Fish posture judgment method based on visual recognition

Abstract

The invention discloses a fish posture determination method based on visual recognition. The method comprises the steps: firstly obtaining a fish image with a to-be-determined posture, carrying out the preprocessing of the fish image, and obtaining a gray image and a binary image; contour extraction is carried out on the binary image to obtain a fish body contour image; drawing circles with the same radius by taking the head and tail positioning points of the fish body as circle centers, judging the area of the fish body in the two circles, and taking the positioning point with the large areaas the direction of the fish head; judging the abdominal and dorsal postures of the fish body, approximately looking the fish body as an ellipse, carrying out ellipse fitting on the contour image of the fish body to obtain an ellipse equation of the contour of the fish body, and regarding the long axis of the ellipse as the central axis of the fish body; taking the central axis as a boundary to define graphs in the elliptical contour in the grayscale image as an R1 region and an R2 region respectively, performing grayscale value averaging calculation on the R1 region and the R2 region respectively, and taking the region with a large average value as a fish belly direction. A database model does not need to be established, contact is not needed, direct judgment can be achieved through visual recognition, and quickness and high efficiency are achieved.

Description

technical field

[0001] The invention belongs to the field of aquatic product processing, and relates to a fish posture judgment method, in particular to a fish posture judgment method based on visual recognition. Background technique

[0002] In the large-scale processing and production of fish, the determination of fish body posture based on image recognition (that is, the head-tail orientation and ventral-back orientation of the fish body) is very important. In the prior art, image recognition methods based on fish body outlines are often used. This method needs to obtain a large amount of raw fish image data, and then carry out modeling through different modeling methods (partial least squares method, neural network, etc.) to obtain such Key profile data for fish. The database containing key contour data is used to compare with the collected images of the fish to be processed, so as to realize the attitude determination of the fish to be processed. However, this method ...

Images