Multispectral 3D Point Cloud Crop State Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current crop state evaluation methods using common monitoring cameras are unable to acquire comprehensive information, leading to inaccurate assessments of crop conditions, which can result in improper cultivation and potential agricultural losses.
Innovation Solution
A training method for an artificial neural network model that utilizes depth information and multispectral information to generate a multispectral three-dimensional point cloud map, processed using the FVNet three-dimensional target detection algorithm, to acquire comprehensive crop feature information and construct an accurate crop state evaluation model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a common monitoring camera is used to collect crop surface characteristics, then the device complexity is low and ease of operation is high, but the measurement precision and information completeness are insufficient
Solution Approach 1:
The patent combines multiple types of sensors (depth sensor, multispectral sensor, visible light sensor) into an integrated monitoring system. This merging of sensing capabilities allows the system to simultaneously capture comprehensive crop information including depth data, spectral characteristics, and visual features, thereby improving measurement precision while managing device complexity through unified system architecture.
Solution Approach 2:
The patent transitions from two-dimensional image capture to three-dimensional point cloud representation by incorporating depth information. This dimensional expansion enables the system to capture spatial structure, surface characteristics, and spectral properties simultaneously, significantly enhancing the completeness and accuracy of crop information without proportionally increasing system complexity.
2Loss of information
If comprehensive sensor data is collected to improve crop information accuracy, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent segments the monitoring system into distinct functional modules: depth sensing module, multispectral sensing module, and visible light sensing module. Each module captures specific types of crop information independently, and the results are subsequently fused to create comprehensive point cloud data. This segmentation reduces overall system complexity by allowing independent optimization and maintenance of each sensing component while achieving comprehensive information coverage.
Solution Approach 2:
The patent creates a universal processing framework that handles multiple sensor types (depth, multispectral, visible light) through a unified point cloud generation and processing system. The FVNet algorithm serves as a universal detector that processes diverse data types to extract comprehensive crop features, reducing the need for separate processing pipelines for each sensor type and thereby managing system complexity.
3Reliability
If artificial neural network models are trained with comprehensive data to improve evaluation accuracy, then crop state evaluation accuracy improves, but training time and computational resources increase
Solution Approach 1:
The patent performs preliminary data processing and feature extraction before model training. The FVNet algorithm pre-processes the point cloud data to extract meaningful crop features and characteristics, creating a optimized training dataset. This preliminary action reduces the complexity of the training process and the amount of raw data that needs to be processed, thereby reducing training time while maintaining high evaluation accuracy.
Solution Approach 2:
The patent optimizes model training by changing key parameters such as learning rate, batch size, and data augmentation levels. The system dynamically adjusts these parameters during training to achieve convergence faster and with fewer computational resources. Additionally, the patent selects optimal feature subsets from the comprehensive sensor data, reducing the dimensionality of training data and accelerating model training while preserving evaluation accuracy.
Data Source
AI summary
The present invention relates to the technical field of field crop cultivation, more particularly to a training method, an evaluation method, an electronic device and a storage medium. According to the present invention, a multispectral three-dimensional point cloud map is obtained through depth information and multispectral information, and the multispectral three-dimensional point cloud map is analyzed by utilizing an FVNet three-dimensional target detection algorithm, thereby acquiring crop feature information. Thus, more comprehensive crop information can be obtained, and a crop state evaluation model constructed based on an artificial neural network can be further trained with the crop feature information.