Publish time: 2022-09-26 00:00:00
This study created a "Phalaenopsis double-spikes rate prediction model." We use AI object detection technology to combine machine learning models to extract the leaf spectrum of hyperspectral images of orchid seedlings and match them with the horticultural traits of orchid plants. This Know-how has significantly improved the traditional assay method.
The application scenario is through the prediction model system developed by our team. For a particular batch of orchid seedlings, input its data, and then output the probability value of the double-spike rates of the whole bunch of orchid plants, sorted from large to small, and take the proportion of the shipment to get the evaluation result. The experimental results show that only 62.17% of the double spikes are correct if the traditional method is used for screening. Still, if this method is used for evaluation, choosing the top 10% of the plants with high scores and 97.50% accuracy of the orchid plants will produce double spikes. If the top 50% of the plants with high scores, 84.63% accuracy, will have double stalks. The proposed method can be proven in experiment results to effectively help flower farmers evaluate and select multi-floral stalk plants, improve the economic value of Phalaenopsis, and increase the competitiveness of Taiwan's export Phalaenopsis industry.
According to the application scenarios, a grouping sorting, screening, and a non-destructive Phalaenopsis grading method are proposed. First, classify Phalaenopsis according to the species and then sort them according to the double-spike rate predicted by the model. Preferentially selecting plants with a high probability of producing double stems for export is considered, and the yield rate of export double-spikes Phalaenopsis can be significantly improved.

Figure1. The model structure we used when taking both hyperspectral images and horticultural traits into consideration.
32A-1110905
Know-how
The technologies can be applied to Phalaenopsis farmers, operators, or traders to forecast or predict the number of spikes of Phalaenopsis. Related technologies include:
1. AI object detection - lock the latest mature leaves:
2. Spectral Information Extraction - Band Information Refinement:.
3. Machine learning prediction model - replaces manual screening and quickly provides a large number of calculation results:
4. Statistical analysis:
5. Website concatenates machine learning models:
Academia sinica: Hsiao, Chiu-Han; Swee-Suak Ko; Yi-Jyun LinNational Taiwan University: Frank Yeong-Sung Lin; KUAN-YI CHEN; Li-Hsuan Chen; WENG, TZU-TING; Chien-Yu Shih; HU, JIA-KAI
If you are interested in this technology, please fill in the information on the webpage under the manufacturer option (Request for Technology Licensing, Material Transfer, or Sponsored Research) on the website of this office

A Prediction Model for Double-Spikes Phalaenopsis Orchid