A Sonar Sensing System for Crop Yield Estimation

Project Information

EchoYieldEstimation

Grant agreement ID: 101100903

DOI

10.3030/101100903

Project closed

EC signature date 22 September 2022

Start date 1 October 2022

End date 31 March 2024

Funded under

• European Research Council (ERC)

Total cost

No data

EU contribution

€ 150 000,00

Coordinated by

TEL AVIV UNIVERSITY
Israel

Periodic Reporting for period 1 - EchoYieldEstimation (A Sonar Sensing System for Crop Yield Estimation)

Reporting period: 2022-10-01 to 2024-03-31

Summary of the context and overall objectives of the project

With the growing world population, accelerated climate changes and global migration, nutritional security becoming one of the major challenges. This situation calls for fast development and utilisation of agricultural technology that will push agricultural yield production to its full potential.
Yield estimation in agricultural production is therefore very important. It is central for both agricultural planning as well as for successful agroeconomic management. It is currently almost solely done manually without assisting technology and it is therefore inaccurate and partial. To meet this challenge, we offer a unique sonar sensing system for estimating vegetable yield. Our technology, as demonstrated in a patent and in a recent peer reviewed paper, has unexampled advantages compared to other sensors. The sonar-based crop estimating sensor has been patented and initial tests show that it is capable of estimating pepper yields with 90% accuracy. In this project, we intend to expand those tests to other vegetables including tomatoes and cucumbers, aiming to provide farmers with a reliable tool for crop estimation. The project will provide the necessary steps for establishing a spin-off company that will exploit this invention and provide a product to the benefit of European and worldwide farmers. Our goal is to collect data from end-users in the EU, to improve the Artificial Intelligence algorithm that analyses the measurements of the sensor and to be ready at the end of the project for raising further investments to develop the full product. Our team is composed of sonar, robotics and business experts. We will collaborate with agri-robotics companies and agricultural stake-holders in Europe to fulfil the goals of the project.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

Since the commence of the project we commenced several activities according to the proposal’s workplan.
1. Data collection – In order to train the CNN (Convolutive Neural Network) one has to collect data from the field. We conducted a vast experiment in a cherry tomato growing greenhouse in Pdeya Israel. We were able to collect and label more then 3200 sonar scans (See Fig. 1).
2. Data processing - We used the data to train a Resnet 18 CNN. We achieved good results in the detection of tomatoes (90% detection) how ever the quantification of the amount of tomatoes was less accurate (about twice better than a random pick). 60% detection in 3 categories (1-7 tomatoes, 8-12 tomatoes, 13-37 tomatoes) and 38% detection in 5 categories (0 tomatoes, 1-5 tomatoes, 6-9 tomatoes, 10-13 tomatoes, 14-37 tomatoes).
3. Prototype development – We made an agreement with a swiss company HighDim GmbH for the development of a sonar sensing system prototype. The sensing system is based on a hybrid FPGA design that has custom modules for signal processing and the ability to use a CNN. The device will collect and process the data and transmit the results via Wi-Fi. HighDim finished designing the device and currently working on the hardware, software and firmware (See Fig. 2).
4. Business development – We designed a pitch-deck for the project and contacted several Business development agencies in Basel area for aiding in the development of Startup companies (Basel Area Swiss, Basel Land Development Support).
5. Collaboration with agricultural experts - We contacted Pepper growers in Italy and will conduct there the next phase of experiments. We prepared a questioner for data collection from farmers in order to estimate the potential market of the future product.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

The current results, as described in the former sub-section, are progressing according to the planned PoC. The development of a product prototype goes beyond the current stage of the research plan. We see many benefits in developing a standalone system that can perform local processing of the sonar data.
In order to fully develop this PoC into a Startup we will need further resources and we intend to apply to EU supporting schemes such as Eurostars in order to obtain them.
In addition to crop estimation, we identified additional application to our sonar technology. The system can be used as a complementary obstacle detection and classification for UAV/UGVs. We intend to submit a PoC project that will exploit the current system in these directions.
Our current technology is patented for agricultural applications. We intend to examine whether this IP can be extended to the new directions we discovered during the current PoC project.

Fig. 2: Functional diagram of the sonar Sensing system.

Fig. 1: A labelled sonar scan of tomatoes containing an image (left) and spectrograms (right)