Cluster Representation of the Structural Description of Images for Effective Classification

Abstract

The problem of image recognition in the computer vision systems is being studied. The results of the development of efficient classification methods, given the figure of processing speed, based on the analysis of the segment representation of the structural description in the form of a set of descriptors are provided. We propose three versions of the classifier according to the following principles: "object-etalon", "object descriptor-etalon" and "vector description of the object-etalon", which are not similar in level of integration of researched data analysis. The options for constructing clusters over the whole set of descriptions of the etalon database, separately for each of the etalons, as well as the optimal method to compare sets of segment centers for the etalons and object, are implemented. An experimental rating of the efficiency of the created classifiers in terms of productivity, processing time, and classification quality has been realized of the applied. The proposed methods classify the set of etalons without error. We have formed the inference about the efficiency of classification approaches based on segment centers. The time of image processing according to the developed methods is hundreds of times less than according to the traditional one, without reducing the accuracy.

Application a Committee of Kohonen Neural Networks to Training of Image Classifier Based on Description of Descriptors Set

“…Data quantization and neural networks are currently powerful tools for knowledge acquisition [5], [6], [7], [16], [19], [20], [21], [22]. Regarding object descriptions as a set of descriptors, the mentioned tools can significantly speed up the image classification process by pre-grouping large amounts of data [9], [13], [20]. This achievement results from high-speed data search when the input data is compared with the etalon data stored in the system's memory [9], [14].…”

“…Regarding object descriptions as a set of descriptors, the mentioned tools can significantly speed up the image classification process by pre-grouping large amounts of data [9], [13], [20]. This achievement results from high-speed data search when the input data is compared with the etalon data stored in the system's memory [9], [14]. One of the practically effective networks used for a wide range of data is the Kohonen network [19], [20].…”

“…Note that along with committee usage, the classification scheme based on determining the relevance of the descriptions of the analyzed object and the etalon remains unchanged [2], [26]. Class vote counting can be organized independently for each etalon [9] or competitively on a set of calculated centroids for the entire database [26].…”

“…At present times, structural classification methods are also being intensively developed, where decision-making is based on describing the image as a set of keypoint descriptors [2], [9], [10]. Keypoints and their descriptors are formed by special filtersdetectors [11], [12].…”

“…Modern structural methods implement the classification process based on the "bag of words" model, where the class of an object is determined by the result of voting of the componentsdescriptors of the recognized object [2], [13]. In the traditional approach, the base set consists of the descriptions of etalons, and in the modifications, the same base is the centers or parameters of the descriptions [9], [14], [15]. Such methods are advantageous over neural networks, as they focus on alignment with the description or parameters of the etalons and do not require long-term training [10].…”