Accelerated Algorithm for Ridge and Valley Detection
Андрей Александрович Зуев, Andrey Ivashko, Denis Lunin · 2023
The paper considers methods for ridge and valley detection in images, including maps of the earth's surface, both real and synthesized. The relevance of this task is due to the fact that the processing of digital elevation maps and the detection of informationally important objects on them has found wide application in agriculture, hydrogeological research, and road construction. The proposed methods make it possible to detect both ridges and valleys, as well as combined structures that include both types of objects. In this case, it is easy to vary both the general sensitivity of the method and the individual sensitivity to the ridge and valley detection. Modifications of the methods have also been proposed that provide a reduction in software and hardware complexity, which will make it possible to implement image processors on microcontrollers and FPGAs and use them in on-board equipment placed on drones. Algorithms can be accelerated, on the one hand, by rounding the coefficients of smoothing filters to powers of two, and, on the other hand, by accelerating the calculation of eigenvalues. The methods have proved their effectiveness in processing digital maps of the earth's surface. The results of the work can be used in the construction of software and hardware image processing systems.