Abstract: Abstract: In order to easily collect images of agricultural diseases and pests and make real-time diagnose, a distributed mobile system was designed with a number of portable image collection devices and one image processing server. Each image collection device consisted of an embedded camera, a stretchable handheld pole and an Android phone equipped with an APP of control capability. The embedded camera was fixed on the end of the handheld pole via universal joints. The handheld pole could extend to about 2 m in length. The embedded camera was built upon a development board with iTOP 4412 and a set of modules, including WIFI control, camera control, image collection, H.264/JPEG coding, RTSP/RTP video transmission, GPS information collection and writing, file transfer, and image preprocessing, which were developed in Linux platform. The mobile application was developed in Android platform with a set of modules, including video streaming preview, network, image browsing and camera control. The image processing sever could receive the images from the image collection devices, record GPS information, diagnose agricultural diseases and pests, and return the diagnosis and control information of agricultural diseases and pests to the mobile phone. Among the components of this system, the handheld pole was used to deliver the embedded camera to some unreachable agricultural disease and pest area, and the mobile phone was used for browsing images and controlling camera to collect the disease and pest images. TCP/UDP protocols and SoftAp technique were used for data exchange among the embedded camera and the mobile phone, which could be independent from cable networks and wireless local area networks. HTTP protocols were used for data exchange and distributed computing among the image collection devices and the image processing server, which can reduce the mobile phone charges and the server overhead. To test the distributed mobile agricultural system, a diagnosis algorithm of damage levels of rice sheath blight was deployed to the image processing server. This algorithm mainly included image feature extraction, disease identification, disease area computation and damage level judgment. The images of rice sheath blight were collected using the image collection device in paddy fields located in China National Rice Research Institute in 2016. After the segmentation of disease area was finished in the embedded camera, the segmented images were uploaded to the image processing server. The diagnosis algorithm in the server was implemented to process these images and the diagnosis results and control information were returned to the mobile phone. The technicians or farmers could control the rice sheath blight based on the diagnosis suggestions. Our experiment indicated that the image collection device could easily collect the images of agricultural diseases and pests, especially on some places where hands and sight were hard to reach. The system could work effectively with low image browse latency, accurate camera control, reliable device-to-server communication and real-time image processing and diagnosis. The accurate rate of 83.5% was achieved to diagnose the damage levels of rice sheath blight based on our algorithm. Therefore, the system is expected to be widely applicable to agricultural disease and pest image collection and diagnosis.