Design And Testing Of A Machine Vision-based Mackerel Body Orientation Arrangement Conveyor

Before the fish can be mechanically dissected and gutted,they usually need to be sorted and transported according to a uniform head,tail and ventral and dorsal orientation.Currently,this is mainly done manually,which is labor intensive and inefficient.In this paper,a machine vision-based mackerel body alignment device was designed and prototyped based on the physical characteristics of mackerel,and tests were conducted to verify the performance of the device.The main research elements of the thesis are as follows:(1)A machine vision-based mackerel body orientation arrangement conveying device was designed and developed.The design of the mackerel carcass orientation conveyor was based on the testing of the physical characteristics of the mackerel carcass.The mackerel body directional arrangement conveying device consists of a fish lifting device,a fish separating device,a fish head,tail and belly directional conveying device,a fish return conveying device and a directional control system.The fish body lifting device enables the single fish body to be separated and conveyed upwards;the fish body separating device guides the fish body down from the lifting device and separates the fish body;the head,tail and ventral dorsal directional conveying device enables the fish body to be oriented according to the set head,tail and ventral dorsal direction;the fish body return conveying device conveys the fish body with the wrong head and tail direction back to the fish body lifting device to adjust the head and tail direction of the fish body.(2)A deep learning-based head,tail and ventral-dorsal orientation detection model was constructed.Developed a deep learning-based model for detecting the head,tail,and ventral-dorsal orientation of mackerel fish.Constructed a mackerel body image dataset and trained the model using YOLOv5 s.The resulting model achieved exceptional performance,with 99.4% accuracy,99.7% recall,and an average detection accuracy of99.4% on the test set.Additionally,the model achieved a detection speed exceeding 50.5fps.The model size was 14.1MB,and it was able to achieve real-time detection of fish bodies.(3)A directional control system is built,mainly including the hardware design and software design of the control system.The Arduino controller is used as the core of the system control,and the main electrical components of the control system are selected;the Arduino IDE is used to complete the programming of the controller,and the program functions mainly include serial communication with the host computer,reading the photoelectric sensor signal and sending the corresponding control instructions to control the cylinder action.(4)A machine vision-based performance test of mackerel body orientation arrangement conveyance device was conducted.The overall device performance was tested,and the results showed that when the speed of the fish lifting device is 0.1m/s,the speed of the fish body separation device is 0.45m/s,and the speed of the head,tail and ventral dorsal directional conveying device is 0.6m/s,the directional arrangement conveying device achieved optimal performance.The success rate of the mackerel head and tail directional was 97.2%,the success rate of the ventral dorsal directional was95.6%,and the efficiency of fish orientation reached 15 fish/min.