Green fruit recognition method

Abstract

The invention discloses a green fruit recognition method, which comprises the steps of first, acquiring images, extracting R channel component images, and performing median filtering and 8-level grayscale discrete operations for the component images; next, carrying out multi-angle grayscale step-scanning for the 8-level grayscale images, and screening scanning results based on heuristic rules; then fusing a variety of information corresponding to screening results, and generating candidate fruit regions; and finally, voting from the candidate fruit regions based on three aspects which are appearance features and block operations, the number of grayscale step directions and the number of grayscale steps, and determining the final real fruit region in accordance with the majority voting principle. The invention introduces the majority voting decision-making mechanism into the recognition process, and fuses the scene image appearance features and mesh grayscale step features through the mechanism, thereby greatly improving the accuracy and robustness of the recognition method, and dramatically facilitating the overall operation efficiency of a harvesting robot.

Description

Technical field: [0001] The invention belongs to the technical field of image recognition, and relates to a green fruit recognition method on a tree based on a network gray scale feature, an adaptive classifier and a majority voting decision mechanism. Background technique: [0002] Picking robots help to improve fruit picking efficiency, reduce damage rates and save labor costs, which has important practical significance. Among them, the fruit recognition strategy based on scene analysis is an important key technology for picking robots to achieve fruit picking. [0003] Machine vision is generally considered to be the best way for picking robots to realize fruit scene analysis, and fruit recognition based on machine vision is a long-term research hotspot in related fields. However, the inherent uncertainty of the unstructured fruit scene poses a huge challenge to the robot's visual perception of fruit and operating environment information. Fruit growth conditions (such a...

AI Technical Summary

Problems solved by technology

After searching the existing literature, it was found that Ferhat et al. published an article entitled "Green Fruit Recognition Method Based on Characteristic Fruit, Color and Ring Gabor Texture Features" in the Computers and Electronics in Agriculture professional journal, Volume 78, Issue 2, 2011 (English name of the article: Green citrus detection using'eigenfruit', color and circularGabor texture features under natural outdoor conditions), this article discloses a green citrus recognition method that integrates multi-category features, multi-scale search strategies and majority voting strategies, but It still has problems such as low fruit recognition rate, high fruit false detection rate, and complex detection algorithm process.

Images