A Method Of Fish Target Detection And Tracking Based On An Identification Sonar

Marine ranching belongs to modern inshore fishery,which has transformed the traditional fishing and farming into a modern industry based upon fish proliferation.Similar to farming on the land,marine ranching is often conducted in a specific underwater area to achieve a high quality yield of marine biological resources.It is constructive to protect the marine ecosystem and maintain a sustainable marine resources.In order to realize that,we must assess fish stock accurately during the marine ranch development.Nowadays,the fish stock assessment mainly relies on fishery-independent catch or statistic data during fishing.These approaches may leverage the acoustic method based upon echo counting or echo integration to roughly estimate the amount of fish in certain area,which is hard to realize target tracking,positioning,and synthetic resource assessments.In addition,Fishery-independent catch does harm to fish resources and provide an inadequate accuracy.The identification sonar is a multi-beam system that uses acoustic lens to form individual beams.It can obtain high-resolution images that are almost as clear as optical images in the dark or turbid waters to identify different targets.Compared with the traditional multi-beam system using time-delay arrays or digital beamforming technology for signal reception and transmission,the identification sonar uses acoustic lens to realize beamforming.It not only reduces the power consumption and volume of the system,but also improve the imaging efficiency significantly.Although identification sonars have been used to do fishery survey like individual target counting,body length measurement and target behavioral observation,there is still no established system of fish stock assessment.Moreover,a dual-frequency identification sonar(DIDSON)and an adaptive resolution imaging sonar(ARIS)were employed recently for the theoretical research and practical application of fishery resource assessment.In this paper,we propose a new method for high accuracy fish stock assessment based on an identification sonar and list the main research contents as follows:Chapter 2 presents the sonar image pre-processing and target extraction algorithms.Coordinate transformation was performed on the collected data to form original images,and the images were then restored to authentic underwater images by using the Inverse Distance Weighted(IDW)interpolation method.Linear transformation was applied to enhance sonar image quality.Recursive least squares(RLS)algorithm was designed to remove the speckle noise in the images.The background removing algorithm based on the signal strength model was designed to remove the image background.The thrice standard error principle was used to determine the threshold to extract targets.For the acoustic images with complex background,the edge detection based on the Sobel operator combined with morphological processing algorithm was proposed for target extraction.Chapter 3 describes the fish tracking algorithm.Due to the complexity of underwater environment,the multi-target tracking algorithms based on data correlation cost a great amount of calculation,especially for the dense targets with strong noise.It is unable to realize the real-time track of multiple targets.Hence,the joint probabilistic data association(JPDA)algorithm,multiple hypothesis tracking(MHT)algorithm,etc.based on the bayesian theory merging state estimation filtering were studied to form a reference.The Sequential Monte Carlo Probability Hypothesis Density(SMCPHD)filtering was applied to the target states,and the Auction track recognition algorithm was used to associate the filtered target states with the identified tracks to achieve the multi-target tracking.With the Visual Studio development tool combined with OpenGL,multi-target trajectories were built in three dimensional(3D)space according to the time sequence.Chapter 4 illustrates the method of evaluating fish amount based on an identification sonar.A method to estimate the fish abundance using volume density or areal density algorithm was proposed.First,the underwater data was collected with a sonar by investigation during navigation.The sonar image processing algorithms were then utilized to extract targets.The nearest neighbor algorithm combined with extended kalman filtering(NN-EKF)was designed for multi-target tracking and counting.The volume or area of swept water was accumulated to obtain the corresponding volume density or surface density.Combined with the water storage capacity or occupied area,the fish abundance was obtained.In order to verify the feasibility of this algorithm,a field experiment was carried out in Qingcaosha reservoir.The fish amount and volume density in the reservoir was accurately estimated using the proposed method.Chapter 5 describes the underwater positioning algorithms based on the identification sonar.Since the information about the target within the detection range acquired from the sonar only includes the distance and azimuth,underwater positioning based on collecting data from one sonar by investigation during navigation,and underwater positioning based on collecting data from two sonars placed in different positions were proposed.Through the field test,the feasibility and accuracy of the underwater positioning with a single sonar has been verified.The positioning errors and the effects due to different motion parameters and positions were analyzed.Combined with the data association algorithm,a field experiment on underwater positioning of fish in Dishui Lake has been conducted,and the three-dimensional(3D)tracking trajectories of targets were obtained.The vertical distribution of fish and features of swimming fish including velocity,direction,etc.in the reservoir were analyzed.Abundance estimation,multi-target tracking and distribution of fish has been studied in this paper.We proposed a new method of fishery resources assessment based on an identification sonar.The new method greatly improve the evaluation accuracy of fishery resources,and forms a technique cornerstone for marine ranching development.