Application of Image Processing for Cutting-Path Generation in Fish Deheading

Abstract

The fish deheading process is a crucial step in the early stages of seafood processing. Traditionally, this process has relied heavily on skilled manual labor to maximize yield. While automated tools and machinery have been developed to assist with fish deheading, these machines still lack the flexibility to adapt their cutting patterns to accommodate the varying sizes and shapes of individual fish. Recent advancements in image processing techniques have led to their widespread adoption in various industries to enhance production efficiency, including quality control, inspection, and object recognition. This research aims to utilize image processing techniques to determine optimal fish-cutting paths to minimize flesh loss. The study involved three main steps: 1) data collection: Images of commonly used canned fish, mackerels, were captured from a top-down perspective under controlled lighting conditions with a reference scale, 2) image processing: Image processing algorithms were applied to extract quantitative information from the fish images, enabling the determination of cutting positions, and 3) analysis and path generation: the results of image processing and statistical analysis were combined to determine cutting positions and generate optimal cutting paths. After processing and analyzing image data of 50 fish samples within the length range of 150 to 250 millimeters, it was found that the average head-to-body length ratio was 0.23, with a standard deviation of 0.006. This ratio was used to establish the position and path for cutting the fish head following a straight-line approach. The cutting path accuracy was evaluated using another 50 sample fish images and the Fréchet distance metric. The average error between the cutting positions generated from image processing and the positions specified by experts was found to be 2.15 millimeters, with a standard deviation of 1.21 millimeters. This corresponds to a mere 1.01% of the body length, indicating a high degree of precision. To assess the practical effectiveness, 15 real fish samples were tested. The average weight loss after cutting the head was found to be 24.47% of the total body weight, with a standard deviation of 3.23%. This falls within the acceptable range for the seafood processing industry. The presented method for cutting fish heads using a straight line approach based on the head-to-body length ratio has proven to be accurate, and efficient, and results in minimal weight loss. The method's effectiveness was validated through both data analysis and real-world testing, making it a promising tool for the seafood processing industry.