Method and system for monitoring residual food in trough

Abstract

The invention provides a method and a system for monitoring the residual food in a trough. The method comprises the steps: carrying out the image semantic segmentation of a target trough image, obtaining a trough sub-image and a cattle body sub-image, and calculating the pixel percentage of the trough sub-image in the target trough image; forming an effective trough image set by the target trough images of which the pixel percentages are greater than a first preset threshold value; inputting effective trough images in the effective trough image set into a residual feed recognition model, and obtaining feed percentages corresponding to the effective trough images; According to the average value of all feed percentages, determining the percentage of residual food in the trough. According to the trough residual food monitoring method and system, semantic segmentation and an image recognition technology based on transfer learning are adopted, the collected pictures in the trough are classified to recognize the residual percentage of the feed, and the trough residual food monitoring method and system have the advantages of being lightweight, easy to deploy, wide in use condition and the like; the digitization degree and the intelligentization level of daily ration monitoring in the cowshed can be effectively improved.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a method and system for monitoring leftover food in a trough. Background technique [0002] With the rapid development of the aquaculture industry, the number of aquaculture continues to increase, the scale of operation continues to expand, and the level of aquaculture technology continues to improve. However, there are still problems such as unreasonable feed addition and high management costs in the entire aquaculture industry. There are generally three rations in the feeding management of the farm: the deployment ration, the feeding ration, and the intake ration. The unification of the three diets is an important way to improve feeding accuracy and reduce management costs. In production, feeding rations are usually used to monitor the remaining material in the trough, and then feed back to adjust the feeding ration, while ensuring sufficient feed, it is also n...

AI Technical Summary

Problems solved by technology

[0005] The manual monitoring method has problems such as large errors, difficult management, and high cost, and lacks effective automatic control methods and systems

The existing machine vision method is to judge the feed part through the RGB color range, and it is necessary to artificially paint the background color of the feed tank as smooth white. The method is bound to be limited, and the feed must be evenly, lightly, and comprehensively put into the white trough. The a priori assumptions for detecting the residual degree of forage grass are too harsh, and there is no plan to exclude the cow body from blocking the feed trough.

This makes the method not adaptable to the field environment and is not suitable for the application in actual production, which limits the promotion of feed remaining detection technology

Images