Compressed sampling plant reconstruction method based on non-local self-similar blocks

Abstract

The invention discloses a compressed sampling plant reconstruction method based on non-local self-similar blocks. The method specifically comprises the following steps: firstly, acquiring a color image and a depth image of a target plant through a KinectV2.0 sensor, and preprocessing the color image and the depth image by utilizing depth data; combining with K-means and Mean-shift clustering algorithms, to extract an effective region of atarget plant i to remove unnecessary interferences in the background; then, considering the non-local characteristics of the image, adopting a weighted lp norm minimization algorithm to solve the low rank optimization problemso that details of the image are saved as much as possible; finally, using a Dog-leg least squares algorithm to replace a fastest descent method for iterative optimization. The research fully considers the local characteristics and structural attributes of the image, provides a new idea for high-precision realization of image acquisition and reconstruction, and lays a foundation for promotion of agricultural IT-based application and intelligentization.

Description

technical field [0001] The invention mainly relates to the fields of compressed sensing (Compressed Sensing, CS) and machine vision, and specifically relates to the field of agricultural plant image acquisition and compressed sensing reconstruction methods. Background technique [0002] With the continuous development of agricultural informatization and automation, image acquisition technology is widely used in modern agricultural production, and farmland information acquisition technology has become one of the key technologies for precision agriculture. The detection and identification of diseases and insect pests is of great significance. How to compress, collect and reconstruct images with high quality and high speed has become a hot spot and focus of institutions at home and abroad. [0003] In the process of detection and recognition of agricultural plants, the recognition of target objects is the primary problem to be solved. Adaptive Drosophila mean aggregation based...

AI Technical Summary

Problems solved by technology

The purpose of the present invention is to solve the problems of low reconstruction accuracy and long time consumption caused by the local and structural features of the signal in traditional image compression sensing methods, which mostly use the sparsity of the signal in a certain feature space for optimization. Propose corresponding solutions, so as to achieve effective observation and analysis of the growth status of some crops, solve the storage and transmission problems caused by the large observation data of crops in the long-term stage, and provide relevant basis for relevant control decisions in later agriculture

Images