• No results found

Feature Extraction and Object Detection

4.2 Object Detection

Object detection is the method for finding and identifying objects in an image or video sequence. Human visual system can recognize multiple objects in images simultaneously with little effort, even though the image of the objects might vary to some extent in different viewpoints, differe nt sizes and scales or may be affected by different image transformations. Objects can even be recognized when they are partially obstructed from view. This task is still a challenge for vision systems. Many approaches to the task have been implemented over multiple decades. The general object detection model system is given in the Figure 4.1. In this, the first step involves in application of image processing techniques for extraction of image features. There are several feature extraction techniques available like, edge detection, corner detection and moment invar ia nt methods. In this chapter several available techniques are discussed along with the proposed methods. Experimental results along with quantitative analysis are presented to validate the outperforming behavior of the suggested methods.

65

Figure 4.1: General model of an object detection system

The problem of object recognition is defined as a labelling problem which is based on known object models. Considering a case where an image containing one or more objects of interest (and background) and a set of labels agreeing to a set of models known to the system are given, the detection system has to assign correct labels to regions, or a set of regions, in the image. This problem is related to the segmentation problem i.e. segmentation is possible only when partial recognition of objects is performed. Similarly, recognition is fully dependent on segmentation. As shown in Figure 4.1, the general model of an object detection system comprising of image acquisition, feature extraction technique, generation of hypothesis, object model database and a matching unit. The details of image acquisition are discussed in Chapter 5. The database comprises of all known object models. The information that describes the object in the database depends on the recognition method. The object representation scheme can be different from a qualitative or functional description to precise geometric representation information. A feature is an important characteristic of the object which helps in describing and recognizing the object with respect to other objects. Size, color, and shape are some commonly used features.

Feature detection techniques are applied on images to extract features and assist in establishing object hypotheses. The type of feature depends on the object type involved in recognition process. The hypothesis generator assigns likelihoods to each object based on the extracted features. This step reduces the search space for the recognizer. The object models are organized by using indexing scheme to facilitate the exclusion of dubious object candidates from likely consideratio n. The pattern matching block, that performs the matching procedure, compares the stored model information with the hypotheses and calculates the likelihood of the object. The system detects the

Detected Object Features Elementary Object Feature Extraction Hypothesis Generation Pattern Matching Object Model Database Image Acquisition

66

object with the highest likelihood as the accurate object. All object recognition systems use models either explicitly or implicitly and employ feature detectors based on object models used in the process. Pattern classification approaches are a good example of this approach.

An object recognition system must select appropriate tools and techniques for the steps discussed above. Many factors must be considered in the selection of appropriate methods for a particular application. The central issues that should be considered in designing an object recognition system are:

 Object representation: This plays an important role in object detection process. The representation of object must be efficient enough that the important attributes or features should be present in the model. It is the case that, geometric representation is suitable for some objects, while, generic or function features are more important for some other type of objects. Hence, object representation method must include all relevant and non-redundant information of objects.

 Feature extraction: The use of appropriate feature extraction technique makes the object detection process efficient. The selection of feature extraction method depends on the type of feature to be extracted. Most features can be extracted in two dimensional images but they are related to three-dimensional characteristics of objects. Due to the nature of the image formation process, some features are easy to compute reliably while others are very difficult. Feature detection issues were discussed in the following sections of this chapter.

 Hypotheses generation: The hypothesis generation step is basically a heuristic to reduce the size of the search space. This step uses knowledge of the object model to assign some kind of probability or confidence measure to different objects present in the image. This measure reflects the likelihood of the presence of objects based on the detected features. Hypothesis generation step is required for such applicatio n environment with large number of objects.

 Object Matching: The presence of each expected object can be verified by using the stored models.

67