Deep learning cordyceps sinensis small target detection method based on image cutting
The invention discloses a small cordyceps sinensis target detection method based on picture cutting, and the method comprises the following steps: 1) cutting a data set picture, putting the data set picture into a deep learning model, and training the data set picture to obtain a model file; (2) a c...
Saved in:
Main Authors | , , , , , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
28.06.2024
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention discloses a small cordyceps sinensis target detection method based on picture cutting, and the method comprises the following steps: 1) cutting a data set picture, putting the data set picture into a deep learning model, and training the data set picture to obtain a model file; (2) a camera is used for collecting grassland pictures, the picture clipping size and the overlapping rate are calculated according to the height of the camera, then the pictures are clipped, and the clipped pictures are input into the model for reasoning; 3) mapping a result frame output by the model to an original image coordinate, and removing repeated frames by using an NMS algorithm; and 4) displaying the result in the original image, and prompting the number of the cordyceps sinensis in real time through voice. The method can solve the problems of limited feature extraction, low small target detection precision and small target scale change in detection of small targets such as cordyceps sinensis.
本发明公开了一种基于切图的虫草小目标 |
---|---|
Bibliography: | Application Number: CN202410348486 |