Title | ||
---|---|---|
Recognition Method Of Thermal Infrared Images Of Plant Canopies Based On The Characteristic Registration Of Heterogeneous Images |
Abstract | ||
---|---|---|
Canopy thermal infrared imaging reflects the temperature change in the crop canopy, which is closely related to the stomatal conductance and water utilization characteristics of the crop. Thermal infrared imaging is an effective and nondestructive way to study the early detection of crop diseases. However, efficiently extracting the thermal infrared canopy region of the crops is an important factor restricting the study of canopy temperature changes. The grayscale distribution of the edges of the crop canopy thermal infrared image is uneven with strong noise and cannot be extracted effectively using the traditional image segmentation method. Thus, a recognition method for the thermal infrared images of plant canopies based on heterogeneous image characteristic registration was proposed to overcome the shortcomings above. First, the Gauss membership function was selected to construct the network recognition rules on the basis of the three-layer backward reasoning of the adaptive BP neural network to recognize the visible light reference images of the plant canopies. Second, the optimal registration parameters of the affine transformation were calculated for registering the canopy region of the reference image and that of the initial thermal infrared image. Third, a recognition model for the thermal infrared images of plant canopies was established based on bilinear mapping factors. Finally, information entropy and mutual information were used to evaluate the effectiveness of the recognition model. The results showed that the initial temperature range of the original thermal infrared image was 20.46-36.40 degrees C. After removing the background interference of the thermal infrared canopy of the crop, the temperature range of the target image was 20.46-26.65 degrees C, and the average temperature after extraction was 2.04 degrees C lower than that before extraction. In addition, the entropy difference between the canopy of the thermal infrared image identified by the proposed model in this study and the standard recognition method was within the range of 0.01-0.06, indicating the effectiveness of the recognition model for the plant canopy. Therefore, this study provided an efficient method for obtaining the canopy temperature characteristics of crops. |
Year | DOI | Venue |
---|---|---|
2020 | 10.1016/j.compag.2020.105678 | COMPUTERS AND ELECTRONICS IN AGRICULTURE |
Keywords | DocType | Volume |
Plant canopies, Thermal infrared image, BP neural network, Affine transformation, Recognition model | Journal | 177 |
ISSN | Citations | PageRank |
0168-1699 | 0 | 0.34 |
References | Authors | |
0 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Meng Liu | 1 | 39 | 18.70 |
Haiou Guan | 2 | 5 | 1.82 |
Xiaodan Ma | 3 | 5 | 1.82 |
Song Yu | 4 | 7 | 3.25 |
Gang Liu | 5 | 98 | 36.92 |