Prediction of the chlorophyll content in pomegranate leaves based on digital image processing technology and stacked sparse autoencoder

Figures & data

Figure 1. Technology roadmap

Figure 2. The process of pomegranate leaf image processing

Figure 3. Image processing result

Figure 4. Sample diagram of steps for maximum inline rectangular area

Figure 5. Autoencoder structure

Figure 6. Stacked sparse autoencoder structure

Table 1. Validation set results under different SSAE structures

Hidden layers	Number of hidden nodes	0.005	0.001	0.0005	0.0001	0.00005	0.00001
3	2048:1024:256	20.135	20.075	18.690	18.425	16.366	17.614
3	2048:1024:512	17.574	18.635	17.890	15.530	13.360	12.474
4	2048:1024:512:256	12.081	10.959	9.272	7.356	5.656	6.470
4	2250:1500:756:256	12.042	13.547	11.772	10.101	6.561	8.880
5	2048:1500:1024:512:256	13.834	11.447	11.181	6.208	6.274	13.719
6	2250:1500:756:512:256	14.169	14.071	13.192	8.131	9.780	10.763
6	2048:1500:1024:756:512:256	17.361	15.592	16.664	14.524	9.418	8.570
6	2250:1750:1250:756:512:256	19.953	20.008	19.834	15.225	13.206	17.549

Figure 7. Test results of different models