Anton Satria Prabuwono

A linear model based on Kalman filter for improving neural network classification performance

This paper proposes a method to improve neural network classification performance.A linear model was used as post processing of neural network.The parameters of linear model was estimated using Kalman filter iteration.The method can be applied to classify an object regardless of the type of feature.The method has been validated with five different datasets. Neural network has been applied in several classification problems such as in medical diagnosis, handwriting recognition, and product inspection, with a good classification performance. The performance of a neural network is characterized by the neural networks structure, transfer function, and learning algorithm. However, a neural network classifier tends to be weak if it uses an inappropriate structure. The neural networks structure depends on the complexity of the relationship between the input and the output. There are no exact rules that can be used to determine the neural networks structure. Therefore, studies in improving neural network classification performance without changing the neural networks structure is a challenging issue. This paper proposes a method to improve neural network classification performance by constructing a linear model based on the Kalman filter as a post processing. The linear model transforms the predicted output of the neural network to a value close to the desired output by using the linear combination of the object features and the predicted output. This simple transformation will reduce the error of neural network and improve classification performance. The Kalman filter iteration is used to estimate the parameters of the linear model. Five datasets from various domains with various characteristics, such as attribute types, the number of attributes, the number of samples, and the number of classes, were used for empirical validation. The validation results show that the linear model based on the Kalman filter can improve the performance of the original neural network.

Monte Carlo Method with Heuristic Adjustment for Irregularly Shaped Food Product Volume Measurement

Volume measurement plays an important role in the production and processing of food products. Various methods have been proposed to measure the volume of food products with irregular shapes based on 3D reconstruction. However, 3D reconstruction comes with a high-priced computational cost. Furthermore, some of the volume measurement methods based on 3D reconstruction have a low accuracy. Another method for measuring volume of objects uses Monte Carlo method. Monte Carlo method performs volume measurements using random points. Monte Carlo method only requires information regarding whether random points fall inside or outside an object and does not require a 3D reconstruction. This paper proposes volume measurement using a computer vision system for irregularly shaped food products without 3D reconstruction based on Monte Carlo method with heuristic adjustment. Five images of food product were captured using five cameras and processed to produce binary images. Monte Carlo integration with heuristic adjustment was performed to measure the volume based on the information extracted from binary images. The experimental results show that the proposed method provided high accuracy and precision compared to the water displacement method. In addition, the proposed method is more accurate and faster than the space carving method.

Adaptive Motion Pattern Analysis for Machine Vision Based Moving Detection from UAV Aerial Images

In order to detect moving object from UAV aerial images motion analysis has started to get attention in recent years where motion of the objects along with moving camera needs to be estimated and compensated by using detection algorithm. Moving object detection from UAV aerial images based on motion analysis involves modeling the pixel value changes over time. Moving object detection with moving cameras from UAV aerial images is still an unsolved issue due to not considering irregular motion of camera and improper estimation of noise, object motion changes and finally unfixed moving object direction. This paper presents a low complexity based motion analysis framework for moving object detection along with camera motion estimation by considering motion change of moving object and unfixed moving object direction. Based on the experimental results it is expected that proposed motion vector estimation performs well for both invariant motion and invariant moving object direction.

Moving Object Detection Using Dynamic Motion Modelling from UAV Aerial Images

Motion analysis based moving object detection from UAV aerial image is still an unsolved issue due to inconsideration of proper motion estimation. Existing moving object detection approaches from UAV aerial images did not deal with motion based pixel intensity measurement to detect moving object robustly. Besides current research on moving object detection from UAV aerial images mostly depends on either frame difference or segmentation approach separately. There are two main purposes for this research: firstly to develop a new motion model called DMM (dynamic motion model) and secondly to apply the proposed segmentation approach SUED (segmentation using edge based dilation) using frame difference embedded together with DMM model. The proposed DMM model provides effective search windows based on the highest pixel intensity to segment only specific area for moving object rather than searching the whole area of the frame using SUED. At each stage of the proposed scheme, experimental fusion of the DMM and SUED produces extracted moving objects faithfully. Experimental result reveals that the proposed DMM and SUED have successfully demonstrated the validity of the proposed methodology.

