Abstract
This study proposes a deep learning based recognition approach from aerial photographs for estimating the yield of watermelons. The precise quantity estimation of the yield is favorable for farmers to negotiate prices at marketing and distribution channel. First of all, obtain high-resolution aerial photograph of watermelon field, Faster R-CNN is used as the deep learning architecture, then label samples for training to get the trained model. Next, detect watermelons on the aerial photograph of watermelon field. But the result is not as satisfactory that there are some watermelons didn't be recognized. Therefore, adding two post-processing to improve the results. With cross scanning, the missing rate reduced from 30% to 18%. After filter by HSV color space, the error rate reduced from 6.1% to 1.4%. Testing different amount of samples, the two approaches can provide higher decrease rate in smaller quantities. This study can get well recognition and yield estimation of watermelon from aerial photographs, and saving considerable time than manually label but also has highly accurate.
Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 4
Claim · model-written summary
Cross scanning reduced missed watermelons from 30% to 18%.
- Taxon
- Watermelon
Citrullus lanatust0 - Treatment
- cross scanninge0
- Outcome
- missing ratee0
- Direction
- reducede0
- Value
- 18%e0
- Qualifier
- 30%e0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“With cross scanning, the missing rate reduced from 30% to 18%”
t0“Watermelon”
Claim · model-written summary
HSV color filtering reduced the recognition error rate from 6.1% to 1.4%.
- Taxon
- WatermelonCitrullus lanatust0
- Treatment
- filter by HSV color spacee0
- Outcome
- error ratee0
- Direction
- reducede0
- Value
- 1.4%e0
- Qualifier
- 6.1%e0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“After filter by HSV color space, the error rate reduced from 6.1% to 1.4%”
t0“Watermelon”
Claim · model-written summary
The two approaches gave a higher decrease rate with fewer samples.
- Taxon
- WatermelonCitrullus lanatust0
- Outcome
- decrease ratee0
- Direction
- highere0
- Comparator
- smaller quantitiese0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Testing different amount of samples, the two approaches can provide higher decrease rate in smaller quantities”
t0“Watermelon”
Claim · model-written summary
Aerial watermelon recognition and yield estimation saved time versus manual labeling.
- Taxon
- watermelonCitrullus lanatuse0
- Outcome
- recognition and yield estimatione0
- Direction
- saving considerable timee0
- Comparator
- manually labele0
- Setting
- aerial photographse0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“This study can get well recognition and yield estimation of watermelon from aerial photographs, and saving considerable time than manually label but also has highly accurate”