261 research outputs found
New developments of the sequential probability ratio test control chart
Get PDFThis thesis aims to enrich the literature on the SPRT control chart following two
major contributions made in the late 1990s and early 2010s. The SPRT chart has
been chosen as the central premise of this thesis due to its strong detection
performance as well as high sampling efficiency. Two major research gaps have been
identified. The first gap is the lack of documentation on the SPRT chart with
estimated process parameters and its statistical design. The second gap is the lack
of a well-motivated SPRT chart for joint monitoring of the mean and dispersion of
a process. To fill the literature gaps, we formulate the theoretical framework for
the SPRT chart with estimated process parameters, as well as develop a new SPRT
chart for joint monitoring of the mean and dispersion. Optimisation designs based
on various industrial objectives have been developed in this thesis. Real industrial
examples involving a variety of destructive and non-destructive tests are also
presented in this thesis to illustrate the implementation of the proposed SPRT
charts. The thesis should serve as a reference to researchers working in the field of
statistical quality control, as well as practitioners seeking to improve the
performance of their processes
Robust object detection and tracking for real-time application
No full textThe report is mainly a detailed documentation of the development of the mean shift method application in Kernel based object tracking. The author will study in depth the concepts of the object detection and object tracking and the implementation of the object tracking and object detection algorithm. The fundamental concepts and background of the mean shift method in object tracking also will be studied in depth. Then the computation of mean shift method in object tracking will be implemented in Matlab. The implementation will started with the basic computation of mean shift tracking method. Next, the author will modify the basic mean shift tracking method. The orientation based approach will be introduced. The tracking results will be discussed and compared. We will obtain the results that below our expectation with the average dropping of 33.75%. The further improvement and modification will be done to get the better tracking results. The modification is done by the implementation of four boxes tracking window algorithm. The improvement of 42.33% of the average performance is obtained after the modification. . In order to better improve the tracking results, the implementation of fusion system is introduced in the last session of the report. The fusion system consists of object tracking algorithm and object detection algorithm. The average tracking results is further improved by 43.69% by fusion system. Finally the recommendation for further research of the project is discussed.Bachelor of Engineerin
Optimal Designs Of The Double Sampling X Chart Based On Parameter Estimation
Get PDFControl charts, viewed as the most powerful and simplest tool in Statistical Process
Control (SPC), are widely used in manufacturing and service industries. The double
sampling (DS) X chart detects small to moderate process mean shifts effectively,
while reduces the sample size. The conventional application of the DS X chart is
usually investigated assuming that the process parameters are known. Nevertheless,
the process parameters are usually unknown in practical applications; thus, they are
estimated from an in-control Phase-I dataset. In this thesis, the effects of parameter
estimation on the DS X chart’s performance are examined. By taking into
consideration of the parameter estimation, the run length properties of the DS X
chart are derived. Since the shape and the skewness of the run length distribution
change with the magnitude of the process mean shift, the number of Phase-I samples
and sample size, the widely applicable performance measure, i.e. the average run
length (ARL) should not be used as a sole measure of a chart’s performance. For this
reason, the ARL, the standard deviation of the run length (SDRL), the median run
length (MRL), the percentiles of the run length distributions and the average sample
size (ASS) are recommended to effectively evaluate the proposed DS X chart with
estimated parameters
Automated human age estimation based on face images
No full textHuman age as an important personal trait can be applied in a variety of settings such as biometric airport security checks or access to product such as alcohol or tobacco in a shop. However, can computers perform age recognition function like what human being did? In this project, the author had developed an age estimation based on facial images system through wrinkles emerging from the facial appearance due to biologic aging. These facial images can either be extract from live webcam or existing digital photo images. The author has made use of local Successive Mean Quantization Transforms (SMQT) to extract feature for face detection follow by Sparse Network of Winnows (SNOW) classifier for face prediction. Upon the detection of face, the image is crop and thereafter spatially localized spectral features will be extracted using Gabor filter. Subsequently, these extracted spectral features are transformed into corresponding Eigen faces using Principle Component Analysis (PCA). This technique allows dimension reduction output in high compression rate for faster estimation. Lastly, results are classify into 4 age groups consisting of “Child”, “Teen”, “Adult” and “Senior adult” with Extreme Learning Machine (ELM) classifier that perform good generalization performance at tremendously fast learning rate.Bachelor of Engineerin
Stereo object detection with the applications to mobile robot
No full textRobust and accurate object detection are needed for the applications to mobile robots.
