With the development of new technology, it becomes possible to widen a range of tasks that can be fulfilled by computers. The tasks that modern researchers and programmers focus on are different in nature, but all of them must have practical significance and be helpful for users. This literature review is intended at studying the works devoted to geometric shape recognition. It focuses on geometric pattern recognition in machine learning discussed in modern articles from high-quality scholarly sources.
The recognition of geometric and other patterns in machine learning did not become a popular research subject a long time ago. In their study devoted to the methods of texture recognition, Ojala et al. focus on the multiresolution approach to the classification of textures in black-and-white images with the help of “uniform local binary patterns” (985). The advantages of the method discussed above are presented by the ease of use, the ability to provide accurate results, and applicability to situations that involve nonuniformly illuminated scenes.
The creation of algorithms that recognize simple geometric shapes is helpful in many fields of activity, including photography, challenge-response tests, and production. As is clear from the works that delve into the history of geometric shape recognition, the ability of computers to distinguish between various shapes is the key aspect of diagram understanding problematized in the twentieth century (Song et al. 936). Shape recognition methods are widely used in modern software systems that are capable of working with both digital and hand-sketched images (Song et al. 936). The variety of approaches to the recognition of geometric shapes allows creating programs with reference to specific tasks.
Speaking about other trends in geometric pattern recognition discussed by modern experimentators, it is pivotal to focus on the Hough transform and its uses. As is stated by Fernández et al., the Hough transform is a popular technique used for shape recognition and the detection of complex elements in 2d images (3901). The technique was invented almost forty years ago, but it is still used in the machine analysis of various images.
When it comes to the beneficial features of the Hough transform, modern researchers note its positive results in working with heterogeneous lighting conditions, low-quality images, or image noise (Mukhopadhyay and Chaudhuri 993). Despite the benefits of this technique for shape recognition, some authors argue that its voting processes need to be improved. For instance, Spratling believes that the need to vote for numerous parameter values results in “spurious peaks and quantization effects” (16). With that in mind, some aspects of the technique need improvement to provide more accurate results.
The analysis of principal components or PCA is another method used to teach computers to differentiate between various geometrical shapes. The mentioned method is based on the extraction of principal components, a number of uncorrelated variables retrieved with the help of orthogonal transformations. PCA forms the basis of numerous approaches to geometric shape recognition in machine learning; for instance, Ahmed and Aradhya use it to test the subspace method of image recognition (10).
As is clear from the review by Song et al., other popular and widely used shape recognition techniques for 2d images include “one-pass detection” and “randomized detection” (937). Additionally, the researchers list the use of global geometric properties of input elements for further filtering and recognition.
In the end, by teaching computers to recognize simple and complex geometric shapes, it is possible to solve a number of visual tasks related to image analysis, the protection of data, and even quality control. Nowadays, the methods of machine learning are widely used to create systems capable of recognizing geometric patterns in various types of images, including binary ones. Popular methods of shape recognition and extraction are based on well-known techniques such as the Hough transform.
Work Cited
Ahmed, Muzameel, and Manjunath Aradhya. “A Study of Sub-Pattern Approach in 2D Shape Recognition Using the PCA and Ridgelet PCA.” International Journal of Rough Sets and Data Analysis (IJRSDA), vol. 2, no. 3, 2016, pp. 10-31.
Fernández, Ariel, et al. “Optical Implementation of the Generalized Hough Transform with Totally Incoherent Light.” Optics Letters, vol. 40, no. 16, 2015, pp. 3901-3904.
Mukhopadhyay, Priyanka, and Bidyut B. Chaudhuri. “A Survey of Hough Transform.” Pattern Recognition, vol. 48, no. 3, 2015, pp. 993-1010.
Ojala, Timo, et al. “Multiresolution Gray-Scale and Rotation Invariant Texture Classification with Local Binary Patterns.” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 7, 2002, pp. 971-987.
Song, Dan, et al. “Retrieving Geometric Information from Images: The Case of Hand-Drawn Diagrams.” Data Mining and Knowledge Discovery, vol. 31, no. 4, 2017, pp. 934-971.
Spratling, Michael W. “A Neural Implementation of the Hough Transform and the Advantages of Explaining Away.” Image and Vision Computing, vol. 52, 2016, pp. 15-24.