Scene Analysis Techniques Research Articles

Parametric Synthetic Aperture Radar Image Recovery for Multiple Linear Structures: An Image Domain Approach

The linear structures in synthetic aperture radar (SAR) images can provide important geometric information regarding the illuminated objects. However, the linear structures of various objects sometimes disappear in traditional SAR images, which severely affects the application in automated scene analysis and information extraction techniques. This paper proposes a parametric SAR image recovery method for linear structures of extended targets. By extracting the spatial phase distribution feature in image domain, the proposed method is used to identify the endpoints among all of the scattered points in SAR images and reconstruct the disappeared linear structures based on the scattering model. The method can be generally divided into three procedures: endpoints and point targets classification, linear structures recognition, and linear structures reconstruction. In the first step, the endpoints of linear structures and point targets are classified by setting a threshold related to the spatial phase distribution feature. Afterwards, the linear structures recognition method is used to determine which two endpoints can be formed into a linear structure. Finally, the parametric scattering models are used to reconstruct the disappeared linear structures. Experiments are conducted on both computer simulations and the data that were acquired by microwave anechoic chamber experiment, tower crane radar experiment, and unmanned vehicle radar experiment in order to validate the effectiveness and robustness of the proposed method.

Registration of Multimodal Remote Sensing Image Based on Deep Fully Convolutional Neural Network

Multimodal image registration is the fundamental technique for scene analysis with series remote sensing images of different spectrum region. Due to the highly nonlinear radiometric relationship, it is quite challenging to find common features between images of different modal types. This paper resorts to the deep neural network, and tries to learn descriptors for multimodal image patch matching, which is the key issue of image registration. A Siamese fully convolutional network is set up and trained with a novel loss function, which adopts the strategy of maximizing the feature distance between positive and hard negative samples. The two branches of the Siamese network are connected by the convolutional operation, resulting in the similarity score between the two input image patches. The similarity score value is used, not only for correspondence point location, but also for outlier identification. A generalized workflow for deep feature based multimodal RS image registration is constructed, including the training data curation, candidate feature point generation, and outlier removal. The proposed network is tested on a variety of optical, near infrared, thermal infrared, SAR, and map images. Experiment results verify the superiority over other state-of-the-art approaches.

Error correction and accuracy scaling for a joint subtraction multiple fundamental frequency estimation algorithm

This talk presents an improved multiple fundamental frequency (F0) estimation algorithm, which is loosely based on a previously developed joint subtraction algorithm [A. Klapuri, Proc. 7th Int. Conf. on Music Information Retrieval, 216–221 (2006)]. The improved algorithm uses some auditory scene analysis techniques to supplement a maximum likelihood based F0 estimator to predict the best combination of F0s. The accuracy of this base algorithm is then improved by use of error detection/correction. Error rates of the algorithm were collected using random mixes of sounds gathered from the University of Iowa Theremin database of musical instrument samples, as well as mixes of continuous audio. The performance of both the base algorithm and the improved algorithm will be presented together. It was found that the improved algorithm decreases error rates by as much as 50%, compared to the base algorithm with minimal loss in computational efficiency. The trade-offs between computational efficiency and accuracy, tested via a tunable parameter built into the algorithm, will be discussed. Trade-offs between computational costs and accuracy as a function of the number of F0s will also be discussed.

Constructing and application of multimedia TV-news archives

This paper addresses an integrated information mining techniques for broadcasting TV-news. This utilizes technique from the fields of acoustic, image, and video analysis, for information on news story title, newsman and scene identification. The goal is to construct a compact yet meaningful abstraction of broadcast TV-news, allowing users to browse through large amounts of data in a non-linear fashion with flexibility and efficiency. By adding acoustic analysis, a news program can be partitioned into news and commercial clips, with 90% accuracy on a data set of 400h TV-news recorded off the air from July 2005 to August 2006. By applying speaker identification and/or image detection techniques, each news stories can be segmented with a better accuracy of 95.92%. On-screen captions or subtitles are recognized by OCR techniques to produce the text title of each news stories. The extracted title words can be used to link or to navigate more related news contents on the WWW. In cooperation with facial and scene analysis and recognition techniques, OCR results can provide users with multimodal query on specific news stories. Some experimental results are presented and discussed for the system reliability, performance evaluation and comparison.

Image edge detection

PurposeThis paper seeks to describe the use of edge‐detection as an approach to support the process of analyzing visual scenes.Design/methodology/approachEdges‐detection is considered as a KM technique task. Several edge‐detection algorithms were discussed but Canny method was chosen in conjunction with JAVA to find the edges of an image.FindingsThe paper provides an overview of knowledge need by the visual scene analyst. It reviews the process of extracting knowledge from image data. Finally, the paper looks at the management aspects of image‐edge detection.Originality/valueExperiments herein have proved the method robustness by considering edge‐detection as a knowledge management technique for visual scene analysis by deriving knowledge from semantic information based on object boundaries shadow.