Real Time Vision Based Object Detection from UAV Aerial Images: A Conceptual Framework

In computer vision research, one of the capabilities of establishing an autonomous UAV is the detection of rigid and non-rigid object. Moving object detection with moving cameras from UAV aerial images is still an unsolved issue due to clutter and rural background contained in the images, even and uneven illumination changes, static and moving objects and motion of camera. This paper presents a conceptual framework for moving object detection with moving camera from UAV aerial images combined with the frame difference and segmentation approach together. Our focus is the human as rigid and vehicle as non rigid object detection where the camera can be mounted on the vehicle or other movable platform. It is expected that the proposed conceptual framework performs well under different situations for uneven environments.

Improved descriptor for dynamic line matching in omnidirectional images

This paper proposes enhancement technique to introduce a robust descriptor for matching the vertical lines between the two moving images in the context of omnidirectional images. The first step is to propose a new descriptor by using signal entropy to determine the number of circular areas for extracted lines. Then, to enhance the orientation histogram, standard deviation over the entropy value and number of circles is used to determine the number of bins in each area. Evaluation results demonstrates the robustness of the proposed descriptor in dynamic line matching (DLM) in omnidirectional images.

Motion analysis for moving object detection from UAV aerial images: A review

Motion analysis for moving object from UAV aerial images is still an unsolved issue in computer vision research field due to fast abrupt motion of object and UAV, low resolution, noisy imagery, cluttered background, low contrast and small target size. The main reason for the inability to handle motion is the weakness of existing approaches for moving object detection. This paper presents critical analysis of the various methods used for motion analysis which states lack of relevancy with motion analysis along with some unsolved problems need to be solved for optimum performance of moving objects detection from UAV aerial images. The overall reviews proposed in this paper have been extensively studied in various research papers which can significantly contribute to computer vision research and can be potential for future development and direction for future research.

Interaction techniques in desktop virtual environment: the study of visual feedback and precise manipulation method

Gesture-based systems allow users to interact with a virtual reality application in a natural way. Visual feedback for the gesture-based interaction technique has an impact on the performance and the hand instability making the manipulation of the object less precise. This paper investigated two new interaction techniques in a virtual environment. It describes the influence of natural and non-natural virtual feedback in the selection process using the GITDVR-G interaction technique, which consists of a grasping visual feedback. The GITDVR-G was evaluated in a virtual knee surgery training system. The results showed that it was effective in terms of the task completion time, and that the participants preferred the natural grasping visual feedback. Besides that, the precise manipulation in a newly-designed interaction technique (Precise GITDVR-G) was evaluated. The Precise GITDVR-G includes a normal manipulation mode and a precise manipulation mode that can be triggered by hand gestures. During the precise manipulation mode, an inset view will appear and move with the selected object to provide a better view to users, while the movements of the virtual hand are scaled down to improve the precision. Four different configurations of the precise manipulation technique were evaluated, and the results showed that the unimanual control method with an inset view performed better in terms of the task performance time and the subjective feedback. The finding suggested that the realistic virtual grasping visual feedback can be applied in a virtual hand interaction technique, and that the inset view feature is helpful in the precise manipulation.

HoMeTrack: RFID-based localization for Hospital Medicine Tracking System

Miscommunication among physicians, nurses, and pharmacists can lead many medication errors. It should be eliminated and medicine information should always be verified. A good medicine management integrated with patients medical records is an important thing to avoid medication errors. A proposed model of Hospital Medicine Tracking System using RFID technology (HoMeTrack) has been presented. The objective of this paper is to track the use of medicine based on given prescription and adjusted with patients medical record in order to reduce medication errors. This tracking starts from the medicine out of the storage area until the medicine arrives at the patient. The prototype of HoMeTrack has been demonstrated by assigning RFID tags to medicines and by employing RFID reader along with web based system to track the use of medicines. Alert system will be activated if the medicines have been expired and stock of the medicines is at the minimum level.

Moment Feature Based Fast Feature Extraction Algorithm for Moving Object Detection Using Aerial Images

Fast and computationally less complex feature extraction for moving object detection using aerial images from unmanned aerial vehicles (UAVs) remains as an elusive goal in the field of computer vision research. The types of features used in current studies concerningmoving object detection are typically chosen based on improving detection rate rather than on providing fast and computationally less complex feature extraction methods. Because moving object detection using aerial images from UAVs involves motion as seen from a certain altitude, effective and fast feature extraction is a vital issue for optimum detection performance. This research proposes a two-layer bucket approach based on a new feature extraction algorithm referred to as the moment-based feature extraction algorithm (MFEA). Because a moment represents thecoherent intensity of pixels and motion estimation is a motion pixel intensity measurement, this research used this relation to develop the proposed algorithm. The experimental results reveal the successful performance of the proposed MFEA algorithm and the proposed methodology.