Limits...
Surface feature based classification of plant organs from 3D laserscanned point clouds for plant phenotyping.

Paulus S, Dupuis J, Mahlein AK, Kuhlmann H - BMC Bioinformatics (2013)

Bottom Line: We introduced an approach using surface feature histograms for automated plant organ parameterization.Highly reliable classification results of about 96% for the separation of grapevine and wheat organs have been obtained.This approach was found to be independent of the point to point distance and applicable to multiple plant species.

View Article: PubMed Central - HTML - PubMed

Affiliation: Institute of Geodesy and Geoinformation - Professorship of Geodesy, University of Bonn, Nussallee 17, 53115 Bonn, Germany. paulus@igg.uni-bonn.de

ABSTRACT

Background: Laserscanning recently has become a powerful and common method for plant parameterization and plant growth observation on nearly every scale range. However, 3D measurements with high accuracy, spatial resolution and speed result in a multitude of points that require processing and analysis. The primary objective of this research has been to establish a reliable and fast technique for high throughput phenotyping using differentiation, segmentation and classification of single plants by a fully automated system. In this report, we introduce a technique for automated classification of point clouds of plants and present the applicability for plant parameterization.

Results: A surface feature histogram based approach from the field of robotics was adapted to close-up laserscans of plants. Local geometric point features describe class characteristics, which were used to distinguish among different plant organs. This approach has been proven and tested on several plant species. Grapevine stems and leaves were classified with an accuracy of up to 98%. The proposed method was successfully transferred to 3D-laserscans of wheat plants for yield estimation. Wheat ears were separated with an accuracy of 96% from other plant organs. Subsequently, the ear volume was calculated and correlated to the ear weight, the kernel weights and the number of kernels. Furthermore the impact of the data resolution was evaluated considering point to point distances between 0.3 and 4.0 mm with respect to the classification accuracy.

Conclusion: We introduced an approach using surface feature histograms for automated plant organ parameterization. Highly reliable classification results of about 96% for the separation of grapevine and wheat organs have been obtained. This approach was found to be independent of the point to point distance and applicable to multiple plant species. Its reliability, flexibility and its high order of automation make this method well suited for the demands of high throughput phenotyping.

Highlights: • Automatic classification of plant organs using geometrical surface information• Transfer of analysis methods for low resolution point clouds to close-up laser measurements of plants• Analysis of 3D-data requirements for automated plant organ classification.

Show MeSH

Related in: MedlinePlus

Dataflow diagram showing the single steps for automated online measurements of wheat yield parameters 1) Laserscanning, 2) Preprocessing, 3) Normal calculation 4) Histogram calculation, 5) Data classification, 6) Region growing and the final 7) Parameter extraction. Arrows indicate the data, while boxes describe the processing steps.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3750309&req=5

Figure 4: Dataflow diagram showing the single steps for automated online measurements of wheat yield parameters 1) Laserscanning, 2) Preprocessing, 3) Normal calculation 4) Histogram calculation, 5) Data classification, 6) Region growing and the final 7) Parameter extraction. Arrows indicate the data, while boxes describe the processing steps.

Mentions: A wheat point cloud with a resolution of 1.0 mm was used for further processing, in accordance to our experience from grapevine plant organ classification. For normal- and histogram calculation rN = 2.5 mm and rH = 12 mm were used. The processing pipeline is as follows 1) laserscanning, 2) pre-processing including cutting of pot points and leaf points, 3) normal calculation, 4) histogram calculation, 5) classification using SVM, 6) region growing and 7) parameter extraction. A visualization of the dataflow is shown in Figure 4. Steps 1 to 5 have been outlined before and were described in the subsection above, therefore we focus on the last two steps to detect the different regions of the labeled point cloud. It was assumed that regions of interest have a significantly bigger size than mislabeled regions. Thus, smaller regions are mislabeled and can be connected to bigger regions next to them. This was done using a region growing algorithm. The results can be seen in Figure 5. The left side shows a characteristic histogram for wheat ears (A) and wheat stems (B) that were calculated out of the training data and used for subsequent SVM classification. Figure 5 (C) shows the results of the classification process of one plant after the region growing step. Separated by colors, the regions are visible. Originally the classification resulted in 39 regions. These were reduced to 8 regions by region growing, clearly dividing 4 ear and 4 stem regions. Overall we reached a mean classification accuracy of 96.56% at a calculation time of 5.40 minutes and 65 thousand points to classify eleven of twelve wheat ears using a leave-one-out cross-classification method. A clear separation of a wheat laserscan was shown using surface feature histograms. The points were allocated to the classes ear and stem and aggregated to a relevant region size. This was done fully automated and enabled the application of a volume estimation algorithms.


