Anxiety caused by the lack of social skills is the biggest problem faced by children with ASD. Playing can improve children’s social skills and relieve anxiety. This study aimed to explore the influence of urban built environments on ASD children’s play behavior. The participants in this study were 57 parents of children with ASD. An anonymous questionnaire was used to collect and analyze data. At the same time, retrospective semi-structured interviews with 31 parents of ASD children were performed to validate the data analysis results. The results showed that lower residential building density, higher residential greening and higher destination accessibility have positive effects on ASD children’s play behavior. Excellent transportation facilities and high NDVI vegetation coverage have positive effects on the play behavior of children with ASD. More recreational facilities and recreational playability have positive impacts on the play behavior of children with ASD. The population density and number of children in the destination, as well as public facilities, influence the play behavior of children with ASD. The research results can promote the integration of this group into urban life and further promote social equity. At the same time, with the social needs of autistic children as an intermediary, it is expected to further explore new directions for sustainable urban development. Finally, combined with the research results, parents of ASD children are given proposals for how to increase the likelihood of children’s play behavior by choosing appropriate urban built environments.

To resolve the problems of deep convolutional neural network models with many parameters and high memory resource consumption, a lightweight network-based algorithm for building detection of Minnan folk light synthetic aperture radar (SAR) images is proposed. Firstly, based on the rotating target detection algorithm R-centernet, the Ghost ResNet network is constructed to reduce the number of model parameters by replacing the traditional convolution in the backbone network with Ghost convolution. Secondly, a channel attention module integrating width and height information is proposed to enhance the network’s ability to accurately locate salient regions in folk light images. Content-aware reassembly of features (CARAFE) up-sampling is used to replace the deconvolution module in the network to fully incorporate feature map information during up-sampling to improve target detection. Finally, the constructed dataset of rotated and annotated light and shadow SAR images is trained and tested using the improved R-centernet algorithm. The experimental results show that the improved algorithm improves the accuracy by 3.8%, the recall by 1.2% and the detection speed by 12 frames/second compared with the original R-centernet algorithm.

The use of urban space by vulnerable groups, especially ASD children with social anxiety, is an important part of building sustainable urban development. In this study, we focus on the play behavior of ASD children from the perspective of urban planning; then, we discussed how the urban built environmental factors affect the social behavior of ASD children. In this paper, 220 parents of ASD children were given questionnaires and 197 valid questionnaires were obtained after removing invalid ones. Stepwise regression was adopted to further accurately analyze the influence of each factor index in the built environment on children’s social behavior. The results showed that multiple urban built environment factors had significant influence on the social behaviors (observation, participation, retreat, and concealment) of children with autism at three stages: before departure, during journey, and arrived at destination. The purpose of study is to fully consider the use of urban space by ASD children when urban researchers or urban planners construct sustainable urban forms, formulate urban design guidelines, and implement old city renewal strategies.

For the problems of deep convolutional neural network model with many parameters and memory resource consumption, a lightweight network-based algorithm for building detection of Minnan folk light SAR images is proposed. Firstly, based on the rotating target detection algorithm R-centernet, the Ghost ResNet network is constructed to reduce the number of model parameters by replacing the traditional convolution in the backbone network with Ghost convolution. Secondly, a channel attention module integrating width and height information is proposed to enhance the network's ability to accurately locate salient regions in folk light images.CARAFE upsampling is used to replace the DCN module in the network to fully incorporate feature map information during upsampling to improve target detection. Finally, the constructed dataset of rotated and annotated light and shadow SAR images is trained and tested using the improved R-centernet algorithm. The experimental results show that the improved algorithm improves the accuracy by 3.8%, the recall by 1.2%, and the detection speed by 12 frames/second compared with the original R-centernet algorithm.