Abstract We have experienced that fish detection in sonar-data from horizontally aligned transducers in shallow water can be difficult. Studying recorded material revealed that echoes from fish often took on high ping to ping variations in variables like pulse length, intensity, shape and phase. Common single-echo detection methods based on a set of echo criteria tended to overlook echoes from fish. At the same time echoes from unwanted targets and fluctuations in the background reverberation was accepted. Hence, tracking and fish counting became difficult. To overcome these problems, we have developed a detection method based on 2-dimensional low-pass filters. By adjusting the cut-off frequency in the time and range domain, noise can be removed, ping to ping variation in signals from fish reduced, and local background reverberation level can be found. By subtracting the echo intensity in the filtered echograms, traces from fish can be detected. A new method for single target detection Helge Balk & Torfinn Lindem University of Oslo, Department of Physics, Pb 1048 Blindern, 0316 Oslo, Norway. Fax: 2856422. Email: [email protected], [email protected] I Problem Split-beam transducer
River cross section Echo sounder Amp-echogram F Single echo detector (SED) Amp echogram presenting the output from the echosounders amplitude detector (Threshold 100dB, 8dB per color step) II Missing echo detections from targets. SED-echogram SED-echogram + Noise echoes surrounding echoes from targets.
= Tracking and classification Debris? Fish? Tracking and classification becomes difficult Stones? See poster 89 for reasons to why the SED tend to fail! We need a better detector! Developing a new detector III Cross-filter detector(CFD) TS Ran
g g am Pin ogr e (m ) (F p m )A h c e - TS versus ping at the range of the fish echoes seen above Specter with a fish in the beam Specter without fish in the beam
a) Fish and background reverberation are overlaid with intensity fluctuations. b) The frequency specter of the echo signal in the ping domain differ with and without a fish in the beam. c) The main echo energy from the fish are located in the lower end of the frequency specter d) In the range domain little or no no difference can be seen in the frequency specter with and with out fish. IV TS Ran ge (m) g Pin
Q1 TS Ran g e (m ) g Pin Q2 1) We can generate an echogram (Q1) with improve fish tracks and reduced noise by applying a lowpass filter in the ping domain. 2) We can generate an echogram(Q2) containing the slowly changing background level by applying a low-pass filter in the range domain 3) Applying each individual sample in Q2 as a threshold in Q1 detects clusters of samples (Q3) from fish and other solid targets. Q1 (m, n)
m2 F ( p, r ) H1(m p 2, n r 1) p m 2 r n 1 m Q2 (m, n) TS n 1 n 25 F ( p, r ) H 2 ( m p, n r 25) p m r n 25 Ran ge
(m) g Pin Q3 F (m, n), Q3 (m, n) 0, Q1 (m, n) Q2 (m, n) 6dB Q1 (m, n) Q2 (m, n) 6dB V noise fish debris 4) Cluster classification Detections are represented as clusters of intensity samples. Reminiscence of noise, passing
targets and stationary targets are detected. Counting samples along the rim of each cluster gives the perimeter. General differences in the perimeter between the three main target types enable us to remove clusters of samples from noise and stationary targets such as stones. Stone Example of detected clusters. Typical rims for noise, fish debris and stone clusters. VI 5) Angular position of is extracted from the echo sounders phase detector 6) All detected echoes are applied in the tracking. However, echoes that would have been removed by a traditional single echo detector are marked as low quality echoes in
order to be removed from sensitive estimation such as targets size and velocity. Phase track from a passing fish. Angle detection is based on the output from the echo sounders phase detector. F Filter 1 Filter 2 Cluster 4 classification Adaptive 3 Q1 thresholding Q2 Angle 5 detection Quality 6 estimation
Q3 The elements of the Cross-filter detector. F represent the Amp-echogram from the echosounder Tracking and classification VII Test data Amp-echogram recorded in River Tana (left) and in the Rimov reservoir (right). No threshold applied. 8 dB per color step. Print from Sonar5 post processing program. Data recorded with a Simrad EY500 echo-sounder equipped with a horizontally aligned ES120-4 transducer. The Tana echogram showes echoes from upstream migrating Atlantic Salmon (Salmo Salar L.), drifting debris and stones. The Rimov echogram showes echoes from spawning Brem (Abramis brama). VIII Detection results from River Tana Amp-echogram. SED-echogram
(Threshold=-50dB) (Threshold=-50dB) CFDechogram. (Adaptive threshold) Output from EY500s Amplitude detector. Output from EY500s built-in single echo detector. Output from the Cross-filter detector. We thank the County governor in Finnmark (No) represented by Kjell Moen, Simrad AS, represented by Frank R. Knudsen, and the Finnish Fish and Game Research Institute (RKTL) for all support with the fieldwork in Tana. IX Detection results from Rimov (Cz) Amp-echogram.
SED-echogram (Threshold=-50dB) (Threshold=-50dB) CFDechogram. (Adaptive threshold) Output from EY500s Amplitude detector. Output from EY500s built-in single echo detector. Output from the Cross-filter detector. We thank Jan Kubecka and Marie Prchalova from the hydrobiological institute Academy of Science of the Czech Rep. Cheske Budejovice for lending us this data set. Setting up the camp in Tana 1999. Equipment and accomodation carravans, transducer, parts of the tripod, weather station on top of the pole, generator house in the background.
Simrad EY500 echo sounder seen inside the caravan. (Blue box to the left) 120 kHz 0.3ms pulses was transmitted with 63w, with a ping rate of 5 ping per sec. Mounting the simrad ES120 4x10deg transducer on the homemade the remote controlled tripod. Tripod with transducer tilt and direction indicators. Tilt=-2deg, direction = near normal to the water current Looking upstream from the site in Tana. Rimov reservoir (Cz) X Conclusion Tests presented here and in the accompanying article demonstrates that the Cross-filter detector (CFD) is better suited for fish detection than the traditional single echo detector.
The CFD detected significantly more echoes from the targets. The CFD significantly reduced the number of noise echo detections. The output from the CFD was better suited for tracking than the output from the SED.
