Fish posture recognition method

Abstract

The invention discloses a fish posture recognition method, and belongs to the field of image processing. The method comprises the following steps: rotating a preprocessed fish image to enable a fish body to be in a horizontal state in the image; cutting the fish in the rotated image, and zooming the cut fish to a set size; making corresponding labels as a training set and a test set according to the fish postures in the images, and training a convolutional neural network to obtain a fish posture recognition model; and performing posture recognition on the fish in the input image by using the fish posture recognition model. According to the invention, the convolutional neural network is adopted to identify the posture of the small and long fish, and when a training set is made, posture correction, cutting and scaling and other operations are performed on the fish image, so that the method can well adapt to the morphological characteristics of the small and long fish (such as pneumatophorus japonicus and little yellow croaker), and the automatic early processing of a fish processing production line is realized; the defects of long time consumption and large workload of the traditional manual processing are overcome, and the invention has the advantages of rapidness, accuracy and high efficiency.

Description

technical field [0001] The invention belongs to the field of image processing, and more particularly relates to a method for fish gesture recognition. Background technique [0002] Fish pose recognition mainly refers to fish target recognition, pose recognition, etc. It has important applications in fish observation research, production and processing, especially for smaller fish. Due to the small size, on the fish processing production line, production line workers are usually required to adjust the fish to a fixed posture for semi-automatic processing, and the processing efficiency of this method is usually relatively low. Therefore, realizing the automatic identification and adjustment of fish posture is of great significance for realizing the full automation of fish processing production line. [0003] Taking mackerel as an example, mackerel is a common edible fish with high nutritional value, and it is one of the important pelagic fishes in my country. Mackerel is wid...

AI Technical Summary

Problems solved by technology

Due to the small size and narrow shape of mackerel, traditional fish processing machines have large errors in mackerel posture and knife point, which leads to a low meat yield of mackerel. Therefore, the current mackerel production and processing industry is still dominated by manual dissection

In a manual work environment, worker productivity is heavily dependent on the proficiency of the worker, and long hours can also lead to lower worker productivity

Images