Computer vision technology has carried out automatic recognition research, and extracted and analyzed the mathematical morphological characteristics of these insects from insect images, such as area, perimeter, horizontal axis length, vertical axis length, shape parameters, leaf shape, sphere, roundness, eccentricity, number of bright spots and so on. Through cluster analysis, the possibility and reliability of the application of these mathematical features in insect classification are discussed from the perspective of mathematical morphology. The results show that there are significant differences in roundness, eccentricity and overall average number of bright spots among Hemiptera, Coleoptera and Lepidoptera, which are suitable for classification and identification of various uses. Generally speaking, the reliability of each project is quasi-circular, eccentric, bright spot number, leaf-shaped, spherical and circular. Through the preliminary study of Scarabaeida, Scarabaeida, Noctuidae, Bombycidae and Papilionidae, the reliability of each item at the family level is less than circular area, circumference, horizontal axis length, sphericity, quasi-roundness and eccentric vertical axis. Circle is more suitable as the classification feature of family order, while shape parameters and bright spots are not suitable as the classification feature of family order. The research on stinkbug, Calypteridae, Noctuidae, Silkworm Moth, Papilionidae, Scarabaeidae and Cerambycidae shows that the reliability of the project features on the family level is roundness, eccentric area, circumference, horizontal axis length, vertical axis length of the ball, circular shape parameters and so on. Quasi-circle and eccentricity are more suitable as classification features at the family level. The study on the insect species of Silkworm moth, Papilionidae and Cerambycidae shows that the reliability of identifying insects is from big to small, which is in turn perimeter area, longitudinal axis, long axis and steep shape parameters, number of round bright spots, quasi-round, spherical and eccentric rate, and leaf shape. From the results, it can be seen that some characteristics are not significantly different in each classification order, such as leaf shape and shape parameters, which shows that these characteristics have strong * * * characteristics in insects and can represent the characteristics of the whole insect, and are not suitable for lower-level classification characteristics of Insecta. Some features are always significantly different in each classification level, such as perimeter and area, which shows that these features are suitable for the classification characteristics of insects in each classification level. Other functions are in between. In addition, the extraction software system of insect image processing and analysis developed according to visual programming language improves the automatic identification level of insect species one by one on the basis of overall accuracy. This paper embodies the following characteristics and innovations in the above research, and discusses the role of mathematical morphological characteristics in insect classification for the first time. This paper discusses the feasibility and reliability of applying computer vision technology to different insect classification purposes. The genetic relationship of insect groups with the same taxonomic order was studied from the perspective of mathematical morphology. The insect image processing and analysis system has been upgraded. Feature extraction of long horizontal axis and long vertical axis is added, and the types of automatic recognition one by one are added. Mathematical phosphorus symbol; Computer vision; Classification of poisoned insects; Automatic identification. I