Resumo

GONZALEZ GONZALEZ, Felipe Anastacio; TORRES MATA, Joaquín; LERMA SANCHEZ, Angel Mario  e  GONZALEZ SANCHEZ, Irma Carolina. Learning to obtain the measure of computational cost of the image recognition algorithms in leaves of the soybean crop. RIDE. Rev. Iberoam. Investig. Desarro. Educ [online]. 2023, vol.14, n.27, e601.  Epub 14-Jun-2024. ISSN 2007-7467.  https://doi.org/10.23913/ride.v14i27.1770.

The objective of this paper is to present the measurement of the computational cost of some image recognition algorithms in soybean leaves based on the experimentation of these algorithms. It is a quantitative and experimental study. Starting from a search of the theoretical foundation of the algorithms, the implementation and experimentation of these algorithms tested with images of soybean crop leaves was carried out.

As for the results obtained, it is found that between the descriptor detection algorithms SIFT and SURF in obtaining descriptors for each image tested there is no clear choice because while one has more processing time in milliseconds, its memory consumption is lower and vice versa. In relation to the Harris and Shi-Tomasi corner search algorithms, there is one that clearly shows that it is better both in the number of corners detected, as well as the time in milliseconds is less and the memory consumption is also lower, in this case it is the Shi-Tomasi. And among the active contour algorithms we have that both the Snake and Chan-Vese algorithms, among them the one with the best response time is the Snake with less time in milliseconds and less memory consumption.

Summarizing the results, it can be suggested that the Shi-Tomasi algorithm for the recognition of leaves in soybean plants would be adequate since it obtains optimal results in terms of time and memory consumption of the computer equipment in comparison with the other mentioned algorithms.

Palavras-chave : Algorithms; image recognition; computational cost.