基于场景自适应的船舶吃水值计算方法

Method for Calculation of Ship Draft Value Based on Scene Adaptation

在船舶吃水值检测工作中,传统的人工水尺读数方式往往会受场景限制,例如到港船只型号各不相同;船体水尺标志倾斜程度以及水尺本身刮擦、锈蚀和残缺的程度也不同;不同气候及时间段采集的光照条件差别很大,这些因素都将影响水尺图像识别读数的结果,导致货物质量计量不准确。针对以上问题,提出了一种基于场景自适应的船舶吃水精确检测的方法。对不同场景下采集的图像用亮度表达值,对图像的明暗进行判断分类,采用不同阈值的修正型伽马校正最大化图像信息;利用改进的语义分割算法对船体、水体和水尺字符进行分割;结合分割后的水体与水尺字符信息对图像进行矫正,对大M字符进行识别,并换算出吃水深度。通过实测表明,该算法能适应不同场景下的船舶水尺吃水值的计算,比例换算为像素级别,与人工读数对比,计算结果更贴近标准值,为水尺计重提供了更精准的数值。

In detection of ship draft value,the traditional manual staff gauge reading method is often limited by scenes,for example,the types of ships arriving at the port are different;the tilt degree of the ship s water gauge logo is different,and the degree of scraping,rusting and incompleteness of the water gauge itself is also different.The light conditions collected in different weather and time periods vary greatly.All these factors will affect the results of the image recognition reading of the water gauge,resulting in inaccurate measurement of the weight of the goods.In view of this,a method of accurate detection of ship draft based on scene adaptation is proposed.Firstly,the images collected in different scenes are classified with their brightness expression values,and the brightness and darkness of the images are judged,and the modified gamma correction with different thresholds is used to maximize the image information.Then,the improved semantic segmentation algorithm is used to segment the hull,water body and water gauge characters.Finally,the image is corrected by combining the segmented water body and water gauge character information,and the large M characters are identified and the draft is calculated.The actual measurement shows that the proposed algorithm can adapt to the calculation of the draft value of the ship s water gauge in different scenes,and the ratio is converted to the pixel level.Compared with the manual reading,the calculation result is closer to the standard value,which provides a more accurate value for the weight measurement by water gauge.