Design of reconnaissance system based on STM32 small multi-rotor real-time image transmission

Jiawei Liu,Lu Zhang,Daosheng Shi

DOI: https://doi.org/10.1088/1742-6596/1237/3/032080

2019-06-01

Journal of Physics: Conference Series

Abstract:Abstract This project analyses the main technical points of aircraft design, makes a preliminary discussion on rotor dynamics and mechanical analysis of four-axis aircraft, analyses the particularity of aircraft structure, control flexibility and control principle, establishes three-dimensional rigid body coordinate axis for aircraft. Describes the relationship between flight attitude and coordinate axis. Chooses a multi-channel attitude fusion algorithm, anti-jamming data transmission mode, adjusts the relationship between frequency and rate of wireless data transmission in detail. At the same time, the principle and arithmetic of the PID controller are analyzed in detail, and it is applied to the speed control of DC motor of aircraft. Through experimental verification and test, a good stable control effect is obtained. The design and implementation of the flight control system and remote control software and hardware are completed, and the test and flight test of the whole aircraft are completed. The flight control stability is good. It has the characteristics of small size, fast response and long communication distance. It can fully meet the reconnaissance requirements of obtaining real-time high-definition images in various complex situations. The upper computer software can quickly decode image data, convert it to video output to ground personnel, and realize FPV reconnaissance.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the kiwifruit image recognition and occlusion problems during the kiwifruit picking process. Specifically, kiwifruits need to be intensively picked during the ripening period, and the current picking process mainly depends on manual labor. As rural labor resources become increasingly tight, the demand for agricultural machinery operation is becoming more and more urgent. In order to make the kiwifruit - picking robot reach a practical level, improving its picking efficiency has become the key. However, in terms of kiwifruit recognition, there are difficulties in image recognition and occlusion problems, and there are relatively few related studies. Therefore, this paper aims to achieve effective recognition and location of kiwifruits through image processing technology, providing a theoretical reference for subsequent kiwifruit picking.

The specific methods mentioned in the paper include:
1. **Image pre - processing**: Use the color model for basic image processing and perform enhancement processing in the frequency domain.
2. **Frequency - domain enhancement**: Apply filters, especially homomorphic filtering, in the frequency domain of the image to enhance the kiwifruit orchard image, highlight the characteristics of the kiwifruit trunk, and reduce the influence of background noise.
3. **Image segmentation**: Use the Otsu algorithm to segment the black trunk and combine it with the region - growing method for more accurate segmentation.
4. **Fruit location**: Use a binocular stereo vision system for fruit location to improve recognition accuracy.

Through these methods, the paper aims to solve the difficulties in kiwifruit image recognition and improve the recognition and location capabilities of the picking robot.