Crop Stem Coordinate Detection From Depth Images Without Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for crop and weed detection in agriculture require complex segmentation and annotation processes, which can be cumbersome and inefficient, especially when dealing with depth information.
Innovation Solution
A method utilizing a trained convolutional neural network to directly extract plant stem coordinates from a multiple-channel image, including depth information, without the need for segmentation, using a response map with intensity peaks corresponding to plant parts, and converting these coordinates into real-world dimensions using depth information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image data is transmitted in original quality without compression, then image quality is maintained, but transmission time and bandwidth consumption increase
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting compression settings based on image characteristics. The system analyzes image complexity and modifies compression parameters (quality factor, block size, transformation type) to achieve optimal balance between image quality and transmission efficiency for different image types
2Productivity
If image data is compressed to reduce transmission time, then transmission efficiency improves, but image quality deteriorates
Solution Approach 1:
The system dynamically changes compression parameters based on image characteristics. For simple images, higher compression is applied while maintaining quality; for complex images, lower compression is used to preserve detail, thus optimizing transmission efficiency without excessive quality loss
Solution Approach 2:
The patent implements dynamic compression where the compression ratio and parameters are adjusted in real-time based on image complexity analysis. This dynamic approach allows the system to adapt compression strength to each image's specific characteristics, improving overall transmission efficiency while maintaining acceptable quality
3Loss of energy
If compression is applied to reduce data size, then transmission bandwidth consumption decreases, but processing complexity increases
Solution Approach 1:
The patent segments the image into multiple blocks or regions and applies different compression parameters to each segment based on its characteristics. This segmentation allows selective compression where simple regions are heavily compressed while complex regions retain more detail, reducing overall data size without requiring uniformly complex processing across the entire image
4Reliability
If image data is transmitted with error correction, then transmission reliability improves, but transmission time increases
Solution Approach 1:
The patent applies partial error correction by selectively applying error correction codes only to critical image data or regions that require high reliability. This partial approach provides sufficient protection for important information while avoiding the time overhead of applying full error correction to all image data
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
An image processing method is disclosed. The method comprises receiving an image (101) of a crop (4), the image comprising an array of elements which includes or is provided with depth information. The method comprises performing detection of coordinates (131) in the image corresponding to a given part of a plant (such as a stem) in the crop or a crop row using a trained convolutional neural network (120) which is arranged to generate a response map. The response map comprises an image comprising intensity values (123) having respective peaks (122) at each detected coordinate.