The Design And Implementation Of Underwater Sea Cucumber Object Detection System Based On Convolutional Neural Network

In recent years,material living standards have been continuously improved.Everyone is willing to purchase products that are rich in nutritional value to maintain the body.As a valuable marine animal,sea cucumber is attracting attention due to its rich nutrition and medicinal value.At this stage,however,fishing method of sea cucumber mainly uses human fishing,which is inefficient,and poses a major threat to personal safety.Therefore,it is an inevitable trend to develop a fully-automatic fishing equipment to replace artificial sea cucumber fishing,which will save a lot of manpower and fishing costs.Whether it can accurately detect the object of underwater sea cucumbers will be the key to fully automatic fishing equipment.Therefore,this paper aims to design an underwater sea cucumber object detection system based on convolutional neural network to accurately detect the underwater sea cucumber object.The main work of this article is as follows:First,introduce the requirements of system,including function and performance,and design according to demand of the system,derive the function of each module,and design the system's detection process.Secondly,by analyzing the problems of the current object detection algorithm based on proposal region and the object detection algorithm based on regression,and weighing the requirements of precision and speed,YOLO v2 algorithm is used to detect the object of underwater sea cucumber.The YOLO v2 algorithm's network structure,coordinate prediction,non-maximal suppression,loss function,and other key elements are analyzed,and corresponding modifications are made according to the specific task of this article.Then,the realization of underwater sea cucumber object detection system based on convolutional neural network is implemented,which is divided into three stages.In the data preparation stage,the construction of data set of the underwater sea cucumber object detection is completed,including: image preprocessing,data annotation,and data screening.Among them,the image pre-processing module uses the ACE algorithm to enhance the underwater sea cucumber image,and restores the color of the underwater sea cucumber image aiming at the color shift problem in the underwater sea cucumber image.During the training stage of the model,the YOLO v2 algorithm model is trained according to the algorithm training method and the specific hyperparameter settings.Underwater sea cucumber object detection stage,introduce the function and interface of the underwater sea cucumber object detection module.Finally,the underwater sea cucumber object detection system based on convolutional neural network is tested.The performance test of the system,by comparing the experimental results,verifies the feasibility and effectiveness of the method used in this article for the underwater sea cucumber object detection.The functional test of the system,through the detection results of the sea cucumber object in the image and video files,verifies whether the system can accurately detect the sea cucumber object in the underwater environment.