Research on Smart Cast for Fish Cage Aquaculture based on image processing

Abstract

In recent years, with the development of fish cage aquaculture technology, the traditional way of bait casting cannot meet the growing demands any more. As fishes feed with certain regularities, it is important to acquire the feeding conditions of the fishes. Through analyzing the images of fishes feed by the way of image processing, the feature of the shape and the area of fishes is extracted. The gross of feeding is setup based on the shape, which reflects the weight of fish. The feeding amount is changed with the variation of the image of fish in feeding at real time. The cast program according the extracted feature and the cage aquaculture experiment is designed to verify the feeding effect. Experimental result shows that the smart cast can setup the bait according to the growing requirement of the fishes. The bait utilization will be increased, and water pollution will be reduced, and the fish quality can be improved as well through the smart cast system. Introduction With the development of marine aquaculture technology, the cage aquaculture develops rapidly for its irreplaceable advantages [1]. Currently fish cage aquaculture is often used feeding machine to cast at a certain time as a certain amount, and this manner of feeding process cannot be modified according to the aquaculture conditions, which wastes too much bait. As the aquaculture technology develops, the feeding machine is needed to cast according to the aquaculture environment and the characteristics of raising species [2]. The amount of bait needed in different growth stages of fishes and the aquaculture environment is different, so the feeding machine should adjust according to the specific situation in real-time [3, 4]. To solve the above problems and needs, this paper studies a way of smart cast used in fish cage aquaculture. To study the regularities of fish feeding by the way of image processing, he control scheme is designed to realize the smart cast which can adapt the regularities of fish growth. The smart cast system can save bait and reduce costs. and it can improve fish quality and protect the marine environment. Research on the regularities of fish feeding In the fish cage aquaculture, fish feeding has some regularity. We can study the regularities through processing the image acquired by the camera [5]. First of all, the image acquisition device is designed to get high quality images for the following image processing; Then process the images through filtering, enhancement, segmentation to separate the fishes from the background of water; At last, the feature of shape and area of feeding fish is extracted to study the variation of the fish feeding process. Image acquisition Using industrial camera and computer to realize the image acquisition, and the camera lens is arranged vertically to reduce the influence caused by the angular difference. The camera is mounted on the bracket to reduce the image blur phenomenon caused by the shake [6]. The images captured by camera are transmitted directly to the computer to process and preserve. Due to the large storage space and the fast memory speed, the computer meets the demand of continuous image acquisition. 2nd Workshop on Advanced Research and Technology in Industry Applications (WARTIA 2016) © 2016. The authors Published by Atlantis Press 1324 Capture the images in the fish breeding base of Ningbo, and the images are showed in Fig. 1. (a) Image before feeding (b) Image in feeding Fig. 1 The original image Image processing Due to the influence of the environment, the images are mixed with the noise signal, which influents the accuracy of image recognition. First of all, image filtering process should be done to reduce the interference of the noise signal; then study the feature of feeding images through the image enhancement and feature extraction [7]. (1) Filtering: The aim of filtering is to reduce the influence of the noise, we use the way of neighborhood average to filter, and the new value of the pixel is equal to the average of the neighborhood pixel value. The formula is: