Grain Spike Counting via Corner Detection and Adaptive Grid Division
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition techniques for predicting the number of grain spikes, such as wheat spikes, face challenges in accurately detecting small target objects due to resize processing, which can obscure the features of the target objects.
Innovation Solution
A spike number prediction device that performs corner detection to grid-divide a region image, excluding outside regions and re-dividing the image in a size appropriate for the target object, thereby enhancing the detection accuracy of grain spikes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image data is divided into prescribed units (resize processing) to enable AI processing, then the image can be processed by AI, but the features of small target objects become unclear or disappear
Solution Approach 1:
The patent applies segmentation by dividing the image into multiple regions based on corner detection, then selectively processing each region. The region containing the target object is identified through corner detection, and only this region is subjected to fine-grained division for AI processing, while other regions are handled differently. This allows AI processing to be applied selectively where needed, preserving target object features while enabling automated processing.
Solution Approach 2:
The patent implements local quality by applying different processing strategies to different regions of the image. The region containing the target object (identified through corner detection) receives fine-grained division and specialized AI processing, while other regions use coarser processing. This localized approach ensures that small target objects maintain their feature clarity in the region of interest while still enabling overall AI processing of the image.
2Measurement precision
If the entire image is divided into many small units to detect small target objects, then detection capability improves, but processing complexity and time increase
Solution Approach 1:
The patent applies preliminary action by first performing corner detection on the entire image to identify the region containing the target object before applying fine-grained division. This preliminary corner detection step guides subsequent processing to focus only on the relevant region, avoiding the need to divide the entire image into many small units and reducing overall processing complexity while maintaining detection accuracy for small targets.
Solution Approach 2:
The patent extracts the region containing the target object from the entire image using corner detection, then applies fine-grained division only to this extracted region. This extraction approach allows the system to maintain high detection accuracy for small target objects by processing only the relevant portion with fine granularity, while avoiding the computational burden of processing the entire image with high resolution.
3Measurement precision
If manual counting of grain spikes is performed to obtain accurate data, then measurement accuracy improves, but labor effort and time consumption increase
Solution Approach 1:
The patent replaces manual mechanical counting with an automated system that uses corner detection and AI-based image processing. The system automatically identifies the region containing grain spikes through corner detection, then uses AI algorithms to count the spikes within that region. This substitution eliminates manual labor while maintaining accurate counting, significantly reducing time consumption while preserving measurement precision.
Solution Approach 2:
The patent creates a digital copy of the image data and processes this copy through corner detection and AI algorithms to perform spike counting. Instead of physically counting spikes in the field, the system analyzes a digital representation of the image, automatically identifying and counting spikes through computational methods. This copying approach enables accurate counting without manual intervention, reducing time while maintaining precision.
Data Source
AI summary
A spike number prediction device (10) includes: a corner detection unit (11) that performs grid division on a region image in which a region including a unit region set in a cultivation field of grain is reflected and detects corners of the region in first divided images; an exclusion unit (12) that acquires a position of the region based on positions of corners and excludes image data of an outside region from image data of the first divided images; a target object detection unit (13) that performs grid division on an image according to a size of spike of grain and detects the spikes from second divided images; and a prediction unit (14) that counts the number of spikes of grain and predicts, based on a size relationship between a target range and the region, and spike number count, the number of spikes in the target range.