Siti Salasiah Mokri

Motion detection using Lucas Kanade algorithm and application enhancement

Currently, computational of the optical flow of a sequence of images still remains a challenge in video processing. There are no specific techniques that can sufficiently generate an accurate and dense optical flow. Computational using local variable such as Lucas Kanade algorithm does not provide a good segmentation which indirectly affects the pattern of the optical flow obtained. In this paper, we will only focus on differential methods which are Lucas Kanade and Horn Schunck. We investigated the difference in standalone Lucas Kanade algorithm and the effect when it is combined with global variable such as number of iteration and smoothing from Horn Schunck algorithm and filtering. Comparison is made based on the optical flow pattern, segmentation of the motion of the images and the processing time. Experiments on the images show that by using the derivation of partial derivative in Lucas Kanade in Horn Schunck algorithm with the smoothing effect and number of iteration along with filters will result in better segmentation and better optical flow. Thus, this shows that the computation of intensity will influence the optical flow.

Motion detection using Horn Schunck algorithm and implementation

Use of unsuitable techniques and parameters in identifying optical flow movement produces poor segmentation and indirectly affects the optical flow pattern. In this paper, emphasis is focused on the production of optical flow image using Horn Schunck technique and finding the optimum parameters. Image flow movement using Horn Schunck technique and its implementation has been researched to comprehend more about the optical flow. Simulation was performed using the simulation software called MATLAB v7.6 by Mathworks Inc. There are three types of displacements used namely small, medium and large displacement. Several important parameters such as iteration, smoothness and density have been identified to achieve the research goal. This paper reports the study on three parameters previously mentioned in combination with three different types of displacements using Horn Schunck algorithm. Based on Horn Schunck algorithm, the results were obtained after considering the segmentation results, field of optical flow, standard derivation, error and processing time. It is then identified that the optimum values of parameters are when the iteration is between 1 to 6, the smoothing is between 0.0001 to 0.002 and the density is equal to 1.

Weed Detection utilizing Quadratic Polynomial and ROI Techniques

Machine vision for selective weeding or selective herbicide spraying relies substantially on the ability of the system to analyze weed images and process the extracted knowledge for decision making prior to implementing the identified control action. To control weed, different weed type would require different herbicide formulation. Consequently the weed must be identified and classified accordingly. In this work, weed images were classified as either broad or narrow weed type. A fundamental problem in weed image recognition using planar curve analysis is to detect curve. It is difficult to successfully extract curve from the image of weed edges since the appropriate scale to use for extraction is not known a priori. As such, this paper considers a curve detection method based on the quadratic polynomial technique which include the use of the region-of- interests (ROI) technique. The ROI technique creates image subsets by selecting regions of the displayed image. The ROIs are typically used to extract statistics for image operations such as classification. As such, the objective of this paper is to present a novel application of curve detection feature extraction technique in weed classification.

Implementation of Robust SIFT-C Technique for Image Classification

This paper describes the development of a robust technique for image classification using scale invariant feature transform (SIFT), abbreviated as SIFT-C. The proposed SIFT-C technique was developed to cater for varying conditions such as lightings, resolution and target range which are known to affect classification accuracies. In this study, the SIFT algorithm is used to extract a set of feature vectors to represent the image and the extracted feature sets are then used for classification of two classes of weed. The weeds are classified as either broad or narrow weed type and the decision will be used in the control strategy of weed infestation in palm oil plantations. The effectiveness of the robust SIFT-C technique was put to test using offline weed images that were captured under various conditions which truly reflect the actual field conditions. A classification accuracy of 95.7% was recorded which implies the effectiveness of the SIFT-C.

Modeling the Varian On-Board Imager (OBI): Cone-beam CT (CBCT) operating modes

Cone-beam CT (CBCT) imaging is heavily being utilized in radiotherapy treatment as means of treating cancer patients. One of the platforms is the Varian On-Board Imager (oBI) where kilo-volts (kV) CBCT imaging is used. In this study, a model of the OBI is developed to simulate the two scan modes that are available, namely: Full-Fan (FF) and Half-Fan (HF) modes. By shifting the same set of 1024×768 detector panels laterally, a larger field-of-view (FOV) is achieved in the HF scan mode. Using a realistic XCAT phantom, the different FOVs are simulated using analytical Feldkamp-Davis-Kress (FDK) and iterative Simultaneous Algebraic Reconstruction Technique (SART) reconstruction methods. As suggested in literatures, ring and radiant artifacts can occur in the HF mode due to its geometry. To implement fast analytical method, an adequate weight factor needs to be applied on the projection data prior to reconstruction. The percentage normalized root mean squared error (NRMSE) value for FF and HF using analytical FDK reconstruction are 7.12% and 16.63% respectively. Using the iterative SART, the respective values are and 2.69% and 4.65%. The simulation model of the Varian OBI: CBCT operating mode is expected to encourage and enhance further studies on image guided radiation therapy (IGRT) during radiotherapy treatment delivery.