Accurate assessment of the number of organisms in aquaculture ponds is an important requirement in aquaculture. Fish counting methods based on machine vision and near-infrared light can quickly and accurately estimate the biomass in the culture pond under better light environmental conditions, but it is affected by fish overlap, water turbidity, light, bubbles, and other factors. Fish counting method based on sonar camera and echo sounder can effectively investigate the amount of biological resources in the ocean or lake, but it is limited by the area and water depth of the aquaculture pond in the industrial aquaculture. A time domain interval selection method based on coefficient of variation was proposed to solve the problem of error in the application of ultrasonic detection in aquaculture ponds with low depth. The ultrasonic transducer with an emission frequency of 40 kHz was used to count the fish in aquaculture ponds in a cylinder with a volume of 0.115 m³. Aiming at the problem of error in the estimation results of ordinary multiple scattering theory, a time-domain interval selection method based on coefficient of variation was proposed. The method was used to estimate the number of 10-40 fishes in the cylinder without and with aeration. The results showed that the MAE was 0.924 and 1.769 fish, RMSE was 1.148 and 2.054, and the CV between estimated and actual values were 0.995 and 0.983, respectively. The results showed that the error was highly correlated with the selection of the time domain interval, and the variation coefficient and the three evaluation indexes kept consistent with the change trend of the time domain interval. By selecting the appropriate time domain interval and increasing the number of measurements, the accuracy of the counting results could be effectively improved. The research showed that the accuracy of aquatic organism counting could be effectively improved after increasing the number of samples although there were bubbles in the environment after oxygenation, which leads to a decrease in the accuracy of estimation. The experiment proved that the method was still applicable in the scene with oxygenation demand. This study provides a new path for assessing fish density in low-depth and small-area aquaculture ponds by ultrasonic technology.