Binocular stereoscopic vision based clustered tomato identification method

Abstract

The invention discloses a binocular stereoscopic vision based clustered tomato identification method, which comprises the following steps: acquiring stereo image pairs of clustered tomatoes by a binocular stereoscopic camera; performing image segmenting for the stereo image pairs, and extracting boundaries of a tomato cluster to be identified; performing multi-result matching and stereo matching in sequence; obtaining a depth map through triangular range finging; and realizing depth map denoising by utilizing an eight-neighborhood median filtering method; realizing clustered style recognition through applying an iterative Otsu method on the depth map; as for the bonding area, directly applying an identification method through margin curvature analysis for the clustered boundaries of the tomatoes to be identified; as for the overlapping area, after the foremost tomato area is segmented, performing operations including area marking again so as to extract the margin of the depth map of the overlapping area of the tomatoes, and followed by segmenting the boundaries of the tomato cluster to be identified, and utilizing the identification method through the margin curvature analysis; and identifying the foremost tomato according to a depth mean of the tomato area. Due to the adoption of the method, different identification methods apply to clustered tomatoes in different types, and the identification for clustered tomatoes under condition of higher overlapping ratio is also realized.

Description

technical field [0001] The invention relates to a clustered tomato recognition method, in particular to a clustered tomato recognition method based on binocular stereo vision. Background technique [0002] The fruit and vegetable picking robot is a solution to realize the automatic picking operation, and the fruit recognition is the prerequisite for the fruit and vegetable picking robot to realize fruit picking. [0003] Tomato clusters refer to multiple tomatoes growing in contact with each other. In the state of natural growth of tomato, the phenomenon of clustering is very common. For example, 87.5% of the tomatoes in a greenhouse grow in clusters. Several common growth states of tomato clusters are as follows: figure 1 shown. Define the tomato overlap rate as the ratio of the area of ​​a certain tomato shaded by other tomatoes to the total area of ​​the tomato in the tomato cluster. According to tomato overlap rate, two types of clustered tomatoes were defined. figu...

AI Technical Summary

Problems solved by technology

When the fruit overlap ratio is large and the edge shape information is insufficient, due to the lack of depth information, it is still difficult to use this type of method to realize the recognition of clustered fruits.

In addition, the above methods do not distinguish the types of clustered fruits, and use a single method for all clustered fruits, so their adaptability to different types of clustered fruits is not ideal.

Images