FvUnderwater sea cucumber identification based on Principal Component Analysis and Support Vector Machine. (February 2019)

• Record Type: Journal Article
• Authors: Qiao, Xi; Bao, Jianhua; Zhang, Hang; Wan, Fanghao; Li, Daoliang
• Is Part Of: Measurement. Volume 133(2019)
• Page Start: 444
• Page End: 455
• Publication Date: 2019-02
• Subjects: Underwater image processing -- Feature extraction -- Feature dimension reduction -- Sea cucumber identification; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2018.10.039
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A method based on principal component analysis and support vector machine for underwater sea cucumber image identification. By the sorted dataset, the stability is increased in PCA-SVM object identification. The superior performance of proposed method over all established methods for classifying sea cucumber and background object. An underwater sea cucumber identification system has been designed with an easy, operation-friendly interface. Abstract: Underwater sea cucumber images are blurred and contain complex backgrounds. To improve the efficiency of sea cucumber identification, a method based on Principal Component Analysis (PCA) and Support Vector Machine (SVM) was proposed. Firstly, colours, textures and shapes of the sample images were extracted. Then, each feature was used separately to train SVM to identify the target. These features were sorted by identification rate. PCA-SVM was used to train the classifier, and the classifier was proposed to identify sea cucumber images. The accuracy of our proposed method was 98.55%, the time taken was 0.73 s. These results were compared with those of Genetic Algorithm (GA)-SVM (97.10%, 19.50 s), Ant Colony Optimization (ACO)-SVM (94.20%, 228.72 s), and Artificial Neural Networks (ANN) (97.10%, 1.25 s). PCA-SVM had the highest accuracy and the shortest time. Thus, PCA-SVM as proposed herein could satisfy the requirement that an underwater robot rapidly and precisely identify sea cucumber objects in a real environment.