Surface feature based classification of plant organs from 3D laserscanned point clouds for plant phenotyping.

Paulus S, Dupuis J, Mahlein AK, Kuhlmann H - BMC Bioinformatics (2013)

Dataflow diagram showing the single steps for automated online measurements of wheat yield parameters 1) Laserscanning, 2) Preprocessing, 3) Normal calculation 4) Histogram calculation, 5) Data classification, 6) Region growing and the final 7) Parameter extraction. Arrows indicate the data, while boxes describe the processing steps.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3750309&req=5

Figure 4: Dataflow diagram showing the single steps for automated online measurements of wheat yield parameters 1) Laserscanning, 2) Preprocessing, 3) Normal calculation 4) Histogram calculation, 5) Data classification, 6) Region growing and the final 7) Parameter extraction. Arrows indicate the data, while boxes describe the processing steps.
Mentions: A wheat point cloud with a resolution of 1.0 mm was used for further processing, in accordance to our experience from grapevine plant organ classification. For normal- and histogram calculation rN = 2.5 mm and rH = 12 mm were used. The processing pipeline is as follows 1) laserscanning, 2) pre-processing including cutting of pot points and leaf points, 3) normal calculation, 4) histogram calculation, 5) classification using SVM, 6) region growing and 7) parameter extraction. A visualization of the dataflow is shown in Figure 4. Steps 1 to 5 have been outlined before and were described in the subsection above, therefore we focus on the last two steps to detect the different regions of the labeled point cloud. It was assumed that regions of interest have a significantly bigger size than mislabeled regions. Thus, smaller regions are mislabeled and can be connected to bigger regions next to them. This was done using a region growing algorithm. The results can be seen in Figure 5. The left side shows a characteristic histogram for wheat ears (A) and wheat stems (B) that were calculated out of the training data and used for subsequent SVM classification. Figure 5 (C) shows the results of the classification process of one plant after the region growing step. Separated by colors, the regions are visible. Originally the classification resulted in 39 regions. These were reduced to 8 regions by region growing, clearly dividing 4 ear and 4 stem regions. Overall we reached a mean classification accuracy of 96.56% at a calculation time of 5.40 minutes and 65 thousand points to classify eleven of twelve wheat ears using a leave-one-out cross-classification method. A clear separation of a wheat laserscan was shown using surface feature histograms. The points were allocated to the classes ear and stem and aggregated to a relevant region size. This was done fully automated and enabled the application of a volume estimation algorithms.

Bottom Line: We introduced an approach using surface feature histograms for automated plant organ parameterization.Highly reliable classification results of about 96% for the separation of grapevine and wheat organs have been obtained.This approach was found to be independent of the point to point distance and applicable to multiple plant species.

View Article: PubMed Central - HTML - PubMed

Affiliation: Institute of Geodesy and Geoinformation - Professorship of Geodesy, University of Bonn, Nussallee 17, 53115 Bonn, Germany. paulus@igg.uni-bonn.de

ABSTRACT

Background: Laserscanning recently has become a powerful and common method for plant parameterization and plant growth observation on nearly every scale range. However, 3D measurements with high accuracy, spatial resolution and speed result in a multitude of points that require processing and analysis. The primary objective of this research has been to establish a reliable and fast technique for high throughput phenotyping using differentiation, segmentation and classification of single plants by a fully automated system. In this report, we introduce a technique for automated classification of point clouds of plants and present the applicability for plant parameterization.

Results: A surface feature histogram based approach from the field of robotics was adapted to close-up laserscans of plants. Local geometric point features describe class characteristics, which were used to distinguish among different plant organs. This approach has been proven and tested on several plant species. Grapevine stems and leaves were classified with an accuracy of up to 98%. The proposed method was successfully transferred to 3D-laserscans of wheat plants for yield estimation. Wheat ears were separated with an accuracy of 96% from other plant organs. Subsequently, the ear volume was calculated and correlated to the ear weight, the kernel weights and the number of kernels. Furthermore the impact of the data resolution was evaluated considering point to point distances between 0.3 and 4.0 mm with respect to the classification accuracy.

Conclusion: We introduced an approach using surface feature histograms for automated plant organ parameterization. Highly reliable classification results of about 96% for the separation of grapevine and wheat organs have been obtained. This approach was found to be independent of the point to point distance and applicable to multiple plant species. Its reliability, flexibility and its high order of automation make this method well suited for the demands of high throughput phenotyping.

Highlights: • Automatic classification of plant organs using geometrical surface information• Transfer of analysis methods for low resolution point clouds to close-up laser measurements of plants• Analysis of 3D-data requirements for automated plant organ classification.

Show MeSH
Related in: MedlinePlus