EVALUATION OF LOW LEVEL FEATURE-BASE METHODS FOR CREATING PANORAMIC MOSAICS OF CROP IMAGES

  • Luciano Jose Senger UEPG
  • Pedro Henrique Soares de Almeida
  • Lilian Tais de Gouveia

Resumo

Unmanned aerial vehicles are making farming more efficient, by allowing farmers to manage crops using georeferenced images. Images acquired from aerial vehicles cameras area used to build image mosaics of the crops. Image mosaicking is the alignment of multiple images into larger compositions which represent portions of a 3D scene. Several image mosaicking algorithms have been proposed over the last two decades. Among all, low-level feature detecting algorithms may be invariant to scale and rotation, among other transformations that commonly occur in agricultural images obtained by unmanned aerial vehicles. This study aimed to evaluate low level feature-based mosaicking methods using agricultural images obtained by unmanned aerial vehicles. The Harris corner detector, the FAST corner detector, the SIFT feature detector and the SURF detector were evaluated according to the computational performance and the quality of the generated mosaics. To assess computing performance, were considered factors such as the detected features average per image, the number of images used to compose the mosaic and the processing times. To assess quality, the mosaics generated by each method were used to estimate the Asian soybean rust severity and a comparison with the commercial software Pix4Dmapper was performed. Regarding quality, there was no significant difference and all methods proved to be on the same level. SURF detector used, on average, only 33.1% of the input images to compose the mosaics. The Harris corner detector achieved the best computing performance results. However, in its final mosaic, the usage of the input images was below than 50%. The FAST corner detector presented the best utilization of the input images, but significant discontinuities of objects where observed in the overlap regions of the resulting mosaics. Besides, the FAST presented the worst computing performance. The SIFT feature detector achieved the second-best processing time, the second-best utilization of the input images and built mosaics without discontinuities in overlapped regions.
Publicado
2019-09-23
Seção
Papers