Program and weed identification device using computer to identify weeds

Abstract

The invention mainly relates to a program for identifying weeds using a computer and a weed identification device, the device includes a camera, an image capture card, a processor and a memory, and also includes a program for performing the following steps: Preprocessing; mark the positions of all plants in the processed images; identify some weeds according to their location features; extract spectral features and classify unidentified plant patterns, match them in the database, and give all patterns according to the matching similarity confidence value. The program also includes extracting shape feature and texture feature information from the pattern and classifying it, matching in the database, and giving a credibility value according to the matching similarity, and this value is weighted with the previous credibility value to obtain a new value. Credibility; if the credibility is above the threshold, the match is considered successful. The invention improves the real-time performance of the system and the recognition accuracy, and improves the recognition efficiency.

Description

technical field [0001] The present invention relates to a program and a weed identification device for identifying weeds using a computer. Background technique [0002] Field weeds refer to those herbs that grow in farmland, harm crops, are widely distributed, and are not artificially cultivated. They survive long-term adaptation to local crops, farming, climate, soil and other ecological conditions and other factors, and are agro-ecological. an integral part of the system. There are many weeds in the field, and there are many kinds of weeds, and the weeds in the field will compete with crops for water and nutrients, encroach on the ground and underground space, affect the photosynthesis of crops, and interfere with the growth of crops. [0003] Although machine vision has a relatively mature theory, the field weed identification technology based on machine vision is still a new field of application research. Generally, the hardware of a machine vision-based weed identific...

AI Technical Summary

Problems solved by technology

[0004] Scholars at home and abroad have conducted many researches and explorations on this, but the existing weed machine vision recognition technology still has the following shortcomings: most of the research is carried out under the ideal situation where a single plant exists, while in the field environment Among them, there is the problem of overlapping or occlusion of plant leaves. The image segmentation function of overlapping crops and weed leaves is extremely cumbersome and difficult. The correct recognition rate decreases with the increase of occlusion degree, and the recognition rate of weeds had a great impact

Images