recognition block and used nearly 200 different video sequences for training and testing the adaptive decision block. Based on measurements of a hardware framework composed of a Raspberry Pi 3 Model B, a Pi NoIR Camera v2.1, and 850 nm LEDs, the proposed system achieves up to 92.7% energy savings with a comparable recognition performance by dynamically changing the frame rate and emitted light intensity based on water turbidity and background illumination level.

Fast Visual Tracking Using Spatial Temporal Background Context Learning

Visual Tracking by now has gained much provenience among researchers in recent years due to its vast variety of applications that occur in daily life. Various applications of visual tracking include counting of cars on a high way, analyzing the crowd intensity in a concert or a football ground or a surveillance camera tracking a single person to track its movements. Various techniques have been proposed and implemented in this research domain where researchers have analyzed various parameters. Still this area has a lot to offer. There are two common approaches that are currently deployed in visual tracking. One is discriminative tracking and the other one is generative tracking. Discriminative tracking requires a pre-trained model that requires the learning of the data and solves the object recognition as a binary classification problem. On the other hand, generative model in tracking makes use of the previous states so that next state can be predicted. In this paper, a novel tacking based on generative tracking method is proposed called as Illumination Inavariant Spatio Temporal Tracker (IISTC). The proposed technique takes into account of the nearby surrounding regions and performs context learning so that the state of the object under consideration and its surrounding regions can be estimated in the next frame. The learning model is deployed both in the spatial domain as well as the temporal domain. Spatial domain part of the tracker takes into consideration the nearby pixels in a frame while the temporal model takes account of the possible change of object location. The proposed tracker was tested on a set of 50 images against other state of the art four trackers. Experimental results reveal that our proposed tracker performs reasonably well as compared with other trackers. The proposed visual tracker is both efficiently with respect to computation power as well as accuracy. The proposed tracker takes only 4 fast Fourier transform computations thus making it reasonably faster. The proposed trackers perform exceptionally well when there is a sudden change in back ground illumination.