With the increasing incidence of autism spectrum disorders (ASDs), the physical and mental status of patients and the economic burden on their families have become a major concern. Individuals with ASD exhibit diverse motor characteristics. The combination of contactless visual sensors and computer vision (CV) technology allows the noncontact and long-term monitoring of these characteristics to extract valuable quantitative information. Therefore, in this paper we systematically review CV technology using visual sensors to obtain motion information from individuals with ASDs with the aim of exploring the application of noncontact perception systems in the diagnosis and treatment of autism. (1) A systematic review of publications indexed on Web of Science, PubMed, and Engineering Village and studies published from January 2015 to March 2023 was conducted. (2) Different publicly available datasets were reviewed to accelerate related research. (3) We summarized the above research results in tables and analyzed the research status, open challenges, and future perspectives. The results of this review show that the use of visual sensors to capture human movement information has wide application value in the diagnosis and treatment of autism.