Ultrasound prostate image segmentation is a very challenging task.At present,traditional detection operators are dificult to identify the parts with insignificant gray contrast.Neural networks ignore the effect of low signal-to-noise ratio of ultrasonic images and consume a lot of computing power.In order to solve the above problems,this paper proposes an efficient segmentation algorithm of ultrasonic prostate images based on spatial constraints of solutions, which transforms the segmentation problem into a boundary point problem.Firstly,the normal vector operator is improved to improve its detection ability.Then the denoising autoencoder is used to overcome the space of the noise optimization solution according to the shape constraint.Finally,iteration operator is introduced to limit the range of solution to a very small area to achieve accurate segmentation.Experiments show that the IoU of this model is 94.4% and the DSC value is about 97.05%,which is higher than the current popular neural network algorithm and lighter.