A VLSI architecture for dynamic scene analysis

Analysis of dynamic scenes to obtain motion parameters, structures and identities of the objects present in the scene and/or descriptions of the events taking place in the scene is an important problem in machine vision. Dynamic scene analysis involves processing and analysis of a sequence of image frames (snapshots of the scene) taken at a regularly spaced time interval. Since a large volume of data is to be processed, dynamic scene analysis techniques are, in general, computationally very intensive which makes them unsuitable for applications which require real-time response. A common approach to obtaining motion related information about a scene involves analyzing the characteristics of the connected regions in the corresponding difference pictures. Previous implementations of the difference-picture-based technique use multicomputer and multiprocessor systems which are uneconomical because of the overheads involved. In this paper, we propose a VLSI architecture that exploits the pipelining and the parallelism possible in the above technique. Each processor is organized as a multistage linear pipeline and the processor architecture is simple enough that the motion detection and classification system can be implemented as a single VLSI chip. Once the difference picture has been generated and segmented into separate regions corresponding to motions of different objects, each processor in our proposed chip will independently analyze a different region to classify the type of motion occuring in that region within the scene. Currently, the processor architecture is being laid out in silicon at the University of South Florida for actual fabrication and testing. The chip is expected to operate at the rate of 20 MHz.

ROBOTIC END EFFECTOR FOR HANDLING GREENHOUSE PLANT MATERIAL

ABSTRACT An end effector was developed to investigate the mechanics and control of robotic handling and manipulation of plant material of a type commonly found in commercial greenhouses. The end effector was shown to handle a wide range of sizes of geranium cuttings with rare indications of damage to the petioles (1.5%) and main stem (2.0%). Several features of the end effector were determined to be important for reliable, non-damaging performance. A two-stage feedback controller which combined position/velocity control with force control was successful in minimizing cycle time while also minimizing impact velocity and resultant impact loads on plant material. A machine vision local scene analysis technique provided an automatic method of obstacle avoidance by the fingers in a plant canopy. Padded fingers with relatively small, curved ends minimized contact area and assisted in decreasing impact forces. In general, results indicate the importance of sensing and interpretation of the sensor data to assist a robot in accommodating the nonuniformity typical of plants.

Textural feature extraction in seismic images

This lecture introduces a set of second‐order statistical features that can be used to discriminate among several textural classes associated with seismic reflection configuration patterns. It is well known in the petroleum geology community that some of these patterns are characteristic of certain types of rock formations. The textural features that we use here are based on an extension of the co‐occurrence matrix approach originally proposed by Haralick et al., in 1973. The feature extraction technique has been applied to a seismic image from offshore West Africa. The results indicate that the method is useful for the local discrimination of rock types deposited during the same geological period. Considerably more elaborate scene analysis techniques are however needed for image understanding purposes. They are not discussed in this talk.

Exploiting Spectral, Spatial and Semantic Constraints in the Segmentation of Landsat Images

A critique of traditional classification techniques for LANDSAT images and consideration of some scene analysis techniques, exploiting spatial organization and meaning, lead to a new approach to computer programs for LANDSAT image understanding. To justify this approach, a program that combines modified maximum likelihood techniques with interpretation-controlled region merging methods to interpret forest cover in LANDSAT images is described. For comparison purposes, a pure supervised classifier using the same data made 43% more errors and produced a segmentation twice as complex.

Quantisation Error in Area Measurement

A method for measuring areas using Visual Scene Analysis techniques is described, and a comparison made with a point counting approach. The theoretical accuracy of such methods is discussed with reference to a previously published analysis, in the light of different randomness criteria.

Quantisation Error in Area Measurement

AbstractA method for measuring areas using Visual Scene Analysis techniques is described, and a comparison made with a point counting approach. The theoretical accuracy of such methods is discussed with reference to a previously published analysis, in the light of different randomness criteria.

A region coloring technique for scene analysis

A method of converting a picture into a “cartoon” or “map” whose regions correspond to differently textured regions is described. Texture edges in the picture are detected, and solid regions surrounded by these (usually broken) edges are “colored in” using a propagation process. The resulting map is cleaned by comparing the region colors with the textures of the corresponding regions in the picture, and also by merging some regions with others according to criteria based on topology and size. The method has been applied to the construction of cloud cover maps from cloud cover pictures obtained by satellites.