Agricultural machinery monitoring data cleaning method and device based on multi-condition time sequence

Abstract

The invention provides a monitoring data cleaning method based on a multi-condition time sequence. The method comprises the following steps: acquiring operation condition data and image snapshot data of a monitoring vehicle in a current land parcel; preprocessing the operation condition data, and dividing the data into a training set and a test set; inputting the working condition time sequence into a multi-condition time sequence model for feature extraction to obtain multi-dimensional spatial-temporal features; fusing the multi-dimensional spatial-temporal characteristics according to channels to obtain working condition state prediction information; and inputting the working condition information corresponding to the snapshot time in the test set into the trained model to obtain a current snapshot image cleaning state prediction result of the vehicle. The problems of calculation interference of the straw coverage rate of protective farming and storage redundancy and transmission congestion of monitoring equipment are solved, and the data quality and the data utilization rate are improved; the influence on the accuracy of subsequent land parcel operation quality evaluation is reduced, and the network pressure caused by data concurrency is relieved.

Description

technical field [0001] The invention relates to the technical field of agricultural engineering, in particular to a method and device for cleaning agricultural machinery monitoring data based on multi-condition time series. Background technique [0002] In the northern region of my country, straw returning to the field and no-tillage sowing is a widely used protective tillage technology for fertilizing the soil. Installing monitoring equipment on the no-tillage planter can use high-definition cameras to capture the straw coverage in the field and calculate the straw coverage rate. Quantitative calculation. In 2020, Jilin Province's agricultural mechanization remote precision operation platform will monitor an area of ​​about 1.2 million hectares of conservation tillage, and obtain about 7.2 million snapshot images. However, due to the reciprocating movement of the seeder between the fields, turning around at both ends or stopping for a long time takes up a lot of farming tim...

AI Technical Summary

Problems solved by technology

However, due to the back and forth movement of the seeder between the fields, turning around at both ends or stopping for a long time takes up a lot of farming time, so that the monitoring range in the captured image is located in less than 40% of the effective area of ​​the field

However, the cleaning method for single-dimensional relational data is relatively systematic, and only part of it can be directly applied to time series data. However, in the image capture system, invalid data accounts for a large proportion, and single-dimensional time series cannot achieve high fitting accuracy. The multidimensional time series method does not make good use of the physical characteristics between the series, for example, the speed and time of the vehicle constrain the area data, and the voltage will affect the magnitude of the current, etc.

Images