Unfortunately, most of the existing object detection approaches cannot satisfy with the real
application due to either slow speed or lower accuracy. Computer Vision has been
increasingly important in enabling smart technologies. In this project, the author aims to
develop an object detection system which can provide high accuracy while reducing missing
detection by using stereo camera and state-of-the-art machine learning. Stereo vision is one of
the options that can be implemented in order to provide more data and parameters besides
visuals provided by a single camera. This will involve the use of the disparity map in order to
obtain depth information of the scene. By combining the use of depth information with a deep
learning framework, fast, robust and accurate stereo object detection can be achieved.Bachelor of Engineerin
The One-Sided Variable Sampling Interval Exponentially Weighted Moving Average x̄ Charts Under the Gamma Distribution
No full textRecently, adaptive quality control charts have been frequently utilised in diverse production and manufacturing industries to ensure process stability and maintain a desirable level of product quality. Among these charts, the variable sampling interval (VSI) exponentially weighted moving average (EWMA) x̄ chart is known for its sensitivity and efficiency in monitoring the process mean shifts. However, the existing literature on the design of the VSI EWMA x̄ chart is predicated on the presumption that the underlying process adheres to a normal distribution. This normality assumption is often violated in manufacturing settings, where many practical processes tend to follow non-normal or skewed distributions. Therefore, this paper investigates the performance of one-sided VSI EWMA x̄ charts designed under the normal distribution model, when the quality characteristics of interest follow a gamma distribution. Our findings indicate that the in-control average time to signal and the standard deviation of the time to signal for the one-sided VSI EWMA x̄ charts are significantly deteriorated under the gamma distribution. To tackle this problem, this paper proposes new charting parameters specifically derived for the one-sided VSI EWMA x̄ charts under the gamma distribution. Besides, comparative analyses show that the proposed one-sided VSI EWMA x̄ charts exhibit the best detection speed compared to the one-sided Shewhart x̄ and EWMA x̄ charts, when the process follows a gamma distribution. An illustrative application of the one-sided VSI EWMA x̄ chart for monitoring the weight of bias tires in scooter manufacturing is provided at the end of this paper
Redesigning the Omnibus SPRT Control Chart for Simultaneous Monitoring of the Mean and Dispersion of Weibull Processes
Get PDFQuality control charts play an important role in distinguishing between abnormal variations and normal variations of a manufacturing process. Generally, unusual variations in a process may arise due to a change in its mean or dispersion, or a simultaneous change in both parameters. In recent literature, the omnibus sequential probability ratio test (OSPRT) control chart has been proven effective for detecting joint shifts in both the process mean and variability. However, one limitation of the proposed scheme lies in its absolute dependence on the validity of the normality assumption, which may not apply to many quality data, such as machine failure times, the strength of plant fibres, etc. In this research, we critically analyze the performances of the OSPRT chart designed for the Normal distribution, in the case where quality data follow the well-known Weibull distribution. Our findings reveal that the in-control average run length and standard deviation of the run length of the OSPRT chart are significantly compromised due to the positive skewness of the Weibull distribution. As a means of tackling the problem, the skewness correction design has been proposed to correct the control limits of the OSPRT chart. The corrected OSPRT chart is found to produce a more satisfactory in-control performance, with an acceptable decline in its sensitivity towards small process shift sizes
Redesigning the Omnibus SPRT Control Chart for Simultaneous Monitoring of the Mean and Dispersion of Weibull Processes
Get PDFQuality control charts play an important role in distinguishing between abnormal variations and normal variations of a manufacturing process. Generally, unusual variations in a process may arise due to a change in its mean or dispersion, or a simultaneous change in both parameters. In recent literature, the omnibus sequential probability ratio test (OSPRT) control chart has been proven effective for detecting joint shifts in both the process mean and variability. However, one limitation of the proposed scheme lies in its absolute dependence on the validity of the normality assumption, which may not apply to many quality data, such as machine failure times, the strength of plant fibres, etc. In this research, we critically analyze the performances of the OSPRT chart designed for the Normal distribution, in the case where quality data follow the well-known Weibull distribution. Our findings reveal that the in-control average run length and standard deviation of the run length of the OSPRT chart are significantly compromised due to the positive skewness of the Weibull distribution. As a means of tackling the problem, the skewness correction design has been proposed to correct the control limits of the OSPRT chart. The corrected OSPRT chart is found to produce a more satisfactory in-control performance, with an acceptable decline in its sensitivity towards small process shift sizes
A study on the run sum X-bar control chart with unknown parameters
No full textIt is well known that the run sum control chart is a simple and powerful statistical process control tool in the monitoring of the process mean. The implementation of the run sum chart is generally based on the assumption that the process parameters are known. However, since the process parameters are usually unknown in practice, they are estimated from an in-control Phase I data set. In this paper, by means of the Markov chain approach, we investigate the effects of parameter estimation on the performance of the run sum X̄ chart with the scores 0, 1, 2 and 4. The results reveal that when the size of the shift and the number of samples from the Phase I process used for the estimation of parameters are both small, the performance of the run sum X̄ chart is significantly deteriorated. Moreover, very large sample sizes are required for the chart with estimated parameters to have a favorable performance like the known parameters case. By virtue of this adverse performance, new charting parameters are proposed for practitioners in the design of the run sum X̄ chart, based on the weights (0, 1, 2, 4) when parameters are estimated. The suggested parameters give a satisfactory performance even when process parameters are estimated from small number of samples
Covariances versus Characteristics in General Equilibrium
Get PDFWe question a deep-ingrained doctrine in asset pricing: If an empirical characteristic-return relation is consistent with investor "rationality," the relation must be "explained" by a risk factor model. The investment approach changes the big picture of asset pricing. Factors formed on characteristics are not necessarily risk factors: Characteristics-based factor models are linear approximations of firm-level investment returns. The evidence that characteristics dominate covariances in horse races does not necessarily mean mispricing: Measurement errors in covariances are more likely to blame. Most important, the investment approach completes the consumption approach in general equilibrium, especially for cross-sectional asset pricing.
- …
