The field of Life Science Engineering (LSE) is rapidly expanding and predicted to grow strongly in the next decades. It covers areas of food and medical research, plant and pests’ research, and environmental research. In each research area, engineers try to find equations that model a certain life science problem. Once found, they research different numerical techniques to solve for the unknown variables of these equations. Afterwards, solution improvement is examined by adopting more accurate conventional techniques, or developing novel algorithms. In particular, signal and image processing techniques are widely used to solve those LSE problems require pattern recognition. However, due to the continuous evolution of the life science problems and their natures, these solution techniques can not cover all aspects, and therefore demanding further enhancement and improvement.
The thesis presents numerical algorithms of digital signal and image processing to help in improving the pattern recognition based solution of some LSE problems. These problems are selected randomly from the different areas covered by LSE, including those involved in hidden animal detection, biological tissues recognition, animal taxonomy, and bioprocess monitoring problems.
Hidden weevils are traditionally detected by pheromone traps which are not able to perform the task in an early stage of infestation. Hardly seen animals such like bats are usually detected through their echolocations, with the need of accurate handling and filtration of the recorded sound streams for correct detection. In this thesis, a signal processing system is developed including the extraction of large list of conventional/unconventional bioacoustics features. The filtration process and application of window functions are investigated, and different algorithms for the selection of distinctive features are proposed. The system is applied to accurately detect the existence of red palm weevils through the analysis of recordings made by an insertion sensor into palm trees.
Meanwhile, automatic recognition of objects in biological tissues is achieved with image processing by detecting objects’ boundaries in over surface picture, x ray image, ultrasound image, or magnetic resonance image. Nevertheless, the main challenge is to apply a robust edge detector. In this direction, a novel edge detector based on the energy and skewness features of the original image is developed. These features behave as smoothed versions of the image and avoid the application of prior smoothing filters and their corrupting influence on the boundaries. Non-maximum suppression approach with sub-pixel accuracy is established to thin out the detected boundaries to one pixel width. Flux equilibrium check is conducted to fill any discontinuities take place in the constructed edges image. With respect to subjective and objective measures, the developed edge detector presents competitive results in comparison to other commonly used approaches. Several two dimensional features are extract from the edges image to completely define the objects boundaries. And the developed edge detector has been applied efficiently to recognize the intramuscular fat contents in non-living animal slices images.
Animal taxonomy can also be performed by image processing through the analysis of their bioacoustics calls represented in spectrogram images. It is a modern method which presents fast and accurate classification down to species levels. However, the collections of bioacoustics calls in animal natural environment add more difficulties to the classification process due to the attached field noise. Many enhancement approaches have been considered to suppress this background noise, but their common challenge is the degradation of the produced temporal and/or spectral accuracies. Hence, an improved spectrogram enhancement approach has been developed. The approach limits the dynamic range of the spectrogram to the enclosed high energy patterns. The crest factor image is extracted as a smoothed version of the spectrogram. The developed edge detector is applied to detect the boundaries of sound patterns, at which their surrounding noise are eliminated. The method is compared to other enhancement methods, and applied successfully to classify some birds/bats species through their bioacoustics calls spectrograms.
Alternatively, several in situ sensors and techniques are applied to monitor bioprocesses. However, some challenges are accompanying regarding the biofouling formation on the sensor surface, limited measuring range, base line drift, cost of application, and calibration complications. Particularly, ultrasound sensors are promising tools to perform online, noncontact, and non-invasive monitoring. The main parameters obtained by these sensors are the time of flight of the propagating echoes, and its corresponding speed of sound. Numerical approaches to calculate these parameters are mostly the threshold method and cross correlation method. Whereas with the first method echoes reach the threshold level sometimes after their exact starts, while with the second method the calculation is highly affected by existent noise spikes. In this thesis, a time of flight and speed of sound calculation approach is presented. The ultrasound signal is restricted to its dominant frequency. The involved power spectrum and phase shift distributions are handled to detect the time of flight between echoes corrected by their individual phase shifts. Afterwards the speed of sound is calculated by the information of the signal path length. Validations and sensitivity analyses are conducted to check the consistency and repeatability of the results. The proposed method is applied to estimate the time of flight and to monitor the speed of sound variation during online yeast fermentation process. Furthermore, the signal features are combined with temperature measurements in an artificial neural network to instantaneously predict the mixture density with high accuracy.
The developed approaches enlighten the passage which help in decreasing the challenges of LSE problems, and open the horizon to think in more improvements for the already existent solutions. Therefore, further applications of these approaches as well as their limitations and constrained are discussed.
«
The field of Life Science Engineering (LSE) is rapidly expanding and predicted to grow strongly in the next decades. It covers areas of food and medical research, plant and pests’ research, and environmental research. In each research area, engineers try to find equations that model a certain life science problem. Once found, they research different numerical techniques to solve for the unknown variables of these equations. Afterwards, solution improvement is examined by adopting more accurate c...
»
Login
BibTeX