AZFP論文：使用U-Net神經(jīng)網(wǎng)絡(luò)對(duì)AZFP多頻魚類浮游動(dòng)物剖面儀回波圖中的鯡魚、鮭魚和氣泡進(jìn)行分類
Abstract: Echosounders are used by fisheries and ocean observatories, but significant manual effort is required to classify species of interest within multifrequency echograms. This article investigates the use of modified U-Net convolutional neural networks for the pixel-level classification of biological and physical data in echogram images with accurate classification of herring and salmon schools, bubbles, and the sea surface. Data were collected on the coast of British Columbia, Canada, over two years using an Acoustic Zooplankton and Fish Profiler at four frequencies (67, 125, 200, 455 kHz). In addition, simulated data (water depth and solar elevation angle) provide spatial and temporal context to improve the quality of predictions. Redundancy is built into the model by using a tiling strategy during training and classification. During training, using a limited set of annotated data, translational augmentation encodes the U-Nets with robust features that enable applications for alternate deployment configurations (lower sampling rates or alternate water depths). To ensure broad applicability, these networks were trained to classify echograms with noise left intact. The best-performing model classifies herring, salmon, and bubble classes with F1 scores of 93.0%, 87.3%, and 86.5%, respectively. The results are accurate even when multiple classes are in close proximity, thus, retaining biological data that would otherwise be discarded due to surface bubble noise.
摘要：漁業(yè)和海洋觀測(cè)站使用回聲測(cè)深儀，但需要大量的人工努力才能在多頻回聲圖中對(duì)感興趣的物種進(jìn)行分類。本文研究了使用改進(jìn)的U-Net卷積神經(jīng)網(wǎng)絡(luò)對(duì)回聲圖像中的生物和物理數(shù)據(jù)進(jìn)行像素級(jí)分類，對(duì)鯡魚和鮭魚魚群、氣泡和海面進(jìn)行準(zhǔn)確分類。在加拿大不列顛哥倫比亞省海岸，使用聲學(xué)浮游動(dòng)物和魚類剖面儀在四個(gè)頻率（67、125、200、455 kHz）上收集了兩年多的數(shù)據(jù)。此外，模擬數(shù)據(jù)（水深和太陽(yáng)仰角）提供了空間和時(shí)間背景，以提高預(yù)測(cè)的質(zhì)量。在訓(xùn)練和分類過程中，通過使用平鋪策略將冗余構(gòu)建到模型中。在訓(xùn)練過程中，使用有限的一組帶注釋的數(shù)據(jù)，平移增強(qiáng)對(duì)U-Net進(jìn)行編碼，使其具有強(qiáng)大的功能，能夠應(yīng)用于替代部署配置（較低的采樣率或替代水深）。為了確保廣泛的適用性，這些網(wǎng)絡(luò)經(jīng)過訓(xùn)練，可以對(duì)噪聲保持不變的回聲圖進(jìn)行分類。表現(xiàn)最佳的模型對(duì)鯡魚、鮭魚和氣泡類進(jìn)行了分類，F1得分分別為93.0%、87.3%和86.5%。即使多個(gè)類別非常接近，結(jié)果也是準(zhǔn)確的，因此保留了由于表面氣泡噪聲而被丟棄的生物數(shù)據(jù)。
